# What I Wish I Knew When Learning Haskell

Version 2.3

% Stephen Diehl % March 2016

Stephen Diehl (@smdiehl )

This is the fourth draft of this document.

#### License

This code and text are dedicated to the public domain. You can copy, modify, distribute and perform the work, even for commercial purposes, all without asking permission.

You may copy and paste any code here verbatim into your codebase, wiki, blog, book or Haskell musical production as you see fit. The Markdown and Haskell source is available on Github. Pull requests are always accepted for changes and additional content. This is a living document.

#### Changelog

**2.4**

**2.3**

- Stack
- Stackage
- ghcid
- Nix (Removed)
- Aeson (Updated)
- Language Extensions (Updated)
- Type Holes (Updated)
- Partial Type Signatures
- Pattern Synonyms (Updated)
- Unboxed Types ( Updated )
- Vim Integration ( Updated )
- Emacs Integration ( Updated )
- Strict Language Extension
- Injective Type Families
- Custom Type Errors
- Language Comparisons
- Recursive Do
- Applicative Do
- LiquidHaskell
- Cpp
- Minimal Pragma
- Typeclass Extensions
- ExtendedDefaultRules
- mmorph
- integer-gmp
- Static Pointers
- spoon
- monad-control
- monad-base
- postgresql-simple
- hedis
- happy/alex
- configurator
- string-conv
- resource-pool
- resourcet
- optparse-applicative
- hastache
- silently
- Mulitiline Strings
- git-embed
- Coercible
- -fdefer-type-errors

**2.2**

Sections that have had been added or seen large changes:

- Irrefutable Patterns
- Hackage
- Exhaustiveness
- Stacktraces
- Laziness
- Skolem Capture
- Foreign Function Pointers
- Attoparsec Parser
- Inline Cmm
- PrimMonad
- Specialization
- unbound-generics
- Editor Integration
- EKG
- Nix
- Haddock
- Corecursion
- Category
- Arrows
- Bifunctors
- ExceptT
- hint / mueval
- Roles
- Higher Kinds
- Kind Polymorphism
- Numeric Tower
- SAT Solvers
- Graph
- Sparks
- Threadscope
- Generic Parsers
- GHC Block Diagram
- GHC Debug Flags
- Core
- Inliner
- Unboxed Types
- Runtime Memory Representation
- ghc-heapview
- STG
- Worker/Wrapper
- Z-Encoding
- Cmm
- Runtime Optimizations
- RTS Profiling
- Algebraic Relations

# Basics

## Cabal

Historically Cabal had a component known as `cabal-install`

that has largely been replaced by Stack. The following use of Cabal sandboxes is left for historical reasons and can often be replaced by modern tools.

Cabal is the build system for Haskell.

For example, to install the parsec package to your system from Hackage, the upstream source of Haskell packages, invoke the `install`

command:

```
$ cabal install parsec # latest version
$ cabal install parsec==3.1.5 # exact version
```

The usual build invocation for Haskell packages is the following:

```
$ cabal get parsec # fetch source
$ cd parsec-3.1.5
$ cabal configure
$ cabal build
$ cabal install
```

To update the package index from Hackage, run:

`$ cabal update`

To start a new Haskell project, run:

```
$ cabal init
$ cabal configure
```

A `.cabal`

file will be created with the configuration options for our new project.

The latest feature of `cabal`

is the addition of Sandboxes, ( in cabal > 1.18 ) which are self contained environments of Haskell packages separate from the global package index stored in the `./.cabal-sandbox`

of our project's root. To create a new `sandbox`

for our `cabal`

project, run:

`$ cabal sandbox init`

Additionally, the `sandbox`

can be torn down:

`$ cabal sandbox delete`

When in the working directory of a project with a `sandbox`

that has a configuration already set up, invoking `cabal`

commands alters the behaviour of cabal itself. For instance, the `cabal install`

command will alter only the install to the local package index, not the global configuration.

To install the dependencies from the `.cabal`

file into the newly created `sandbox`

, run:

`$ cabal install --only-dependencies`

Dependencies can also be built in parallel by passing `-j<n>`

where `n`

is the number of concurrent builds.

`$ cabal install -j4 --only-dependencies`

Let's look at an example `.cabal`

file. There are two main entry points that any package may provide: a `library`

and an `executable`

. Multiple executables can be defined, but only one library. In addition, there is a special form of executable entry point `Test-Suite`

, which defines an interface for invoking unit tests from `cabal`

.

For a library, the `exposed-modules`

field in the `.cabal`

file indicates which modules within the package structure will be publicly visible when the package is installed. These modules are the user-facing APIs that we wish to expose to downstream consumers.

For an executable, the `main-is`

field indicates the module that exports the `main`

function running the executable logic of the application. Every module in the package must be listed in one of `other-modules`

, `exposed-modules`

or `main-is`

fields.

```
name: mylibrary
version: 0.1
cabal-version: >= 1.10
author: Paul Atreides
license: MIT
license-file: LICENSE
synopsis: The code must flow.
category: Math
tested-with: GHC
build-type: Simple
library
exposed-modules:
Library.ExampleModule1
Library.ExampleModule2
build-depends:
base >= 4 && < 5
default-language: Haskell2010
ghc-options: -O2 -Wall -fwarn-tabs
executable "example"
build-depends:
base >= 4 && < 5,
mylibrary == 0.1
default-language: Haskell2010
main-is: Main.hs
Test-Suite test
type: exitcode-stdio-1.0
main-is: Test.hs
default-language: Haskell2010
build-depends:
base >= 4 && < 5,
mylibrary == 0.1
```

To run an "executable" for a project under the `cabal`

`sandbox`

:

```
$ cabal run
$ cabal run <name> # when there are several executables in a project
```

To load the "library" into a GHCi shell under `cabal`

`sandbox`

:

```
$ cabal repl
$ cabal repl <name>
```

The `<name>`

metavariable is either one of the executable or library declarations in the `.cabal`

file and can optionally be disambiguated by the prefix `exe:<name>`

or `lib:<name>`

respectively.

To build the package locally into the `./dist/build`

folder, execute the `build`

command:

`$ cabal build`

To run the tests, our package must itself be reconfigured with the `--enable-tests`

and the `build-depends`

options. The `Test-Suite`

must be installed manually, if not already present.

```
$ cabal install --only-dependencies --enable-tests
$ cabal configure --enable-tests
$ cabal test
$ cabal test <name>
```

Moreover, arbitrary shell commands can be invoked with the GHC environmental variables set up for the `sandbox`

. Quite common is to invoke a new shell with this command such that the `ghc`

and `ghci`

commands use the `sandbox`

. ( They don't by default, which is a common source of frustration. ).

```
$ cabal exec
$ cabal exec sh # launch a shell with GHC sandbox path set.
```

The haddock documentation can be generated for the local project by executing the `haddock`

command. The documentation will be built to the `./dist`

folder.

`$ cabal haddock`

When we're finally ready to upload to Hackage ( presuming we have a Hackage account set up ), then we can build the tarball and upload with the following commands:

```
$ cabal sdist
$ cabal upload dist/mylibrary-0.1.tar.gz
```

Sometimes you'd also like to add a library from a local project into a `sandbox`

. In this case, run the `add-source`

command to bring the library into the `sandbox`

from a local directory:

`$ cabal sandbox add-source /path/to/project`

The current state of a `sandbox`

can be frozen with all current package constraints enumerated:

`$ cabal freeze`

This will create a file `cabal.config`

with the constraint set.

```
constraints: mtl ==2.2.1,
text ==1.1.1.3,
transformers ==0.4.1.0
```

Using the `cabal repl`

and `cabal run`

commands is preferable, but sometimes we'd like to manually perform their equivalents at the shell. Several useful aliases rely on shell directory expansion to find the package database in the current working directory and launch GHC with the appropriate flags:

```
alias ghc-sandbox="ghc -no-user-package-db -package-db .cabal-sandbox/*-packages.conf.d"
alias ghci-sandbox="ghci -no-user-package-db -package-db .cabal-sandbox/*-packages.conf.d"
alias runhaskell-sandbox="runhaskell -no-user-package-db -package-db .cabal-sandbox/*-packages.conf.d"
```

There is also a zsh script to show the sandbox status of the current working directory in our shell:

```
function cabal_sandbox_info() {
cabal_files=(*.cabal(N))
if [ $#cabal_files -gt 0 ]; then
if [ -f cabal.sandbox.config ]; then
echo "%{$fg[green]%}sandboxed%{$reset_color%}"
else
echo "%{$fg[red]%}not sandboxed%{$reset_color%}"
fi
fi
}
RPROMPT="\$(cabal_sandbox_info) $RPROMPT"
```

The `cabal`

configuration is stored in `$HOME/.cabal/config`

and contains various options including credential information for Hackage upload. One addition to configuration is to completely disallow the installation of packages outside of sandboxes to prevent accidental collisions.

```
-- Don't allow global install of packages.
require-sandbox: True
```

A library can also be compiled with runtime profiling information enabled. More on this is discussed in the section on Concurrency and Profiling.

`library-profiling: True`

Another common flag to enable is `documentation`

which forces the local build of Haddock documentation, which can be useful for offline reference. On a Linux filesystem these are built to the `/usr/share/doc/ghc-doc/html/libraries/`

directory.

`documentation: True`

If GHC is currently installed, the documentation for the Prelude and Base libraries should be available at this local link:

/usr/share/doc/ghc-doc/html/libraries/index.html

See:

## Stack

Stack is a new approach to Haskell package structure that emerged in 2015. Instead of using a rolling build like `cabal-install`

, `stack`

breaks up sets of packages into release blocks that guarantee internal compatibility between sets of packages. The package solver for `stack`

uses a different, more robust strategy for resolving dependencies than `cabal-install`

has historically used.

Contrary to much misinformation, **Stack does not replace Cabal as the build system** and uses it under the hood. Stack simply streamlines integration with third-party packages and the resolution of their dependencies.

#### Install

To install `stack`

on Ubuntu Linux, run:

```
sudo apt-key adv --keyserver keyserver.ubuntu.com --recv-keys 575159689BEFB442 # get fp complete key
echo 'deb http://download.fpcomplete.com/ubuntu trusty main'|sudo tee /etc/apt/sources.list.d/fpco.list # add appropriate source repo
sudo apt-get update && sudo apt-get install stack -y
```

For other operating systems, see the official install directions.

#### Usage

Once `stack`

is installed, it is possible to setup a build environment on top of your existing project's `cabal`

file by running:

`stack init`

An example `stack.yaml`

file for GHC 7.10.3 would look like:

```
resolver: lts-6.4
flags: {}
extra-package-dbs: []
packages: []
extra-deps: []
```

Most of the common libraries used in everyday development are already in the Stackage repository. The `extra-deps`

field can be used to add Hackage dependencies that are not in the Stackage repository. They are specified by the package and the version key. For instance, the `zenc`

package could be added to the `stack`

build:

```
extra-deps:
- zenc-0.1.1
```

The `stack`

command can be used to install packages and executables into either the current build environment or the global environment. For example, the following command installs the executable for `hlint`

, a popular linting tool for Haskell, and places it in the PATH:

`$ stack install hlint`

To check the set of dependencies, run:

`$ stack list-dependencies`

Just as with `cabal`

, the build and debug process can be orchestrated using `stack`

commands:

```
$ stack build # Build a cabal target
$ stack repl # Launch ghci
$ stack ghc # Invoke the standalone compiler in stack environment
$ stack exec bash # Execute a shell command with the stack GHC environment variables
$ stack build --file-watch # Build on every filesystem change
```

To visualize the dependency graph, use the dot command piped first into graphviz, then piped again into your favorite image viewer:

`$ stack dot --external | dot -Tpng | feh -`

## Flags

Enabling GHC compiler flags grants the user more control in detecting common code errors. The most frequently used flags are:

Flag | Description |
---|---|

-fwarn-tabs | Emit warnings of tabs instead of spaces in the source code |

-fwarn-unused-imports | Warn about libraries imported without being used |

-fwarn-name-shadowing | Warn on duplicate names in nested bindings |

-fwarn-incomplete-uni-patterns | Emit warnings for incomplete patterns in lambdas or pattern bindings |

-fwarn-incomplete-patterns | Warn on non-exhaustive patterns |

-fwarn-overlapping-patterns | Warn on pattern matching branches that overlap |

-fwarn-incomplete-record-updates | Warn when records are not instantiated with all fields |

-fdefer-type-errors | Turn type errors into warnings |

-fwarn-missing-signatures | Warn about toplevel missing type signatures |

-fwarn-monomorphism-restriction | Warn when the monomorphism restriction is applied implicitly |

-fwarn-orphans | Warn on orphan typeclass instances |

-fforce-recomp | Force recompilation regardless of timestamp |

-fno-code | Omit code generation, just parse and typecheck |

-fobject-code | Generate object code |

Like most compilers, GHC takes the `-Wall`

flag to enable all warnings. However, a few of the enabled warnings are highly verbose. For example, `-fwarn-unused-do-bind`

and `-fwarn-unused-matches`

typically would not correspond to errors or failures.

Any of these flags can be added to the `ghc-options`

section of a project's `.cabal`

file. For example:

```
library mylib
ghc-options:
-fwarn-tabs
-fwarn-unused-imports
-fwarn-missing-signatures
-fwarn-name-shadowing
-fwarn-incomplete-patterns
```

The flags described above are simply the most useful. See the official reference for the complete set of GHC's supported flags.

For information on debugging GHC internals, see the commentary on GHC internals.

## Hackage

Hackage is the upstream source of Free and/or Open Source Haskell packages. With Haskell's continuing evolution, Hackage has become many things to developers, but there seem to be two dominant philosophies of uploaded libraries.

**Reusable Code / Building Blocks**

In the first philosophy, libraries exist as reliable, community-supported building blocks for constructing higher level functionality on top of a common, stable edifice. In development communities where this method is the dominant philosophy, the author(s) of libraries have written them as a means of packaging up their understanding of a problem domain so that others can build on their understanding and expertise.

**A Staging Area / Request for Comments**

In contrast to the previous method of packaging, a common philosophy in the Haskell community is that Hackage is a place to upload experimental libraries as a means of getting community feedback and making the code publicly available. Library author(s) often rationalize putting these kind of libraries up undocumented, often without indication of what the library actually does, by simply stating that they intend to tear the code down and rewrite it later. This approach unfortunately means a lot of Hackage namespace has become polluted with dead-end, bit-rotting code. Sometimes packages are also uploaded purely for internal use within an organisation, to accompany a paper, or just to integrate with the `cabal`

build system. These packages are often left undocumented as well.

For developers coming to Haskell from other language ecosystems that favor the former philsophy (e.g., Python, Javascript, Ruby), seeing *thousands of libraries without the slightest hint of documentation or description of purpose* can be unnerving. It is an open question whether the current cultural state of Hackage is sustainable in light of these philsophical differences.

Needless to say, there is a lot of very low-quality Haskell code and documentation out there today, so being conservative in library assessment is a necessary skill. That said, there are also quite a few phenomenal libraries on Hackage that are highly curated by many people.

As a general rule, if the Haddock documentation for the library does not have a **minimal worked example**, it is usually safe to assume that it is an RFC-style library and probably should be avoided in production-grade code.

Similarly, if the library **predates the text library** (released circa 2007), it probably should be avoided in production code. The way we write Haskell has changed drastically since the early days.

## GHCi

GHCi is the interactive shell for the GHC compiler. GHCi is where we will spend most of our time in every day development.

Command | Shortcut | Action |
---|---|---|

`:reload` |
`:r` |
Code reload |

`:type` |
`:t` |
Type inspection |

`:kind` |
`:k` |
Kind inspection |

`:info` |
`:i` |
Information |

`:print` |
`:p` |
Print the expression |

`:edit` |
`:e` |
Load file in system editor |

`:load` |
`:l` |
Set the active Main module in the REPL |

`:add` |
`:ad` |
Load a file into the REPL namespace |

`:browse` |
`:bro` |
Browse all available symbols in the REPL namespace |

The introspection commands are an essential part of debugging and interacting with Haskell code:

```
λ: :type 3
3 :: Num a => a
```

```
λ: :kind Either
Either :: * -> * -> *
```

```
λ: :info Functor
class Functor f where
fmap :: (a -> b) -> f a -> f b
(<$) :: a -> f b -> f a
-- Defined in `GHC.Base'
...
```

```
λ: :i (:)
data [] a = ... | a : [a] -- Defined in `GHC.Types'
infixr 5 :
```

Querying the current state of the global environment in the shell is also possible. For example, to view module-level bindings and types in GHCi, run:

```
λ: :browse
λ: :show bindings
```

Examining module-level imports, execute:

```
λ: :show imports
import Prelude -- implicit
import Data.Eq
import Control.Monad
```

To see compiler-level flags and pragmas, use:

```
λ: :set
options currently set: none.
base language is: Haskell2010
with the following modifiers:
-XNoDatatypeContexts
-XNondecreasingIndentation
GHCi-specific dynamic flag settings:
other dynamic, non-language, flag settings:
-fimplicit-import-qualified
warning settings:
λ: :showi language
base language is: Haskell2010
with the following modifiers:
-XNoDatatypeContexts
-XNondecreasingIndentation
-XExtendedDefaultRules
```

Language extensions and compiler pragmas can be set at the prompt. See the Flag Reference for the vast collection of compiler flag options.

Several commands for the interactive shell have shortcuts:

Function | |
---|---|

`+t` |
Show types of evaluated expressions |

`+s` |
Show timing and memory usage |

`+m` |
Enable multi-line expression delimited by `:{` and `:}` . |

```
λ: :set +t
λ: []
[]
it :: [a]
```

```
λ: :set +s
λ: foldr (+) 0 [1..25]
325
it :: Prelude.Integer
(0.02 secs, 4900952 bytes)
```

```
λ: :{
λ:| let foo = do
λ:| putStrLn "hello ghci"
λ:| :}
λ: foo
"hello ghci"
```

The configuration for the GHCi shell can be customized globally by defining a `ghci.conf`

in `$HOME/.ghc/`

or in the current working directory as `./.ghci.conf`

.

For example, we can add a command to use the Hoogle type search from within GHCi. First, install `hoogle`

:

`cabal install hoogle`

Then, we can enable the search functionality by adding a command to our `ghci.conf`

:

```
:set prompt "λ: "
:def hlint const . return $ ":! hlint \"src\""
:def hoogle \s -> return $ ":! hoogle --count=15 \"" ++ s ++ "\""
```

```
λ: :hoogle (a -> b) -> f a -> f b
Data.Traversable fmapDefault :: Traversable t => (a -> b) -> t a -> t b
Prelude fmap :: Functor f => (a -> b) -> f a -> f b
```

For reasons of sexiness, it is desirable to set your GHC prompt to a `λ`

or a `λΠ`

. Only if you're into that lifestyle, though.

```
:set prompt "λ: "
:set prompt "ΠΣ: "
```

#### GHCi Performance

For large projects, GHCi with the default flags can use quite a bit of memory and take a long time to compile. To speed compilation by keeping artificats for compiled modules around, we can enable object code compilation instead of bytecode.

`:set -fobject-code`

Enabling object code compliation may complicate type inference, since type information provided to the shell can sometimes be less informative than source-loaded code. This under specificity can result in breakage with some langauge extensions. In that case, you can temporarily reenable bytecode compilation on a per module basis with the `-fbyte-code`

flag.

```
:set -fbyte-code
:load MyModule.hs
```

If you all you need is to typecheck your code in the interactive shell, then disabling code generation entirely makes reloading code almost instantaneous:

`:set -fno-code`

## Editor Integration

Haskell has a variety of editor tools that can be used to provide interactive development feedback and functionality such as querying types of subexpressions, linting, type checking, and code completion.

Several prepackaged setups exist to expedite the process of setting up many of the programmer editors for Haskell development. In particular, using `ghc-mod`

can remarkably improve programmer efficiency and productivity because the project attempts to implement features common to modern IDEs.

**Vim**

**Emacs**

**Atom**

## Bottoms

The bottom is a singular value that inhabits every type. When this value is evaluated, the semantics of Haskell no longer yield a meaningful value. In other words, further operations on the value cannot be defined in Haskell. A bottom value is usually written as the symbol ⊥, ( i.e. the compiler flipping you off ). Several ways exist to express bottoms in Haskell code.

For instance, `undefined`

is an easily called example of a bottom value. This function has type `a`

but lacks any type constraints in its type signature. Thus, `undefined`

is able to stand in for any type in a function body, allowing type checking to succeed, even if the function is incomplete or lacking a definition entirely. The `undefined`

function is extremely practical for debugging or to accommodate writing incomplete programs.

```
undefined :: a
mean :: Num a => Vector a -> a
mean nums = (total / count) where -- Partially defined function
total = undefined
count = undefined
addThreeNums :: Num a => a -> a -> a -> a
addThreeNums n m j = undefined -- No function body declared at all
f :: a -> Complicated Type
f = undefined -- Write tomorrow, typecheck today!
-- Arbitrarily complicated types
-- welcome!
```

Another example of a bottom value comes from the evaluation of the `error`

function, which takes a `String`

and returns something that can be of any type. This property is quite similar to `undefined`

, which also can also stand in for any type.

Calling `error`

in a function causes the compiler to throw an exception, halt the program, and print the specified error message. In the `divByY`

function below, passing the function `0`

as the divisor results in this function results in such an exception.

```
error :: String -> a -- Takes an error message of type
-- String and returns whatever type
-- is needed
```

```
-- Annotated code that features use of the error function.
divByY:: (Num a, Eq a, Fractional a) => a -> a -> a
divByY _ 0 = error "Divide by zero error" -- Dividing by 0 causes an error
divByY dividend divisor = dividend / divisor -- Handles defined division
```

A third type way to express a bottom is with an infinitely looping term:

```
f :: a
f = let x = x in x
```

Examples of actual Haskell code that use this looping syntax live in the source code of the GHC.Prim module. These bottoms exist because the operations cannot be defined in native Haskell. Such operations are baked into the compiler at a very low level. However, this module exists so that Haddock can generate documentation for these primitive operations, while the looping syntax serves as a placeholder for the actual implementation of the primops.

Perhaps the most common introduction to bottoms is writing a partial function that does not have exhaustive pattern matching defined. For example, the following code has non-exhaustive pattern matching because the `case`

expression, lacks a definition of what to do with a `B`

:

```
data F = A | B
case x of
A -> ()
```

The code snippet above is translated into the following GHC Core output. The compiler inserts an exception to account for the non-exhaustive patterns:

```
case x of _ {
A -> ();
B -> patError "<interactive>:3:11-31|case"
}
```

GHC can be made more vocal about incomplete patterns using the `-fwarn-incomplete-patterns`

and `-fwarn-incomplete-uni-patterns`

flags.

A similar situation can arise with records. Although constructing a record with missing fields is rarely useful, it is still possible.

```
data Foo = Foo { example1 :: Int }
f = Foo {} -- Record defined with a missing field
```

When the developer omits a field's definition, the compiler inserts an exception in the GHC Core representation:

`Foo (recConError "<interactive>:4:9-12|a")`

Fortunately, GHC will warn us by default about missing record fields.

Bottoms are used extensively throughout the Prelude, although this fact may not be immediately apparent. The reasons for including bottoms are either practical or historical.

The canonical example is the `head`

function which has type `[a] -> a`

. This function could not be well-typed without the bottom.

```
import GHC.Err
import Prelude hiding (head, (!!), undefined)
-- degenerate functions
undefined :: a
undefined = error "Prelude.undefined"
head :: [a] -> a
head (x:_) = x
head [] = error "Prelude.head: empty list"
(!!) :: [a] -> Int -> a
xs !! n | n < 0 = error "Prelude.!!: negative index"
[] !! _ = error "Prelude.!!: index too large"
(x:_) !! 0 = x
(_:xs) !! n = xs !! (n-1)
```

It is rare to see these partial functions thrown around carelessly in production code because they cause the program to halt. The preferred method for handling exceptions is to combine the use of safe variants provided in `Data.Maybe`

with the usual fold functions `maybe`

and `either`

.

Another method is to use pattern matching, as shown in `listToMaybe`

, a safer version of `head`

described below:

```
listToMaybe :: [a] -> Maybe a
listToMaybe [] = Nothing -- An empty list returns Nothing
listToMaybe (a:_) = Just a -- A non-empty list returns the first element
-- wrapped in the Just context.
```

Invoking a bottom defined in terms of `error`

typically will not generate any position information. However, `assert`

, which is used to provide assertions, can be short-circuited to generate position information in the place of either `undefined`

or `error`

calls.

```
import GHC.Base
foo :: a
foo = undefined
-- *** Exception: Prelude.undefined
bar :: a
bar = assert False undefined
-- *** Exception: src/fail.hs:8:7-12: Assertion failed
```

See: Avoiding Partial Functions

## Exhaustiveness

Pattern matching in Haskell allows for the possibility of non-exhaustive patterns. For example, passing Nothing to `unsafe`

will cause the program to crash at runtime. However, this function is an otherwise valid, type-checked program.

```
unsafe :: Num a => Maybe a -> Maybe a
unsafe (Just x) = Just $ x + 1
```

Since `unsafe`

takes a `Maybe a`

value as its argument, two possible values are valid input: `Nothing`

and `Just a`

. Since the case of a `Nothing`

was not defined in `unsafe`

, we say that the pattern matching within that function is *non-exhaustive*. In other words, the function does not implement appropriate handling of all valid inputs. Instead of yielding a value, such a function will halt from an incomplete match.

Partial functions from non-exhaustivity are a controversial subject, and frequent use of non-exhaustive patterns is considered a dangerous code smell. However, the complete removal of non-exhaustive patterns from the language would itself be too restrictive and forbid too many valid programs.

Several flags exist that we can pass to the compiler to warn us about such patterns or forbid them entirely either locally or globally.

```
$ ghc -c -Wall -Werror A.hs
A.hs:3:1:
Warning: Pattern match(es) are non-exhaustive
In an equation for `unsafe': Patterns not matched: Nothing
```

The `-Wall`

or `-fwarn-incomplete-patterns`

flag can also be added on a per-module basis by using the `OPTIONS_GHC`

pragma.

```
{-# OPTIONS_GHC -Wall #-}
{-# OPTIONS_GHC -fwarn-incomplete-patterns #-}
```

A more subtle case of non-exhaustivity is the use of implicit pattern matching with a single *uni-pattern* in a lambda expression. In a manner similar to the `unsafe`

function above, a uni-pattern cannot handle all types of valid input. For instance, the function `boom`

will fail when given a Nothing, even though the type of the lambda expression's argument is a `Maybe a`

.

`boom = \(Just a) -> something`

Non-exhaustivity arising from uni-patterns in lambda expressions occurs frequently in `let`

or `do`

-blocks after desugaring, because such code is translated into lambda expressions similar to `boom`

.

```
boom2 = let
Just a = something
boom3 = do
Just a <- something
```

GHC can warn about these cases of non-exhaustivity with the `-fwarn-incomplete-uni-patterns`

flag.

Grossly speaking, any non-trivial program will use some measure of partial functions. It is simply a fact. Thus, there exist obligations for the programmer than cannot be manifest in the Haskell type system.

## Debugger

Since GHCi version 6.8.1, a built-in debugger has been available, although its use is somewhat rare. Debugging uncaught exceptions from bottoms or asynchronous exceptions is in similar style to debugging segfaults with gdb.

```
λ: :set -fbreak-on-exception -- Sets option for evaluation to stop on exception
λ: :break 2 15 -- Sets a break point at line 2, column 15
λ: :trace main -- Run a function to generate a sequence of evaluation steps
λ: :hist -- Step backwards from a breakpoint through previous steps of evaluation
λ: :back -- Step backwards a single step at a time through the history
λ: :forward -- Step forward a single step at a time through the history
```

## Stack Traces

With runtime profiling enabled, GHC can also print a stack trace when a diverging bottom term (error, undefined) is hit. This action, though, requires a special flag and profiling to be enabled, both of which are disabled by default. So, for example:

```
import Control.Exception
f x = g x
g x = error (show x)
main = try (evaluate (f ())) :: IO (Either SomeException ())
```

```
$ ghc -O0 -rtsopts=all -prof -auto-all --make stacktrace.hs
./stacktrace +RTS -xc
```

And indeed, the runtime tells us that the exception occurred in the function `g`

and enumerates the call stack.

```
*** Exception (reporting due to +RTS -xc): (THUNK_2_0), stack trace:
Main.g,
called from Main.f,
called from Main.main,
called from Main.CAF
--> evaluated by: Main.main,
called from Main.CAF
```

It is best to run this code without optimizations applied `-O0`

so as to preserve the original call stack as represented in the source. With optimizations applied, GHC will rearrange the program in rather drastic ways, resulting in what may be an entirely different call stack.

See:

## Trace

Since Haskell is a pure language, it has the unique property that most code is introspectable on its own. As such, using printf to display the state of the program at critical times throughout execution is often unnecessary because we can simply open GHCi and test the function. Nevertheless, Haskell does come with an unsafe `trace`

function which can be used to perform arbitrary print statements outside of the IO monad.

```
import Debug.Trace
example1 :: Int
example1 = trace "impure print" 1
example2 :: Int
example2 = traceShow "tracing" 2
example3 :: [Int]
example3 = [trace "will not be called" 3]
main :: IO ()
main = do
print example1
print example2
print $ length example3
-- impure print
-- 1
-- "tracing"
-- 2
-- 1
```

Trace uses `unsafePerformIO`

under the hood and should **not** be used in stable code.

In addition to the `trace`

function, several monadic `trace`

variants are quite common.

```
import Text.Printf
import Debug.Trace
traceM :: (Monad m) => String -> m ()
traceM string = trace string $ return ()
traceShowM :: (Show a, Monad m) => a -> m ()
traceShowM = traceM . show
tracePrintfM :: (Monad m, PrintfArg a) => String -> a -> m ()
tracePrintfM s = traceM . printf s
```

## Type Holes

Since the release of GHC 7.8, *typed holes* allow for debugging incomplete programs. By placing an underscore on any value on the right hand-side of a declaration, GHC will throw an error during type-checking. Such an error reflects what type the value in the position of the type hole could be in order for the program to type-check successfully.

```
instance Functor [] where
fmap f (x:xs) = f x : fmap f _
```

```
[1 of 1] Compiling Main ( src/typedhole.hs, interpreted )
src/typedhole.hs:7:32:
Found hole ‘_’ with type: [a]
Where: ‘a’ is a rigid type variable bound by
the type signature for fmap :: (a -> b) -> [a] -> [b]
at src/typedhole.hs:7:3
Relevant bindings include
xs :: [a] (bound at src/typedhole.hs:7:13)
x :: a (bound at src/typedhole.hs:7:11)
f :: a -> b (bound at src/typedhole.hs:7:8)
fmap :: (a -> b) -> [a] -> [b] (bound at src/typedhole.hs:7:3)
In the second argument of ‘fmap’, namely ‘_’
In the second argument of ‘(:)’, namely ‘fmap f _’
In the expression: f x : fmap f _
Failed, modules loaded: none.
```

GHC has rightly suggested that the expression needed to finish the program is `xs :: [a]`

.

## Deferred Type Errors

Since the release of version 7.8, GHC supports the option of treating type errors as runtime errors. With this option enabled, programs will run, but they will fail when a mistyped expression is evaluated. This feature is enabled with the `-fdefer-type-errors`

flag in three ways: at the module level, when compiled from the command line, or inside of a GHCi interactive session.

For instance, the program below will compile:

```
{-# OPTIONS_GHC -fdefer-type-errors #-} -- Enable deferred type
-- errors at module level
x :: ()
x = print 3
y :: Char
y = 0
z :: Int
z = 0 + "foo"
main :: IO ()
main = do
print x
```

However, when a pathological term is evaluated at runtime, we'll see a message like:

```
defer: defer.hs:4:5:
Couldn't match expected type ‘()’ with actual type ‘IO ()’
In the expression: print 3
In an equation for ‘x’: x = print 3
(deferred type error)
```

This error tells us that while `x`

has a declared type of `()`

, the body of the function `print 3`

has a type of `IO ()`

. However, if the term is never evaluated, GHC will not throw an exception.

## ghcid

ghcid is a lightweight IDE hook that allows continuous feedback whenever code is updated. It can be run from the command line in the root of the `cabal`

project directory by specifying a command to run (e.g. `ghci`

, `cabal repl`

, or `stack repl`

).

```
ghcid --command="cabal repl" # Run cabal repl under ghcid
ghcid --command="stack repl" # Run stack repl under ghcid
ghcid --command="ghci baz.hs" # Open baz.hs under ghcid
```

When a Haskell module is loaded into `ghcid`

, the code is evaluated in order to provide the user with any errors or warnings that would happen at compile time. When the developer edits and saves code loaded into `ghcid`

, the program automatically reloads and evaluates the code for errors and warnings.

## Haddock

Haddock is the automatic documentation generation tool for Haskell source code. It integrates with the usual `cabal`

toolchain. In this section, we will explore how to document code so that Haddock can generate documentation successfully.

Several frequent comment patterns are used to document code for Haddock. The first of these methods uses `-- |`

to delineate the beginning of a comment:

```
-- | Documentation for f
f :: a -> a
f = ...
```

Multiline comments are also possible:

```
-- | Multiline documentation for the function
-- f with multiple arguments.
fmap :: Functor f =>
=> (a -> b) -- ^ function
-> f a -- ^ input
-> f b -- ^ output
```

`-- ^`

is also used to comment Constructors or Record fields:

```
data T a b
= A a -- ^ Documentation for A
| B b -- ^ Documentation for B
data R a b = R
{ f1 :: a -- ^ Documentation for the field f1
, f2 :: b -- ^ Documentation for the field f2
}
```

Elements within a module (i.e. value, types, classes) can be hyperlinked by enclosing the identifier in single quotes:

```
data T a b
= A a -- ^ Documentation for 'A'
| B b -- ^ Documentation for 'B'
```

Modules themselves can be referenced by enclosing them in double quotes:

```
-- | Here we use the "Data.Text" library and import
-- the 'Data.Text.pack' function.
```

`haddock`

also allows the user to include blocks of code within the generated documentation. Two methods of demarcating the code blocks exist in `haddock`

. For example, enclosing a code snippet in `@`

symbols marks it as a code block:

```
-- | An example of a code block.
--
-- @
-- f x = f (f x)
-- @
```

Similarly, it's possible to use bird tracks (`>`

) in a comment line to set off a code block. This usage is very similar to Bird style Literate Haskell.

```
-- | A similar code block example that uses bird tracks (i.e. '>')
-- > f x = f (f x)
```

Snippets of interactive shell sessions can also be included in `haddock`

documentation. In order to denote the beginning of code intended to be run in a REPL, the `>>>`

symbol is used:

```
-- | Example of an interactive shell session embedded within documentation
--
-- >>> factorial 5
-- 120
```

Headers for specific blocks can be added by prefacing the comment in the module block with a `*`

:

```
module Foo (
-- * My Header
example1,
example2
)
```

Sections can also be delineated by `$`

blocks that pertain to references in the body of the module:

```
module Foo (
-- $section1
example1,
example2
)
-- $section1
-- Here is the documentation section that describes the symbols
-- 'example1' and 'example2'.
```

Links can be added with the following syntax:

`<url text>`

Images can also be included, so long as the path is either absolute or relative to the directory in which `haddock`

is run.

`<<diagram.png title>>`

`haddock`

options can also be specified with pragmas in the source, either at the module or project level.

`{-# OPTIONS_HADDOCK show-extensions, ignore-exports #-}`

Option | Description |
---|---|

ignore-exports | Ignores the export list and includes all signatures in scope. |

not-home | Module will not be considered in the root documentation. |

show-extensions | Annotates the documentation with the language extensions used. |

hide | Forces the module to be hidden from Haddock. |

prune | Omits definitions with no annotations. |

# Monads

## Eightfold Path to Monad Satori

Much ink has been spilled waxing lyrical about the supposed mystique of monads. Instead, I suggest a path to enlightenment:

- Don't read the monad tutorials.
- No really, don't read the monad tutorials.
- Learn about Haskell types.
- Learn what a typeclass is.
- Read the Typeclassopedia.
- Read the monad definitions.
- Use monads in real code.
- Don't write monad-analogy tutorials.

In other words, the only path to understanding monads is to read the fine source, fire up GHC, and write some code. Analogies and metaphors will not lead to understanding.

## Monadic Myths

The following are all **false**:

- Monads are impure.
- Monads are about effects.
- Monads are about state.
- Monads are about imperative sequencing.
- Monads are about IO.
- Monads are dependent on laziness.
- Monads are a "back-door" in the language to perform side-effects.
- Monads are an embedded imperative language inside Haskell.
- Monads require knowing abstract mathematics.
- Monads are unique to Haskell.

See: What a Monad Is Not

## Monadic Methods

Monads are not complicated. They are implemented as a typeclass with two methods, `return`

and `(>>=)`

(pronounced "bind"). In order to implement a Monad instance, these two functions must be defined in accordance with the arity described in the typeclass definition:

```
class Monad m where
return :: a -> m a -- N.B. 'm' refers to a type constructor
-- (e.g., Maybe, Either, etc.) that
-- implements the Monad typeclass
(>>=) :: m a -> (a -> m b) -> m b
```

The first type signature in the Monad class definition is for `return`

. Any preconceptions one might have for the word "return" should be discarded: It has an entirely different meaning in the context of Haskell and acts very differently than in languages like C, Python, or Java. Instead of being the final arbiter of what value a function produces, `return`

in Haskell injects a value of type `a`

into a monadic context (e.g., Maybe, Either, etc.), which is denoted as `m a`

.

The other function essential to implementing a Monad instance is `(>>=)`

. This infix takes two arguments. On its left side is a value with type `m a`

, while on the right side is a function with type `(a -> m b)`

. The bind operation results in a final value of type `m b`

.

A third, auxiliary function (`(>>)`

) is defined in terms of the bind operation that discards its argument.

```
(>>) :: Monad m => m a -> m b -> m b
m >> k = m >>= \_ -> k
```

This definition says that (>>) has a left and right argument which are monadic with types `m a`

and `m b`

respectively, while the infix returns a value of type `m b`

. The actual implementation of (>>) says that when `m`

is passed to `(>>)`

with `k`

on the right, the value `k`

will always be returned.

## Laws

In addition to specific implementations of `(>>=)`

and `return`

, all monad instances must satisfy three laws.

**Law 1**

The first law says that when `return a`

is passed through a `(>>=)`

into a function `f`

, this expression is exactly equivalent to `f a`

.

`return a >>= f ≡ f a -- N.B. 'a' refers to a value, not a type`

In discussing the next two laws, we'll refer to a value `m`

. This notation is shorthand for value wrapped in a monadic context. Such a value has type `m a`

, and could be represented more concretely by values like `Nothing`

, `Just x`

, or `Right x`

. It is important to note that some of these concrete instantiations of the value `m`

have multiple components. In discussing the second and third monad laws, we'll see some examples of how this plays out.

**Law 2**

The second law states that a monadic value `m`

passed through `(>>=)`

into `return`

is exactly equivalent to itself. In other words, using bind to pass a monadic value to `return`

does not change the initial value.

`m >>= return ≡ m -- 'm' here refers to a value that has type 'm a'`

A more explicit way to write the second Monad law exists. In this following example code, the first expression shows how the second law applies to values represented by non-nullary type constructors. The second snippet shows how a value represented by a nullary type constructor works within the context of the second law.

```
(SomeMonad val) >>= return ≡ SomeMonad val -- 'SomeMonad val' has type 'm a' just
-- like 'm' from the first example of the
-- second law
NullaryMonadType >>= return ≡ NullaryMonadType
```

**Law 3**

While the first two laws are relatively clear, the third law may be more difficult to understand. This law states that when a monadic value `m`

is passed through `(>>=)`

to the function `f`

and then the result of that expression is passed to `>>= g`

, the entire expression is exactly equivalent to passing `m`

to a lambda expression that takes one parameter `x`

and outputs the function `f`

applied to `x`

. By the definition of bind, `f x`

*must* return a value wrapped in the same Monad. Because of this property, the resultant value of that expression can be passed through `(>>=)`

to the function `g`

, which also returns a monadic value.

```
(m >>= f) >>= g ≡ m >>= (\x -> f x >>= g) -- Like in the last law, 'm' has
-- has type 'm a'. The functions 'f'
-- and 'g' have types '(a -> m b)'
-- and '(b -> m c)' respectively
```

Again, it is possible to write this law with more explicit code. Like in the explicit examples for law 2, `m`

has been replaced by `SomeMonad val`

in order to be very clear that there can be multiple components to a monadic value. Although little has changed in the code, it is easier to see what value--namely, `val`

--corresponds to the `x`

in the lambda expression. After `SomeMonad val`

is passed through `(>>=)`

to `f`

, the function `f`

operates on `val`

and returns a result still wrapped in the `SomeMonad`

type constructor. We can call this new value `SomeMonad newVal`

. Since it is still wrapped in the monadic context, `SomeMonad newVal`

can thus be passed through the bind operation into the function `g`

.

`((SomeMonad val) >>= f) >>= g ≡ (SomeMonad val) >>= (\x -> f x >>= g)`

See: Monad Laws

## Do Notation

Monadic syntax in Haskell is written in a sugared form, known as `do`

notation. The advantages of this special syntax are that it is easier to write and is entirely equivalent to just applications of the monad operations. The desugaring is defined recursively by the rules:

```
do { a <- f ; m } ≡ f >>= \a -> do { m } -- bind 'f' to a, proceed to desugar
-- 'm'
do { f ; m } ≡ f >> do { m } -- evaluate 'f', then proceed to
-- desugar m
do { m } ≡ m
```

Thus, through the application of the desugaring rules, the following expressions are equivalent:

```
do
a <- f -- f, g, and h are bound to the names a,
b <- g -- b, and c. These names are then passed
c <- h -- to 'return' to ensure that all values
return (a, b, c) -- are wrapped in the appropriate monadic
-- context
do { -- N.B. '{}' and ';' characters are
a <- f; -- rarely used in do-notation
b <- g;
c <- h;
return (a, b, c)
}
f >>= \a ->
g >>= \b ->
h >>= \c ->
return (a, b, c)
```

If one were to write the bind operator as an uncurried function ( this is not how Haskell uses it ) the same desugaring might look something like the following chain of nested binds with lambdas.

```
bindMonad(f, lambda a:
bindMonad(g, lambda b:
bindMonad(h, lambda c:
returnMonad (a,b,c))))
```

In the do-notation, the monad laws from above are equivalently written:

**Law 1**

```
do y <- return x
f y
= do f x
```

**Law 2**

```
do x <- m
return x
= do m
```

**Law 3**

```
do b <- do a <- m
f a
g b
= do a <- m
b <- f a
g b
= do a <- m
do b <- f a
g b
```

See: Haskell 2010: Do Expressions

## Maybe

The *Maybe* monad is the simplest first example of a monad instance. The Maybe monad models computations which fail to yield a value at any point during computation.

The Maybe type has two value constructors. The first, `Just`

, is a unary constructor representing a successful computation, while the second, `Nothing`

, is a nullary constructor that represents failure.

`data Maybe a = Just a | Nothing`

The monad instance describes the implementation of `(>>=)`

for `Maybe`

by pattern matching on the possible inputs that could be passed to the bind operation (i.e., `Nothing`

or `Just x`

). The instance declaration also provides an implementation of `return`

, which in this case is simply `Just`

.

```
instance Monad Maybe where
(Just x) >>= k = k x -- 'k' is a function with type (a -> Maybe a)
Nothing >>= k = Nothing
return = Just -- Just's type signature is (a -> Maybe a), in
-- other words, extremely similar to the
-- type of 'return' in the typeclass
-- declaration above.
```

The following code shows some simple operations to do within the Maybe monad.

In the first example, The value `Just 3`

is passed via `(>>=)`

to the lambda function `\x -> return (x + 1)`

. `x`

refers to the `Int`

portion of `Just 3`

, and we can use `x`

in the second half of the lambda expression, where `return (x + 1)`

evaluates to `Just 4`

, indicating a successful computation.

```
(Just 3) >>= (\x -> return (x + 1))
-- Just 4
```

In the second example, the value `Nothing`

is passed via `(>>=)`

to the same lambda function as in the previous example. However, according to the `Maybe`

Monad instance, whenever `Nothing`

is bound to a function, the expression's result will be `Nothing`

.

```
Nothing >>= (\x -> return (x + 1))
-- Nothing
```

In the next example, `return`

is applied to `4`

and returns `Just 4`

.

```
return 4 :: Maybe Int
-- Just 4
```

The next code examples show the use of `do`

notation within the Maybe monad to do addition that might fail. Desugared examples are provided as well.

```
example1 :: Maybe Int
example1 = do
a <- Just 3 -- Bind 3 to name a
b <- Just 4 -- Bind 4 to name b
return $ a + b -- Evaluate (a + b), then use 'return' to ensure
-- the result is in the Maybe monad in order to
-- satisfy the type signature
-- Just 7
desugared1 :: Maybe Int
desugared1 = Just 3 >>= \a -> -- This example is the desugared
Just 4 >>= \b -> -- equivalent to example1
return $ a + b
-- Just 7
example2 :: Maybe Int
example2 = do
a <- Just 3 -- Bind 3 to name a
b <- Nothing -- Bind Nothing to name b
return $ a + b
-- Nothing -- This result might be somewhat surprising, since
-- addition within the Maybe monad can actually
-- return 'Nothing'. This result occurs because one
-- of the values, Nothing, indicates computational
-- failure. Since the computation failed at one
-- step within the process, the whole computation
-- fails, leaving us with 'Nothing' as the final
-- result.
desugared2 :: Maybe Int
desugared2 = Just 3 >>= \a -> -- This example is the desugared
Nothing >>= \b -> -- equivalent to example2
return $ a + b
-- Nothing
```

## List

The *List* monad is the second simplest example of a monad instance. As always, this monad implements both `(>>=)`

and `return`

. The definition of bind says that when the list `m`

is bound to a function `f`

, the result is a concatenation of `map f`

over the list `m`

. The `return`

method simply takes a single value `x`

and injects into a singleton list `[x]`

.

```
instance Monad [] where
m >>= f = concat (map f m) -- 'm' is a list
return x = [x]
```

In order to demonstrate the `List`

monad's methods, we will define two functions: `m`

and `f`

. `m`

is a simple list, while `f`

is a function that takes a single `Int`

and returns a two element list `[1, 0]`

.

```
m :: [Int]
m = [1,2,3,4]
f :: Int -> [Int]
f = \x -> [1,0] -- 'f' always returns [1, 0]
```

The evaluation proceeds as follows:

```
m >>= f
==> [1,2,3,4] >>= \x -> [1,0]
==> concat (map (\x -> [1,0]) [1,2,3,4])
==> concat ([[1,0],[1,0],[1,0],[1,0]])
==> [1,0,1,0,1,0,1,0]
```

The list comprehension syntax in Haskell can be implemented in terms of the list monad. List comprehensions can be considered syntactic sugar for more obviously monadic implementations. Examples `a`

and `b`

illustrate these use cases.

The first example (`a`

) illustrates how to write a list comprehension. Although the syntax looks strange at first, there are elements of it that may look familiar. For instance, the use of `<-`

is just like bind in a `do`

notation: It binds an element of a list to a name. However, one major difference is apparent: `a`

seems to lack a call to `return`

. Not to worry, though, the `[]`

fills this role. This syntax can be easily desugared by the compiler to an explicit invocation of `return`

. Furthermore, it serves to remind the user that the computation takes place in the List monad.

```
a = [
f x y | -- Corresponds to 'f x y' in example b
x <- xs,
y <- ys,
x == y -- Corresponds to 'guard $ x == y' in example b
]
```

The second example (`b`

) shows the list comprehension above rewritten with `do`

notation:

```
-- Identical to `a`
b = do
x <- xs
y <- ys
guard $ x == y -- Corresponds to 'x == y' in example a
return $ f x y -- Corresponds to the '[]' and 'f x y' in example a
```

The final examples are further illustrations of the List monad. The functions below each return a list of 3-tuples which contain the possible combinations of the three lists that get bound the names `a`

, `b`

, and `c`

. N.B.: Only values in the list bound to `a`

can be used in `a`

position of the tuple; the same fact holds true for the lists bound to `b`

and `c`

.

```
example :: [(Int, Int, Int)]
example = do
a <- [1,2]
b <- [10,20]
c <- [100,200]
return (a,b,c)
-- [(1,10,100),(1,10,200),(1,20,100),(1,20,200),(2,10,100),(2,10,200),(2,20,100),(2,20,200)]
desugared :: [(Int, Int, Int)]
desugared = [1, 2] >>= \a ->
[10, 20] >>= \b ->
[100, 200] >>= \c ->
return (a, b, c)
-- [(1,10,100),(1,10,200),(1,20,100),(1,20,200),(2,10,100),(2,10,200),(2,20,100),(2,20,200)]
```

## IO

Perhaps the most (in)famous example in Haskell of a type that forms a monad is `IO`

. A value of type `IO a`

is a computation which, when performed, does some I/O before returning a value of type `a`

. These computations are called actions. IO actions executed in `main`

are the means by which a program can operate on or access information in the external world. IO actions allow the program to do many things, including, but not limited to:

- Print a
`String`

to the terminal - Read and parse input from the terminal
- Read from or write to a file on the system
- Establish an
`ssh`

connection to a remote computer - Take input from a radio antenna for singal processing

Conceptualizing I/O as a monad enables the developer to access information outside the program, but operate on the data with pure functions. The following examples will show how we can use IO actions and IO values to receive input from stdin and print to stdout.

Perhaps the most immediately useful function for doing I/O in Haskell is `putStrLn`

. This function takes a `String`

and returns an `IO ()`

. Calling it from `main`

will result in the `String`

being printed to stdout followed by a newline character.

`putStrLn :: String -> IO ()`

Here is some code that prints a couple of lines to the terminal. The first invocation of `putStrLn`

is executed, causing the `String`

to be printed to stdout. The result is bound to a lambda expression that discards its argument, and then the next `putStrLn`

is executed.

```
main :: IO ()
main = putStrLn "Vesihiisi sihisi hississäään." >>=
\_ -> putStrLn "Or in English: 'The water devil was hissing
in her elevator'."
-- Sugared code, written with do notation
main :: IO ()
main = do putStrLn "Vesihiisi sihisi hississäään."
putStrLn "Or in English: 'The water devil was hissing in her
elevator'."
```

Another useful function is `getLine`

which has type `IO String`

. This function gets a line of input from stdin. The developer can then bind this line to a name in order to operate on the value within the program.

`getLine :: IO String`

The code below demonstrates a simple combination of these two functions as well as desugaring `IO`

code. First, `putStrLn`

prints a `String`

to stdout to ask the user to supply their name, with the result being bound to a lambda that discards it argument. Then, `getLine`

is executed, supplying a prompt to the user for entering their name. Next, the resultant `IO String`

is bound to `name`

and passed to `putStrLn`

. Finally, the program prints the name to the terminal.

```
main :: IO ()
main = do putStrLn "What is your name: "
name <- getLine
putStrLn name
```

The next code block is the desugared equivalent of the previous example; however, the uses of `(>>=)`

are made explict.

```
main :: IO ()
main = putStrLn "What is your name:" >>=
\_ -> getLine >>=
\name -> putStrLn name
```

Our final example executes in the same way as the previous two examples. This example, though, uses the special `(>>)`

operator to take the place of binding a result to the lamda that discards its argument.

```
main :: IO ()
main = putStrLn "What is your name: " >> (getLine >>= (\name -> putStrLn name))
```

See: Haskell 2010: Basic/Input Output

## What's the point?

Although it is difficult, if not impossible, to touch, see, or otherwise physically interact with a monad, this construct has some very interesting implications for programmers. For instance, consider the non-intuitive fact that we now have a uniform interface for talking about three very different, but foundational ideas for programming: *Failure*, *Collections* and *Effects*.

Let's write down a new function called `sequence`

which folds a function `mcons`

over a list of monadic computations. We can think of `mcons`

as analogous to the list constructor (i.e. `(a : b : [])`

) except it pulls the two list elements out of two monadic values (`p`

,`q`

) by means of bind. The bound values are then joined with the list constructor `:`

, before finally being rewrapped in the appropriate monadic context with `return`

.

```
sequence :: Monad m => [m a] -> m [a]
sequence = foldr mcons (return [])
mcons :: Monad m => m t -> m [t] -> m [t]
mcons p q = do
x <- p -- 'x' refers to a singleton value
y <- q -- 'y' refers to a list. Because of this fact, 'x' can be
return (x:y) -- prepended to it
```

What does this function mean in terms of each of the monads discussed above?

**Maybe**

Sequencing a list of values within the `Maybe`

context allows us to collect the results of a series of computations which can possibly fail. However, `sequence`

yields the aggregated values only if each computation succeeds. In other words, if even one of the `Maybe`

values in the initial list passed to `sequence`

is a `Nothing`

, the result of `sequence`

will also be `Nothing`

.

`sequence :: [Maybe a] -> Maybe [a]`

```
sequence [Just 3, Just 4]
-- Just [3,4]
sequence [Just 3, Just 4, Nothing] -- Since one of the results is Nothing,
-- Nothing -- the whole computation fails
```

**List**

The bind operation for the list monad forms the pairwise list of elements from the two operands. Thus, folding the binds contained in `mcons`

over a list of lists with `sequence`

implements the general Cartesian product for an arbitrary number of lists.

`sequence :: [[a]] -> [[a]]`

```
sequence [[1,2,3],[10,20,30]]
-- [[1,10],[1,20],[1,30],[2,10],[2,20],[2,30],[3,10],[3,20],[3,30]]
```

**IO**

Applying `sequence`

within the IO context results in still a different result. The function takes a list of IO actions, performs them sequentially, and then returns the list of resulting values in the order sequenced.

`sequence :: [IO a] -> IO [a]`

```
sequence [getLine, getLine, getLine]
-- a -- a, b, and 9 are the inputs given by the
-- b -- user at the prompt
-- 9
-- ["a", "b", "9"] -- All inputs are returned in a list as
-- an IO [String].
```

So there we have it, three fundamental concepts of computation that are normally defined independently of each other actually all share this similar structure. This unifying pattern can be abstracted out and reused to build higher abstractions that work for all current and future implementations. If you want a motivating reason for understanding monads, this is it! These insights are the essence of what I wish I knew about monads looking back.

See: Control.Monad

## Reader Monad

The reader monad lets us access shared immutable state within a monadic context.

```
ask :: Reader r r
asks :: (r -> a) -> Reader r a
local :: (r -> r) -> Reader r a -> Reader r a
runReader :: Reader r a -> r -> a
```

```
import Control.Monad.Reader
data MyContext = MyContext
{ foo :: String
, bar :: Int
} deriving (Show)
computation :: Reader MyContext (Maybe String)
computation = do
n <- asks bar
x <- asks foo
if n > 0
then return (Just x)
else return Nothing
ex1 :: Maybe String
ex1 = runReader computation $ MyContext "hello" 1
ex2 :: Maybe String
ex2 = runReader computation $ MyContext "haskell" 0
```

A simple implementation of the Reader monad:

```
newtype Reader r a = Reader { runReader :: r -> a }
instance Monad (Reader r) where
return a = Reader $ \_ -> a
m >>= k = Reader $ \r -> runReader (k (runReader m r)) r
ask :: Reader a a
ask = Reader id
asks :: (r -> a) -> Reader r a
asks f = Reader f
local :: (r -> r) -> Reader r a -> Reader r a
local f m = Reader $ runReader m . f
```

## Writer Monad

The writer monad lets us emit a lazy stream of values from within a monadic context.

```
tell :: w -> Writer w ()
execWriter :: Writer w a -> w
runWriter :: Writer w a -> (a, w)
```

```
import Control.Monad.Writer
type MyWriter = Writer [Int] String
example :: MyWriter
example = do
tell [1..3]
tell [3..5]
return "foo"
output :: (String, [Int])
output = runWriter example
-- ("foo", [1, 2, 3, 3, 4, 5])
```

A simple implementation of the Writer monad:

```
import Data.Monoid
newtype Writer w a = Writer { runWriter :: (a, w) }
instance Monoid w => Monad (Writer w) where
return a = Writer (a, mempty)
m >>= k = Writer $ let
(a, w) = runWriter m
(b, w') = runWriter (k a)
in (b, w `mappend` w')
execWriter :: Writer w a -> w
execWriter m = snd (runWriter m)
tell :: w -> Writer w ()
tell w = Writer ((), w)
```

This implementation is lazy, so some care must be taken that one actually wants to only generate a stream of thunks. Most often the lazy writer is not suitable for use, instead implement the equivalent structure by embedding some monomial object inside a StateT monad, or using the strict version.

`import Control.Monad.Writer.Strict`

## State Monad

The state monad allows functions within a stateful monadic context to access and modify shared state.

```
runState :: State s a -> s -> (a, s)
evalState :: State s a -> s -> a
execState :: State s a -> s -> s
```

```
import Control.Monad.State
test :: State Int Int
test = do
put 3
modify (+1)
get
main :: IO ()
main = print $ execState test 0
```

The state monad is often mistakenly described as being impure, but it is in fact entirely pure and the same effect could be achieved by explicitly passing state. A simple implementation of the State monad takes only a few lines:

```
newtype State s a = State { runState :: s -> (a,s) }
instance Monad (State s) where
return a = State $ \s -> (a, s)
State act >>= k = State $ \s ->
let (a, s') = act s
in runState (k a) s'
get :: State s s
get = State $ \s -> (s, s)
put :: s -> State s ()
put s = State $ \_ -> ((), s)
modify :: (s -> s) -> State s ()
modify f = get >>= \x -> put (f x)
evalState :: State s a -> s -> a
evalState act = fst . runState act
execState :: State s a -> s -> s
execState act = snd . runState act
```

## Monad Tutorials

So many monad tutorials have been written that it begs the question: what makes monads so difficult when first learning Haskell? I hypothesize there are three aspects to why this is so:

*There are several levels on indirection with desugaring.*

A lot of the Haskell we write is radically rearranged and transformed into an entirely new form under the hood.

Most monad tutorials will not manually expand out the do-sugar. This leaves the beginner thinking that monads are a way of dropping into a pseudo-imperative language inside of code and further fuels that misconception that specific instances like IO are monads in their full generality.

```
main = do
x <- getLine
putStrLn x
return ()
```

Being able to manually desugar is crucial to understanding.

```
main =
getLine >>= \x ->
putStrLn x >>= \_ ->
return ()
```

*Asymmetric binary infix operators for higher order functions are not common in other languages.*

`(>>=) :: Monad m => m a -> (a -> m b) -> m b`

On the left hand side of the operator we have an `m a`

and on the right we have `a -> m b`

. Although some languages do have infix operators that are themselves higher order functions, it is still a rather rare occurrence.

So with a function desugared, it can be confusing that `(>>=)`

operator is in fact building up a much larger function by composing functions together.

```
main =
getLine >>= \x ->
putStrLn >>= \_ ->
return ()
```

Written in prefix form, it becomes a little bit more digestible.

```
main =
(>>=) getLine (\x ->
(>>=) putStrLn (\_ ->
return ()
)
)
```

Perhaps even removing the operator entirely might be more intuitive coming from other languages.

```
main = bind getLine (\x -> bind putStrLn (\_ -> return ()))
where
bind x y = x >>= y
```

*Ad-hoc polymorphism is not commonplace in other languages.*

Haskell's implementation of overloading can be unintuitive if one is not familiar with type inference. It is abstracted away from the user, but the `(>>=)`

or `bind`

function is really a function of 3 arguments with the extra typeclass dictionary argument (`$dMonad`

) implicitly threaded around.

`main $dMonad = bind $dMonad getLine (\x -> bind $dMonad putStrLn (\_ -> return $dMonad ()))`

Except in the case where the parameter of the monad class is unified ( through inference ) with a concrete class instance, in which case the instance dictionary (`$dMonadIO`

) is instead spliced throughout.

```
main :: IO ()
main = bind $dMonadIO getLine (\x -> bind $dMonadIO putStrLn (\_ -> return $dMonadIO ()))
```

Now, all of these transformations are trivial once we understand them, they're just typically not discussed. In my opinion the fundamental fallacy of monad tutorials is not that intuition for monads is hard to convey ( nor are metaphors required! ), but that novices often come to monads with an incomplete understanding of points (1), (2), and (3) and then trip on the simple fact that monads are the first example of a Haskell construct that is the confluence of all three.

# Monad Transformers

## mtl / transformers

So, the descriptions of Monads in the previous chapter are a bit of a white lie. Modern Haskell monad libraries typically use a more general form of these, written in terms of monad transformers which allow us to compose monads together to form composite monads. The monads mentioned previously are subsumed by the special case of the transformer form composed with the Identity monad.

Monad | Transformer | Type | Transformed Type |
---|---|---|---|

Maybe | MaybeT | `Maybe a` |
`m (Maybe a)` |

Reader | ReaderT | `r -> a` |
`r -> m a` |

Writer | WriterT | `(a,w)` |
`m (a,w)` |

State | StateT | `s -> (a,s)` |
`s -> m (a,s)` |

```
type State s = StateT s Identity
type Writer w = WriterT w Identity
type Reader r = ReaderT r Identity
instance Monad m => MonadState s (StateT s m)
instance Monad m => MonadReader r (ReaderT r m)
instance (Monoid w, Monad m) => MonadWriter w (WriterT w m)
```

In terms of generality the mtl library is the most common general interface for these monads, which itself depends on the transformers library which generalizes the "basic" monads described above into transformers.

## Transformers

At their core monad transformers allow us to nest monadic computations in a stack with an interface to exchange values between the levels, called `lift`

.

```
lift :: (Monad m, MonadTrans t) => m a -> t m a
liftIO :: MonadIO m => IO a -> m a
```

```
class MonadTrans t where
lift :: Monad m => m a -> t m a
class (Monad m) => MonadIO m where
liftIO :: IO a -> m a
instance MonadIO IO where
liftIO = id
```

Just as the base monad class has laws, monad transformers also have several laws:

**Law #1**

`lift . return = return`

**Law #2**

`lift (m >>= f) = lift m >>= (lift . f)`

Or equivalently:

**Law #1**

```
lift (return x)
= return x
```

**Law #2**

```
do x <- lift m
lift (f x)
= lift $ do x <- m
f x
```

It's useful to remember that transformers compose *outside-in* but are *unrolled inside out*.

See: Monad Transformers: Step-By-Step

## Basics

The most basic use requires us to use the T-variants for each of the monad transformers in the outer layers and to explicitly `lift`

and `return`

values between the layers. Monads have kind `(* -> *)`

, so monad transformers which take monads to monads have `((* -> *) -> * -> *)`

:

```
Monad (m :: * -> *)
MonadTrans (t :: (* -> *) -> * -> *)
```

So, for example, if we wanted to form a composite computation using both the Reader and Maybe monads we can now put the Maybe inside of a `ReaderT`

to form `ReaderT t Maybe a`

.

```
import Control.Monad.Reader
type Env = [(String, Int)]
type Eval a = ReaderT Env Maybe a
data Expr
= Val Int
| Add Expr Expr
| Var String
deriving (Show)
eval :: Expr -> Eval Int
eval ex = case ex of
Val n -> return n
Add x y -> do
a <- eval x
b <- eval y
return (a+b)
Var x -> do
env <- ask
val <- lift (lookup x env)
return val
env :: Env
env = [("x", 2), ("y", 5)]
ex1 :: Eval Int
ex1 = eval (Add (Val 2) (Add (Val 1) (Var "x")))
example1, example2 :: Maybe Int
example1 = runReaderT ex1 env
example2 = runReaderT ex1 []
```

The fundamental limitation of this approach is that we find ourselves `lift.lift.lift`

ing and `return.return.return`

ing a lot.

## ReaderT

For example, there exist three possible forms of the Reader monad. The first is the Haskell 98 version that no longer exists, but is useful for understanding the underlying ideas. The other two are the *transformers* and *mtl* variants.

*Reader*

```
newtype Reader r a = Reader { runReader :: r -> a }
instance MonadReader r (Reader r) where
ask = Reader id
local f m = Reader (runReader m . f)
```

*ReaderT*

```
newtype ReaderT r m a = ReaderT { runReaderT :: r -> m a }
instance (Monad m) => Monad (ReaderT r m) where
return a = ReaderT $ \_ -> return a
m >>= k = ReaderT $ \r -> do
a <- runReaderT m r
runReaderT (k a) r
instance MonadTrans (ReaderT r) where
lift m = ReaderT $ \_ -> m
```

*MonadReader*

```
class (Monad m) => MonadReader r m | m -> r where
ask :: m r
local :: (r -> r) -> m a -> m a
instance (Monad m) => MonadReader r (ReaderT r m) where
ask = ReaderT return
local f m = ReaderT $ \r -> runReaderT m (f r)
```

So, hypothetically the three variants of ask would be:

```
ask :: Reader r r
ask :: Monad m => ReaderT r m r
ask :: MonadReader r m => m r
```

In practice only the last one is used in modern Haskell.

## Newtype Deriving

Newtypes let us reference a data type with a single constructor as a new distinct type, with no runtime overhead from boxing, unlike an algebraic datatype with a single constructor. Newtype wrappers around strings and numeric types can often drastically reduce accidental errors.

Consider the case of using a newtype to distinguish between two different text blobs with different semantics. Both have the same runtime representation as a text object, but are distinguished statically, so that plaintext can not be accidentally interchanged with encrypted text.

```
newtype Plaintext = Plaintext Text
newtype Crytpotext = Cryptotext Text
encrypt :: Key -> Plaintext -> Cryptotext
decrypt :: Key -> Cryptotext -> Plaintext
```

The other common use case is using newtypes to derive logic for deriving custom monad transformers in our business logic. Using `-XGeneralizedNewtypeDeriving`

we can recover the functionality of instances of the underlying types composed in our transformer stack.

```
{-# LANGUAGE GeneralizedNewtypeDeriving #-}
newtype Velocity = Velocity { unVelocity :: Double }
deriving (Eq, Ord)
v :: Velocity
v = Velocity 2.718
x :: Double
x = 2.718
-- Type error is caught at compile time even though
-- they are the same value at runtime!
err = v + x
newtype Quantity v a = Quantity a
deriving (Eq, Ord, Num, Show)
data Haskeller
type Haskellers = Quantity Haskeller Int
a = Quantity 2 :: Haskellers
b = Quantity 6 :: Haskellers
totalHaskellers :: Haskellers
totalHaskellers = a + b
```

```
Couldn't match type `Double' with `Velocity'
Expected type: Velocity
Actual type: Double
In the second argument of `(+)', namely `x'
In the expression: v + x
```

Using newtype deriving with the mtl library typeclasses we can produce flattened transformer types that don't require explicit lifting in the transform stack. For example, here is a little stack machine involving the Reader, Writer and State monads.

```
{-# LANGUAGE GeneralizedNewtypeDeriving #-}
import Control.Monad.Reader
import Control.Monad.Writer
import Control.Monad.State
type Stack = [Int]
type Output = [Int]
type Program = [Instr]
type VM a = ReaderT Program (WriterT Output (State Stack)) a
newtype Comp a = Comp { unComp :: VM a }
deriving (Monad, MonadReader Program, MonadWriter Output, MonadState Stack)
data Instr = Push Int | Pop | Puts
evalInstr :: Instr -> Comp ()
evalInstr instr = case instr of
Pop -> modify tail
Push n -> modify (n:)
Puts -> do
tos <- gets head
tell [tos]
eval :: Comp ()
eval = do
instr <- ask
case instr of
[] -> return ()
(i:is) -> evalInstr i >> local (const is) eval
execVM :: Program -> Output
execVM = flip evalState [] . execWriterT . runReaderT (unComp eval)
program :: Program
program = [
Push 42,
Push 27,
Puts,
Pop,
Puts,
Pop
]
main :: IO ()
main = mapM_ print $ execVM program
```

Pattern matching on a newtype constructor compiles into nothing. For example the`extractB`

function does not scrutinize the `MkB`

constructor like the `extractA`

does, because `MkB`

does not exist at runtime, it is purely a compile-time construct.

```
data A = MkA Int
newtype B = MkB Int
extractA :: A -> Int
extractA (MkA x) = x
extractB :: B -> Int
extractB (MkB x) = x
```

## Efficiency

The second monad transformer law guarantees that sequencing consecutive lift operations is semantically equivalent to lifting the results into the outer monad.

```
do x <- lift m == lift $ do x <- m
lift (f x) f x
```

Although they are guaranteed to yield the same result, the operation of lifting the results between the monad levels is not without cost and crops up frequently when working with the monad traversal and looping functions. For example, all three of the functions on the left below are less efficient than the right hand side which performs the bind in the base monad instead of lifting on each iteration.

```
-- Less Efficient More Efficient
forever (lift m) == lift (forever m)
mapM_ (lift . f) xs == lift (mapM_ f xs)
forM_ xs (lift . f) == lift (forM_ xs f)
```

## Monad Morphisms

The base monad transformer package provides a `MonadTrans`

class for lifting to another monad:

`lift :: Monad m => m a -> t m a`

But often times we need to work with and manipulate our monad transformer stack to either produce new transformers, modify existing ones or extend an upstream library with new layers. The `mmorph`

library provides the capacity to compose monad morphism transformation directly on transformer stacks. The equivalent of type transformer type-level map is the `hoist`

function.

`hoist :: Monad m => (forall a. m a -> n a) -> t m b -> t n b`

Hoist takes a *monad morphism* (a mapping from a `m a`

to a `n a`

) and applies in on the inner value monad of a transformer stack, transforming the value under the outer layer.

The monad morphism `generalize`

takes an Identity monad into any another monad `m`

.

`generalize :: Monad m => Identity a -> m a`

For example, it generalizes `State s a`

(which is `StateT s Identity a`

) to `StateT s m a`

.

So we can generalize an existing transformer to lift an IO layer onto it.

```
import Control.Monad.State
import Control.Monad.Morph
type Eval a = State [Int] a
runEval :: [Int] -> Eval a -> a
runEval = flip evalState
pop :: Eval Int
pop = do
top <- gets head
modify tail
return top
push :: Int -> Eval ()
push x = modify (x:)
ev1 :: Eval Int
ev1 = do
push 3
push 4
pop
pop
ev2 :: StateT [Int] IO ()
ev2 = do
result <- hoist generalize ev1
liftIO $ putStrLn $ "Result: " ++ show result
```

See: mmorph

# Language Extensions

It's important to distinguish between different categories of language extensions *general* and *specialized*.

The inherent problem with classifying the extensions into the general and specialized categories is that it's a subjective classification. Haskellers who do type system research will have a very different interpretation of Haskell than people who do web programming. As such this is a conservative assessment, as an arbitrary baseline let's consider `FlexibleInstances`

and `OverloadedStrings`

"everyday" while `GADTs`

and `TypeFamilies`

are "specialized".

**Key**

*Benign*implies both that importing the extension won't change the semantics of the module if not used and that enabling it makes it no easier to shoot yourself in the foot.*Historical*implies that one shouldn't use this extension, it's in GHC purely for backwards compatibility. Sometimes these are dangerous to enable.*Steals syntax*means that enabling this extension means that certain code valid in vanilla Haskell will no longer be accepted. For example,`f $(a)`

is the same as`f $ (a)`

in Haskell98, but`TemplateHaskell`

will interpret`$(a)`

as a splice.

Benign | Historical | Steals Syntax | Use | Use | GHC Reference | Reference | |

AllowAmbiguousTypes | Specialized | Typelevel Programming | Ref | ||||

Arrows | ✓ | Specialized | Syntax Extension | Ref | Arrows | ||

AutoDeriveTypeable | ✓ | Specialized | Deriving | Ref | |||

BangPatterns | ✓ | ✓ | General | Strictness Annotation | Ref | Strictness Annotations | |

ApplicativeDo | Specialized | FFI | Ref | Applicative Do | |||

CApiFFI | Specialized | FFI | Ref | ||||

ConstrainedClassMethods | ✓ | Specialized | Typelevel Programming | Ref | |||

ConstraintKinds | ✓ | Specialized | Typelevel Programming | Ref | Constraint Kinds | ||

CPP | ✓ | General | Preprocessor | Ref | Cpp | ||

DataKinds | ✓ | Specialized | Typelevel Programming | Ref | Data Kinds | ||

DatatypeContexts | ✓ | Deprecated | Deprecated | Ref | |||

DefaultSignatures | ✓ | Specialized | Generic Programming | Ref | Generic | ||

DeriveAnyClass | ✓ | General | Deriving | Ref | |||

DeriveDataTypeable | ✓ | General | Deriving | Ref | Typeable | ||

DeriveFoldable | ✓ | General | Deriving | Ref | Foldable / Traversable | ||

DeriveFunctor | ✓ | General | Deriving | Ref | |||

DeriveGeneric | ✓ | General | Deriving | Ref | Generic | ||

DeriveLift | ✓ | General | Deriving | Ref | Template Haskell | ||

DeriveTraversable | ✓ | General | Deriving | Ref | |||

DisambiguateRecordFields | ✓ | ✓ | Specialized | Syntax Extension | Ref | ||

DuplicateRecordFields | ✓ | ✓ | Specialized | Syntax Extension | Ref | DuplicateRecordFields | |

DoRec | ✓ | ✓ | Specialized | Syntax Extension | Ref | Recursive Do | |

EmptyCase | ✓ | Specialized | Syntax Extension | Ref | EmptyCase | ||

EmptyDataDecls | ✓ | General | Syntax Extension | Ref | Void | ||

ExistentialQuantification | ✓ | Specialized | Typelevel Programming | Ref | Existential Quantification | ||

ExplicitForAll | Specialized | Typelevel Programming | Ref | Universal Quantification | |||

ExplicitNamespaces | ✓ | ✓ | Specialized | Syntax Disambiguation | Ref | ||

ExtendedDefaultRules | Specialized | Type Disambiguation | Ref | ||||

FlexibleContexts | ✓ | General | Typeclass Extension | Ref | Flexible Contexts | ||

FlexibleInstances | ✓ | General | Typeclass Extension | Ref | Flexible Instances | ||

ForeignFunctionInterface | ✓ | General | FFI | Ref | FFI | ||

FunctionalDependencies | ✓ | General | Typeclass Extension | Ref | Multiparam Typeclasses | ||

GADTs | ✓ | General | Typelevel Programming | Ref | GADTs | ||

GADTSyntax | ✓ | General | Syntax Extension | Ref | GADTs | ||

GeneralizedNewtypeDeriving | ✓ | General | Typeclass Extension | Ref | Newtype Deriving | ||

GHCForeignImportPrim | Specialized | FFI | Ref | Cmm | |||

ImplicitParams | ✓ | Specialized | Typelevel Programming | Ref | |||

ImpredicativeTypes | ✓ | Specialized | Typelevel Programming | Ref | Impredicative Types | ||

IncoherentInstances | Specialized | Typelevel Programming | Ref | Incoherent Instances | |||

InstanceSigs | ✓ | Specialized | Typelevel Programming | Ref | |||

InterruptibleFFI | Specialized | FFI | Ref | FFI | |||

KindSignatures | ✓ | Specialized | Typelevel Programming | Ref | Kind Signatures | ||

LambdaCase | ✓ | General | Syntax Extension | Ref | Lambda Case | ||

LiberalTypeSynonyms | ✓ | Specialized | Typeclass Extension | Ref | |||

MagicHash | ✓ | Specialized | GHC Internals | Ref | Unboxed Types | ||

MonadComprehensions | ✓ | Specialized | Syntax Extension | Ref | |||

MonoLocalBinds | General | Type Disambiguation | Ref | ||||

MonoPatBinds | Specialized | Type Disambiguation | Ref | ||||

MultiParamTypeClasses | ✓ | General | Typeclass Extension | Ref | Multiparam Typeclasses | ||

MultiWayIf | ✓ | Specialized | Syntax Extension | Ref | MultiWawyIf | ||

NamedFieldPuns | ✓ | Specialized | Syntax Extension | Ref | Named Field Puns | ||

NegativeLiterals | General | Type Disambiguation | Ref | ||||

NoImplicitPrelude | Specialized | Import Disambiguation | Ref | Custom Prelude | |||

NoMonomorphismRestriction | General | Type Disambiguation | Ref | Monomorphism Restriction | |||

NPlusKPatterns | ✓ | ✓ | Deprecated | Deprecated | Ref | ||

NullaryTypeClasses | Specialized | Typeclass Extension | Ref | Multiparam Typeclasses | |||

NumDecimals | General | Type Disambiguation | Ref | NumDecimals | |||

OverlappingInstances | Specialized | Typeclass Extension | Ref | Overlapping Instances | |||

OverloadedLabels | ✓ | General | Type Disambiguation | Ref | Overloaded Labels | ||

OverloadedRecordFields | ✓ | General | Syntax Extension | Ref | Overloaded Labels | ||

OverloadedLists | General | Syntax Extension | Ref | Overloaded Lists | |||

OverloadedStrings | General | Syntax Extension | Ref | Overloaded Strings | |||

PackageImports | ✓ | General | Import Disambiguation | Ref | Package Imports | ||

ParallelArrays | Specialized | Data Parallel Haskell | Ref | ||||

ParallelListComp | ✓ | ✓ | General | Syntax Extension | Ref | ||

PartialTypeSignatures | ✓ | General | Interactive Typing | Ref | Partial Type Signatures | ||

PatternGuards | ✓ | ✓ | General | Syntax Extension | Ref | Pattern Guards | |

PatternSynonyms | ✓ | ✓ | General | Syntax Extension | Ref | Pattern Synonyms | |

PolyKinds | Specialized | Typelevel Programming | Ref | Kind Polymorphism | |||

PolymorphicComponents | ✓ | Specialized | Deprecated | Ref | |||

PostfixOperators | ✓ | ✓ | Specialized | Syntax Extension | Ref | ||

QuasiQuotes | ✓ | Specialized | Metaprogramming | Ref | QuasiQuotation | ||

Rank2Types | ✓ | Specialized | Historical Artifact | Ref | Rank N Types | ||

RankNTypes | ✓ | Specialized | Typelevel Programming | Ref | Rank N Types | ||

RebindableSyntax | ✓ | Specialized | Metaprogramming | Ref | Indexed Monads | ||

RecordWildCards | ✓ | ✓ | General | Syntax Extension | Ref | Record Wildcards | |

RecursiveDo | ✓ | Specialized | Syntax Extension | Ref | MonadFix | ||

RelaxedPolyRec | Specialized | Type Disambiguation | Ref | ||||

RoleAnnotations | ✓ | Specialized | Type Disambiguation | Ref | Roles | ||

Safe | Specialized | Security Auditing | Ref | Safe Haskell | |||

SafeImports | Specialized | Security Auditing | Ref | Safe Haskell | |||

ScopedTypeVariables | ✓ | Specialized | Typelevel Programming | Ref | Scoped Type Variables | ||

StandaloneDeriving | ✓ | ✓ | General | Typeclass Extension | Ref | ||

StaticPointers | ✓ | ✓ | General | Distributed Programming | Ref | ||

Strict | General | Strictness Annotations | Ref | Strict Haskell | |||

StrictData | General | Strictness Annotations | Ref | Strict Haskell | |||

TemplateHaskell | ✓ | ✓ | Specialized | Metaprogramming | Ref | Template Haskell | |

TraditionalRecordSyntax | ✓ | ✓ | Specialized | Historical Artifact | Ref | Historical Extensions | |

TransformListComp | ✓ | Specialized | Syntax Extension | Ref | |||

Trustworthy | Specialized | Security Auditing | Ref | Safe Haskell | |||

TupleSections | ✓ | General | Syntax Extension | Ref | Tuple Sections | ||

TypeApplications | ✓ | ✓ | Specialized | Typelevel Programming | Ref | ||

TypeFamilies | ✓ | Specialized | Typelevel Programming | Ref | Type Families | ||

TypeHoles | ✓ | General | Interactive Typing | Ref | Type Holes | ||

TypeInType | Specialized | Typelevel Programming | Ref | ||||

TypeOperators | ✓ | Specialized | Typelevel Programming | Ref | Manual Proofs | ||

TypeSynonymInstances | ✓ | General | Typeclass Extension | Ref | Type Synonym Instances | ||

UnboxedTuples | ✓ | ✓ | Specialized | FFI | Ref | ||

UndecidableInstances | Specialized | Typelevel Programming | Ref | Multiparam Typeclasses | |||

UnicodeSyntax | ✓ | Specialized | Syntax Extension | Ref | |||

UnliftedFFITypes | Specialized | FFI | Ref | Cmm | |||

Unsafe | Specialized | Security Auditing | Ref | Safe Haskell | |||

ViewPatterns | ✓ | ✓ | General | Syntax Extension | Ref | View Patterns |

## The Benign

It's not obvious which extensions are the most common but it's fairly safe to say that these extensions are benign and are safely used extensively:

- OverloadedStrings
- FlexibleContexts
- FlexibleInstances
- GeneralizedNewtypeDeriving
- TypeSynonymInstances
- MultiParamTypeClasses
- FunctionalDependencies
- NoMonomorphismRestriction
- GADTs
- BangPatterns
- DeriveGeneric
- ScopedTypeVariables

## The Dangerous

GHC's typechecker sometimes just casually tells us to enable language extensions when it can't solve certain problems. These include:

- DatatypeContexts
- OverlappingInstances
- IncoherentInstances
- ImpredicativeTypes
- AllowAmbigiousTypes

These almost always indicate a design flaw and shouldn't be turned on to remedy the error at hand, as much as GHC might suggest otherwise!

## Type Inference

Inference in Haskell is usually precise, although there are several boundary cases where inference is difficult or impossible to infer a principal type of an expression. There a two common cases:

#### Mutually Recursive Binding Groups

```
f x = const x g
g y = f 'A'
```

The inferred type signatures are correct in their usage, but don't represent the most general signatures. When GHC analyzes the module it analyzes the dependencies of expressions on each other, groups them together, and applies substitutions from unification across mutually defined groups. As such the inferred types may not be the most general types possible, and an explicit signature may be desired.

```
-- Inferred types
f :: Char -> Char
g :: t -> Char
-- Most general types
f :: a -> a
g :: a -> Char
```

#### Polymorphic recursion

```
data Tree a = Leaf | Bin a (Tree (a, a))
size Leaf = 0
size (Bin _ t) = 1 + 2 * size t
```

The problem with this expression is because the inferred type variable `a`

in `size`

spans two possible types (`a`

and `(a,a)`

), the recursion is polymorphic. These two types won't pass the occurs-check of the typechecker and it yields an incorrect inferred type.

```
Occurs check: cannot construct the infinite type: t0 = (t0, t0)
Expected type: Tree t0
Actual type: Tree (t0, t0)
In the first argument of `size', namely `t'
In the second argument of `(*)', namely `size t'
In the second argument of `(+)', namely `2 * size t'
```

Simply adding an explicit type signature corrects this. Type inference using polymorphic recursion is undecidable in the general case.

```
size :: Tree a -> Int
size Leaf = 0
size (Bin _ t) = 1 + 2 * size t
```

See: Static Semantics of Function and Pattern Bindings

## Monomorphism Restriction

The most common edge case of the inference is known as the dreaded *monomorphism restriction*.

When the toplevel declarations of a module are generalized the monomorphism restricts that toplevel values (i.e. expressions not under a lambda ) whose type contains the subclass of the `Num`

type from the Prelude are not generalized and instead are instantiated with a monotype tried sequentially from the list specified by the `default`

which is normally `Integer`

, then `Double`

.

```
-- Double is inferred by type inferencer.
example1 :: Double
example1 = 3.14
-- In the presense of a lambda, a different type is inferred!
example2 :: Fractional a => t -> a
example2 _ = 3.14
default (Integer, Double)
```

As of GHC 7.8, the monomorphism restriction is switched off by default in GHCi.

```
λ: set +t
λ: 3
3
it :: Num a => a
λ: default (Double)
λ: 3
3.0
it :: Num a => a
```

## Extended Defaulting

Haskell normally applies several defaulting rules for ambigious literals in the absence of an explicit type signature. When an ambiguous literal is typechecked if at least one of its typeclass constraints is numeric and all of its classes are standard library classes, the module's default list is consulted, and the first type from the list that will satisfy the context of the type variable is instantiated. So for instance given the following default rules.

`default (C1 a,...,Cn a)`

The following set of heuristics is used to determine what to instnatiate the ambiguous type variable to.

- The type variable a appears in no other constraints
- All the classes Ci are standard.
- At least one of the classes Ci is numeric.

The default default is (Integer, Double)

This is normally fine, but sometimes we'd like more granular control over defaulting. The `-XExtendedDefaultRules`

loosens the restriction that we're constrained with working on Numerical typeclasses and the constraint that we can only work with standard library classes. If we'd like to have our string literals (using -XOverlodaedStrings) automatically default to the more efficient `Text`

implementation instead of `String`

we can twiddle the flag and GHC will perform the right substitution without the need for an explicit annotation on every string literal.

```
{-# LANGUAGE OverloadedStrings #-}
{-# LANGUAGE ExtendedDefaultRules #-}
import qualified Data.Text as T
default (T.Text)
example = "foo"
```

For code typed at the GHCi prompt, the `-XExtendedDefaultRules`

flag is always on, and cannot be switched off.

## Safe Haskell

As everyone eventually finds out there are several functions within the implementation of GHC ( not the Haskell language ) that can be used to subvert the type-system, they are marked with the prefix `unsafe`

. These functions exist only for when one can manually prove the soundness of an expression but can't express this property in the type-system or externalities to Haskell.

```
unsafeCoerce :: a -> b
unsafePerformIO :: IO a -> a
```

Using these functions to subvert the Haskell typesystem will cause all measure of undefined behavior with unimaginable pain and suffering, and are strongly discouraged. When initially starting out with Haskell there are no legitimate reason to use these functions at all, period.

The Safe Haskell language extensions allow us to restrict the use of unsafe language features using `-XSafe`

which restricts the import of modules which are themselves marked as Safe. It also forbids the use of certain language extensions (`-XTemplateHaskell`

) which can be used to produce unsafe code. The primary use case of these extensions is security auditing.

```
{-# LANGUAGE Safe #-}
{-# LANGUAGE Trustworthy #-}
```

```
{-# LANGUAGE Safe #-}
import Unsafe.Coerce
import System.IO.Unsafe
bad1 :: String
bad1 = unsafePerformIO getLine
bad2 :: a
bad2 = unsafeCoerce 3.14 ()
```

```
Unsafe.Coerce: Can't be safely imported!
The module itself isn't safe.
```

See: Safe Haskell

## Partial Type Signatures

The same hole technique can be applied at the toplevel for signatures:

```
const' :: _
const' x y = x
```

```
[1 of 1] Compiling Main ( src/typedhole.hs, interpreted )
typedhole.hs:3:11:
Found hole ‘_’ with type: t1 -> t -> t1
Where: ‘t’ is a rigid type variable bound by
the inferred type of const' :: t1 -> t -> t1 at foo.hs:4:1
‘t1’ is a rigid type variable bound by
the inferred type of const' :: t1 -> t -> t1 at foo.hs:4:1
To use the inferred type, enable PartialTypeSignatures
In the type signature for ‘const'’: _
Failed, modules loaded: none.
```

Pattern wildcards can also be given explicit names so that GHC will use when reporting the inferred type in the resulting message.

```
foo :: _a -> _a
foo _ = False
```

```
typedhole.hs:6:9:
Couldn't match expected type ‘_a’ with actual type ‘Bool’
‘_a’ is a rigid type variable bound by
the type signature for foo :: _a -> _a at foo.hs:5:8
Relevant bindings include foo :: _a -> _a (bound at foo.hs:6:1)
In the expression: False
In an equation for ‘foo’: foo _ = False
Failed, modules loaded: none.
```

The same wildcards can be used in type contexts to dump out inferred type class constraints:

```
succ' :: _ => a -> a
succ' x = x + 1
```

```
typedhole.hs:3:10:
Found hole ‘_’ with inferred constraints: (Num a)
To use the inferred type, enable PartialTypeSignatures
In the type signature for ‘succ'’: _ => a -> a
Failed, modules loaded: none.
```

When the flag `-XPartialTypeSignatures`

is passed to GHC and the inferred type is unambiguous, GHC will let us leave the holes in place and the compilation will proceed.

```
typedhole.hs:3:10: Warning:
Found hole ‘_’ with type: w_
Where: ‘w_’ is a rigid type variable bound by
the inferred type of succ' :: w_ -> w_1 -> w_ at foo.hs:4:1
In the type signature for ‘succ'’: _ -> _ -> _
```

## Recursive Do

Recursive do notation allows to use to self-reference expressions on both sides of a monadic bind. For instance the following uses lazy evaluation to generate a infinite list. This is sometimes used for instantiating cyclic datatypes inside of a monadic context that need to hold a reference to themselves.

```
{-# LANGUAGE RecursiveDo #-}
justOnes :: Maybe [Int]
justOnes = do
rec xs <- Just (1:xs)
return (map negate xs)
```

## Applicative Do

By default GHC desugars do-notation to use implicit invocations of bind and return.

```
test :: Monad m => m (a, b, c)
test = do
a <- f
b <- g
c <- h
return (a, b, c)
```

Desugars into:

```
test :: Monad m => m (a, b, c)
test =
f >>= \a ->
g >>= \b ->
h >>= \c ->
return (a, b, c)
```

With `ApplicativeDo`

this instead desugars into use of applicative combinators and a laxer Applicative constraint.

```
test :: Applicative m => m (a, b, c)
test = (,,) <$> f <*> g <*> h
```

## Pattern Guards

Pattern guards are an extension to the pattern matching syntax. Given a `<-`

pattern qualifier, the right hand side is evaluated and matched against the pattern on the left. If the match fails then the whole guard fails and the next equation is tried. If it succeeds, then the appropriate binding takes place, and the next qualifier is matched, in the augmented environment.

```
{-# LANGUAGE PatternGuards #-}
combine env x y
| Just a <- lookup x env
, Just b <- lookup y env
= Just $ a + b
| otherwise = Nothing
```

## ViewPatterns

View patterns are like pattern guards that can be nested inside of other patterns. They are a convenient way of pattern-matching against values of algebraic data types.

```
{-# LANGUAGE ViewPatterns #-}
{-# LANGUAGE NoMonomorphismRestriction #-}
import Safe
lookupDefault :: Eq a => a -> b -> [(a,b)] -> b
lookupDefault k _ (lookup k -> Just s) = s
lookupDefault _ d _ = d
headTup :: (a, [t]) -> [t]
headTup (headMay . snd -> Just n) = [n]
headTup _ = []
headNil :: [a] -> [a]
headNil (headMay -> Just x) = [x]
headNil _ = []
```

## TupleSections

```
{-# LANGUAGE TupleSections #-}
first :: a -> (a, Bool)
first = (,True)
second :: a -> (Bool, a)
second = (True,)
```

```
f :: t -> t1 -> t2 -> t3 -> (t, (), t1, (), (), t2, t3)
f = (,(),,(),(),,)
```

## MultiWayIf

```
{-# LANGUAGE MultiWayIf #-}
bmiTell :: Float -> String
bmiTell bmi = if
| bmi <= 18.5 -> "You're underweight."
| bmi <= 25.0 -> "You're average weight."
| bmi <= 30.0 -> "You're overewight."
| otherwise -> "You're a whale."
```

## EmptyCase

GHC normally requires at least one pattern branch in case statement this restriction can be relaxed with -XEmptyCase. The case statement then immediately yields a `Non-exhaustive patterns in case`

if evaluated.

`test = case of`

## LambdaCase

For case statements, LambdaCase allows the elimination of redundant free variables introduced purely for the case of pattern matching on.

```
\case
p1 -> 32
p2 -> 32
```

```
\temp -> case temp of
p1 -> 32
p2 -> 32
```

```
{-# LANGUAGE LambdaCase #-}
data Exp a
= Lam a (Exp a)
| Var a
| App (Exp a) (Exp a)
example :: Exp a -> a
example = \case
Lam a b -> a
Var a -> a
App a b -> example a
```

## NumDecimals

NumDecimals allows the use of exponential notation for integral literals that are not necessarily floats. Without it, any use of expontial notation induces a Fractional class constraint.

```
googol :: Fractional a => a
googol = 1e100
```

```
{-# LANGUAGE NumDecimals #-}
googol :: Num a => a
googol = 1e100
```

## PackageImports

Package imports allows us to disambiguate hierarchical package names by their respective package key. This is useful in the case where you have to imported packages that expose the same module. In practice most of the common libraries have taken care to avoid conflicts in the namespace and this is not usually a problem in most modern Haskell.

For example we could explicitly ask GHC to resolve that `Control.Monad.Error`

package be drawn from the `mtl`

library.

```
import qualified "mtl" Control.Monad.Error as Error
import qualified "mtl" Control.Monad.State as State
import qualified "mtl" Control.Monad.Reader as Reader
```

## RecordWildCards

Record wild cards allow us to expand out the names of a record as variables scoped as the labels of the record implicitly. The extension can be used to extract variables names into a scope or to assign to variables in a record drawing, aligning the record's labels with the variables in scope for the assignment. The syntax introduced is the `{..}`

pattern selector.

```
{-# LANGUAGE RecordWildCards #-}
{-# LANGUAGE OverloadedStrings #-}
import Data.Text
data Example = Example
{ e1 :: Int
, e2 :: Text
, e3 :: Text
} deriving (Show)
-- Extracting from a record using wildcards.
scope :: Example -> (Int, Text, Text)
scope Example {..} = (e1, e2, e3)
-- Assign to a record using wildcards.
assign :: Example
assign = Example {..}
where
(e1, e2, e3) = (1, "Kirk", "Picard")
```

## NamedFieldPuns

Provides alternative syntax for accessing record fields in a pattern match.

```
data D = D {a :: Int, b :: Int}
f :: D -> Int
f D {a, b} = a - b
-- Order doesn't matter
g :: D -> Int
g D {b, a} = a - b
```

## PatternSynonyms

Suppose we were writing a typechecker, it would be very common to include a distinct `TArr`

term to ease the telescoping of function signatures, this is what GHC does in its Core language. Even though technically it could be written in terms of more basic application of the `(->)`

constructor.

```
data Type
= TVar TVar
| TCon TyCon
| TApp Type Type
| TArr Type Type
deriving (Show, Eq, Ord)
```

With pattern synonyms we can eliminate the extraneous constructor without losing the convenience of pattern matching on arrow types.

```
{-# LANGUAGE PatternSynonyms #-}
pattern TArr t1 t2 = TApp (TApp (TCon "(->)") t1) t2
```

So now we can write an eliminator and constructor for arrow type very naturally.

```
{-# LANGUAGE PatternSynonyms #-}
import Data.List (foldl1')
type Name = String
type TVar = String
type TyCon = String
data Type
= TVar TVar
| TCon TyCon
| TApp Type Type
deriving (Show, Eq, Ord)
pattern TArr t1 t2 = TApp (TApp (TCon "(->)") t1) t2
tapp :: TyCon -> [Type] -> Type
tapp tcon args = foldl TApp (TCon tcon) args
arr :: [Type] -> Type
arr ts = foldl1' (\t1 t2 -> tapp "(->)" [t1, t2]) ts
elimTArr :: Type -> [Type]
elimTArr (TArr (TArr t1 t2) t3) = t1 : t2 : elimTArr t3
elimTArr (TArr t1 t2) = t1 : elimTArr t2
elimTArr t = [t]
-- (->) a ((->) b a)
-- a -> b -> a
to :: Type
to = arr [TVar "a", TVar "b", TVar "a"]
from :: [Type]
from = elimTArr to
```

Pattern synonyms can be exported from a module like any other definition by prefixing them with the prefix `pattern`

.

```
module MyModule (
pattern Elt
) where
pattern Elt = [a]
```

## DeriveTraversable

## DeriveFoldable

## DeriveFunctor

## DeriveGeneric

## DeriveAnyClass

With `-XDeriveAnyClass`

we can derive any class. The deriving logic s generates an instance declaration for the type with no explicitly-defined methods. If the typeclass implements a default for each method then this will be well-defined and give rise to an automatic instances.

## StaticPointers

## DuplicateRecordFields

GHC 8.0 introduced the DuplicateRecordFields extensions which loosens GHC's restriction on records in the same module with identical accessors. The precise type that is being projected into is now deferred to the callsite.

```
{-# LANGUAGE DuplicateRecordFields #-}
data Person = Person { id :: Int }
data Animal = Animal { id :: Int }
data Vegetable = Vegetable { id :: Int }
test :: (Person, Animal, Vegetable)
test = (Person {id = 1}, Animal {id = 2}, Vegetable {id = 3})
```

Using just DuplicateRecordFields, projection is still not supported so the following will not work. OverloadedLabels fixes this to some extent.

```
test :: (Person, Animal, Vegetable)
test = (id (Person 1), id (Animal 2), id (Animal 3))
```

## OverloadedLabels

GHC 8.0 also introduced the OverloadedLabels extension which allows a limited form of polymorphism over labels that share the same name.

To work with overloaded labels types we need to enable several language extensions to work with promoted strings and multiparam typeclasses that underly it's implementation.

```
extract :: IsLabel "id" t => t
extract = #id
```

```
{-# LANGUAGE OverloadedLabels #-}
{-# LANGUAGE FlexibleInstances #-}
{-# LANGUAGE MultiParamTypeClasses #-}
{-# LANGUAGE DuplicateRecordFields #-}
{-# LANGUAGE ExistentialQuantification #-}
import GHC.Records (HasField(..))
import GHC.OverloadedLabels (IsLabel(..))
data S = MkS { foo :: Int }
data T x y z = forall b . MkT { foo :: y, bar :: b }
instance HasField x r a => IsLabel x (r -> a) where
fromLabel = getField
main :: IO ()
main = do
print (#foo (MkS 42))
print (#foo (MkT True False))
```

See:

## Cpp

The C++ preprocessor is the fallback whenever we really need to separate out logic that has to span multiple versions of GHC and language changes while maintaining backwards compatibility. It can dispatch on the version of GHC being used to compile a module.

```
{-# LANGUAGE CPP #-}
#if (__GLASGOW_HASKELL__ > 710)
-- Imports for GHC 7.10.x
#else
-- Imports for other GHC
#endif
```

To demarcate code based on the operating system compiled on.

```
{-# LANGUAGE CPP #-}
#ifdef OS_Linux
-- Linux specific logic
#else
# ifdef OS_Win32
-- Windows specific logic
# else
# ifdef OS_Mac
-- Macintosh specific logic
# else
-- Other operating systems
# endif
# endif
#endif
```

Or on the version of the base library used.

```
#if !MIN_VERSION_base(4,6,0)
-- Base specific logic
#endif
```

It can also be abused to do terrible things like metaprogrammming with strings, but please don't do this.

## Historical Extensions

Several language extensions have either been absorbed into the core language or become deprecated in favor of others. Others are just considered misfeatures.

- Rank2Types - Rank2Types has been subsumed by RankNTypes
- XPolymorphicComponents - Was an implementation detail of higher-rank polymorphism that no longer exists.
- NPlusKPatterns - These were largely considered an ugly edge-case of pattern matching language that was best removed.
- TraditionalRecordSyntax - Traditional record syntax was an extension to the Haskell 98 specification for what we now consider standard record syntax.
- OverlappingInstances - Subsumed by explicit OVERLAPPING pragmas.
- IncoherentInstances - Subsumed by explicit INCOHERENT pragmas.
- NullaryTypeClasses - Subsumed by explicit Multiparameter Typeclasses with no parameters.

# Type Classes

## Minimal Annotations

In the presence of default implementations of typeclasses methods, there may be several ways to implement a typeclass. For instance Eq is entirely defined by either defining when two values are equal or not equal by implying taking the negation of the other. We can define equality in terms of non-equality and vice-versa.

```
class Eq a where
(==), (/=) :: a -> a -> Bool
x == y = not (x /= y)
x /= y = not (x == y)
```

Before 7.6.1 there was no way to specify what was the "minimal" definition required to implement a typeclass

```
class Eq a where
(==), (/=) :: a -> a -> Bool
x == y = not (x /= y)
x /= y = not (x == y)
{-# MINIMAL (==) #-}
{-# MINIMAL (/=) #-}
```

Minimal pragmas are boolean expressions, with `|`

as logical `OR`

, *either* definition must be defined). Comma indicates logical `AND`

where both sides *both* definitions must be defined.

```
{-# MINIMAL (==) | (/=) #-} -- Either (==) or (/=)
{-# MINIMAL (==) , (/=) #-} -- Both (==) and (/=)
```

Compiling the `-Wmissing-methods`

will warn when a instance is defined that does not meet the minimal criterion.

## FlexibleInstances

```
{-# LANGUAGE FlexibleInstances #-}
class MyClass a
-- Without flexible instances, all instance heads must be type variable. The
-- following would be legal.
instance MyClass (Maybe a)
-- With flexible instances, typeclass heads can be arbitrary nested types. The
-- following would be forbidden without it.
instance MyClass (Maybe Int)
```

## FlexibleContexts

```
{-# LANGUAGE FlexibleContexts #-}
class MyClass a
-- Without flexible contexts, all contexts must be type variable. The
-- following would be legal.
instance (MyClass a) => MyClass (Either a b)
-- With flexible contexts, typeclass contexts can be arbitrary nested types. The
-- following would be forbidden without it.
instance (MyClass (Maybe a)) => MyClass (Either a b)
```

## OverlappingInstances

Typeclasses are normally globally coherent, there is only ever one instance that can be resolved for a type unambiguously for a type at any call site in the program. There are however extensions to loosen this restriction and perform more manual direction of the instance search.

Overlapping instances loosens the coherent condition (there can be multiple instances) but introduces a criterion that it will resolve to the most specific one.

```
{-# LANGUAGE FlexibleInstances #-}
{-# LANGUAGE OverlappingInstances #-}
{-# LANGUAGE MultiParamTypeClasses #-}
class MyClass a b where
fn :: (a,b)
instance MyClass Int b where
fn = error "b"
instance MyClass a Int where
fn = error "a"
instance MyClass Int Int where
fn = error "c"
example :: (Int, Int)
example = fn
```

Historically enabling this on module-level was not the best idea, since generally we define multiple classes in a module only a subset of which may be incoherent. So as of 7.10 we now have the capacity to just annotate instances with the OVERLAPPING and INCOHERENT pragmas.

```
{-# LANGUAGE FlexibleInstances #-}
{-# LANGUAGE MultiParamTypeClasses #-}
class MyClass a b where
fn :: (a,b)
instance {-# OVERLAPPING #-} MyClass Int b where
fn = error "b"
instance {-# OVERLAPPING #-} MyClass a Int where
fn = error "a"
instance {-# OVERLAPPING #-} MyClass Int Int where
fn = error "c"
example :: (Int, Int)
example = fn
```

## IncoherentInstances

Incoherent instance loosens the restriction that there be only one specific instance, will choose one arbitrarily (based on the arbitrary sorting of it's internal representation ) and the resulting program will typecheck. This is generally pretty crazy and usually a sign of poor design.

```
{-# LANGUAGE FlexibleInstances #-}
{-# LANGUAGE IncoherentInstances #-}
{-# LANGUAGE MultiParamTypeClasses #-}
class MyClass a b where
fn :: (a,b)
instance MyClass Int b where
fn = error "a"
instance MyClass a Int where
fn = error "b"
example :: (Int, Int)
example = fn
```

There is also an incoherent instance.

```
{-# LANGUAGE FlexibleInstances #-}
{-# LANGUAGE MultiParamTypeClasses #-}
class MyClass a b where
fn :: (a,b)
instance {-# INCOHERENT #-} MyClass a Int where
fn = error "general"
instance {-# INCOHERENT #-} MyClass Int Int where
fn = error "specific"
example :: (Int, Int)
example = fn
```

## TypeSynonymInstances

```
{-# LANGUAGE TypeSynonymInstances #-}
{-# LANGUAGE FlexibleInstances #-}
type IntList = [Int]
class MyClass a
-- Without type synonym instances, we're forced to manually expand out type
-- synonyms in the typeclass head.
instance MyClass [Int]
-- With it GHC will do this for us automatically. Type synonyms still need to
-- be fully applied.
instance MyClass IntList
```

# Laziness

Again, a subject on which *much* ink has been spilled. There is an ongoing discussion in the land of Haskell about the compromises between lazy and strict evaluation, and there are nuanced arguments for having either paradigm be the default. Haskell takes a hybrid approach and allows strict evaluation when needed and uses laziness by default. Needless to say, we can always find examples where strict evaluation exhibits worse behavior than lazy evaluation and vice versa.

The primary advantage of lazy evaluation in the large is that algorithms that operate over both unbounded and bounded data structures can inhabit the same type signatures and be composed without additional need to restructure their logic or force intermediate computations. Languages that attempt to bolt laziness on to a strict evaluation model often bifurcate classes of algorithms into ones that are hand-adjusted to consume unbounded structures and those which operate over bounded structures. In strict languages mixing and matching between lazy vs strict processing often necessitates manifesting large intermediate structures in memory when such composition would "just work" in a lazy language.

By virtue of Haskell being the only language to actually explore this point in the design space to the point of being industrial strength; knowledge about lazy evaluation is not widely absorbed into the collective programmer consciousness and can often be non-intuitive to the novice. This doesn't reflect on the model itself, merely on the need for more instruction material and research on optimizing lazy compilers.

The paradox of Haskell is that it explores so many definably unique ideas ( laziness, purity, typeclasses ) that it becomes difficult to separate out the discussion of any one from the gestalt of the whole implementation.

See:

- Oh My Laziness!
- Reasoning about Laziness
- Lazy Evaluation of Haskell
- More Points For Lazy Evaluation
- How Lazy Evaluation Works in Haskell

## Strictness

There are several evaluation models for the lambda calculus:

- Strict - Evaluation is said to be strict if all arguments are evaluated before the body of a function.
- Non-strict - Evaluation is non-strict if the arguments are not necessarily evaluated before entering the body of a function.

These ideas give rise to several models, Haskell itself use the *call-by-need* model.

Model | Strictness | Description |
---|---|---|

Call-by-value | Strict | arguments evaluated before function entered |

Call-by-name | Non-strict | arguments passed unevaluated |

Call-by-need | Non-strict | arguments passed unevaluated but an expression is only evaluated once (sharing) |

## Seq and WHNF

A term is said to be in *weak head normal-form* if the outermost constructor or lambda cannot be reduced further. A term is said to be in *normal form* if it is fully evaluated and all sub-expressions and thunks contained within are evaluated.

```
-- In Normal Form
42
(2, "foo")
\x -> x + 1
-- Not in Normal Form
1 + 2
(\x -> x + 1) 2
"foo" ++ "bar"
(1 + 1, "foo")
-- In Weak Head Normal Form
(1 + 1, "foo")
\x -> 2 + 2
'f' : ("oo" ++ "bar")
-- Not In Weak Head Normal Form
1 + 1
(\x -> x + 1) 2
"foo" ++ "bar"
```

In Haskell normal evaluation only occurs at the outer constructor of case-statements in Core. If we pattern match on a list we don't implicitly force all values in the list. An element in a data structure is only evaluated up to the most outer constructor. For example, to evaluate the length of a list we need only scrutinize the outer Cons constructors without regard for their inner values.

```
λ: length [undefined, 1]
2
λ: head [undefined, 1]
Prelude.undefined
λ: snd (undefined, 1)
1
λ: fst (undefined, 1)
Prelude.undefined
```

For example, in a lazy language the following program terminates even though it contains diverging terms.

```
ignore :: a -> Int
ignore x = 0
loop :: a
loop = loop
main :: IO ()
main = print $ ignore loop
```

In a strict language like OCaml ( ignoring its suspensions for the moment ), the same program diverges.

```
let ignore x = 0;;
let rec loop a = loop a;;
print_int (ignore (loop ()));
```

In Haskell a *thunk* is created to stand for an unevaluated computation. Evaluation of a thunk is called *forcing* the thunk. The result is an *update*, a referentially transparent effect, which replaces the memory representation of the thunk with the computed value. The fundamental idea is that a thunk is only updated once ( although it may be forced simultaneously in a multi-threaded environment ) and its resulting value is shared when referenced subsequently.

The command `:sprint`

can be used to introspect the state of unevaluated thunks inside an expression without forcing evaluation. For instance:

```
λ: let a = [1..] :: [Integer]
λ: let b = map (+ 1) a
λ: :sprint a
a = _
λ: :sprint b
b = _
λ: a !! 4
5
λ: :sprint a
a = 1 : 2 : 3 : 4 : 5 : _
λ: b !! 10
12
λ: :sprint a
a = 1 : 2 : 3 : 4 : 5 : 6 : 7 : 8 : 9 : 10 : 11 : _
λ: :sprint b
b = _ : _ : _ : _ : _ : _ : _ : _ : _ : _ : 12 : _
```

While a thunk is being computed its memory representation is replaced with a special form known as *blackhole* which indicates that computation is ongoing and allows for a short circuit for when a computation might depend on itself to complete. The implementation of this is some of the more subtle details of the GHC runtime.

The `seq`

function introduces an artificial dependence on the evaluation of order of two terms by requiring that the first argument be evaluated to WHNF before the evaluation of the second. The implementation of the `seq`

function is an implementation detail of GHC.

```
seq :: a -> b -> b
⊥ `seq` a = ⊥
a `seq` b = b
```

The infamous `foldl`

is well-known to leak space when used carelessly and without several compiler optimizations applied. The strict `foldl'`

variant uses seq to overcome this.

```
foldl :: (a -> b -> a) -> a -> [b] -> a
foldl f z [] = z
foldl f z (x:xs) = foldl f (f z x) xs
```

```
foldl' :: (a -> b -> a) -> a -> [b] -> a
foldl' _ z [] = z
foldl' f z (x:xs) = let z' = f z x in z' `seq` foldl' f z' xs
```

In practice, a combination between the strictness analyzer and the inliner on `-O2`

will ensure that the strict variant of `foldl`

is used whenever the function is inlinable at call site so manually using `foldl'`

is most often not required.

Of important note is that GHCi runs without any optimizations applied so the same program that performs poorly in GHCi may not have the same performance characteristics when compiled with GHC.

## Strictness Annotations

The extension `BangPatterns`

allows an alternative syntax to force arguments to functions to be wrapped in seq. A bang operator on an arguments forces its evaluation to weak head normal form before performing the pattern match. This can be used to keep specific arguments evaluated throughout recursion instead of creating a giant chain of thunks.

```
{-# LANGUAGE BangPatterns #-}
sum :: Num a => [a] -> a
sum = go 0
where
go !acc (x:xs) = go (acc + x) xs
go acc [] = acc
```

This is desugared into code effectively equivalent to the following:

```
sum :: Num a => [a] -> a
sum = go 0
where
go acc _ | acc `seq` False = undefined
go acc (x:xs) = go (acc + x) xs
go acc [] = acc
```

Function application to seq'd arguments is common enough that it has a special operator.

```
($!) :: (a -> b) -> a -> b
f $! x = let !vx = x in f vx
```

## Strict Haskell

As of GHC 8.0 strictness annotations can be applied to all definitions in a module automatically. In previous versions it was necessary to definitions via explicit syntactic annotations at all sites.

#### StrictData

Enabling StrictData makes constructor fields strict by default on any module it is enabled on.

```
{-# LANGUAGE StrictData #-}
data Employee = Employee
{ name :: T.Text
, age :: Int
}
```

Is equivalent to:

```
data Employee = Employee
{ name :: !T.Text
, age :: !Int
}
```

#### Strict

Strict implies `-XStrictData`

and extends strictness annotations to all arguments of functions.

`f x y = x + y`

Is equivalent to the following function declaration with explicit bang patterns:

`f !x !y = x + y`

On a module-level this effectively makes Haskell a call-by-value language with some caveats. All arguments to functions are now explicitly evaluated and all data in constructors within this module are in head normal form by construction. However there are some subtle points to this that are better explained in the language guide.

## Deepseq

There are often times when for performance reasons we need to deeply evaluate a data structure to normal form leaving no terms unevaluated. The `deepseq`

library performs this task.

The typeclass `NFData`

(Normal Form Data) allows us to seq all elements of a structure across any subtypes which themselves implement NFData.

```
class NFData a where
rnf :: a -> ()
rnf a = a `seq` ()
deepseq :: NFData a => a -> b -> b
($!!) :: (NFData a) => (a -> b) -> a -> b
```

```
instance NFData Int
instance NFData (a -> b)
instance NFData a => NFData (Maybe a) where
rnf Nothing = ()
rnf (Just x) = rnf x
instance NFData a => NFData [a] where
rnf [] = ()
rnf (x:xs) = rnf x `seq` rnf xs
```

```
[1, undefined] `seq` ()
-- ()
[1, undefined] `deepseq` ()
-- Prelude.undefined
```

To force a data structure itself to be fully evaluated we share the same argument in both positions of deepseq.

```
force :: NFData a => a -> a
force x = x `deepseq` x
```

## Irrefutable Patterns

A lazy pattern doesn't require a match on the outer constructor, instead it lazily calls the accessors of the values as needed. In the presence of a bottom, we fail at the usage site instead of the outer pattern match.

```
f :: (a, b) -> Int
f (a,b) = const 1 a
g :: (a, b) -> Int
g ~(a,b) = const 1 a
-- λ: f undefined
-- *** Exception: Prelude.undefined
-- λ: g undefined
-- 1
j :: Maybe t -> t
j ~(Just x) = x
k :: Maybe t -> t
k (Just x) = x
-- λ: j Nothing
-- *** Exception: src/05-laziness/lazy_patterns.hs:15:1-15: Irrefutable pattern failed for pattern (Just x)
--
-- λ: k Nothing
-- *** Exception: src/05-laziness/lazy_patterns.hs:18:1-14: Non-exhaustive patterns in function k
```

# Prelude

## What to Avoid?

Haskell being a 25 year old language has witnessed several revolutions in the way we structure and compose functional programs. Yet as a result several portions of the Prelude still reflect old schools of thought that simply can't be removed without breaking significant parts of the ecosystem.

Currently it really only exists in folklore which parts to use and which not to use, although this is a topic that almost all introductory books don't mention and instead make extensive use of the Prelude for simplicity's sake.

The short version of the advice on the Prelude is:

- Avoid String.
- Use
`fmap`

instead of`map`

. - Use Foldable and Traversable instead of the Control.Monad, and Data.List versions of traversals.
- Avoid partial functions like
`head`

and`read`

or use their total variants. - Avoid exceptions, use ExceptT or Either instead.
- Avoid boolean blind functions.

The instances of Foldable for the list type often conflict with the monomorphic versions in the Prelude which are left in for historical reasons. So often times it is desirable to explicitly mask these functions from implicit import and force the use of Foldable and Traversable instead.

Of course often times one wishes only to use the Prelude explicitly and one can explicitly import it qualified and use the pieces as desired without the implicit import of the whole namespace.

`import qualified Prelude as P`

## What Should be in Base

To get work done you probably need.

- async
- bytestring
- containers
- mtl
- stm
- text
- transformers
- unordered-containers
- vector
- filepath
- directory
- process
- unix
- deepseq
- optparse-applicative

## Custom Preludes

The default Prelude can be disabled in it's entirety by twiddling the `-XNoImplicitPrelude`

flag.

`{-# LANGUAGE NoImplicitPrelude #-}`

We are then free to build an equivalent Prelude that is more to our liking. Using module reexporting we can pluck the good parts of the prelude and libraries like `safe`

to build up a more industrial focused set of default functions. For example:

```
module Custom (
module Exports,
) where
import Data.Int as Exports
import Data.Tuple as Exports
import Data.Maybe as Exports
import Data.String as Exports
import Data.Foldable as Exports
import Data.Traversable as Exports
import Control.Monad.Trans.Except
as Exports
(ExceptT(ExceptT), Except, except, runExcept, runExceptT,
mapExcept, mapExceptT, withExcept, withExceptT)
```

The Prelude itself is entirely replicable as well, presuming that an entire project is compiled without the implicit Prelude. Several packages have arisen that supply much of the same functionality in a way that appeals to more modern design principles.

## Protolude

Protolude is a minimalist Prelude which provides many sensible defaults for writing modern Haskell and is compatible with existing code.

```
{-# LANGUAGE NoImplicitPrelude #-}
import Protolude
```

Other examples for alternative Preludes include (your mileage may vary with these):

## Partial Functions

A *partial function* is a function which doesn't terminate and yield a value for all given inputs. Conversely a *total function* terminates and is always defined for all inputs. As mentioned previously, certain historical parts of the Prelude are full of partial functions.

The difference between partial and total functions is the compiler can't reason about the runtime safety of partial functions purely from the information specified in the language and as such the proof of safety is left to the user to guarantee. They are safe to use in the case where the user can guarantee that invalid inputs cannot occur, but like any unchecked property its safety or not-safety is going to depend on the diligence of the programmer. This very much goes against the overall philosophy of Haskell and as such they are discouraged when not necessary.

```
head :: [a] -> a
read :: Read a => String -> a
(!!) :: [a] -> Int -> a
```

## Safe

The Prelude has total variants of the historical partial functions (i.e. `Text.Read.readMaybe`

)in some cases, but often these are found in the various utility libraries like `safe`

.

The total versions provided fall into three cases:

`May`

- return Nothing when the function is not defined for the inputs`Def`

- provide a default value when the function is not defined for the inputs`Note`

- call`error`

with a custom error message when the function is not defined for the inputs. This is not safe, but slightly easier to debug!

```
-- Total
headMay :: [a] -> Maybe a
readMay :: Read a => String -> Maybe a
atMay :: [a] -> Int -> Maybe a
-- Total
headDef :: a -> [a] -> a
readDef :: Read a => a -> String -> a
atDef :: a -> [a] -> Int -> a
-- Partial
headNote :: String -> [a] -> a
readNote :: Read a => String -> String -> a
atNote :: String -> [a] -> Int -> a
```

## Boolean Blindness

```
data Bool = True | False
isJust :: Maybe a -> Bool
isJust (Just x) = True
isJust Nothing = False
```

The problem with the boolean type is that there is effectively no difference between True and False at the type level. A proposition taking a value to a Bool takes any information given and destroys it. To reason about the behavior we have to trace the provenance of the proposition we're getting the boolean answer from, and this introduces a whole slew of possibilities for misinterpretation. In the worst case, the only way to reason about safe and unsafe use of a function is by trusting that a predicate's lexical name reflects its provenance!

For instance, testing some proposition over a Bool value representing whether the branch can perform the computation safely in the presence of a null is subject to accidental interchange. Consider that in a language like C or Python testing whether a value is null is indistinguishable to the language from testing whether the value is *not null*. Which of these programs encodes safe usage and which segfaults?

```
# This one?
if p(x):
# use x
elif not p(x):
# don't use x
# Or this one?
if p(x):
# don't use x
elif not p(x):
# use x
```

From inspection we can't tell without knowing how p is defined, the compiler can't distinguish the two either and thus the language won't save us if we happen to mix them up. Instead of making invalid states *unrepresentable* we've made the invalid state *indistinguishable* from the valid one!

The more desirable practice is to match on terms which explicitly witness the proposition as a type ( often in a sum type ) and won't typecheck otherwise.

```
case x of
Just a -> use x
Nothing -> don't use x
-- not ideal
case p x of
True -> use x
False -> don't use x
-- not ideal
if p x
then use x
else don't use x
```

To be fair though, many popular languages completely lack the notion of sum types ( the source of many woes in my opinion ) and only have product types, so this type of reasoning sometimes has no direct equivalence for those not familiar with ML family languages.

In Haskell, the Prelude provides functions like `isJust`

and `fromJust`

both of which can be used to subvert this kind of reasoning and make it easy to introduce bugs and should often be avoided.

## Foldable / Traversable

If coming from an imperative background retraining one's self to think about iteration over lists in terms of maps, folds, and scans can be challenging.

```
Prelude.foldl :: (a -> b -> a) -> a -> [b] -> a
Prelude.foldr :: (a -> b -> b) -> b -> [a] -> b
-- pseudocode
foldr f z [a...] = f a (f b ( ... (f y z) ... ))
foldl f z [a...] = f ... (f (f z a) b) ... y
```

For a concrete consider the simple arithmetic sequence over the binary operator `(+)`

:

```
-- foldr (+) 1 [2..]
(1 + (2 + (3 + (4 + ...))))
```

```
-- foldl (+) 1 [2..]
((((1 + 2) + 3) + 4) + ...)
```

Foldable and Traversable are the general interface for all traversals and folds of any data structure which is parameterized over its element type ( List, Map, Set, Maybe, ...). These two classes are used everywhere in modern Haskell and are extremely important.

A foldable instance allows us to apply functions to data types of monoidal values that collapse the structure using some logic over `mappend`

.

A traversable instance allows us to apply functions to data types that walk the structure left-to-right within an applicative context.

```
class (Functor f, Foldable f) => Traversable f where
traverse :: Applicative g => (a -> g b) -> f a -> g (f b)
class Foldable f where
foldMap :: Monoid m => (a -> m) -> f a -> m
```

The `foldMap`

function is extremely general and non-intuitively many of the monomorphic list folds can themselves be written in terms of this single polymorphic function.

`foldMap`

takes a function of values to a monoidal quantity, a functor over the values and collapses the functor into the monoid. For instance for the trivial Sum monoid:

```
λ: foldMap Sum [1..10]
Sum {getSum = 55}
```

For instance if we wanted to map a list of some abstract element types into a hashtable of elements based on pattern matching we could use it.

```
import Data.Foldable
import qualified Data.Map as Map
data Elt
= Elt Int Double
| Nil
foo :: [Elt] -> Map.Map Int Double
foo = foldMap go
where
go (Elt x y) = Map.singleton x y
go Nil = Map.empty
```

The full Foldable class (with all default implementations) contains a variety of derived functions which themselves can be written in terms of `foldMap`

and `Endo`

.

```
newtype Endo a = Endo {appEndo :: a -> a}
instance Monoid (Endo a) where
mempty = Endo id
Endo f `mappend` Endo g = Endo (f . g)
```

```
class Foldable t where
fold :: Monoid m => t m -> m
foldMap :: Monoid m => (a -> m) -> t a -> m
foldr :: (a -> b -> b) -> b -> t a -> b
foldr' :: (a -> b -> b) -> b -> t a -> b
foldl :: (b -> a -> b) -> b -> t a -> b
foldl' :: (b -> a -> b) -> b -> t a -> b
foldr1 :: (a -> a -> a) -> t a -> a
foldl1 :: (a -> a -> a) -> t a -> a
```

For example:

```
foldr :: (a -> b -> b) -> b -> t a -> b
foldr f z t = appEndo (foldMap (Endo . f) t) z
```

Most of the operations over lists can be generalized in terms of combinations of Foldable and Traversable to derive more general functions that work over all data structures implementing Foldable.

```
Data.Foldable.elem :: (Eq a, Foldable t) => a -> t a -> Bool
Data.Foldable.sum :: (Num a, Foldable t) => t a -> a
Data.Foldable.minimum :: (Ord a, Foldable t) => t a -> a
Data.Traversable.mapM :: (Monad m, Traversable t) => (a -> m b) -> t a -> m (t b)
```

Unfortunately for historical reasons the names exported by foldable quite often conflict with ones defined in the Prelude, either import them qualified or just disable the Prelude. The operations in the Foldable all specialize to the same and behave the same as the ones in Prelude for List types.

```
import Data.Monoid
import Data.Foldable
import Data.Traversable
import Control.Applicative
import Control.Monad.Identity (runIdentity)
import Prelude hiding (mapM_, foldr)
-- Rose Tree
data Tree a = Node a [Tree a] deriving (Show)
instance Functor Tree where
fmap f (Node x ts) = Node (f x) (fmap (fmap f) ts)
instance Traversable Tree where
traverse f (Node x ts) = Node <$> f x <*> traverse (traverse f) ts
instance Foldable Tree where
foldMap f (Node x ts) = f x `mappend` foldMap (foldMap f) ts
tree :: Tree Integer
tree = Node 1 [Node 1 [], Node 2 [] ,Node 3 []]
example1 :: IO ()
example1 = mapM_ print tree
example2 :: Integer
example2 = foldr (+) 0 tree
example3 :: Maybe (Tree Integer)
example3 = traverse (\x -> if x > 2 then Just x else Nothing) tree
example4 :: Tree Integer
example4 = runIdentity $ traverse (\x -> pure (x+1)) tree
```

The instances we defined above can also be automatically derived by GHC using several language extensions. The automatic instances are identical to the hand-written versions above.

```
{-# LANGUAGE DeriveFunctor #-}
{-# LANGUAGE DeriveFoldable #-}
{-# LANGUAGE DeriveTraversable #-}
data Tree a = Node a [Tree a]
deriving (Show, Functor, Foldable, Traversable)
```

See: Typeclassopedia

## Corecursion

`unfoldr :: (b -> Maybe (a, b)) -> b -> [a]`

A recursive function consumes data and eventually terminates, a corecursive function generates data and **coterminates**. A corecursive function is said to be *productive* if it can always evaluate more of the resulting value in bounded time.

```
import Data.List
f :: Int -> Maybe (Int, Int)
f 0 = Nothing
f x = Just (x, x-1)
rev :: [Int]
rev = unfoldr f 10
fibs :: [Int]
fibs = unfoldr (\(a,b) -> Just (a,(b,a+b))) (0,1)
```

## split

The split package provides a variety of missing functions for splitting list and string types.

```
import Data.List.Split
example1 :: [String]
example1 = splitOn "." "foo.bar.baz"
-- ["foo","bar","baz"]
example2 :: [String]
example2 = chunksOf 10 "To be or not to be that is the question."
-- ["To be or n","ot to be t","hat is the"," question."]
```

## monad-loops

The monad-loops package provides a variety of missing functions for control logic in monadic contexts.

```
whileM :: Monad m => m Bool -> m a -> m [a]
untilM :: Monad m => m a -> m Bool -> m [a]
iterateUntilM :: Monad m => (a -> Bool) -> (a -> m a) -> a -> m a
whileJust :: Monad m => m (Maybe a) -> (a -> m b) -> m [b]
```

## Foundation

TODO

See: Foundation

# Strings

## String

**The default String type is broken and should be avoided whenever possible.** Unfortunately for historical reasons large portions of GHC and Base depend on String.

The default Haskell string type is implemented as a naive linked list of characters, this is terrible for most purposes but no one knows how to fix it without rewriting large portions of all code that exists and nobody can commit the time to fix it. So it remains broken, likely forever.

`type String = [Char]`

For more performance sensitive cases there are two libraries for processing textual data: `text`

and `bytestring`

.

**text**- Used for handling unicode data.**bytestring**- Used for handling ASCII data that needs to interchanged with C code or network protocols.

For each of these there are two variants for both text and bytestring.

**lazy**Lazy text objects are encoded as lazy lists of strict chunks of bytes.**strict**Byte vectors are encoded as strict Word8 arrays of bytes or code points

Giving rise to the four types.

Variant | Module |
---|---|

strict text |
Data.Text |

lazy text |
Data.Text.Lazy |

strict bytestring |
Data.ByteString |

lazy bytestring |
Data.ByteString.Lazy |

#### Conversions

Conversions between strings types ( from : left column, to : top row ) are done with several functions across the bytestring and text libraries. The mapping between text and bytestring is inherently lossy so there is some degree of freedom in choosing the encoding. We'll just consider utf-8 for simplicity.

Data.Text | Data.Text.Lazy | Data.ByteString | Data.ByteString.Lazy | |
---|---|---|---|---|

Data.Text | id | fromStrict | encodeUtf8 | encodeUtf8 |

Data.Text.Lazy | toStrict | id | encodeUtf8 | encodeUtf8 |

Data.ByteString | decodeUtf8 | decodeUtf8 | id | fromStrict |

Data.ByteString.Lazy | decodeUtf8 | decodeUtf8 | toStrict | id |

#### Overloaded Strings

With the `-XOverloadedStrings`

extension string literals can be overloaded without the need for explicit packing and can be written as string literals in the Haskell source and overloaded via a typeclass `IsString`

. Sometimes this is desirable.

```
class IsString a where
fromString :: String -> a
```

For instance:

```
λ: :type "foo"
"foo" :: [Char]
λ: :set -XOverloadedStrings
λ: :type "foo"
"foo" :: IsString a => a
```

We can also derive IsString for newtypes using `GeneralizedNewtypeDeriving`

, although much of the safety of the newtype is then lost if it is interchangeable with other strings.

```
newtype Cat = Cat Text
deriving (IsString)
fluffy :: Cat
fluffy = "Fluffy"
```

## Import Conventions

```
import qualified Data.Text as T
import qualified Data.Text.Lazy as TL
import qualified Data.ByteString as BS
import qualified Data.ByteString.Lazy as BL
import qualified Data.ByteString.Char8 as C
import qualified Data.ByteString.Lazy.Char8 as CL
```

```
import qualified Data.Text.IO as TIO
import qualified Data.Text.Lazy.IO as TLIO
```

```
import qualified Data.Text.Encoding as TE
import qualified Data.Text.Lazy.Encoding as TLE
```

## Text

A `Text`

type is a packed blob of Unicode characters.

```
pack :: String -> Text
unpack :: Text -> String
```

```
{-# LANGUAGE OverloadedStrings #-}
import qualified Data.Text as T
-- From pack
myTStr1 :: T.Text
myTStr1 = T.pack ("foo" :: String)
-- From overloaded string literal.
myTStr2 :: T.Text
myTStr2 = "bar"
```

See: Text

## Text.Builder

```
toLazyText :: Builder -> Data.Text.Lazy.Internal.Text
fromLazyText :: Data.Text.Lazy.Internal.Text -> Builder
```

The Text.Builder allows the efficient monoidal construction of lazy Text types without having to go through inefficient forms like String or List types as intermediates.

```
{-# LANGUAGE OverloadedStrings #-}
import Data.Monoid (mconcat, (<>))
import Data.Text.Lazy.Builder (Builder, toLazyText)
import Data.Text.Lazy.Builder.Int (decimal)
import qualified Data.Text.Lazy.IO as L
beer :: Int -> Builder
beer n = decimal n <> " bottles of beer on the wall.\n"
wall :: Builder
wall = mconcat $ fmap beer [1..1000]
main :: IO ()
main = L.putStrLn $ toLazyText wall
```

## ByteString

ByteStrings are arrays of unboxed characters with either strict or lazy evaluation.

```
pack :: String -> ByteString
unpack :: ByteString -> String
```

```
{-# LANGUAGE OverloadedStrings #-}
import qualified Data.ByteString as S
import qualified Data.ByteString.Char8 as S8
-- From pack
bstr1 :: S.ByteString
bstr1 = S.pack ("foo" :: String)
-- From overloaded string literal.
bstr2 :: S.ByteString
bstr2 = "bar"
```

See:

## utf8-string

See: utf8-string

## base64-bytestring

## Printf

Haskell also has a variadic `printf`

function in the style of C.

```
import Data.Text
import Text.Printf
a :: Int
a = 3
b :: Double
b = 3.14159
c :: String
c = "haskell"
example :: String
example = printf "(%i, %f, %s)" a b c
-- "(3, 3.14159, haskell)"
```

## Overloaded Lists

It is ubiquitous for data structure libraries to expose `toList`

and `fromList`

functions to construct various structures out of lists. As of GHC 7.8 we now have the ability to overload the list syntax in the surface language with a typeclass `IsList`

.

```
class IsList l where
type Item l
fromList :: [Item l] -> l
toList :: l -> [Item l]
instance IsList [a] where
type Item [a] = a
fromList = id
toList = id
```

```
λ: :type [1,2,3]
[1,2,3] :: (Num (Item l), IsList l) => l
```

```
{-# LANGUAGE OverloadedLists #-}
{-# LANGUAGE TypeFamilies #-}
import qualified Data.Map as Map
import GHC.Exts (IsList(..))
instance (Ord k) => IsList (Map.Map k v) where
type Item (Map.Map k v) = (k,v)
fromList = Map.fromList
toList = Map.toList
example1 :: Map.Map String Int
example1 = [("a", 1), ("b", 2)]
```

## String Conversions

Playing "type-tetris" to convert between Strings explicitly can be frustrating, fortunately there are several packages that automate the conversion using typeclasses to automatically convert between any two common string representations automatically. We can then write generic comparison and concatenation operators that automatically convert types of operands to a like form.

```
{-# LANGUAGE OverloadedStrings #-}
import Data.String.Conv
import qualified Data.Text as T
import qualified Data.Text.Lazy.IO as TL
import qualified Data.ByteString as B
import qualified Data.ByteString.Lazy as BL
import Data.Monoid
a :: String
a = "Gödel"
b :: BL.ByteString
b = "Einstein"
c :: T.Text
c = "Feynmann"
d :: B.ByteString
d = "Schrödinger"
-- Compare unlike strings
(==~) :: (Eq a, StringConv b a) => a -> b -> Bool
(==~) a b = a == toS b
-- Concat unlike strings
(<>~) :: (Monoid a, StringConv b a) => a -> b -> a
(<>~) a b = a <> toS b
main :: IO ()
main = do
putStrLn (toS a)
TL.putStrLn (toS b)
print (a ==~ b)
print (c ==~ d)
print (c ==~ c)
print (b <>~ c)
```

# Applicatives

Like monads Applicatives are an abstract structure for a wide class of computations that sit between functors and monads in terms of generality.

```
pure :: Applicative f => a -> f a
(<$>) :: Functor f => (a -> b) -> f a -> f b
(<*>) :: f (a -> b) -> f a -> f b
```

As of GHC 7.6, Applicative is defined as:

```
class Functor f => Applicative f where
pure :: a -> f a
(<*>) :: f (a -> b) -> f a -> f b
(<$>) :: Functor f => (a -> b) -> f a -> f b
(<$>) = fmap
```

With the following laws:

```
pure id <*> v = v
pure f <*> pure x = pure (f x)
u <*> pure y = pure ($ y) <*> u
u <*> (v <*> w) = pure (.) <*> u <*> v <*> w
```

As an example, consider the instance for Maybe:

```
instance Applicative Maybe where
pure = Just
Nothing <*> _ = Nothing
_ <*> Nothing = Nothing
Just f <*> Just x = Just (f x)
```

As a rule of thumb, whenever we would use `m >>= return . f`

what we probably want is an applicative functor, and not a monad.

```
import Network.HTTP
import Control.Applicative ((<$>),(<*>))
example1 :: Maybe Integer
example1 = (+) <$> m1 <*> m2
where
m1 = Just 3
m2 = Nothing
-- Nothing
example2 :: [(Int, Int, Int)]
example2 = (,,) <$> m1 <*> m2 <*> m3
where
m1 = [1,2]
m2 = [10,20]
m3 = [100,200]
-- [(1,10,100),(1,10,200),(1,20,100),(1,20,200),(2,10,100),(2,10,200),(2,20,100),(2,20,200)]
example3 :: IO String
example3 = (++) <$> fetch1 <*> fetch2
where
fetch1 = simpleHTTP (getRequest "http://www.fpcomplete.com/") >>= getResponseBody
fetch2 = simpleHTTP (getRequest "http://www.haskell.org/") >>= getResponseBody
```

The pattern `f <$> a <*> b ...`

shows up so frequently that there are a family of functions to lift applicatives of a fixed number arguments. This pattern also shows up frequently with monads (`liftM`

, `liftM2`

, `liftM3`

).

```
liftA :: Applicative f => (a -> b) -> f a -> f b
liftA f a = pure f <*> a
liftA2 :: Applicative f => (a -> b -> c) -> f a -> f b -> f c
liftA2 f a b = f <$> a <*> b
liftA3 :: Applicative f => (a -> b -> c -> d) -> f a -> f b -> f c -> f d
liftA3 f a b c = f <$> a <*> b <*> c
```

Applicative also has functions `*>`

and `<*`

that sequence applicative actions while discarding the value of one of the arguments. The operator `*>`

discard the left while `<*`

discards the right. For example in a monadic parser combinator library the `*>`

would parse with first parser argument but return the second.

The Applicative functions `<$>`

and `<*>`

are generalized by `liftM`

and `ap`

for monads.

```
import Control.Monad
import Control.Applicative
data C a b = C a b
mnd :: Monad m => m a -> m b -> m (C a b)
mnd a b = C `liftM` a `ap` b
apl :: Applicative f => f a -> f b -> f (C a b)
apl a b = C <$> a <*> b
```

See: Applicative Programming with Effects

## Alternative

Alternative is an extension of the Applicative class with a zero element and an associative binary operation respecting the zero.

```
class Applicative f => Alternative f where
-- | The identity of '<|>'
empty :: f a
-- | An associative binary operation
(<|>) :: f a -> f a -> f a
-- | One or more.
some :: f a -> f [a]
-- | Zero or more.
many :: f a -> f [a]
optional :: Alternative f => f a -> f (Maybe a)
```

```
instance Alternative Maybe where
empty = Nothing
Nothing <|> r = r
l <|> _ = l
instance Alternative [] where
empty = []
(<|>) = (++)
```

```
λ: foldl1 (<|>) [Nothing, Just 5, Just 3]
Just 5
```

These instances show up very frequently in parsers where the alternative operator can model alternative parse branches.

## Arrows

A category is an algebraic structure that includes a notion of an identity and a composition operation that is associative and preserves identities.

```
class Category cat where
id :: cat a a
(.) :: cat b c -> cat a b -> cat a c
```

```
instance Category (->) where
id = Prelude.id
(.) = (Prelude..)
```

```
(<<<) :: Category cat => cat b c -> cat a b -> cat a c
(<<<) = (.)
(>>>) :: Category cat => cat a b -> cat b c -> cat a c
f >>> g = g . f
```

Arrows are an extension of categories with the notion of products.

```
class Category a => Arrow a where
arr :: (b -> c) -> a b c
first :: a b c -> a (b,d) (c,d)
second :: a b c -> a (d,b) (d,c)
(***) :: a b c -> a b' c' -> a (b,b') (c,c')
(&&&) :: a b c -> a b c' -> a b (c,c')
```

The canonical example is for functions.

```
instance Arrow (->) where
arr f = f
first f = f *** id
second f = id *** f
(***) f g ~(x,y) = (f x, g y)
```

In this form functions of multiple arguments can be threaded around using the arrow combinators in a much more pointfree form. For instance a histogram function has a nice one-liner.

```
import Data.List (group, sort)
histogram :: Ord a => [a] -> [(a, Int)]
histogram = map (head &&& length) . group . sort
```

```
λ: histogram "Hello world"
[(' ',1),('H',1),('d',1),('e',1),('l',3),('o',2),('r',1),('w',1)]
```

**Arrow notation**

GHC has builtin syntax for composing arrows using `proc`

notation. The following are equivalent after desugaring:

```
{-# LANGUAGE Arrows #-}
addA :: Arrow a => a b Int -> a b Int -> a b Int
addA f g = proc x -> do
y <- f -< x
z <- g -< x
returnA -< y + z
```

```
addA f g = arr (\ x -> (x, x)) >>>
first f >>> arr (\ (y, x) -> (x, y)) >>>
first g >>> arr (\ (z, y) -> y + z)
```

`addA f g = f &&& g >>> arr (\ (y, z) -> y + z)`

In practice this notation is not often used and may become deprecated in the future.

See: Arrow Notation

## Bifunctors

Bifunctors are a generalization of functors to include types parameterized by two parameters and include two map functions for each parameter.

```
class Bifunctor p where
bimap :: (a -> b) -> (c -> d) -> p a c -> p b d
first :: (a -> b) -> p a c -> p b c
second :: (b -> c) -> p a b -> p a c
```

The bifunctor laws are a natural generalization of the usual functor. Namely they respect identities and composition in the usual way:

```
bimap id id ≡ id
first id ≡ id
second id ≡ id
```

`bimap f g ≡ first f . second g`

The canonical example is for 2-tuples.

```
λ: first (+1) (1,2)
(2,2)
λ: second (+1) (1,2)
(1,3)
λ: bimap (+1) (+1) (1,2)
(2,3)
λ: first (+1) (Left 3)
Left 4
λ: second (+1) (Left 3)
Left 3
λ: second (+1) (Right 3)
Right 4
```

## Polyvariadic Functions

One surprising application of typeclasses is the ability to construct functions which take an arbitrary number of arguments by defining instances over function types. The arguments may be of arbitrary type, but the resulting collected arguments must either converted into a single type or unpacked into a sum type.

```
{-# LANGUAGE FlexibleInstances #-}
class Arg a where
collect' :: [String] -> a
-- extract to IO
instance Arg (IO ()) where
collect' acc = mapM_ putStrLn acc
-- extract to [String]
instance Arg [String] where
collect' acc = acc
instance (Show a, Arg r) => Arg (a -> r) where
collect' acc = \x -> collect' (acc ++ [show x])
collect :: Arg t => t
collect = collect' []
example1 :: [String]
example1 = collect 'a' 2 3.0
example2 :: IO ()
example2 = collect () "foo" [1,2,3]
```

# Error Handling

## Control.Exception

The low-level (and most dangerous) way to handle errors is to use the `throw`

and `catch`

functions which allow us to throw extensible exceptions in pure code but catch the resulting exception within IO. Of specific note is that return value of the `throw`

inhabits all types. There's no reason to use this for custom code that doesn't use low-level system operations.

```
throw :: Exception e => e -> a
catch :: Exception e => IO a -> (e -> IO a) -> IO a
try :: Exception e => IO a -> IO (Either e a)
evaluate :: a -> IO a
```

```
{-# LANGUAGE DeriveDataTypeable #-}
import Data.Typeable
import Control.Exception
data MyException = MyException
deriving (Show, Typeable)
instance Exception MyException
evil :: [Int]
evil = [throw MyException]
example1 :: Int
example1 = head evil
example2 :: Int
example2 = length evil
main :: IO ()
main = do
a <- try (evaluate example1) :: IO (Either MyException Int)
print a
b <- try (return example2) :: IO (Either MyException Int)
print b
```

Because a value will not be evaluated unless needed, if one desires to know for sure that an exception is either caught or not it can be deeply forced into head normal form before invoking catch. The `strictCatch`

is not provided by standard library but has a simple implementation in terms of `deepseq`

.

```
strictCatch :: (NFData a, Exception e) => IO a -> (e -> IO a) -> IO a
strictCatch = catch . (toNF =<<)
```

## Exceptions

The problem with the previous approach is having to rely on GHC's asynchronous exception handling inside of IO to handle basic operations. The `exceptions`

provides the same API as `Control.Exception`

but loosens the dependency on IO.

```
{-# LANGUAGE DeriveDataTypeable #-}
import Data.Typeable
import Control.Monad.Catch
import Control.Monad.Identity
data MyException = MyException
deriving (Show, Typeable)
instance Exception MyException
example :: MonadCatch m => Int -> Int -> m Int
example x y | y == 0 = throwM MyException
| otherwise = return $ x `div` y
pure :: MonadCatch m => m (Either MyException Int)
pure = do
a <- try (example 1 2)
b <- try (example 1 0)
return (a >> b)
```

See: exceptions

## ExceptT

As of mtl 2.2 or higher, the `ErrorT`

class has been replaced by the `ExceptT`

. At transformers level.

```
newtype ExceptT e m a = ExceptT (m (Either e a))
runExceptT :: ExceptT e m a -> m (Either e a)
runExceptT (ExceptT m) = m
instance (Monad m) => Monad (ExceptT e m) where
return a = ExceptT $ return (Right a)
m >>= k = ExceptT $ do
a <- runExceptT m
case a of
Left e -> return (Left e)
Right x -> runExceptT (k x)
fail = ExceptT . fail
throwE :: (Monad m) => e -> ExceptT e m a
throwE = ExceptT . return . Left
catchE :: (Monad m) =>
ExceptT e m a -- ^ the inner computation
-> (e -> ExceptT e' m a) -- ^ a handler for exceptions in the inner
-- computation
-> ExceptT e' m a
m `catchE` h = ExceptT $ do
a <- runExceptT m
case a of
Left l -> runExceptT (h l)
Right r -> return (Right r)
```

Using mtl:

```
instance MonadTrans (ExceptT e) where
lift = ExceptT . liftM Right
class (Monad m) => MonadError e m | m -> e where
throwError :: e -> m a
catchError :: m a -> (e -> m a) -> m a
instance MonadError IOException IO where
throwError = ioError
catchError = catch
instance MonadError e (Either e) where
throwError = Left
Left l `catchError` h = h l
Right r `catchError` _ = Right r
```

See:

## spoon

Sometimes you'll be forced to deal with seemingly pure functions that can throw up at any point. There are many functions in the standard library like this, and many more on Hackage. You'd like to be handle this logic purely as if it were returning a proper `Maybe a`

but to catch the logic you'd need to install a IO handler inside IO to catch it. Spoon allows us to safely (and "purely", although it uses a referentially transparent invocation of unsafePerformIO) to catch these exceptions and put them in Maybe where they belong.

The `spoon`

function evaluates its argument to head normal form, while `teaspoon`

evaluates to weak head normal form.

```
import Control.Spoon
goBoom :: Int -> Int -> Int
goBoom x y = x `div` y
-- evaluate to normal form
test1 :: Maybe [Int]
test1 = spoon [1, 2, undefined]
-- evaluate to weak head normal form
test2 :: Maybe [Int]
test2 = teaspoon [1, 2, undefined]
main :: IO ()
main = do
maybe (putStrLn "Nothing") (print . length) test1
maybe (putStrLn "Nothing") (print . length) test2
```

See:

## safe-exceptions

TODO

# Advanced Monads

## Function Monad

If one writes Haskell long enough one might eventually encounter the curious beast that is the `((->) r)`

monad instance. It generally tends to be non-intuitive to work with, but is quite simple when one considers it as an unwrapped Reader monad.

```
instance Functor ((->) r) where
fmap = (.)
instance Monad ((->) r) where
return = const
f >>= k = \r -> k (f r) r
```

This just uses a prefix form of the arrow type operator.

```
import Control.Monad
id' :: (->) a a
id' = id
const' :: (->) a ((->) b a)
const' = const
-- Monad m => a -> m a
fret :: a -> b -> a
fret = return
-- Monad m => m a -> (a -> m b) -> m b
fbind :: (r -> a) -> (a -> (r -> b)) -> (r -> b)
fbind f k = f >>= k
-- Monad m => m (m a) -> m a
fjoin :: (r -> (r -> a)) -> (r -> a)
fjoin = join
fid :: a -> a
fid = const >>= id
-- Functor f => (a -> b) -> f a -> f b
fcompose :: (a -> b) -> (r -> a) -> (r -> b)
fcompose = (.)
```

```
type Reader r = (->) r -- pseudocode
instance Monad (Reader r) where
return a = \_ -> a
f >>= k = \r -> k (f r) r
ask' :: r -> r
ask' = id
asks' :: (r -> a) -> (r -> a)
asks' f = id . f
runReader' :: (r -> a) -> r -> a
runReader' = id
```

## RWS Monad

The RWS monad combines the functionality of the three monads discussed above, the **R**eader, **W**riter, and **S**tate. There is also a `RWST`

transformer.

```
runReader :: Reader r a -> r -> a
runWriter :: Writer w a -> (a, w)
runState :: State s a -> s -> (a, s)
```

These three eval functions are now combined into the following functions:

```
runRWS :: RWS r w s a -> r -> s -> (a, s, w)
execRWS :: RWS r w s a -> r -> s -> (s, w)
evalRWS :: RWS r w s a -> r -> s -> (a, w)
```

```
import Control.Monad.RWS
type R = Int
type W = [Int]
type S = Int
computation :: RWS R W S ()
computation = do
e <- ask
a <- get
let b = a + e
put b
tell [b]
example = runRWS computation 2 3
```

The usual caveat about Writer laziness also applies to RWS.

## Cont

```
runCont :: Cont r a -> (a -> r) -> r
callCC :: MonadCont m => ((a -> m b) -> m a) -> m a
cont :: ((a -> r) -> r) -> Cont r a
```

In continuation passing style, composite computations are built up from sequences of nested computations which are terminated by a final continuation which yields the result of the full computation by passing a function into the continuation chain.

```
add :: Int -> Int -> Int
add x y = x + y
add :: Int -> Int -> (Int -> r) -> r
add x y k = k (x + y)
```

```
import Control.Monad
import Control.Monad.Cont
add :: Int -> Int -> Cont k Int
add x y = return $ x + y
mult :: Int -> Int -> Cont k Int
mult x y = return $ x * y
contt :: ContT () IO ()
contt = do
k <- do
callCC $ \exit -> do
lift $ putStrLn "Entry"
exit $ \_ -> do
putStrLn "Exit"
lift $ putStrLn "Inside"
lift $ k ()
callcc :: Cont String Integer
callcc = do
a <- return 1
b <- callCC (\k -> k 2)
return $ a+b
ex1 :: IO ()
ex1 = print $ runCont (f >>= g) id
where
f = add 1 2
g = mult 3
-- 9
ex2 :: IO ()
ex2 = print $ runCont callcc show
-- "3"
ex3 :: IO ()
ex3 = runContT contt print
-- Entry
-- Inside
-- Exit
main :: IO ()
main = do
ex1
ex2
ex3
```

```
newtype Cont r a = Cont { runCont :: ((a -> r) -> r) }
instance Monad (Cont r) where
return a = Cont $ \k -> k a
(Cont c) >>= f = Cont $ \k -> c (\a -> runCont (f a) k)
class (Monad m) => MonadCont m where
callCC :: ((a -> m b) -> m a) -> m a
instance MonadCont (Cont r) where
callCC f = Cont $ \k -> runCont (f (\a -> Cont $ \_ -> k a)) k
```

## MonadPlus

Choice and failure.

```
class Monad m => MonadPlus m where
mzero :: m a
mplus :: m a -> m a -> m a
instance MonadPlus [] where
mzero = []
mplus = (++)
instance MonadPlus Maybe where
mzero = Nothing
Nothing `mplus` ys = ys
xs `mplus` _ys = xs
```

MonadPlus forms a monoid with

```
mzero `mplus` a = a
a `mplus` mzero = a
(a `mplus` b) `mplus` c = a `mplus` (b `mplus` c)
```

```
when :: (Monad m) => Bool -> m () -> m ()
when p s = if p then s else return ()
guard :: MonadPlus m => Bool -> m ()
guard True = return ()
guard False = mzero
msum :: MonadPlus m => [m a] -> m a
msum = foldr mplus mzero
```

```
import Safe
import Control.Monad
list1 :: [(Int,Int)]
list1 = [(a,b) | a <- [1..25], b <- [1..25], a < b]
list2 :: [(Int,Int)]
list2 = do
a <- [1..25]
b <- [1..25]
guard (a < b)
return $ (a,b)
maybe1 :: String -> String -> Maybe Double
maybe1 a b = do
a' <- readMay a
b' <- readMay b
guard (b' /= 0.0)
return $ a'/b'
maybe2 :: Maybe Int
maybe2 = msum [Nothing, Nothing, Just 3, Just 4]
```

```
import Control.Monad
range :: MonadPlus m => [a] -> m a
range [] = mzero
range (x:xs) = range xs `mplus` return x
pyth :: Integer -> [(Integer,Integer,Integer)]
pyth n = do
x <- range [1..n]
y <- range [1..n]
z <- range [1..n]
if x*x + y*y == z*z then return (x,y,z) else mzero
main :: IO ()
main = print $ pyth 15
{-
[ ( 12 , 9 , 15 )
, ( 12 , 5 , 13 )
, ( 9 , 12 , 15 )
, ( 8 , 6 , 10 )
, ( 6 , 8 , 10 )
, ( 5 , 12 , 13 )
, ( 4 , 3 , 5 )
, ( 3 , 4 , 5 )
]
-}
```

## MonadFix

The fixed point of a monadic computation. `mfix f`

executes the action `f`

only once, with the eventual output fed back as the input.

```
fix :: (a -> a) -> a
fix f = let x = f x in x
mfix :: (a -> m a) -> m a
```

```
class Monad m => MonadFix m where
mfix :: (a -> m a) -> m a
instance MonadFix Maybe where
mfix f = let a = f (unJust a) in a
where unJust (Just x) = x
unJust Nothing = error "mfix Maybe: Nothing"
```

The regular do-notation can also be extended with `-XRecursiveDo`

to accommodate recursive monadic bindings.

```
{-# LANGUAGE RecursiveDo #-}
import Control.Applicative
import Control.Monad.Fix
stream1 :: Maybe [Int]
stream1 = do
rec xs <- Just (1:xs)
return (map negate xs)
stream2 :: Maybe [Int]
stream2 = mfix $ \xs -> do
xs' <- Just (1:xs)
return (map negate xs')
```

## ST Monad

The ST monad models "threads" of stateful computations which can manipulate mutable references but are restricted to only return pure values when evaluated and are statically confined to the ST monad of a `s`

thread.

```
runST :: (forall s. ST s a) -> a
newSTRef :: a -> ST s (STRef s a)
readSTRef :: STRef s a -> ST s a
writeSTRef :: STRef s a -> a -> ST s ()
```

```
import Data.STRef
import Control.Monad
import Control.Monad.ST
import Control.Monad.State.Strict
example1 :: Int
example1 = runST $ do
x <- newSTRef 0
forM_ [1..1000] $ \j -> do
writeSTRef x j
readSTRef x
example2 :: Int
example2 = runST $ do
count <- newSTRef 0
replicateM_ (10^6) $ modifySTRef' count (+1)
readSTRef count
example3 :: Int
example3 = flip evalState 0 $ do
replicateM_ (10^6) $ modify' (+1)
get
modify' :: MonadState a m => (a -> a) -> m ()
modify' f = get >>= (\x -> put $! f x)
```

Using the ST monad we can create a class of efficient purely functional data structures that use mutable references in a referentially transparent way.

## Free Monads

```
Pure :: a -> Free f a
Free :: f (Free f a) -> Free f a
liftF :: (Functor f, MonadFree f m) => f a -> m a
retract :: Monad f => Free f a -> f a
```

Free monads are monads which instead of having a `join`

operation that combines computations, instead forms composite computations from application of a functor.

```
join :: Monad m => m (m a) -> m a
wrap :: MonadFree f m => f (m a) -> m a
```

One of the best examples is the Partiality monad which models computations which can diverge. Haskell allows unbounded recursion, but for example we can create a free monad from the `Maybe`

functor which can be used to fix the call-depth of, for example the Ackermann function.

```
import Control.Monad.Fix
import Control.Monad.Free
type Partiality a = Free Maybe a
-- Non-termination.
never :: Partiality a
never = fix (Free . Just)
fromMaybe :: Maybe a -> Partiality a
fromMaybe (Just x) = Pure x
fromMaybe Nothing = Free Nothing
runPartiality :: Int -> Partiality a -> Maybe a
runPartiality 0 _ = Nothing
runPartiality _ (Pure a) = Just a
runPartiality _ (Free Nothing) = Nothing
runPartiality n (Free (Just a)) = runPartiality (n-1) a
ack :: Int -> Int -> Partiality Int
ack 0 n = Pure $ n + 1
ack m 0 = Free $ Just $ ack (m-1) 1
ack m n = Free $ Just $ ack m (n-1) >>= ack (m-1)
main :: IO ()
main = do
let diverge = never :: Partiality ()
print $ runPartiality 1000 diverge
print $ runPartiality 1000 (ack 3 4)
print $ runPartiality 5500 (ack 3 4)
```

The other common use for free monads is to build embedded domain-specific languages to describe computations. We can model a subset of the IO monad by building up a pure description of the computation inside of the IOFree monad and then using the free monad to encode the translation to an effectful IO computation.

```
{-# LANGUAGE DeriveFunctor #-}
import System.Exit
import Control.Monad.Free
data Interaction x
= Puts String x
| Gets (Char -> x)
| Exit
deriving Functor
type IOFree a = Free Interaction a
puts :: String -> IOFree ()
puts s = liftF $ Puts s ()
get :: IOFree Char
get = liftF $ Gets id
exit :: IOFree r
exit = liftF Exit
gets :: IOFree String
gets = do
c <- get
if c == '\n'
then return ""
else gets >>= \line -> return (c : line)
-- Collapse our IOFree DSL into IO monad actions.
interp :: IOFree a -> IO a
interp (Pure r) = return r
interp (Free x) = case x of
Puts s t -> putStrLn s >> interp t
Gets f -> getChar >>= interp . f
Exit -> exitSuccess
echo :: IOFree ()
echo = do
puts "Enter your name:"
str <- gets
puts str
if length str > 10
then puts "You have a long name."
else puts "You have a short name."
exit
main :: IO ()
main = interp echo
```

An implementation such as the one found in free might look like the following:

```
{-# LANGUAGE MultiParamTypeClasses #-}
import Control.Applicative
data Free f a
= Pure a
| Free (f (Free f a))
instance Functor f => Monad (Free f) where
return a = Pure a
Pure a >>= f = f a
Free f >>= g = Free (fmap (>>= g) f)
class Monad m => MonadFree f m where
wrap :: f (m a) -> m a
liftF :: (Functor f, MonadFree f m) => f a -> m a
liftF = wrap . fmap return
iter :: Functor f => (f a -> a) -> Free f a -> a
iter _ (Pure a) = a
iter phi (Free m) = phi (iter phi <$> m)
retract :: Monad f => Free f a -> f a
retract (Pure a) = return a
retract (Free as) = as >>= retract
```

See:

## Indexed Monads

Indexed monads are a generalisation of monads that adds an additional type parameter to the class that carries information about the computation or structure of the monadic implementation.

```
class IxMonad md where
return :: a -> md i i a
(>>=) :: md i m a -> (a -> md m o b) -> md i o b
```

The canonical use-case is a variant of the vanilla State which allows type-changing on the state for intermediate steps inside of the monad. This indeed turns out to be very useful for handling a class of problems involving resource management since the extra index parameter gives us space to statically enforce the sequence of monadic actions by allowing and restricting certain state transitions on the index parameter at compile-time.

To make this more usable we'll use the somewhat esoteric `-XRebindableSyntax`

allowing us to overload the do-notation and if-then-else syntax by providing alternative definitions local to the module.

```
{-# LANGUAGE RebindableSyntax #-}
{-# LANGUAGE ScopedTypeVariables #-}
{-# LANGUAGE NoMonomorphismRestriction #-}
import Data.IORef
import Data.Char
import Prelude hiding (fmap, (>>=), (>>), return)
import Control.Applicative
newtype IState i o a = IState { runIState :: i -> (a, o) }
evalIState :: IState i o a -> i -> a
evalIState st i = fst $ runIState st i
execIState :: IState i o a -> i -> o
execIState st i = snd $ runIState st i
ifThenElse :: Bool -> a -> a -> a
ifThenElse b i j = case b of
True -> i
False -> j
return :: a -> IState s s a
return a = IState $ \s -> (a, s)
fmap :: (a -> b) -> IState i o a -> IState i o b
fmap f v = IState $ \i -> let (a, o) = runIState v i
in (f a, o)
join :: IState i m (IState m o a) -> IState i o a
join v = IState $ \i -> let (w, m) = runIState v i
in runIState w m
(>>=) :: IState i m a -> (a -> IState m o b) -> IState i o b
v >>= f = IState $ \i -> let (a, m) = runIState v i
in runIState (f a) m
(>>) :: IState i m a -> IState m o b -> IState i o b
v >> w = v >>= \_ -> w
get :: IState s s s
get = IState $ \s -> (s, s)
gets :: (a -> o) -> IState a o a
gets f = IState $ \s -> (s, f s)
put :: o -> IState i o ()
put o = IState $ \_ -> ((), o)
modify :: (i -> o) -> IState i o ()
modify f = IState $ \i -> ((), f i)
data Locked = Locked
data Unlocked = Unlocked
type Stateful a = IState a Unlocked a
acquire :: IState i Locked ()
acquire = put Locked
-- Can only release the lock if it's held, try release the lock
-- that's not held is a now a type error.
release :: IState Locked Unlocked ()
release = put Unlocked
-- Statically forbids improper handling of resources.
lockExample :: Stateful a
lockExample = do
ptr <- get :: IState a a a
acquire :: IState a Locked ()
-- ...
release :: IState Locked Unlocked ()
return ptr
-- Couldn't match type `Locked' with `Unlocked'
-- In a stmt of a 'do' block: return ptr
failure1 :: Stateful a
failure1 = do
ptr <- get
acquire
return ptr -- didn't release
-- Couldn't match type `a' with `Locked'
-- In a stmt of a 'do' block: release
failure2 :: Stateful a
failure2 = do
ptr <- get
release -- didn't acquire
return ptr
-- Evaluate the resulting state, statically ensuring that the
-- lock is released when finished.
evalReleased :: IState i Unlocked a -> i -> a
evalReleased f st = evalIState f st
example :: IO (IORef Integer)
example = evalReleased <$> pure lockExample <*> newIORef 0
```

## lifted-base

The default prelude predates a lot of the work on monad transformers and as such many of the common functions for handling errors and interacting with IO are bound strictly to the IO monad and not to functions implementing stacks on top of IO or ST. The lifted-base provides generic control operations such as `catch`

can be lifted from IO or any other base monad.

#### monad-base

Monad base provides an abstraction over `liftIO`

and other functions to explicitly lift into a "privileged" layer of the transformer stack. It's implemented a multiparamater typeclass with the "base" monad as the parameter b.

```
-- | Lift a computation from the base monad
class (Applicative b, Applicative m, Monad b, Monad m)
=> MonadBase b m | m -> b where
liftBase ∷ b a -> m a
```

#### monad-control

Monad control builds on top of monad-base to extended lifting operation to control operations like `catch`

and `bracket`

can be written generically in terms of any transformer with a base layer supporting these operations. Generic operations can then be expressed in terms of a `MonadBaseControl`

and written in terms of the combinator `control`

which handles the bracket and automatic handler lifting.

`control :: MonadBaseControl b m => (RunInBase m b -> b (StM m a)) -> m a`

For example the function catch provided by `Control.Exception`

is normally locked into IO.

`catch :: Exception e => IO a -> (e -> IO a) -> IO a`

By composing it in terms of control we can construct a generic version which automatically lifts inside of any combination of the usual transformer stacks that has `MonadBaseControl`

instance.

```
catch
:: (MonadBaseControl IO m, Exception e)
=> m a -- ^ Computation
-> (e -> m a) -- ^ Handler
-> m a
catch a handler = control $ \runInIO ->
E.catch (runInIO a)
(\e -> runInIO $ handler e)
```

# Quantification

This is an advanced section, and is not typically necessary to write Haskell.

## Universal Quantification

Universal quantification the primary mechanism of encoding polymorphism in Haskell. The essence of universal quantification is that we can express functions which operate the same way for a set of types and whose function behavior is entirely determined *only* by the behavior of all types in this span.

```
{-# LANGUAGE ExplicitForAll #-}
-- ∀a. [a]
example1 :: forall a. [a]
example1 = []
-- ∀a. [a]
example2 :: forall a. [a]
example2 = [undefined]
-- ∀a. ∀b. (a → b) → [a] → [b]
map' :: forall a. forall b. (a -> b) -> [a] -> [b]
map' f = foldr ((:) . f) []
-- ∀a. [a] → [a]
reverse' :: forall a. [a] -> [a]
reverse' = foldl (flip (:)) []
```

Normally quantifiers are omitted in type signatures since in Haskell's vanilla surface language it is unambiguous to assume to that free type variables are universally quantified.

## Free theorems

A universally quantified type-variable actually implies quite a few rather deep properties about the implementation of a function that can be deduced from its type signature. For instance the identity function in Haskell is guaranteed to only have one implementation since the only information that the information that can present in the body

```
id :: forall a. a -> a
id x = x
```

`fmap :: Functor f => (a -> b) -> f a -> f b`

The free theorem of fmap:

`forall f g. fmap f . fmap g = fmap (f . g)`

See: Theorems for Free

## Type Systems

**Hindley-Milner type system**

The Hindley-Milner type system is historically important as one of the first typed lambda calculi that admitted both polymorphism and a variety of inference techniques that could always decide principal types.

```
e : x
| λx:t.e -- value abstraction
| e1 e2 -- application
| let x = e1 in e2 -- let
t : t -> t -- function types
| a -- type variables
σ : ∀ a . t -- type scheme
```

In an implementation, the function `generalize`

converts all type variables within the type into polymorphic type variables yielding a type scheme. The function `instantiate`

maps a scheme to a type, but with any polymorphic variables converted into unbound type variables.

## Rank-N Types

System-F is the type system that underlies Haskell. System-F subsumes the HM type system in the sense that every type expressible in HM can be expressed within System-F. System-F is sometimes referred to in texts as the *Girald-Reynolds polymorphic lambda calculus* or *second-order lambda calculus*.

```
t : t -> t -- function types
| a -- type variables
| ∀ a . t -- forall
e : x -- variables
| λ(x:t).e -- value abstraction
| e1 e2 -- value application
| Λa.e -- type abstraction
| e_t -- type application
```

An example with equivalents of GHC Core in comments:

```
id : ∀ t. t -> t
id = Λt. λx:t. x
-- id :: forall t. t -> t
-- id = \ (@ t) (x :: t) -> x
tr : ∀ a. ∀ b. a -> b -> a
tr = Λa. Λb. λx:a. λy:b. x
-- tr :: forall a b. a -> b -> a
-- tr = \ (@ a) (@ b) (x :: a) (y :: b) -> x
fl : ∀ a. ∀ b. a -> b -> b
fl = Λa. Λb. λx:a. λy:b. y
-- fl :: forall a b. a -> b -> b
-- fl = \ (@ a) (@ b) (x :: a) (y :: b) -> y
nil : ∀ a. [a]
nil = Λa. Λb. λz:b. λf:(a -> b -> b). z
-- nil :: forall a. [a]
-- nil = \ (@ a) (@ b) (z :: b) (f :: a -> b -> b) -> z
cons : ∀ a. a -> [a] -> [a]
cons = Λa. λx:a. λxs:(∀ b. b -> (a -> b -> b) -> b).
Λb. λz:b. λf : (a -> b -> b). f x (xs_b z f)
-- cons :: forall a. a
-- -> (forall b. (a -> b -> b) -> b) -> (forall b. (a -> b -> b) -> b)
-- cons = \ (@ a) (x :: a) (xs :: forall b. (a -> b -> b) -> b)
-- (@ b) (z :: b) (f :: a -> b -> b) -> f x (xs @ b z f)
```

Normally when Haskell's typechecker infers a type signature it places all quantifiers of type variables at the outermost position such that no quantifiers appear within the body of the type expression, called the prenex restriction. This restricts an entire class of type signatures that would otherwise be expressible within System-F, but has the benefit of making inference much easier.

`-XRankNTypes`

loosens the prenex restriction such that we may explicitly place quantifiers within the body of the type. The bad news is that the general problem of inference in this relaxed system is undecidable in general, so we're required to explicitly annotate functions which use RankNTypes or they are otherwise inferred as rank 1 and may not typecheck at all.

```
{-# LANGUAGE RankNTypes #-}
-- Can't unify ( Bool ~ Char )
rank1 :: forall a. (a -> a) -> (Bool, Char)
rank1 f = (f True, f 'a')
rank2 :: (forall a. a -> a) -> (Bool, Char)
rank2 f = (f True, f 'a')
auto :: (forall a. a -> a) -> (forall b. b -> b)
auto x = x
xauto :: forall a. (forall b. b -> b) -> a -> a
xauto f = f
```

```
Monomorphic Rank 0: t
Polymorphic Rank 1: forall a. a -> t
Polymorphic Rank 2: (forall a. a -> t) -> t
Polymorphic Rank 3: ((forall a. a -> t) -> t) -> t
```

Of important note is that the type variables bound by an explicit quantifier in a higher ranked type may not escape their enclosing scope. The typechecker will explicitly enforce this by enforcing that variables bound inside of rank-n types (called skolem constants) will not unify with free meta type variables inferred by the inference engine.

```
{-# LANGUAGE RankNTypes #-}
escape :: (forall a. a -> a) -> Int
escape f = f 0
g x = escape (\a -> x)
```

In this example in order for the expression to be well typed, `f`

would necessarily have (`Int -> Int`

) which implies that `a ~ Int`

over the whole type, but since `a`

is bound under the quantifier it must not be unified with `Int`

and so the typechecker must fail with a skolem capture error.

```
Couldn't match expected type `a' with actual type `t'
`a' is a rigid type variable bound by a type expected by the context: a -> a
`t' is a rigid type variable bound by the inferred type of g :: t -> Int
In the expression: x In the first argument of `escape', namely `(\ a -> x)'
In the expression: escape (\ a -> x)
```

This can actually be used for our advantage to enforce several types of invariants about scope and use of specific type variables. For example the ST monad uses a second rank type to prevent the capture of references between ST monads with separate state threads where the `s`

type variable is bound within a rank-2 type and cannot escape, statically guaranteeing that the implementation details of the ST internals can't leak out and thus ensuring its referential transparency.

## Existential Quantification

An existential type is a pair of a type and a term with a special set of packing and unpacking semantics. The type of the value encoded in the existential is known by the producer but not by the consumer of the existential value.

```
{-# LANGUAGE ExistentialQuantification #-}
{-# LANGUAGE RankNTypes #-}
-- ∃ t. (t, t → t, t → String)
data Box = forall a. Box a (a -> a) (a -> String)
boxa :: Box
boxa = Box 1 negate show
boxb :: Box
boxb = Box "foo" reverse show
apply :: Box -> String
apply (Box x f p) = p (f x)
-- ∃ t. Show t => t
data SBox = forall a. Show a => SBox a
boxes :: [SBox]
boxes = [SBox (), SBox 2, SBox "foo"]
showBox :: SBox -> String
showBox (SBox a) = show a
main :: IO ()
main = mapM_ (putStrLn . showBox) boxes
-- ()
-- 2
-- "foo"
```

The existential over `SBox`

gathers a collection of values defined purely in terms of their Show interface and an opaque pointer, no other information is available about the values and they can't be accessed or unpacked in any other way.

Passing around existential types allows us to hide information from consumers of data types and restrict the behavior that functions can use. Passing records around with existential variables allows a type to be "bundled" with a fixed set of functions that operate over its hidden internals.

```
{-# LANGUAGE RankNTypes #-}
{-# LANGUAGE ExistentialQuantification #-}
-- a b are existentially bound type variables, m is a free type variable
data MonadI m = MonadI
{ _return :: forall a . a -> m a
, _bind :: forall a b . m a -> (a -> m b) -> m b
}
monadMaybe:: MonadI Maybe
monadMaybe = MonadI
{ _return = Just
, _bind = \m f -> case m of
Nothing -> Nothing
Just x -> f x
}
```

## Impredicative Types

This is an advanced section, and is not typically necessary to write Haskell.

Although extremely brittle, GHC also has limited support for impredicative polymorphism which allows instantiating type variable with a polymorphic type. Implied is that this loosens the restriction that quantifiers must precede arrow types and now they may be placed inside of type-constructors.

```
-- Can't unify ( Int ~ Char )
revUni :: forall a. Maybe ([a] -> [a]) -> Maybe ([Int], [Char])
revUni (Just g) = Just (g [3], g "hello")
revUni Nothing = Nothing
```

```
{-# LANGUAGE ImpredicativeTypes #-}
-- Uses higher-ranked polymorphism.
f :: (forall a. [a] -> a) -> (Int, Char)
f get = (get [1,2], get ['a', 'b', 'c'])
-- Uses impredicative polymorphism.
g :: Maybe (forall a. [a] -> a) -> (Int, Char)
g Nothing = (0, '0')
g (Just get) = (get [1,2], get ['a','b','c'])
```

Use of this extension is very rare, and there is some consideration that `-XImpredicativeTypes`

is fundamentally broken. Although GHC is very liberal about telling us to enable it when one accidentally makes a typo in a type signature!

Some notable trivia, the `($)`

operator is wired into GHC in a very special way as to allow impredicative instantiation of `runST`

to be applied via `($)`

by special-casing the `($)`

operator only when used for the ST monad. If this sounds like an ugly hack it's because it is, but a rather convenient hack.

For example if we define a function `apply`

which should behave identically to `($)`

we'll get an error about polymorphic instantiation even though they are defined identically!

```
{-# LANGUAGE RankNTypes #-}
import Control.Monad.ST
f `apply` x = f x
foo :: (forall s. ST s a) -> a
foo st = runST $ st
bar :: (forall s. ST s a) -> a
bar st = runST `apply` st
```

```
Couldn't match expected type `forall s. ST s a'
with actual type `ST s0 a'
In the second argument of `apply', namely `st'
In the expression: runST `apply` st
In an equation for `bar': bar st = runST `apply` st
```

See:

## Scoped Type Variables

Normally the type variables used within the toplevel signature for a function are only scoped to the type-signature and not the body of the function and its rigid signatures over terms and let/where clauses. Enabling `-XScopedTypeVariables`

loosens this restriction allowing the type variables mentioned in the toplevel to be scoped within the value-level body of a function and all signatures contained therein.

```
{-# LANGUAGE ExplicitForAll #-}
{-# LANGUAGE ScopedTypeVariables #-}
poly :: forall a b c. a -> b -> c -> (a, a)
poly x y z = (f x y, f x z)
where
-- second argument is universally quantified from inference
-- f :: forall t0 t1. t0 -> t1 -> t0
f x' _ = x'
mono :: forall a b c. a -> b -> c -> (a, a)
mono x y z = (f x y, f x z)
where
-- b is not implictly universally quantified because it is in scope
f :: a -> b -> a
f x' _ = x'
example :: IO ()
example = do
x :: [Int] <- readLn
print x
```

# GADTs

## GADTs

*Generalized Algebraic Data types* (GADTs) are an extension to algebraic datatypes that allow us to qualify the constructors to datatypes with type equality constraints, allowing a class of types that are not expressible using vanilla ADTs.

`-XGADTs`

implicitly enables an alternative syntax for datatype declarations ( `-XGADTSyntax`

) such that the following declarations are equivalent:

```
-- Vanilla
data List a
= Empty
| Cons a (List a)
-- GADTSyntax
data List a where
Empty :: List a
Cons :: a -> List a -> List a
```

For an example use consider the data type `Term`

, we have a term in which we `Succ`

which takes a `Term`

parameterized by `a`

which span all types. Problems arise between the clash whether (`a ~ Bool`

) or (`a ~ Int`

) when trying to write the evaluator.

```
data Term a
= Lit a
| Succ (Term a)
| IsZero (Term a)
-- can't be well-typed :(
eval (Lit i) = i
eval (Succ t) = 1 + eval t
eval (IsZero i) = eval i == 0
```

And we admit the construction of meaningless terms which forces more error handling cases.

```
-- This is a valid type.
failure = Succ ( Lit True )
```

Using a GADT we can express the type invariants for our language (i.e. only type-safe expressions are representable). Pattern matching on this GADTs then carries type equality constraints without the need for explicit tags.

```
{-# Language GADTs #-}
data Term a where
Lit :: a -> Term a
Succ :: Term Int -> Term Int
IsZero :: Term Int -> Term Bool
If :: Term Bool -> Term a -> Term a -> Term a
eval :: Term a -> a
eval (Lit i) = i -- Term a
eval (Succ t) = 1 + eval t -- Term (a ~ Int)
eval (IsZero i) = eval i == 0 -- Term (a ~ Int)
eval (If b e1 e2) = if eval b then eval e1 else eval e2 -- Term (a ~ Bool)
example :: Int
example = eval (Succ (Succ (Lit 3)))
```

This time around:

```
-- This is rejected at compile-time.
failure = Succ ( Lit True )
```

Explicit equality constraints (`a ~ b`

) can be added to a function's context. For example the following expand out to the same types.

```
f :: a -> a -> (a, a)
f :: (a ~ b) => a -> b -> (a,b)
```

```
(Int ~ Int) => ...
(a ~ Int) => ...
(Int ~ a) => ...
(a ~ b) => ...
(Int ~ Bool) => ... -- Will not typecheck.
```

This is effectively the implementation detail of what GHC is doing behind the scenes to implement GADTs ( implicitly passing and threading equality terms around ). If we wanted we could do the same setup that GHC does just using equality constraints and existential quantification. Indeed, the internal representation of GADTs is as regular algebraic datatypes that carry coercion evidence as arguments.

```
{-# LANGUAGE GADTs #-}
{-# LANGUAGE ExistentialQuantification #-}
-- Using Constraints
data Exp a
= (a ~ Int) => LitInt a
| (a ~ Bool) => LitBool a
| forall b. (b ~ Bool) => If (Exp b) (Exp a) (Exp a)
-- Using GADTs
-- data Exp a where
-- LitInt :: Int -> Exp Int
-- LitBool :: Bool -> Exp Bool
-- If :: Exp Bool -> Exp a -> Exp a -> Exp a
eval :: Exp a -> a
eval e = case e of
LitInt i -> i
LitBool b -> b
If b tr fl -> if eval b then eval tr else eval fl
```

In the presence of GADTs inference becomes intractable in many cases, often requiring an explicit annotation. For example `f`

can either have `T a -> [a]`

or `T a -> [Int]`

and neither is principal.

```
data T :: * -> * where
T1 :: Int -> T Int
T2 :: T a
f (T1 n) = [n]
f T2 = []
```

## Kind Signatures

Haskell's kind system (i.e. the "type of the types") is a system consisting the single kind `*`

and an arrow kind `->`

.

```
κ : *
| κ -> κ
```

```
Int :: *
Maybe :: * -> *
Either :: * -> * -> *
```

There are in fact some extensions to this system that will be covered later ( see: PolyKinds and Unboxed types in later sections ) but most kinds in everyday code are simply either stars or arrows.

With the KindSignatures extension enabled we can now annotate top level type signatures with their explicit kinds, bypassing the normal kind inference procedures.

```
{-# LANGUAGE KindSignatures #-}
id :: forall (a :: *). a -> a
id x = x
```

On top of default GADT declaration we can also constrain the parameters of the GADT to specific kinds. For basic usage Haskell's kind inference can deduce this reasonably well, but combined with some other type system extensions that extend the kind system this becomes essential.

```
{-# Language GADTs #-}
{-# LANGUAGE KindSignatures #-}
data Term a :: * where
Lit :: a -> Term a
Succ :: Term Int -> Term Int
IsZero :: Term Int -> Term Bool
If :: Term Bool -> Term a -> Term a -> Term a
data Vec :: * -> * -> * where
Nil :: Vec n a
Cons :: a -> Vec n a -> Vec n a
data Fix :: (* -> *) -> * where
In :: f (Fix f) -> Fix f
```

## Void

The Void type is the type with no inhabitants. It unifies only with itself.

Using a newtype wrapper we can create a type where recursion makes it impossible to construct an inhabitant.

```
-- Void :: Void -> Void
newtype Void = Void Void
```

Or using `-XEmptyDataDecls`

we can also construct the uninhabited type equivalently as a data declaration with no constructors.

`data Void`

The only inhabitant of both of these types is a diverging term like (`undefined`

).

## Phantom Types

Phantom types are parameters that appear on the left hand side of a type declaration but which are not constrained by the values of the types inhabitants. They are effectively slots for us to encode additional information at the type-level.

```
import Data.Void
data Foo tag a = Foo a
combine :: Num a => Foo tag a -> Foo tag a -> Foo tag a
combine (Foo a) (Foo b) = Foo (a+b)
-- All identical at the value level, but differ at the type level.
a :: Foo () Int
a = Foo 1
b :: Foo t Int
b = Foo 1
c :: Foo Void Int
c = Foo 1
-- () ~ ()
example1 :: Foo () Int
example1 = combine a a
-- t ~ ()
example2 :: Foo () Int
example2 = combine a b
-- t0 ~ t1
example3 :: Foo t Int
example3 = combine b b
-- Couldn't match type `t' with `Void'
example4 :: Foo t Int
example4 = combine b c
```

Notice the type variable `tag`

does not appear in the right hand side of the declaration. Using this allows us to express invariants at the type-level that need not manifest at the value-level. We're effectively programming by adding extra information at the type-level.

Consider the case of using newtypes to statically distinguish between plaintext and cryptotext.

```
newtype Plaintext = Plaintext Text
newtype Crytpotext = Cryptotext Text
encrypt :: Key -> Plaintext -> Cryptotext
decrypt :: Key -> Cryptotext -> Plaintext
```

Using phantom types we use an extra parameter.

```
import Data.Text
data Cryptotext
data Plaintext
data Msg a = Msg Text
encrypt :: Msg Plaintext -> Msg Cryptotext
encrypt = undefined
decrypt :: Msg Cryptotext -> Msg Plaintext
decrypt = undefined
```

Using `-XEmptyDataDecls`

can be a powerful combination with phantom types that contain no value inhabitants and are "anonymous types".

```
{-# LANGUAGE EmptyDataDecls #-}
data Token a
```

## Typelevel Operations

This is an advanced section, and is not typically necessary to write Haskell.

With a richer language for datatypes we can express terms that witness the relationship between terms in the constructors, for example we can now express a term which expresses propositional equality between two types.

The type `Eql a b`

is a proof that types `a`

and `b`

are equal, by pattern matching on the single `Refl`

constructor we introduce the equality constraint into the body of the pattern match.

```
{-# LANGUAGE GADTs #-}
{-# LANGUAGE ExplicitForAll #-}
-- a ≡ b
data Eql a b where
Refl :: Eql a a
-- Congruence
-- (f : A → B) {x y} → x ≡ y → f x ≡ f y
cong :: Eql a b -> Eql (f a) (f b)
cong Refl = Refl
-- Symmetry
-- {a b : A} → a ≡ b → a ≡ b
sym :: Eql a b -> Eql b a
sym Refl = Refl
-- Transitivity
-- {a b c : A} → a ≡ b → b ≡ c → a ≡ c
trans :: Eql a b -> Eql b c -> Eql a c
trans Refl Refl = Refl
-- Coerce one type to another given a proof of their equality.
-- {a b : A} → a ≡ b → a → b
castWith :: Eql a b -> a -> b
castWith Refl = id
-- Trivial cases
a :: forall n. Eql n n
a = Refl
b :: forall. Eql () ()
b = Refl
```

As of GHC 7.8 these constructors and functions are included in the Prelude in the Data.Type.Equality module.

# Interpreters

The lambda calculus forms the theoretical and practical foundation for many languages. At the heart of every calculus is three components:

**Var**- A variable**Lam**- A lambda abstraction**App**- An application

There are many different ways of modeling these constructions and data structure representations, but they all more or less contain these three elements. For example, a lambda calculus that uses String names on lambda binders and variables might be written like the following:

```
type Name = String
data Exp
= Var Name
| Lam Name Exp
| App Exp Exp
```

A lambda expression in which all variables that appear in the body of the expression are referenced in an outer lambda binder is said to be *closed* while an expression with unbound free variables is *open*.

## HOAS

Higher Order Abstract Syntax (*HOAS*) is a technique for implementing the lambda calculus in a language where the binders of the lambda expression map directly onto lambda binders of the host language ( i.e. Haskell ) to give us substitution machinery in our custom language by exploiting Haskell's implementation.

```
{-# LANGUAGE GADTs #-}
data Expr a where
Con :: a -> Expr a
Lam :: (Expr a -> Expr b) -> Expr (a -> b)
App :: Expr (a -> b) -> Expr a -> Expr b
i :: Expr (a -> a)
i = Lam (\x -> x)
k :: Expr (a -> b -> a)
k = Lam (\x -> Lam (\y -> x))
s :: Expr ((a -> b -> c) -> (a -> b) -> (a -> c))
s = Lam (\x -> Lam (\y -> Lam (\z -> App (App x z) (App y z))))
eval :: Expr a -> a
eval (Con v) = v
eval (Lam f) = \x -> eval (f (Con x))
eval (App e1 e2) = (eval e1) (eval e2)
skk :: Expr (a -> a)
skk = App (App s k) k
example :: Integer
example = eval skk 1
-- 1
```

Pretty printing HOAS terms can also be quite complicated since the body of the function is under a Haskell lambda binder.

## PHOAS

A slightly different form of HOAS called PHOAS uses lambda datatype parameterized over the binder type. In this form evaluation requires unpacking into a separate Value type to wrap the lambda expression.

```
{-# LANGUAGE RankNTypes #-}
data ExprP a
= VarP a
| AppP (ExprP a) (ExprP a)
| LamP (a -> ExprP a)
| LitP Integer
data Value
= VLit Integer
| VFun (Value -> Value)
fromVFun :: Value -> (Value -> Value)
fromVFun val = case val of
VFun f -> f
_ -> error "not a function"
fromVLit :: Value -> Integer
fromVLit val = case val of
VLit n -> n
_ -> error "not a integer"
newtype Expr = Expr { unExpr :: forall a . ExprP a }
eval :: Expr -> Value
eval e = ev (unExpr e) where
ev (LamP f) = VFun(ev . f)
ev (VarP v) = v
ev (AppP e1 e2) = fromVFun (ev e1) (ev e2)
ev (LitP n) = VLit n
i :: ExprP a
i = LamP (\a -> VarP a)
k :: ExprP a
k = LamP (\x -> LamP (\y -> VarP x))
s :: ExprP a
s = LamP (\x -> LamP (\y -> LamP (\z -> AppP (AppP (VarP x) (VarP z)) (AppP (VarP y) (VarP z)))))
skk :: ExprP a
skk = AppP (AppP s k) k
example :: Integer
example = fromVLit $ eval $ Expr (AppP skk (LitP 3))
```

See:

## Final Interpreters

Using typeclasses we can implement a *final interpreter* which models a set of extensible terms using functions bound to typeclasses rather than data constructors. Instances of the typeclass form interpreters over these terms.

For example we can write a small language that includes basic arithmetic, and then retroactively extend our expression language with a multiplication operator without changing the base. At the same time our interpreter logic remains invariant under extension with new expressions.

```
{-# LANGUAGE FlexibleInstances #-}
{-# LANGUAGE FlexibleContexts #-}
{-# LANGUAGE TypeSynonymInstances #-}
{-# LANGUAGE NoMonomorphismRestriction #-}
class Expr repr where
lit :: Int -> repr
neg :: repr -> repr
add :: repr -> repr -> repr
mul :: repr -> repr -> repr
instance Expr Int where
lit n = n
neg a = -a
add a b = a + b
mul a b = a * b
instance Expr String where
lit n = show n
neg a = "(-" ++ a ++ ")"
add a b = "(" ++ a ++ " + " ++ b ++ ")"
mul a b = "(" ++ a ++ " * " ++ b ++ ")"
class BoolExpr repr where
eq :: repr -> repr -> repr
tr :: repr
fl :: repr
instance BoolExpr Int where
eq a b = if a == b then tr else fl
tr = 1
fl = 0
instance BoolExpr String where
eq a b = "(" ++ a ++ " == " ++ b ++ ")"
tr = "true"
fl = "false"
eval :: Int -> Int
eval = id
render :: String -> String
render = id
expr :: (BoolExpr repr, Expr repr) => repr
expr = eq (add (lit 1) (lit 2)) (lit 3)
result :: Int
result = eval expr
-- 1
string :: String
string = render expr
-- "((1 + 2) == 3)"
```

## Finally Tagless

Writing an evaluator for the lambda calculus can likewise also be modeled with a final interpreter and a Identity functor.

```
import Prelude hiding (id)
class Expr rep where
lam :: (rep a -> rep b) -> rep (a -> b)
app :: rep (a -> b) -> (rep a -> rep b)
lit :: a -> rep a
newtype Interpret a = R { reify :: a }
instance Expr Interpret where
lam f = R $ reify . f . R
app f a = R $ reify f $ reify a
lit = R
eval :: Interpret a -> a
eval e = reify e
e1 :: Expr rep => rep Int
e1 = app (lam (\x -> x)) (lit 3)
e2 :: Expr rep => rep Int
e2 = app (lam (\x -> lit 4)) (lam $ \x -> lam $ \y -> y)
example1 :: Int
example1 = eval e1
-- 3
example2 :: Int
example2 = eval e2
-- 4
```

See: Typed Tagless Interpretations and Typed Compilation

## Datatypes

The usual hand-wavy of describing algebraic datatypes is to indicate the how natural correspondence between sum types, product types, and polynomial expressions arises.

```
data Void -- 0
data Unit = Unit -- 1
data Sum a b = Inl a | Inr b -- a + b
data Prod a b = Prod a b -- a * b
type (->) a b = a -> b -- b ^ a
```

Intuitively it follows the notion that the cardinality of set of inhabitants of a type can always be given as a function of the number of its holes. A product type admits a number of inhabitants as a function of the product (i.e. cardinality of the Cartesian product), a sum type as the sum of its holes and a function type as the exponential of the span of the domain and codomain.

```
-- 1 + A
data Maybe a = Nothing | Just a
```

Recursive types are correspond to infinite series of these terms.

```
-- pseudocode
-- μX. 1 + X
data Nat a = Z | S Nat
Nat a = μ a. 1 + a
= 1 + (1 + (1 + ...))
-- μX. 1 + A * X
data List a = Nil | Cons a (List a)
List a = μ a. 1 + a * (List a)
= 1 + a + a^2 + a^3 + a^4 ...
-- μX. A + A*X*X
data Tree a f = Leaf a | Tree a f f
Tree a = μ a. 1 + a * (List a)
= 1 + a^2 + a^4 + a^6 + a^8 ...
```

See: Species and Functors and Types, Oh My!

## F-Algebras

This is an advanced section, and is not typically necessary to write Haskell.

The *initial algebra* approach differs from the final interpreter approach in that we now represent our terms as algebraic datatypes and the interpreter implements recursion and evaluation occurs through pattern matching.

```
type Algebra f a = f a -> a
type Coalgebra f a = a -> f a
newtype Fix f = Fix { unFix :: f (Fix f) }
cata :: Functor f => Algebra f a -> Fix f -> a
ana :: Functor f => Coalgebra f a -> a -> Fix f
hylo :: Functor f => Algebra f b -> Coalgebra f a -> a -> b
```

In Haskell a F-algebra is a functor `f a`

together with a function `f a -> a`

. A coalgebra reverses the function. For a functor `f`

we can form its recursive unrolling using the recursive `Fix`

newtype wrapper.

```
newtype Fix f = Fix { unFix :: f (Fix f) }
Fix :: f (Fix f) -> Fix f
unFix :: Fix f -> f (Fix f)
```

```
Fix f = f (f (f (f (f (f ( ... ))))))
newtype T b a = T (a -> b)
Fix (T a)
Fix T -> a
(Fix T -> a) -> a
(Fix T -> a) -> a -> a
...
```

In this form we can write down a generalized fold/unfold function that are datatype generic and written purely in terms of the recursing under the functor.

```
cata :: Functor f => Algebra f a -> Fix f -> a
cata alg = alg . fmap (cata alg) . unFix
ana :: Functor f => Coalgebra f a -> a -> Fix f
ana coalg = Fix . fmap (ana coalg) . coalg
```

We call these functions *catamorphisms* and *anamorphisms*. Notice especially that the types of these two functions simply reverse the direction of arrows. Interpreted in another way they transform an algebra/coalgebra which defines a flat structure-preserving mapping between `Fix f`

`f`

into a function which either rolls or unrolls the fixpoint. What is particularly nice about this approach is that the recursion is abstracted away inside the functor definition and we are free to just implement the flat transformation logic!

For example a construction of the natural numbers in this form:

```
{-# LANGUAGE TypeOperators #-}
{-# LANGUAGE DeriveFunctor #-}
{-# LANGUAGE StandaloneDeriving #-}
{-# LANGUAGE FlexibleInstances #-}
{-# LANGUAGE UndecidableInstances #-}
type Algebra f a = f a -> a
type Coalgebra f a = a -> f a
newtype Fix f = Fix { unFix :: f (Fix f) }
-- catamorphism
cata :: Functor f => Algebra f a -> Fix f -> a
cata alg = alg . fmap (cata alg) . unFix
-- anamorphism
ana :: Functor f => Coalgebra f a -> a -> Fix f
ana coalg = Fix . fmap (ana coalg) . coalg
-- hylomorphism
hylo :: Functor f => Algebra f b -> Coalgebra f a -> a -> b
hylo f g = cata f . ana g
type Nat = Fix NatF
data NatF a = S a | Z deriving (Eq,Show)
instance Functor NatF where
fmap f Z = Z
fmap f (S x) = S (f x)
plus :: Nat -> Nat -> Nat
plus n = cata phi where
phi Z = n
phi (S m) = s m
times :: Nat -> Nat -> Nat
times n = cata phi where
phi Z = z
phi (S m) = plus n m
int :: Nat -> Int
int = cata phi where
phi Z = 0
phi (S f) = 1 + f
nat :: Integer -> Nat
nat = ana (psi Z S) where
psi f _ 0 = f
psi _ f n = f (n-1)
z :: Nat
z = Fix Z
s :: Nat -> Nat
s = Fix . S
type Str = Fix StrF
data StrF x = Cons Char x | Nil
instance Functor StrF where
fmap f (Cons a as) = Cons a (f as)
fmap f Nil = Nil
nil :: Str
nil = Fix Nil
cons :: Char -> Str -> Str
cons x xs = Fix (Cons x xs)
str :: Str -> String
str = cata phi where
phi Nil = []
phi (Cons x xs) = x : xs
str' :: String -> Str
str' = ana (psi Nil Cons) where
psi f _ [] = f
psi _ f (a:as) = f a as
map' :: (Char -> Char) -> Str -> Str
map' f = hylo g unFix
where
g Nil = Fix Nil
g (Cons a x) = Fix $ Cons (f a) x
type Tree a = Fix (TreeF a)
data TreeF a f = Leaf a | Tree a f f deriving (Show)
instance Functor (TreeF a) where
fmap f (Leaf a) = Leaf a
fmap f (Tree a b c) = Tree a (f b) (f c)
depth :: Tree a -> Int
depth = cata phi where
phi (Leaf _) = 0
phi (Tree _ l r) = 1 + max l r
example1 :: Int
example1 = int (plus (nat 125) (nat 25))
-- 150
```

Or for example an interpreter for a small expression language that depends on a scoping dictionary.

```
{-# LANGUAGE GADTs #-}
{-# LANGUAGE DeriveFunctor #-}
{-# LANGUAGE StandaloneDeriving #-}
{-# LANGUAGE FlexibleInstances #-}
{-# LANGUAGE UndecidableInstances #-}
import Control.Applicative
import qualified Data.Map as M
type Algebra f a = f a -> a
type Coalgebra f a = a -> f a
newtype Fix f = Fix { unFix :: f (Fix f) }
cata :: Functor f => Algebra f a -> Fix f -> a
cata alg = alg . fmap (cata alg) . unFix
ana :: Functor f => Coalgebra f a -> a -> Fix f
ana coalg = Fix . fmap (ana coalg) . coalg
hylo :: Functor f => Algebra f b -> Coalgebra f a -> a -> b
hylo f g = cata f . ana g
type Id = String
type Env = M.Map Id Int
type Expr = Fix ExprF
data ExprF a
= Lit Int
| Var Id
| Add a a
| Mul a a
deriving (Show, Eq, Ord, Functor)
deriving instance Eq (f (Fix f)) => Eq (Fix f)
deriving instance Ord (f (Fix f)) => Ord (Fix f)
deriving instance Show (f (Fix f)) => Show (Fix f)
eval :: M.Map Id Int -> Fix ExprF -> Maybe Int
eval env = cata phi where
phi ex = case ex of
Lit c -> pure c
Var i -> M.lookup i env
Add x y -> liftA2 (+) x y
Mul x y -> liftA2 (*) x y
expr :: Expr
expr = Fix (Mul n (Fix (Add x y)))
where
n = Fix (Lit 10)
x = Fix (Var "x")
y = Fix (Var "y")
env :: M.Map Id Int
env = M.fromList [("x", 1), ("y", 2)]
compose :: (f (Fix f) -> c) -> (a -> Fix f) -> a -> c
compose x y = x . unFix . y
example :: Maybe Int
example = eval env expr
-- Just 30
```

What's especially nice about this approach is how naturally catamorphisms compose into efficient composite transformations.

```
compose :: Functor f => (f (Fix f) -> c) -> (a -> Fix f) -> a -> c
compose f g = f . unFix . g
```

## recursion-schemes

This is an advanced section, and is not typically necessary to write Haskell.

The code from the F-algebra examples above is implemented in an off-the-shelf library called `recursion-schemes`

.

```
{-# LANGUAGE TypeFamilies #-}
{-# LANGUAGE DeriveFunctor #-}
import Data.Functor.Foldable
type Var = String
data Exp
= Var Var
| App Exp Exp
| Lam [Var] Exp
deriving Show
data ExpF a
= VarF Var
| AppF a a
| LamF [Var] a
deriving Functor
type instance Base Exp = ExpF
instance Foldable Exp where
project (Var a) = VarF a
project (App a b) = AppF a b
project (Lam a b) = LamF a b
instance Unfoldable Exp where
embed (VarF a) = Var a
embed (AppF a b) = App a b
embed (LamF a b) = Lam a b
fvs :: Exp -> [Var]
fvs = cata phi
where phi (VarF a) = [a]
phi (AppF a b) = a ++ b
phi (LamF a b) = foldr (filter . (/=)) a b
```

An example of usage:

```
{-# LANGUAGE DeriveFunctor #-}
{-# LANGUAGE KindSignatures #-}
{-# LANGUAGE FlexibleInstances #-}
{-# LANGUAGE TypeSynonymInstances #-}
import Data.Traversable
import Control.Monad hiding (forM_, mapM, sequence)
import Prelude hiding (mapM)
import qualified Data.Map as M
newtype Fix (f :: * -> *) = Fix { outF :: f (Fix f) }
-- Catamorphism
cata :: Functor f => (f a -> a) -> Fix f -> a
cata f = f . fmap (cata f) . outF
-- Monadic catamorphism
cataM :: (Traversable f, Monad m) => (f a -> m a) -> Fix f -> m a
cataM f = f <=< mapM (cataM f) . outF
data ExprF r
= EVar String
| EApp r r
| ELam r r
deriving (Show, Eq, Ord, Functor)
type Expr = Fix ExprF
instance Show (Fix ExprF) where
show (Fix f) = show f
instance Eq (Fix ExprF) where
Fix x == Fix y = x == y
instance Ord (Fix ExprF) where
compare (Fix x) (Fix y) = compare x y
mkApp :: Fix ExprF -> Fix ExprF -> Fix ExprF
mkApp x y = Fix (EApp x y)
mkVar :: String -> Fix ExprF
mkVar x = Fix (EVar x)
mkLam :: Fix ExprF -> Fix ExprF -> Fix ExprF
mkLam x y = Fix (ELam x y)
i :: Fix ExprF
i = mkLam (mkVar "x") (mkVar "x")
k :: Fix ExprF
k = mkLam (mkVar "x") $ mkLam (mkVar "y") $ (mkVar "x")
subst :: M.Map String (ExprF Expr) -> Expr -> Expr
subst env = cata alg where
alg (EVar x) | Just e <- M.lookup x env = Fix e
alg e = Fix e
```

See:

## Data types à la carte

TODO

## Hint and Mueval

This is an advanced section, and is not typically necessary to write Haskell.

GHC itself can actually interpret arbitrary Haskell source on the fly by hooking into the GHC's bytecode interpreter ( the same used for GHCi ). The hint package allows us to parse, typecheck, and evaluate arbitrary strings into arbitrary Haskell programs and evaluate them.

```
import Language.Haskell.Interpreter
foo :: Interpreter String
foo = eval "(\\x -> x) 1"
example :: IO (Either InterpreterError String)
example = runInterpreter foo
```

This is generally not a wise thing to build a library around, unless of course the purpose of the program is itself to evaluate arbitrary Haskell code ( something like an online Haskell shell or the likes ).

Both hint and mueval do effectively the same task, designed around slightly different internals of the GHC Api.

See:

# Testing

Contrary to a lot of misinformation, unit testing in Haskell is quite common and robust. Although generally speaking unit tests tend to be of less importance in Haskell since the type system makes an enormous amount of invalid programs completely inexpressible by construction. Unit tests tend to be written later in the development lifecycle and generally tend to be about the core logic of the program and not the intermediate plumbing.

A prominent school of thought on Haskell library design tends to favor constructing programs built around strong equation laws which guarantee strong invariants about program behavior under composition. Many of the testing tools are built around this style of design.

## QuickCheck

Probably the most famous Haskell library, QuickCheck is a testing framework for generating large random tests for arbitrary functions automatically based on the types of their arguments.

```
quickCheck :: Testable prop => prop -> IO ()
(==>) :: Testable prop => Bool -> prop -> Property
forAll :: (Show a, Testable prop) => Gen a -> (a -> prop) -> Property
choose :: Random a => (a, a) -> Gen a
```

```
import Test.QuickCheck
qsort :: [Int] -> [Int]
qsort [] = []
qsort (x:xs) = qsort lhs ++ [x] ++ qsort rhs
where lhs = filter (< x) xs
rhs = filter (>= x) xs
prop_maximum :: [Int] -> Property
prop_maximum xs = not (null xs) ==>
last (qsort xs) == maximum xs
main :: IO ()
main = quickCheck prop_maximum
```

```
$ runhaskell qcheck.hs
*** Failed! Falsifiable (after 3 tests and 4 shrinks):
[0]
[1]
$ runhaskell qcheck.hs
+++ OK, passed 1000 tests.
```

The test data generator can be extended with custom types and refined with predicates that restrict the domain of cases to test.

```
import Test.QuickCheck
data Color = Red | Green | Blue deriving Show
instance Arbitrary Color where
arbitrary = do
n <- choose (0,2) :: Gen Int
return $ case n of
0 -> Red
1 -> Green
2 -> Blue
example1 :: IO [Color]
example1 = sample' arbitrary
-- [Red,Green,Red,Blue,Red,Red,Red,Blue,Green,Red,Red]
```

See: QuickCheck: An Automatic Testing Tool for Haskell

## SmallCheck

Like QuickCheck, SmallCheck is a property testing system but instead of producing random arbitrary test data it instead enumerates a deterministic series of test data to a fixed depth.

```
smallCheck :: Testable IO a => Depth -> a -> IO ()
list :: Depth -> Series Identity a -> [a]
sample' :: Gen a -> IO [a]
```

```
λ: list 3 series :: [Int]
[0,1,-1,2,-2,3,-3]
λ: list 3 series :: [Double]
[0.0,1.0,-1.0,2.0,0.5,-2.0,4.0,0.25,-0.5,-4.0,-0.25]
λ: list 3 series :: [(Int, String)]
[(0,""),(1,""),(0,"a"),(-1,""),(0,"b"),(1,"a"),(2,""),(1,"b"),(-1,"a"),(-2,""),(-1,"b"),(2,"a"),(-2,"a"),(2,"b"),(-2,"b")]
```

It is useful to generate test cases over *all* possible inputs of a program up to some depth.

```
import Test.SmallCheck
distrib :: Int -> Int -> Int -> Bool
distrib a b c = a * (b + c) == a * b + a * c
cauchy :: [Double] -> [Double] -> Bool
cauchy xs ys = (abs (dot xs ys))^2 <= (dot xs xs) * (dot ys ys)
failure :: [Double] -> [Double] -> Bool
failure xs ys = abs (dot xs ys) <= (dot xs xs) * (dot ys ys)
dot :: Num a => [a] -> [a] -> a
dot xs ys = sum (zipWith (*) xs ys)
main :: IO ()
main = do
putStrLn "Testing distributivity..."
smallCheck 25 distrib
putStrLn "Testing Cauchy-Schwarz..."
smallCheck 4 cauchy
putStrLn "Testing invalid Cauchy-Schwarz..."
smallCheck 4 failure
```

```
$ runhaskell smallcheck.hs
Testing distributivity...
Completed 132651 tests without failure.
Testing Cauchy-Schwarz...
Completed 27556 tests without failure.
Testing invalid Cauchy-Schwarz...
Failed test no. 349.
there exist [1.0] [0.5] such that
condition is false
```

Just like for QuickCheck we can implement series instances for our custom datatypes. For example there is no default instance for Vector, so let's implement one:

```
{-# LANGUAGE FlexibleInstances #-}
{-# LANGUAGE MultiParamTypeClasses #-}
import Test.SmallCheck
import Test.SmallCheck.Series
import Control.Applicative
import qualified Data.Vector as V
dot :: Num a => V.Vector a -> V.Vector a -> a
dot xs ys = V.sum (V.zipWith (*) xs ys)
cauchy :: V.Vector Double -> V.Vector Double -> Bool
cauchy xs ys = (abs (dot xs ys))^2 <= (dot xs xs) * (dot ys ys)
instance (Serial m a, Monad m) => Serial m (V.Vector a) where
series = V.fromList <$> series
main :: IO ()
main = smallCheck 4 cauchy
```

SmallCheck can also use Generics to derive Serial instances, for example to enumerate all trees of a certain depth we might use:

```
{-# LANGUAGE FlexibleInstances #-}
{-# LANGUAGE MultiParamTypeClasses #-}
{-# LANGUAGE DeriveGeneric #-}
import GHC.Generics
import Test.SmallCheck.Series
data Tree a = Null | Fork (Tree a) a (Tree a)
deriving (Show, Generic)
instance Serial m a => Serial m (Tree a)
example :: [Tree ()]
example = list 3 series
main = print example
```

## QuickSpec

Using the QuickCheck arbitrary machinery we can also rather remarkably enumerate a large number of combinations of functions to try and deduce algebraic laws from trying out inputs for small cases.

Of course the fundamental limitation of this approach is that a function may not exhibit any interesting properties for small cases or for simple function compositions. So in general case this approach won't work, but practically it still quite useful.

```
{-# LANGUAGE TypeOperators #-}
{-# LANGUAGE ConstraintKinds #-}
{-# LANGUAGE ScopedTypeVariables #-}
import Data.List
import Data.Typeable
import Test.QuickSpec hiding (lists, bools, arith)
import Test.QuickCheck
type Var k a = (Typeable a, Arbitrary a, CoArbitrary a, k a)
listCons :: forall a. Var Ord a => a -> Sig
listCons a = background
[
"[]" `fun0` ([] :: [a]),
":" `fun2` ((:) :: a -> [a] -> [a])
]
lists :: forall a. Var Ord a => a -> [Sig]
lists a =
[
-- Names to print arbitrary variables
funs',
funvars',
vars',
-- Ambient definitions
listCons a,
-- Expressions to deduce properties of
"sort" `fun1` (sort :: [a] -> [a]),
"map" `fun2` (map :: (a -> a) -> [a] -> [a]),
"id" `fun1` (id :: [a] -> [a]),
"reverse" `fun1` (reverse :: [a] -> [a]),
"minimum" `fun1` (minimum :: [a] -> a),
"length" `fun1` (length :: [a] -> Int),
"++" `fun2` ((++) :: [a] -> [a] -> [a])
]
where
funs' = funs (undefined :: a)
funvars' = vars ["f", "g", "h"] (undefined :: a -> a)
vars' = ["xs", "ys", "zs"] `vars` (undefined :: [a])
tvar :: A
tvar = undefined
main :: IO ()
main = quickSpec (lists tvar)
```

Running this we rather see it is able to deduce most of the laws for list functions.

```
$ runhaskell src/quickspec.hs
== API ==
-- functions --
map :: (A -> A) -> [A] -> [A]
minimum :: [A] -> A
(++) :: [A] -> [A] -> [A]
length :: [A] -> Int
sort, id, reverse :: [A] -> [A]
-- background functions --
id :: A -> A
(:) :: A -> [A] -> [A]
(.) :: (A -> A) -> (A -> A) -> A -> A
[] :: [A]
-- variables --
f, g, h :: A -> A
xs, ys, zs :: [A]
-- the following types are using non-standard equality --
A -> A
-- WARNING: there are no variables of the following types; consider adding some --
A
== Testing ==
Depth 1: 12 terms, 4 tests, 24 evaluations, 12 classes, 0 raw equations.
Depth 2: 80 terms, 500 tests, 18673 evaluations, 52 classes, 28 raw equations.
Depth 3: 1553 terms, 500 tests, 255056 evaluations, 1234 classes, 319 raw equations.
319 raw equations; 1234 terms in universe.
== Equations about map ==
1: map f [] == []
2: map id xs == xs
3: map (f.g) xs == map f (map g xs)
== Equations about minimum ==
4: minimum [] == undefined
== Equations about (++) ==
5: xs++[] == xs
6: []++xs == xs
7: (xs++ys)++zs == xs++(ys++zs)
== Equations about sort ==
8: sort [] == []
9: sort (sort xs) == sort xs
== Equations about id ==
10: id xs == xs
== Equations about reverse ==
11: reverse [] == []
12: reverse (reverse xs) == xs
== Equations about several functions ==
13: minimum (xs++ys) == minimum (ys++xs)
14: length (map f xs) == length xs
15: length (xs++ys) == length (ys++xs)
16: sort (xs++ys) == sort (ys++xs)
17: map f (reverse xs) == reverse (map f xs)
18: minimum (sort xs) == minimum xs
19: minimum (reverse xs) == minimum xs
20: minimum (xs++xs) == minimum xs
21: length (sort xs) == length xs
22: length (reverse xs) == length xs
23: sort (reverse xs) == sort xs
24: map f xs++map f ys == map f (xs++ys)
25: reverse xs++reverse ys == reverse (ys++xs)
```

Keep in mind the rather remarkable fact that this is all deduced automatically from the types alone!

## Criterion

Criterion is a statistically aware benchmarking tool.

```
whnf :: (a -> b) -> a -> Pure
nf :: NFData b => (a -> b) -> a -> Pure
nfIO :: NFData a => IO a -> IO ()
bench :: Benchmarkable b => String -> b -> Benchmark
```

```
import Criterion.Main
import Criterion.Config
-- Naive recursion for fibonacci numbers.
fib1 :: Int -> Int
fib1 0 = 0
fib1 1 = 1
fib1 n = fib1 (n-1) + fib1 (n-2)
-- Use the De Moivre closed form for fibonacci numbers.
fib2 :: Int -> Int
fib2 x = truncate $ ( 1 / sqrt 5 ) * ( phi ^ x - psi ^ x )
where
phi = ( 1 + sqrt 5 ) / 2
psi = ( 1 - sqrt 5 ) / 2
suite :: [Benchmark]
suite = [
bgroup "naive" [
bench "fib 10" $ whnf fib1 5
, bench "fib 20" $ whnf fib1 10
],
bgroup "de moivre" [
bench "fib 10" $ whnf fib2 5
, bench "fib 20" $ whnf fib2 10
]
]
main :: IO ()
main = defaultMain suite
```

```
$ runhaskell criterion.hs
warming up
estimating clock resolution...
mean is 2.349801 us (320001 iterations)
found 1788 outliers among 319999 samples (0.6%)
1373 (0.4%) high severe
estimating cost of a clock call...
mean is 65.52118 ns (23 iterations)
found 1 outliers among 23 samples (4.3%)
1 (4.3%) high severe
benchmarking naive/fib 10
mean: 9.903067 us, lb 9.885143 us, ub 9.924404 us, ci 0.950
std dev: 100.4508 ns, lb 85.04638 ns, ub 123.1707 ns, ci 0.950
benchmarking naive/fib 20
mean: 120.7269 us, lb 120.5470 us, ub 120.9459 us, ci 0.950
std dev: 1.014556 us, lb 858.6037 ns, ub 1.296920 us, ci 0.950
benchmarking de moivre/fib 10
mean: 7.699219 us, lb 7.671107 us, ub 7.802116 us, ci 0.950
std dev: 247.3021 ns, lb 61.66586 ns, ub 572.1260 ns, ci 0.950
found 4 outliers among 100 samples (4.0%)
2 (2.0%) high mild
2 (2.0%) high severe
variance introduced by outliers: 27.726%
variance is moderately inflated by outliers
benchmarking de moivre/fib 20
mean: 8.082639 us, lb 8.018560 us, ub 8.350159 us, ci 0.950
std dev: 595.2161 ns, lb 77.46251 ns, ub 1.408784 us, ci 0.950
found 8 outliers among 100 samples (8.0%)
4 (4.0%) high mild
4 (4.0%) high severe
variance introduced by outliers: 67.628%
variance is severely inflated by outliers
```

Criterion can also generate a HTML page containing the benchmark results plotted

```
$ ghc -O2 --make criterion.hs
$ ./criterion -o bench.html
```

## Tasty

Tasty combines all of the testing frameworks into a common API for forming runnable batches of tests and collecting the results.

```
import Test.Tasty
import Test.Tasty.HUnit
import Test.Tasty.QuickCheck
import qualified Test.Tasty.SmallCheck as SC
arith :: Integer -> Integer -> Property
arith x y = (x > 0) && (y > 0) ==> (x+y)^2 > x^2 + y^2
negation :: Integer -> Bool
negation x = abs (x^2) >= x
suite :: TestTree
suite = testGroup "Test Suite" [
testGroup "Units"
[ testCase "Equality" $ True @=? True
, testCase "Assertion" $ assert $ (length [1,2,3]) == 3
],
testGroup "QuickCheck tests"
[ testProperty "Quickcheck test" arith
],
testGroup "SmallCheck tests"
[ SC.testProperty "Negation" negation
]
]
main :: IO ()
main = defaultMain suite
```

```
$ runhaskell TestSuite.hs
Unit tests
Units
Equality: OK
Assertion: OK
QuickCheck tests
Quickcheck test: OK
+++ OK, passed 100 tests.
SmallCheck tests
Negation: OK
11 tests completed
```

## silently

Often in the process of testing IO heavy code we'll need to redirect stdout to compare it some known quantity. The `silently`

package allows us to capture anything done to stdout across any library inside of IO block and return the result to the test runner.

`capture :: IO a -> IO (String, a)`

```
import Test.Tasty
import Test.Tasty.HUnit
import System.IO.Silently
test :: Int -> IO ()
test n = print (n * n)
testCapture n = do
(stdout, result) <- capture (test n)
assert (stdout == show (n*n) ++ "\n")
suite :: TestTree
suite = testGroup "Test Suite" [
testGroup "Units"
[ testCase "Equality" $ testCapture 10
]
]
main :: IO ()
main = defaultMain suite
```

# Type Families

## MultiParam Typeclasses

Resolution of vanilla Haskell 98 typeclasses proceeds via very simple context reduction that minimizes interdependency between predicates, resolves superclasses, and reduces the types to head normal form. For example:

```
(Eq [a], Ord [a]) => [a]
==> Ord a => [a]
```

If a single parameter typeclass expresses a property of a type ( i.e. it's in a class or not in class ) then a multiparameter typeclass expresses relationships between types. For example if we wanted to express the relation a type can be converted to another type we might use a class like:

```
{-# LANGUAGE MultiParamTypeClasses #-}
import Data.Char
class Convertible a b where
convert :: a -> b
instance Convertible Int Integer where
convert = toInteger
instance Convertible Int Char where
convert = chr
instance Convertible Char Int where
convert = ord
```

Of course now our instances for `Convertible Int`

are not unique anymore, so there no longer exists a nice procedure for determining the inferred type of `b`

from just `a`

. To remedy this let's add a functional dependency `a -> b`

, which tells GHC that an instance `a`

uniquely determines the instance that b can be. So we'll see that our two instances relating `Int`

to both `Integer`

and `Char`

conflict.

```
{-# LANGUAGE MultiParamTypeClasses #-}
{-# LANGUAGE FunctionalDependencies #-}
import Data.Char
class Convertible a b | a -> b where
convert :: a -> b
instance Convertible Int Char where
convert = chr
instance Convertible Char Int where
convert = ord
```

```
Functional dependencies conflict between instance declarations:
instance Convertible Int Integer
instance Convertible Int Char
```

Now there's a simpler procedure for determining instances uniquely and multiparameter typeclasses become more usable and inferable again. Effectively a functional dependency `| a -> b`

says that we can't define multiple multiparamater typeclass instances with the same `a`

but different `b`

.

```
λ: convert (42 :: Int)
'*'
λ: convert '*'
42
```

Now let's make things not so simple. Turning on `UndecidableInstances`

loosens the constraint on context reduction that can only allow constraints of the class to become structural smaller than its head. As a result implicit computation can now occur *within in the type class instance search*. Combined with a type-level representation of Peano numbers we find that we can encode basic arithmetic at the type-level.

```
{-# LANGUAGE FlexibleContexts #-}
{-# LANGUAGE FlexibleInstances #-}
{-# LANGUAGE MultiParamTypeClasses #-}
{-# LANGUAGE FunctionalDependencies #-}
{-# LANGUAGE UndecidableInstances #-}
data Z
data S n
type Zero = Z
type One = S Zero
type Two = S One
type Three = S Two
type Four = S Three
zero :: Zero
zero = undefined
one :: One
one = undefined
two :: Two
two = undefined
three :: Three
three = undefined
four :: Four
four = undefined
class Eval a where
eval :: a -> Int
instance Eval Zero where
eval _ = 0
instance Eval n => Eval (S n) where
eval m = 1 + eval (prev m)
class Pred a b | a -> b where
prev :: a -> b
instance Pred Zero Zero where
prev = undefined
instance Pred (S n) n where
prev = undefined
class Add a b c | a b -> c where
add :: a -> b -> c
instance Add Zero a a where
add = undefined
instance Add a b c => Add (S a) b (S c) where
add = undefined
f :: Three
f = add one two
g :: S (S (S (S Z)))
g = add two two
h :: Int
h = eval (add three four)
```

If the typeclass contexts look similar to Prolog you're not wrong, if one reads the contexts qualifier `(=>)`

backwards as turnstiles `:-`

then it's precisely the same equations.

```
add(0, A, A).
add(s(A), B, s(C)) :- add(A, B, C).
pred(0, 0).
pred(S(A), A).
```

This is kind of abusing typeclasses and if used carelessly it can fail to terminate or overflow at compile-time. `UndecidableInstances`

shouldn't be turned on without careful forethought about what it implies.

```
<interactive>:1:1:
Context reduction stack overflow; size = 201
```

## Type Families

Type families allows us to write functions in the type domain which take types as arguments which can yield either types or values indexed on their arguments which are evaluated at compile-time in during typechecking. Type families come in two varieties: **data families** and **type synonym families**.

**type families**are named function on types**data families**are type-indexed data types

First let's look at *type synonym families*, there are two equivalent syntactic ways of constructing them. Either as *associated* type families declared within a typeclass or as standalone declarations at the toplevel. The following forms are semantically equivalent, although the unassociated form is strictly more general:

```
-- (1) Unassociated form
type family Rep a
type instance Rep Int = Char
type instance Rep Char = Int
class Convertible a where
convert :: a -> Rep a
instance Convertible Int where
convert = chr
instance Convertible Char where
convert = ord
-- (2) Associated form
class Convertible a where
type Rep a
convert :: a -> Rep a
instance Convertible Int where
type Rep Int = Char
convert = chr
instance Convertible Char where
type Rep Char = Int
convert = ord
```

Using the same example we used for multiparameter + functional dependencies illustration we see that there is a direct translation between the type family approach and functional dependencies. These two approaches have the same expressive power.

An associated type family can be queried using the `:kind!`

command in GHCi.

```
λ: :kind! Rep Int
Rep Int :: *
= Char
λ: :kind! Rep Char
Rep Char :: *
= Int
```

*Data families* on the other hand allow us to create new type parameterized data constructors. Normally we can only define typeclasses functions whose behavior results in a uniform result which is purely a result of the typeclasses arguments. With data families we can allow specialized behavior indexed on the type.

For example if we wanted to create more complicated vector structures ( bit-masked vectors, vectors of tuples, ... ) that exposed a uniform API but internally handled the differences in their data layout we can use data families to accomplish this:

```
{-# LANGUAGE TypeFamilies #-}
import qualified Data.Vector.Unboxed as V
data family Array a
data instance Array Int = IArray (V.Vector Int)
data instance Array Bool = BArray (V.Vector Bool)
data instance Array (a,b) = PArray (Array a) (Array b)
data instance Array (Maybe a) = MArray (V.Vector Bool) (Array a)
class IArray a where
index :: Array a -> Int -> a
instance IArray Int where
index (IArray xs) i = xs V.! i
instance IArray Bool where
index (BArray xs) i = xs V.! i
-- Vector of pairs
instance (IArray a, IArray b) => IArray (a, b) where
index (PArray xs ys) i = (index xs i, index ys i)
-- Vector of missing values
instance (IArray a) => IArray (Maybe a) where
index (MArray bm xs) i =
case bm V.! i of
True -> Nothing
False -> Just $ index xs i
```

## Injectivity

The type level functions defined by type-families are not necessarily *injective*, the function may map two distinct input types to the same output type. This differs from the behavior of type constructors ( which are also type-level functions ) which are injective.

For example for the constructor `Maybe`

, `Maybe t1 = Maybe t2`

implies that `t1 = t2`

.

```
data Maybe a = Nothing | Just a
-- Maybe a ~ Maybe b implies a ~ b
type instance F Int = Bool
type instance F Char = Bool
-- F a ~ F b does not imply a ~ b, in general
```

## Roles

This is an advanced section, and is not typically necessary to write Haskell.

Roles are a further level of specification for type variables parameters of datatypes.

`nominal`

`representational`

`phantom`

They were added to the language to address a rather nasty and long-standing bug around the correspondence between a newtype and its runtime representation. The fundamental distinction that roles introduce is there are two notions of type equality. Two types are *nominally equal* when they have the same name. This is the usual equality in Haskell or Core. Two types are *representationally equal* when they have the same representation. (If a type is higher-kinded, all nominally equal instantiations lead to representationally equal types.)

`nominal`

- Two types are the same.`representational`

- Two types have the same runtime representation.

```
{-# LANGUAGE TypeFamilies #-}
{-# LANGUAGE StandaloneDeriving #-}
{-# LANGUAGE GeneralizedNewtypeDeriving #-}
newtype Age = MkAge { unAge :: Int }
type family Inspect x
type instance Inspect Age = Int
type instance Inspect Int = Bool
class Boom a where
boom :: a -> Inspect a
instance Boom Int where
boom = (== 0)
deriving instance Boom Age
-- GHC 7.6.3 exhibits undefined behavior
failure = boom (MkAge 3)
-- -6341068275333450897
```

Roles are normally inferred automatically, but with the `RoleAnnotations`

extension they can be manually annotated. Except in rare cases this should not be necessary although it is helpful to know what is going on under the hood.

```
{-# LANGUAGE GADTs #-}
{-# LANGUAGE PolyKinds #-}
{-# LANGUAGE DataKinds #-}
{-# LANGUAGE KindSignatures #-}
{-# LANGUAGE RoleAnnotations #-}
data Nat = Zero | Suc Nat
type role Vec nominal representational
data Vec :: Nat -> * -> * where
Nil :: Vec Zero a
(:*) :: a -> Vec n a -> Vec (Suc n) a
type role App representational nominal
data App (f :: k -> *) (a :: k) = App (f a)
type role Mu nominal nominal
data Mu (f :: (k -> *) -> k -> *) (a :: k) = Roll (f (Mu f) a)
type role Proxy phantom
data Proxy (a :: k) = Proxy
```

See:

## Monotraversable

Using type families, mono-traversable generalizes the notion of Functor, Foldable, and Traversable to include both monomorphic and polymorphic types.

```
omap :: MonoFunctor mono => (Element mono -> Element mono) -> mono -> mono
otraverse :: (Applicative f, MonoTraversable mono)
=> (Element mono -> f (Element mono)) -> mono -> f mono
ofoldMap :: (Monoid m, MonoFoldable mono)
=> (Element mono -> m) -> mono -> m
ofoldl' :: MonoFoldable mono
=> (a -> Element mono -> a) -> a -> mono -> a
ofoldr :: MonoFoldable mono
=> (Element mono -> b -> b) -> b -> mono -> b
```

For example the text type normally does not admit any of these type-classes since, but now we can write down the instances that model the interface of Foldable and Traversable.

```
{-# LANGUAGE TypeFamilies #-}
{-# LANGUAGE OverloadedStrings #-}
import Data.Text
import Data.Char
import Data.Monoid
import Data.MonoTraversable
import Control.Applicative
bs :: Text
bs = "Hello Haskell."
shift :: Text
shift = omap (chr . (+1) . ord) bs
-- "Ifmmp!Ibtlfmm/"
backwards :: [Char]
backwards = ofoldl' (flip (:)) "" bs
-- ".lleksaH olleH"
data MyMonoType = MNil | MCons Int MyMonoType deriving Show
type instance Element MyMonoType = Int
instance MonoFunctor MyMonoType where
omap f MNil = MNil
omap f (MCons x xs) = f x `MCons` omap f xs
instance MonoFoldable MyMonoType where
ofoldMap f = ofoldr (mappend . f) mempty
ofoldr = mfoldr
ofoldl' = mfoldl'
ofoldr1Ex f = ofoldr1Ex f . mtoList
ofoldl1Ex' f = ofoldl1Ex' f . mtoList
instance MonoTraversable MyMonoType where
omapM f xs = mapM f (mtoList xs) >>= return . mfromList
otraverse f = ofoldr acons (pure MNil)
where acons x ys = MCons <$> f x <*> ys
mtoList :: MyMonoType -> [Int]
mtoList (MNil) = []
mtoList (MCons x xs) = x : (mtoList xs)
mfromList :: [Int] -> MyMonoType
mfromList [] = MNil
mfromList (x:xs) = MCons x (mfromList xs)
mfoldr :: (Int -> a -> a) -> a -> MyMonoType -> a
mfoldr f z MNil = z
mfoldr f z (MCons x xs) = f x (mfoldr f z xs)
mfoldl' :: (a -> Int -> a) -> a -> MyMonoType -> a
mfoldl' f z MNil = z
mfoldl' f z (MCons x xs) = let z' = z `f` x
in seq z' $ mfoldl' f z' xs
ex1 :: Int
ex1 = mfoldl' (+) 0 (mfromList [1..25])
ex2 :: MyMonoType
ex2 = omap (+1) (mfromList [1..25])
```

See: From Semigroups to Monads

## NonEmpty

Rather than having degenerate (and often partial) cases of many of the Prelude functions to accommodate the null case of lists, it is sometimes preferable to statically enforce empty lists from even being constructed as an inhabitant of a type.

```
infixr 5 :|, <|
data NonEmpty a = a :| [a]
head :: NonEmpty a -> a
toList :: NonEmpty a -> [a]
fromList :: [a] -> NonEmpty a
```

```
head :: NonEmpty a -> a
head ~(a :| _) = a
```

```
import Data.List.NonEmpty
import Prelude hiding (head, tail, foldl1)
import Data.Foldable (foldl1)
a :: NonEmpty Integer
a = fromList [1,2,3]
-- 1 :| [2,3]
b :: NonEmpty Integer
b = 1 :| [2,3]
-- 1 :| [2,3]
c :: NonEmpty Integer
c = fromList []
-- *** Exception: NonEmpty.fromList: empty list
d :: Integer
d = foldl1 (+) $ fromList [1..100]
-- 5050
```

## Overloaded Lists

In GHC 7.8 `-XOverloadedLists`

can be used to avoid the extraneous `fromList`

and `toList`

conversions.

## Manual Proofs

This is an advanced section, and is not typically necessary to write Haskell.

One of most deep results in computer science, the Curry–Howard correspondence, is the relation that logical propositions can be modeled by types and instantiating those types constitute proofs of these propositions. Programs are proofs and proofs are programs.

Types | Logic |
---|---|

`A` |
proposition |

`a : A` |
proof |

`B(x)` |
predicate |

`Void` |
⊥ |

`Unit` |
⊤ |

`A + B` |
A ∨ B |

`A × B` |
A ∧ B |

`A -> B` |
A ⇒ B |

In dependently typed languages we can exploit this result to its full extent, in Haskell we don't have the strength that dependent types provide but can still prove trivial results. For example, now we can model a type level function for addition and provide a small proof that zero is an additive identity.

```
P 0 [ base step ]
∀n. P n → P (1+n) [ inductive step ]
-------------------
∀n. P(n)
```

```
Axiom 1: a + 0 = a
Axiom 2: a + suc b = suc (a + b)
0 + suc a
= suc (0 + a) [by Axiom 2]
= suc a [Induction hypothesis]
∎
```

Translated into Haskell our axioms are simply type definitions and recursing over the inductive datatype constitutes the inductive step of our proof.

```
{-# LANGUAGE GADTs #-}
{-# LANGUAGE TypeFamilies #-}
{-# LANGUAGE ExplicitForAll #-}
{-# LANGUAGE TypeOperators #-}
data Z
data S n
data SNat n where
Zero :: SNat Z
Succ :: SNat n -> SNat (S n)
data Eql a b where
Refl :: Eql a a
type family Add m n
type instance Add Z n = n
type instance Add (S m) n = S (Add m n)
add :: SNat n -> SNat m -> SNat (Add n m)
add Zero m = m
add (Succ n) m = Succ (add n m)
cong :: Eql a b -> Eql (f a) (f b)
cong Refl = Refl
-- ∀n. 0 + suc n = suc n
plus_suc :: forall n. SNat n
-> Eql (Add Z (S n)) (S n)
plus_suc Zero = Refl
plus_suc (Succ n) = cong (plus_suc n)
-- ∀n. 0 + n = n
plus_zero :: forall n. SNat n
-> Eql (Add Z n) n
plus_zero Zero = Refl
plus_zero (Succ n) = cong (plus_zero n)
```

Using the `TypeOperators`

extension we can also use infix notation at the type-level.

```
data a :=: b where
Refl :: a :=: a
cong :: a :=: b -> (f a) :=: (f b)
cong Refl = Refl
type family (n :: Nat) :+ (m :: Nat) :: Nat
type instance Zero :+ m = m
type instance (Succ n) :+ m = Succ (n :+ m)
plus_suc :: forall n m. SNat n -> SNat m -> (n :+ (S m)) :=: (S (n :+ m))
plus_suc Zero m = Refl
plus_suc (Succ n) m = cong (plus_suc n m)
```

## Constraint Kinds

This is an advanced section, and is not typically necessary to write Haskell.

GHC's implementation also exposes the predicates that bound quantifiers in Haskell as types themselves, with the `-XConstraintKinds`

extension enabled. Using this extension we work with constraints as first class types.

```
Num :: * -> Constraint
Odd :: * -> Constraint
```

`type T1 a = (Num a, Ord a)`

The empty constraint set is indicated by `() :: Constraint`

.

For a contrived example if we wanted to create a generic `Sized`

class that carried with it constraints on the elements of the container in question we could achieve this quite simply using type families.

```
{-# LANGUAGE TypeFamilies #-}
{-# LANGUAGE ConstraintKinds #-}
import GHC.Exts (Constraint)
import Data.Hashable
import Data.HashSet
type family Con a :: Constraint
type instance Con [a] = (Ord a, Eq a)
type instance Con (HashSet a) = (Hashable a)
class Sized a where
gsize :: Con a => a -> Int
instance Sized [a] where
gsize = length
instance Sized (HashSet a) where
gsize = size
```

One use-case of this is to capture the typeclass dictionary constrained by a function and reify it as a value.

```
{-# LANGUAGE GADTs #-}
{-# LANGUAGE ConstraintKinds #-}
{-# LANGUAGE KindSignatures #-}
import GHC.Exts (Constraint)
data Dict :: Constraint -> * where
Dict :: (c) => Dict c
dShow :: Dict (Show a) -> a -> String
dShow Dict x = show x
dEqNum :: Dict (Eq a, Num a) -> a -> Bool
dEqNum Dict x = x == 0
fShow :: String
fShow = dShow Dict 10
fEqual :: Bool
fEqual = dEqNum Dict 0
```

## TypeFamilyDependencies

Type families historically have not been injective, i.e. they are not guaranteed to maps distinct elements of its arguments to the same element of its result. The syntax is similar to the multiparmater typeclass functional dependencies in that the resulting type is uniquely determined by a set of the type families parameters.

```
{-# LANGUAGE XTypeFamilyDependencies #-}
type family F a b c = (result :: k) | result -> a b c
type instance F Int Char Bool = Bool
type instance F Char Bool Int = Int
type instance F Bool Int Char = Char
```

See:

# Promotion

## Higher Kinded Types

What are higher kinded types?

The kind system in Haskell is unique by contrast with most other languages in that it allows datatypes to be constructed which take types and type constructor to other types. Such a system is said to support *higher kinded types*.

All kind annotations in Haskell necessarily result in a kind `*`

although any terms to the left may be higher-kinded (`* -> *`

).

The common example is the Monad which has kind `* -> *`

. But we have also seen this higher-kindedness in free monads.

```
data Free f a where
Pure :: a -> Free f a
Free :: f (Free f a) -> Free f a
data Cofree f a where
Cofree :: a -> f (Cofree f a) -> Cofree f a
```

```
Free :: (* -> *) -> * -> *
Cofree :: (* -> *) -> * -> *
```

For instance `Cofree Maybe a`

for some monokinded type `a`

models a non-empty list with `Maybe :: * -> *`

.

```
-- Cofree Maybe a is a non-empty list
testCofree :: Cofree Maybe Int
testCofree = (Cofree 1 (Just (Cofree 2 Nothing)))
```

## Kind Polymorphism

This is an advanced section, knowledge of kind polymorphism is not typically necessary to write Haskell.

The regular value level function which takes a function and applies it to an argument is universally generalized over in the usual Hindley-Milner way.

```
app :: forall a b. (a -> b) -> a -> b
app f a = f a
```

But when we do the same thing at the type-level we see we lose information about the polymorphism of the constructor applied.

```
-- TApp :: (* -> *) -> * -> *
data TApp f a = MkTApp (f a)
```

Turning on `-XPolyKinds`

allows polymorphic variables at the kind level as well.

```
-- Default: (* -> *) -> * -> *
-- PolyKinds: (k -> *) -> k -> *
data TApp f a = MkTApp (f a)
-- Default: ((* -> *) -> (* -> *)) -> (* -> *)
-- PolyKinds: ((k -> *) -> (k -> *)) -> (k -> *)
data Mu f a = Roll (f (Mu f) a)
-- Default: * -> *
-- PolyKinds: k -> *
data Proxy a = Proxy
```

Using the polykinded `Proxy`

type allows us to write down type class functions over constructors of arbitrary kind arity.

```
{-# LANGUAGE PolyKinds #-}
{-# LANGUAGE GADTs #-}
{-# LANGUAGE KindSignatures #-}
data Proxy a = Proxy
data Rep = Rep
class PolyClass a where
foo :: Proxy a -> Rep
foo = const Rep
-- () :: *
-- [] :: * -> *
-- Either :: * -> * -> *
instance PolyClass ()
instance PolyClass []
instance PolyClass Either
```

For example we can write down the polymorphic `S`

`K`

combinators at the type level now.

```
{-# LANGUAGE PolyKinds #-}
newtype I (a :: *) = I a
newtype K (a :: *) (b :: k) = K a
newtype Flip (f :: k1 -> k2 -> *) (x :: k2) (y :: k1) = Flip (f y x)
unI :: I a -> a
unI (I x) = x
unK :: K a b -> a
unK (K x) = x
unFlip :: Flip f x y -> f y x
unFlip (Flip x) = x
```

## Data Kinds

This is an advanced section, knowledge of kind data kinds is not typically necessary to write Haskell.

The `-XDataKinds`

extension allows us to use refer to constructors at the value level and the type level. Consider a simple sum type:

```
data S a b = L a | R b
-- S :: * -> * -> *
-- L :: a -> S a b
-- R :: b -> S a b
```

With the extension enabled we see that our type constructors are now automatically promoted so that `L`

or `R`

can be viewed as both a data constructor of the type `S`

or as the type `L`

with kind `S`

.

```
{-# LANGUAGE DataKinds #-}
data S a b = L a | R b
-- S :: * -> * -> *
-- L :: * -> S * *
-- R :: * -> S * *
```

Promoted data constructors can referred to in type signatures by prefixing them with a single quote. Also of importance is that these promoted constructors are not exported with a module by default, but type synonym instances can be created for the ticked promoted types and exported directly.

```
data Foo = Bar | Baz
type Bar = 'Bar
type Baz = 'Baz
```

Combining this with type families we see we can write meaningful, meaningful type-level functions by lifting types to the kind level.

```
{-# LANGUAGE TypeFamilies #-}
{-# LANGUAGE DataKinds #-}
import Prelude hiding (Bool(..))
data Bool = True | False
type family Not (a :: Bool) :: Bool
type instance Not True = False
type instance Not False = True
false :: Not True ~ False => a
false = undefined
true :: Not False ~ True => a
true = undefined
-- Fails at compile time.
-- Couldn't match type 'False with 'True
invalid :: Not True ~ True => a
invalid = undefined
```

## Size-Indexed Vectors

Using this new structure we can create a `Vec`

type which is parameterized by its length as well as its element type now that we have a kind language rich enough to encode the successor type in the kind signature of the generalized algebraic datatype.

```
{-# LANGUAGE GADTs #-}
{-# LANGUAGE DataKinds #-}
{-# LANGUAGE KindSignatures #-}
{-# LANGUAGE FlexibleInstances #-}
{-# LANGUAGE FlexibleContexts #-}
data Nat = Z | S Nat deriving (Eq, Show)
type Zero = Z
type One = S Zero
type Two = S One
type Three = S Two
type Four = S Three
type Five = S Four
data Vec :: Nat -> * -> * where
Nil :: Vec Z a
Cons :: a -> Vec n a -> Vec (S n) a
instance Show a => Show (Vec n a) where
show Nil = "Nil"
show (Cons x xs) = "Cons " ++ show x ++ " (" ++ show xs ++ ")"
class FromList n where
fromList :: [a] -> Vec n a
instance FromList Z where
fromList [] = Nil
instance FromList n => FromList (S n) where
fromList (x:xs) = Cons x $ fromList xs
lengthVec :: Vec n a -> Nat
lengthVec Nil = Z
lengthVec (Cons x xs) = S (lengthVec xs)
zipVec :: Vec n a -> Vec n b -> Vec n (a,b)
zipVec Nil Nil = Nil
zipVec (Cons x xs) (Cons y ys) = Cons (x,y) (zipVec xs ys)
vec4 :: Vec Four Int
vec4 = fromList [0, 1, 2, 3]
vec5 :: Vec Five Int
vec5 = fromList [0, 1, 2, 3, 4]
example1 :: Nat
example1 = lengthVec vec4
-- S (S (S (S Z)))
example2 :: Vec Four (Int, Int)
example2 = zipVec vec4 vec4
-- Cons (0,0) (Cons (1,1) (Cons (2,2) (Cons (3,3) (Nil))))
```

So now if we try to zip two `Vec`

types with the wrong shape then we get an error at compile-time about the off-by-one error.

```
example2 = zipVec vec4 vec5
-- Couldn't match type 'S 'Z with 'Z
-- Expected type: Vec Four Int
-- Actual type: Vec Five Int
```

The same technique we can use to create a container which is statically indexed by an empty or non-empty flag, such that if we try to take the head of an empty list we'll get a compile-time error, or stated equivalently we have an obligation to prove to the compiler that the argument we hand to the head function is non-empty.

```
{-# LANGUAGE GADTs #-}
{-# LANGUAGE DataKinds #-}
{-# LANGUAGE KindSignatures #-}
{-# LANGUAGE FlexibleInstances #-}
{-# LANGUAGE FlexibleContexts #-}
data Size = Empty | NonEmpty
data List a b where
Nil :: List Empty a
Cons :: a -> List b a -> List NonEmpty a
head' :: List NonEmpty a -> a
head' (Cons x _) = x
example1 :: Int
example1 = head' (1 `Cons` (2 `Cons` Nil))
-- Cannot match type Empty with NonEmpty
example2 :: Int
example2 = head' Nil
```

```
Couldn't match type None with Many
Expected type: List NonEmpty Int
Actual type: List Empty Int
```

See:

## Typelevel Numbers

GHC's type literals can also be used in place of explicit Peano arithmetic.

GHC 7.6 is very conservative about performing reduction, GHC 7.8 is much less so and will can solve many typelevel constraints involving natural numbers but sometimes still needs a little coaxing.

```
{-# LANGUAGE GADTs #-}
{-# LANGUAGE DataKinds #-}
{-# LANGUAGE KindSignatures #-}
{-# LANGUAGE TypeOperators #-}
import GHC.TypeLits
data Vec :: Nat -> * -> * where
Nil :: Vec 0 a
Cons :: a -> Vec n a -> Vec (1 + n) a
-- GHC 7.6 will not reduce
-- vec3 :: Vec (1 + (1 + (1 + 0))) Int
vec3 :: Vec 3 Int
vec3 = 0 `Cons` (1 `Cons` (2 `Cons` Nil))
```

```
{-# LANGUAGE GADTs #-}
{-# LANGUAGE DataKinds #-}
{-# LANGUAGE KindSignatures #-}
{-# LANGUAGE TypeOperators #-}
{-# LANGUAGE FlexibleContexts #-}
import GHC.TypeLits
import Data.Type.Equality
data Foo :: Nat -> * where
Small :: (n <= 2) => Foo n
Big :: (3 <= n) => Foo n
Empty :: ((n == 0) ~ True) => Foo n
NonEmpty :: ((n == 0) ~ False) => Foo n
big :: Foo 10
big = Big
small :: Foo 2
small = Small
empty :: Foo 0
empty = Empty
nonempty :: Foo 3
nonempty = NonEmpty
```

See: Type-Level Literals

## Typelevel Strings

## Custom Errors

As of GHC 8.0 we have the capacity to provide custom type error using type families. The messages themselves hook into GHC and expressed using the small datatype found in `GHC.TypeLits`

```
data ErrorMessage where
Text :: Symbol -> ErrorMessage
ShowType :: t -> ErrorMessage
-- Put two messages next to each other
(:<>:) :: ErrorMessage -> ErrorMessage -> ErrorMessage
-- Put two messages on top of each other
(:$$:) :: ErrorMessage -> ErrorMessage -> ErrorMessage
```

If one of these expressions is found in the signature of an expression GHC reports an error message of the form:

```
example.hs:1:1: error:
• My custom error message line 1.
• My custom error message line 2.
• In the expression: example
In an equation for ‘foo’: foo = ECoerce (EFloat 3) (EInt 4)
```

```
{-# LANGUAGE DataKinds #-}
{-# LANGUAGE TypeOperators #-}
{-# LANGUAGE UndecidableInstances #-}
import GHC.TypeLits
instance
-- Error Message
TypeError (Text "Equality is not defined for functions"
:$$:
(ShowType a :<>: Text " -> " :<>: ShowType b))
-- Instance head
=> Eq (a -> b) where (==) = undefined
-- Fail when we try to equate two functions
example = id == id
```

A less contrived example would be creating a type-safe embedded DSL that enforces invariants about the semantics at the type-level. We've been able to do this sort of thing using GADTs and type-families for a while but the error reporting has been horrible. With 8.0 we can have type-families that emit useful type errors that reflect what actually goes wrong and integrate this inside of GHC.

```
{-# LANGUAGE GADTs #-}
{-# LANGUAGE DataKinds #-}
{-# LANGUAGE TypeFamilies #-}
{-# LANGUAGE TypeOperators #-}
{-# LANGUAGE UndecidableInstances #-}
import GHC.TypeLits
type family Coerce a b where
Coerce Int Int = Int
Coerce Float Float = Float
Coerce Int Float = Float
Coerce Float Int = TypeError (Text "Cannot cast to smaller type")
data Expr a where
EInt :: Int -> Expr Int
EFloat :: Float -> Expr Float
ECoerce :: Expr b -> Expr c -> Expr (Coerce b c)
foo :: Expr Int
foo = ECoerce (EFloat 3) (EInt 4)
```

## Type Equality

Continuing with the theme of building more elaborate proofs in Haskell, GHC 7.8 recently shipped with the `Data.Type.Equality`

module which provides us with an extended set of type-level operations for expressing the equality of types as values, constraints, and promoted booleans.

```
(~) :: k -> k -> Constraint
(==) :: k -> k -> Bool
(<=) :: Nat -> Nat -> Constraint
(<=?) :: Nat -> Nat -> Bool
(+) :: Nat -> Nat -> Nat
(-) :: Nat -> Nat -> Nat
(*) :: Nat -> Nat -> Nat
(^) :: Nat -> Nat -> Nat
```

```
(:~:) :: k -> k -> *
Refl :: a1 :~: a1
sym :: (a :~: b) -> b :~: a
trans :: (a :~: b) -> (b :~: c) -> a :~: c
castWith :: (a :~: b) -> a -> b
gcastWith :: (a :~: b) -> (a ~ b => r) -> r
```

With this we have a much stronger language for writing restrictions that can be checked at a compile-time, and a mechanism that will later allow us to write more advanced proofs.

```
{-# LANGUAGE GADTs #-}
{-# LANGUAGE DataKinds #-}
{-# LANGUAGE TypeOperators #-}
{-# LANGUAGE ConstraintKinds #-}
import GHC.TypeLits
import Data.Type.Equality
type Not a b = ((b == a) ~ False)
restrictUnit :: Not () a => a -> a
restrictUnit = id
restrictChar :: Not Char a => a -> a
restrictChar = id
```

## Proxies

Using kind polymorphism with phantom types allows us to express the Proxy type which is inhabited by a single constructor with no arguments but with a polykinded phantom type variable which carries an arbitrary type.

```
{-# LANGUAGE PolyKinds #-}
-- | A concrete, poly-kinded proxy type
data Proxy t = Proxy
```

```
import Data.Proxy
a :: Proxy ()
a = Proxy
b :: Proxy 3
b = Proxy
c :: Proxy "symbol"
c = Proxy
d :: Proxy Maybe
d = Proxy
e :: Proxy (Maybe ())
e = Proxy
```

In cases where we'd normally pass around a `undefined`

as a witness of a typeclass dictionary, we can instead pass a Proxy object which carries the phantom type without the need for the bottom. Using scoped type variables we can then operate with the phantom paramater and manipulate wherever is needed.

```
t1 :: a
t1 = (undefined :: a)
t2 :: Proxy a
t2 Proxy :: Proxy a
```

## Promoted Syntax

We've seen constructors promoted using DataKinds, but just like at the value-level GHC also allows us some syntactic sugar for list and tuples instead of explicit cons'ing and pair'ing. This is enabled with the `-XTypeOperators`

extension, which introduces list syntax and tuples of arbitrary arity at the type-level.

```
data HList :: [*] -> * where
HNil :: HList '[]
HCons :: a -> HList t -> HList (a ': t)
data Tuple :: (*,*) -> * where
Tuple :: a -> b -> Tuple '(a,b)
```

Using this we can construct all variety of composite type-level objects.

```
λ: :kind 1
1 :: Nat
λ: :kind "foo"
"foo" :: Symbol
λ: :kind [1,2,3]
[1,2,3] :: [Nat]
λ: :kind [Int, Bool, Char]
[Int, Bool, Char] :: [*]
λ: :kind Just [Int, Bool, Char]
Just [Int, Bool, Char] :: Maybe [*]
λ: :kind '("a", Int)
(,) Symbol *
λ: :kind [ '("a", Int), '("b", Bool) ]
[ '("a", Int), '("b", Bool) ] :: [(,) Symbol *]
```

## Singleton Types

This is an advanced section, knowledge of singletons is not typically necessary to write Haskell.

A singleton type is a type with a single value inhabitant. Singleton types can be constructed in a variety of ways using GADTs or with data families.

```
data instance Sing (a :: Nat) where
SZ :: Sing 'Z
SS :: Sing n -> Sing ('S n)
data instance Sing (a :: Maybe k) where
SNothing :: Sing 'Nothing
SJust :: Sing x -> Sing ('Just x)
data instance Sing (a :: Bool) where
STrue :: Sing True
SFalse :: Sing False
```

**Promoted Naturals**

Value-level | Type-level | Models |
---|---|---|

SZ | Sing 'Z | 0 |

SS SZ | Sing ('S 'Z) | 1 |

SS (SS SZ) | Sing ('S ('S 'Z)) | 2 |

**Promoted Booleans**

Value-level | Type-level | Models |
---|---|---|

STrue | Sing 'False | False |

SFalse | Sing 'True | True |

**Promoted Maybe**

Value-level | Type-level | Models |
---|---|---|

SJust a | Sing (SJust 'a) | Just a |

SNothing | Sing Nothing | Nothing |

Singleton types are an integral part of the small cottage industry of faking dependent types in Haskell, i.e. constructing types with terms predicated upon values. Singleton types are a way of "cheating" by modeling the map between types and values as a structural property of the type.

```
{-# LANGUAGE GADTs #-}
{-# LANGUAGE RankNTypes #-}
{-# LANGUAGE DataKinds #-}
{-# LANGUAGE PolyKinds #-}
{-# LANGUAGE KindSignatures #-}
{-# LANGUAGE TypeFamilies #-}
{-# LANGUAGE TypeOperators #-}
{-# LANGUAGE StandaloneDeriving #-}
{-# LANGUAGE TypeSynonymInstances #-}
{-# LANGUAGE FlexibleContexts #-}
{-# LANGUAGE FlexibleInstances #-}
{-# LANGUAGE UndecidableInstances #-}
import Data.Proxy
import GHC.Exts (Any)
import Prelude hiding (succ)
data Nat = Z | S Nat
-- kind-indexed data family
data family Sing (a :: k)
data instance Sing (a :: Nat) where
SZ :: Sing 'Z
SS :: Sing n -> Sing ('S n)
data instance Sing (a :: Maybe k) where
SNothing :: Sing 'Nothing
SJust :: Sing x -> Sing ('Just x)
data instance Sing (a :: Bool) where
STrue :: Sing True
SFalse :: Sing False
data Fin (n :: Nat) where
FZ :: Fin (S n)
FS :: Fin n -> Fin (S n)
data Vec a n where
Nil :: Vec a Z
Cons :: a -> Vec a n -> Vec a (S n)
class SingI (a :: k) where
sing :: Sing a
instance SingI Z where
sing = SZ
instance SingI n => SingI (S n) where
sing = SS sing
deriving instance Show Nat
deriving instance Show (SNat a)
deriving instance Show (SBool a)
deriving instance Show (Fin a)
deriving instance Show a => Show (Vec a n)
type family (m :: Nat) :+ (n :: Nat) :: Nat where
Z :+ n = n
S m :+ n = S (m :+ n)
type SNat (k :: Nat) = Sing k
type SBool (k :: Bool) = Sing k
type SMaybe (b :: a) (k :: Maybe a) = Sing k
size :: Vec a n -> SNat n
size Nil = SZ
size (Cons x xs) = SS (size xs)
forget :: SNat n -> Nat
forget SZ = Z
forget (SS n) = S (forget n)
natToInt :: Integral n => Nat -> n
natToInt Z = 0
natToInt (S n) = natToInt n + 1
intToNat :: (Integral a, Ord a) => a -> Nat
intToNat 0 = Z
intToNat n = S $ intToNat (n - 1)
sNatToInt :: Num n => SNat x -> n
sNatToInt SZ = 0
sNatToInt (SS n) = sNatToInt n + 1
index :: Fin n -> Vec a n -> a
index FZ (Cons x _) = x
index (FS n) (Cons _ xs) = index n xs
test1 :: Fin (S (S (S Z)))
test1 = FS (FS FZ)
test2 :: Int
test2 = index FZ (1 `Cons` (2 `Cons` Nil))
test3 :: Sing ('Just ('S ('S Z)))
test3 = SJust (SS (SS SZ))
test4 :: Sing ('S ('S Z))
test4 = SS (SS SZ)
-- polymorphic constructor SingI
test5 :: Sing ('S ('S Z))
test5 = sing
```

The builtin singleton types provided in `GHC.TypeLits`

have the useful implementation that type-level values can be reflected to the value-level and back up to the type-level, albeit under an existential.

```
someNatVal :: Integer -> Maybe SomeNat
someSymbolVal :: String -> SomeSymbol
natVal :: KnownNat n => proxy n -> Integer
symbolVal :: KnownSymbol n => proxy n -> String
```

```
{-# LANGUAGE PolyKinds #-}
{-# LANGUAGE DataKinds #-}
{-# LANGUAGE TypeOperators #-}
import Data.Proxy
import GHC.TypeLits
a :: Integer
a = natVal (Proxy :: Proxy 1)
-- 1
b :: String
b = symbolVal (Proxy :: Proxy "foo")
-- "foo"
c :: Integer
c = natVal (Proxy :: Proxy (2 + 3))
-- 5
```

## Closed Type Families

In the type families we've used so far (called open type families) there is no notion of ordering of the equations involved in the type-level function. The type family can be extended at any point in the code resolution simply proceeds sequentially through the available definitions. Closed type-families allow an alternative declaration that allows for a base case for the resolution allowing us to actually write recursive functions over types.

For example consider if we wanted to write a function which counts the arguments in the type of a function and reifies at the value-level.

```
{-# LANGUAGE TypeFamilies #-}
{-# LANGUAGE DataKinds #-}
{-# LANGUAGE TypeOperators #-}
{-# LANGUAGE UndecidableInstances #-}
import Data.Proxy
import GHC.TypeLits
type family Count (f :: *) :: Nat where
Count (a -> b) = 1 + (Count b)
Count x = 1
type Fn1 = Int -> Int
type Fn2 = Int -> Int -> Int -> Int
fn1 :: Integer
fn1 = natVal (Proxy :: Proxy (Count Fn1))
-- 2
fn2 :: Integer
fn2 = natVal (Proxy :: Proxy (Count Fn2))
-- 4
```

The variety of functions we can now write down are rather remarkable, allowing us to write meaningful logic at the type level.

```
{-# LANGUAGE DataKinds #-}
{-# LANGUAGE PolyKinds #-}
{-# LANGUAGE TypeFamilies #-}
{-# LANGUAGE TypeOperators #-}
{-# LANGUAGE ScopedTypeVariables #-}
{-# LANGUAGE UndecidableInstances #-}
import GHC.TypeLits
import Data.Proxy
import Data.Type.Equality
-- Type-level functions over type-level lists.
type family Reverse (xs :: [k]) :: [k] where
Reverse '[] = '[]
Reverse xs = Rev xs '[]
type family Rev (xs :: [k]) (ys :: [k]) :: [k] where
Rev '[] i = i
Rev (x ': xs) i = Rev xs (x ': i)
type family Length (as :: [k]) :: Nat where
Length '[] = 0
Length (x ': xs) = 1 + Length xs
type family If (p :: Bool) (a :: k) (b :: k) :: k where
If True a b = a
If False a b = b
type family Concat (as :: [k]) (bs :: [k]) :: [k] where
Concat a '[] = a
Concat '[] b = b
Concat (a ': as) bs = a ': Concat as bs
type family Map (f :: a -> b) (as :: [a]) :: [b] where
Map f '[] = '[]
Map f (x ': xs) = f x ': Map f xs
type family Sum (xs :: [Nat]) :: Nat where
Sum '[] = 0
Sum (x ': xs) = x + Sum xs
ex1 :: Reverse [1,2,3] ~ [3,2,1] => Proxy a
ex1 = Proxy
ex2 :: Length [1,2,3] ~ 3 => Proxy a
ex2 = Proxy
ex3 :: (Length [1,2,3]) ~ (Length (Reverse [1,2,3])) => Proxy a
ex3 = Proxy
-- Reflecting type level computations back to the value level.
ex4 :: Integer
ex4 = natVal (Proxy :: Proxy (Length (Concat [1,2,3] [4,5,6])))
-- 6
ex5 :: Integer
ex5 = natVal (Proxy :: Proxy (Sum [1,2,3]))
-- 6
-- Couldn't match type ‘2’ with ‘1’
ex6 :: Reverse [1,2,3] ~ [3,1,2] => Proxy a
ex6 = Proxy
```

The results of type family functions need not necessarily be kinded as `(*)`

either. For example using Nat or Constraint is permitted.

```
type family Elem (a :: k) (bs :: [k]) :: Constraint where
Elem a (a ': bs) = (() :: Constraint)
Elem a (b ': bs) = a `Elem` bs
type family Sum (ns :: [Nat]) :: Nat where
Sum '[] = 0
Sum (n ': ns) = n + Sum ns
```

## Kind Indexed Type Families

This is an advanced section, and is not typically necessary to write Haskell.

Just as typeclasses are normally indexed on types, type families can also be indexed on kinds with the kinds given as explicit kind signatures on type variables.

```
type family (a :: k) == (b :: k) :: Bool
type instance a == b = EqStar a b
type instance a == b = EqArrow a b
type instance a == b = EqBool a b
type family EqStar (a :: *) (b :: *) where
EqStar a a = True
EqStar a b = False
type family EqArrow (a :: k1 -> k2) (b :: k1 -> k2) where
EqArrow a a = True
EqArrow a b = False
type family EqBool a b where
EqBool True True = True
EqBool False False = True
EqBool a b = False
type family EqList a b where
EqList '[] '[] = True
EqList (h1 ': t1) (h2 ': t2) = (h1 == h2) && (t1 == t2)
EqList a b = False
type family a && b where
True && True = True
a && a = False
```

## Promoted Symbols

```
{-# LANGUAGE DataKinds #-}
{-# LANGUAGE PolyKinds #-}
{-# LANGUAGE FlexibleInstances #-}
{-# LANGUAGE FlexibleContexts #-}
{-# LANGUAGE FunctionalDependencies #-}
{-# LANGUAGE TypeOperators #-}
{-# LANGUAGE ConstraintKinds #-}
import GHC.TypeLits
import Data.Type.Equality
data Label (l :: Symbol) = Get
class Has a l b | a l -> b where
from :: a -> Label l -> b
data Point2D = Point2 Double Double deriving Show
data Point3D = Point3 Double Double Double deriving Show
instance Has Point2D "x" Double where
from (Point2 x _) _ = x
instance Has Point2D "y" Double where
from (Point2 _ y) _ = y
instance Has Point3D "x" Double where
from (Point3 x _ _) _ = x
instance Has Point3D "y" Double where
from (Point3 _ y _) _ = y
instance Has Point3D "z" Double where
from (Point3 _ _ z) _ = z
infixl 6 #
(#) :: a -> (a -> b) -> b
(#) = flip ($)
_x :: Has a "x" b => a -> b
_x pnt = from pnt (Get :: Label "x")
_y :: Has a "y" b => a -> b
_y pnt = from pnt (Get :: Label "y")
_z :: Has a "z" b => a -> b
_z pnt = from pnt (Get :: Label "z")
type Point a r = (Has a "x" r, Has a "y" r)
distance :: (Point a r, Point b r, Floating r) => a -> b -> r
distance p1 p2 = sqrt (d1^2 + d2^2)
where
d1 = (p1 # _x) + (p1 # _y)
d2 = (p2 # _x) + (p2 # _y)
main :: IO ()
main = do
print $ (Point2 10 20) # _x
-- Fails with: No instance for (Has Point2D "z" a0)
-- print $ (Point2 10 20) # _z
print $ (Point3 10 20 30) # _x
print $ (Point3 10 20 30) # _z
print $ distance (Point2 1 3) (Point2 2 7)
print $ distance (Point2 1 3) (Point3 2 7 4)
print $ distance (Point3 1 3 5) (Point3 2 7 3)
```

Since record is fundamentally no different from the tuple we can also do the same kind of construction over record field names.

```
{-# LANGUAGE DataKinds #-}
{-# LANGUAGE KindSignatures #-}
{-# LANGUAGE MultiParamTypeClasses #-}
{-# LANGUAGE FunctionalDependencies #-}
{-# LANGUAGE FlexibleInstances #-}
{-# LANGUAGE FlexibleContexts #-}
{-# LANGUAGE StandaloneDeriving #-}
{-# LANGUAGE ExistentialQuantification #-}
{-# LANGUAGE ConstraintKinds #-}
import GHC.TypeLits
newtype Field (n :: Symbol) v = Field { unField :: v }
deriving Show
data Person1 = Person1
{ _age :: Field "age" Int
, _name :: Field "name" String
}
data Person2 = Person2
{ _age' :: Field "age" Int
, _name' :: Field "name" String
, _lib' :: Field "lib" String
}
deriving instance Show Person1
deriving instance Show Person2
data Label (l :: Symbol) = Get
class Has a l b | a l -> b where
from :: a -> Label l -> b
instance Has Person1 "age" Int where
from (Person1 a _) _ = unField a
instance Has Person1 "name" String where
from (Person1 _ a) _ = unField a
instance Has Person2 "age" Int where
from (Person2 a _ _) _ = unField a
instance Has Person2 "name" String where
from (Person2 _ a _) _ = unField a
age :: Has a "age" b => a -> b
age pnt = from pnt (Get :: Label "age")
name :: Has a "name" b => a -> b
name pnt = from pnt (Get :: Label "name")
-- Parameterized constraint kind for "Simon-ness" of a record.
type Simon a = (Has a "name" String, Has a "age" Int)
spj :: Person1
spj = Person1 (Field 56) (Field "Simon Peyton Jones")
smarlow :: Person2
smarlow = Person2 (Field 38) (Field "Simon Marlow") (Field "rts")
catNames :: (Simon a, Simon b) => a -> b -> String
catNames a b = name a ++ name b
addAges :: (Simon a, Simon b) => a -> b -> Int
addAges a b = age a + age b
names :: String
names = name smarlow ++ "," ++ name spj
-- "Simon Marlow,Simon Peyton Jones"
ages :: Int
ages = age spj + age smarlow
-- 94
```

Notably this approach is mostly just all boilerplate class instantiation which could be abstracted away using TemplateHaskell or a Generic deriving.

## HLists

This is an advanced section, and is not typically necessary to write Haskell.

A heterogeneous list is a cons list whose type statically encodes the ordered types of its values.

```
{-# LANGUAGE DataKinds #-}
{-# LANGUAGE GADTs #-}
{-# LANGUAGE TypeOperators #-}
{-# LANGUAGE KindSignatures #-}
infixr 5 :::
data HList (ts :: [ * ]) where
Nil :: HList '[]
(:::) :: t -> HList ts -> HList (t ': ts)
-- Take the head of a non-empty list with the first value as Bool type.
headBool :: HList (Bool ': xs) -> Bool
headBool hlist = case hlist of
(a ::: _) -> a
hlength :: HList x -> Int
hlength Nil = 0
hlength (_ ::: b) = 1 + (hlength b)
tuple :: (Bool, (String, (Double, ())))
tuple = (True, ("foo", (3.14, ())))
hlist :: HList '[Bool, String , Double , ()]
hlist = True ::: "foo" ::: 3.14 ::: () ::: Nil
```

Of course this immediately begs the question of how to print such a list out to a string in the presence of type-heterogeneity. In this case we can use type-families combined with constraint kinds to apply the Show over the HLists parameters to generate the aggregate constraint that all types in the HList are Showable, and then derive the Show instance.

```
{-# LANGUAGE GADTs #-}
{-# LANGUAGE TypeFamilies #-}
{-# LANGUAGE TypeOperators #-}
{-# LANGUAGE DataKinds #-}
{-# LANGUAGE PolyKinds #-}
{-# LANGUAGE KindSignatures #-}
{-# LANGUAGE ConstraintKinds #-}
{-# LANGUAGE UndecidableInstances #-}
import GHC.Exts (Constraint)
infixr 5 :::
data HList (ts :: [ * ]) where
Nil :: HList '[]
(:::) :: t -> HList ts -> HList (t ': ts)
type family Map (f :: a -> b) (xs :: [a]) :: [b]
type instance Map f '[] = '[]
type instance Map f (x ': xs) = f x ': Map f xs
type family Constraints (cs :: [Constraint]) :: Constraint
type instance Constraints '[] = ()
type instance Constraints (c ': cs) = (c, Constraints cs)
type AllHave (c :: k -> Constraint) (xs :: [k]) = Constraints (Map c xs)
showHList :: AllHave Show xs => HList xs -> [String]
showHList Nil = []
showHList (x ::: xs) = (show x) : showHList xs
instance AllHave Show xs => Show (HList xs) where
show = show . showHList
example1 :: HList '[Bool, String , Double , ()]
example1 = True ::: "foo" ::: 3.14 ::: () ::: Nil
-- ["True","\"foo\"","3.14","()"]
```

## Typelevel Dictionaries

Much of this discussion of promotion begs the question whether we can create data structures at the type-level to store information at compile-time. For example a type-level association list can be used to model a map between type-level symbols and any other promotable types. Together with type-families we can write down type-level traversal and lookup functions.

```
{-# LANGUAGE GADTs #-}
{-# LANGUAGE DataKinds #-}
{-# LANGUAGE PolyKinds #-}
{-# LANGUAGE RankNTypes #-}
{-# LANGUAGE TypeOperators #-}
{-# LANGUAGE TypeFamilies #-}
{-# LANGUAGE KindSignatures #-}
{-# LANGUAGE ConstraintKinds #-}
{-# LANGUAGE UndecidableInstances #-}
import GHC.TypeLits
import Data.Proxy
import Data.Type.Equality
type family If (p :: Bool) (a :: k) (b :: k) :: k where
If True a b = a
If False a b = b
type family Lookup (k :: a) (ls :: [(a, b)]) :: Maybe b where
Lookup k '[] = 'Nothing
Lookup k ('(a, b) ': xs) = If (a == k) ('Just b) (Lookup k xs)
type M = [
'("a", 1)
, '("b", 2)
, '("c", 3)
, '("d", 4)
]
type K = "a"
type (!!) m (k :: Symbol) a = (Lookup k m) ~ Just a
value :: Integer
value = natVal ( Proxy :: (M !! "a") a => Proxy a )
```

If we ask GHC to expand out the type signature we can view the explicit implementation of the type-level map lookup function.

```
(!!)
:: If
(GHC.TypeLits.EqSymbol "a" k)
('Just 1)
(If
(GHC.TypeLits.EqSymbol "b" k)
('Just 2)
(If
(GHC.TypeLits.EqSymbol "c" k)
('Just 3)
(If (GHC.TypeLits.EqSymbol "d" k) ('Just 4) 'Nothing)))
~ 'Just v =>
Proxy k -> Proxy v
```

## Advanced Proofs

This is an advanced section, and is not typically necessary to write Haskell.

Now that we have the length-indexed vector let's go write the reverse function, how hard could it be?

So we go and write down something like this:

```
reverseNaive :: forall n a. Vec a n -> Vec a n
reverseNaive xs = go Nil xs -- Error: n + 0 != n
where
go :: Vec a m -> Vec a n -> Vec a (n :+ m)
go acc Nil = acc
go acc (Cons x xs) = go (Cons x acc) xs -- Error: n + succ m != succ (n + m)
```

Running this we find that GHC is unhappy about two lines in the code:

```
Couldn't match type ‘n’ with ‘n :+ 'Z’
Expected type: Vec a n
Actual type: Vec a (n :+ 'Z)
Could not deduce ((n1 :+ 'S m) ~ 'S (n1 :+ m))
Expected type: Vec a1 (k :+ m)
Actual type: Vec a1 (n1 :+ 'S m)
```

As we unfold elements out of the vector we'll end up doing a lot of type-level arithmetic over indices as we combine the subparts of the vector backwards, but as a consequence we find that GHC will run into some unification errors because it doesn't know about basic arithmetic properties of the natural numbers. Namely that `forall n. n + 0 = 0`

and `forall n m. n + (1 + m) = 1 + (n + m)`

. Which of course it really shouldn't be given that we've constructed a system at the type-level which intuitively *models* arithmetic but GHC is just a dumb compiler, it can't automatically deduce the isomorphism between natural numbers and Peano numbers.

So at each of these call sites we now have a proof obligation to construct proof terms. Recall from our discussion of propositional equality from GADTs that we actually have such machinery to construct this now.

```
{-# LANGUAGE GADTs #-}
{-# LANGUAGE PolyKinds #-}
{-# LANGUAGE DataKinds #-}
{-# LANGUAGE TypeFamilies #-}
{-# LANGUAGE TypeOperators #-}
{-# LANGUAGE KindSignatures #-}
{-# LANGUAGE ExplicitForAll #-}
import Data.Type.Equality
data Nat = Z | S Nat
data SNat n where
Zero :: SNat Z
Succ :: SNat n -> SNat (S n)
data Vec :: * -> Nat -> * where
Nil :: Vec a Z
Cons :: a -> Vec a n -> Vec a (S n)
instance Show a => Show (Vec a n) where
show Nil = "Nil"
show (Cons x xs) = "Cons " ++ show x ++ " (" ++ show xs ++ ")"
type family (m :: Nat) :+ (n :: Nat) :: Nat where
Z :+ n = n
S m :+ n = S (m :+ n)
-- (a ~ b) implies (f a ~ f b)
cong :: a :~: b -> f a :~: f b
cong Refl = Refl
-- (a ~ b) implies (f a) implies (f b)
subst :: a :~: b -> f a -> f b
subst Refl = id
plus_zero :: forall n. SNat n -> (n :+ Z) :~: n
plus_zero Zero = Refl
plus_zero (Succ n) = cong (plus_zero n)
plus_suc :: forall n m. SNat n -> SNat m -> (n :+ (S m)) :~: (S (n :+ m))
plus_suc Zero m = Refl
plus_suc (Succ n) m = cong (plus_suc n m)
size :: Vec a n -> SNat n
size Nil = Zero
size (Cons _ xs) = Succ $ size xs
reverse :: forall n a. Vec a n -> Vec a n
reverse xs = subst (plus_zero (size xs)) $ go Nil xs
where
go :: Vec a m -> Vec a k -> Vec a (k :+ m)
go acc Nil = acc
go acc (Cons x xs) = subst (plus_suc (size xs) (size acc)) $ go (Cons x acc) xs
append :: Vec a n -> Vec a m -> Vec a (n :+ m)
append (Cons x xs) ys = Cons x (append xs ys)
append Nil ys = ys
vec :: Vec Int (S (S (S Z)))
vec = 1 `Cons` (2 `Cons` (3 `Cons` Nil))
test :: Vec Int (S (S (S Z)))
test = Main.reverse vec
```

One might consider whether we could avoid using the singleton trick and just use type-level natural numbers, and technically this approach should be feasible although it seems that the natural number solver in GHC 7.8 can decide some properties but not the ones needed to complete the natural number proofs for the reverse functions.

```
{-# LANGUAGE DataKinds #-}
{-# LANGUAGE ExplicitForAll #-}
{-# LANGUAGE TypeFamilies #-}
{-# LANGUAGE TypeOperators #-}
{-# LANGUAGE UndecidableInstances #-}
import Prelude hiding (Eq)
import GHC.TypeLits
import Data.Type.Equality
type Z = 0
type family S (n :: Nat) :: Nat where
S n = n + 1
-- Yes!
eq_zero :: Z :~: Z
eq_zero = Refl
-- Yes!
zero_plus_one :: (Z + 1) :~: (1 + Z)
zero_plus_one = Refl
-- Yes!
plus_zero :: forall n. (n + Z) :~: n
plus_zero = Refl
-- Yes!
plus_one :: forall n. (n + S Z) :~: S n
plus_one = Refl
-- No.
plus_suc :: forall n m. (n + (S m)) :~: (S (n + m))
plus_suc = Refl
```

Caveat should be that there might be a way to do this in GHC 7.6 that I'm not aware of. In GHC 7.10 there are some planned changes to solver that should be able to resolve these issues. In particular there are plans to allow pluggable type system extensions that could outsource these kind of problems to third party SMT solvers which can solve these kind of numeric relations and return this information back to GHC's typechecker.

As an aside this is a direct transliteration of the equivalent proof in Agda, which is accomplished via the same method but without the song and dance to get around the lack of dependent types.

```
module Vector where
infixr 10 _∷_
data ℕ : Set where
zero : ℕ
suc : ℕ → ℕ
{-# BUILTIN NATURAL ℕ #-}
{-# BUILTIN ZERO zero #-}
{-# BUILTIN SUC suc #-}
infixl 6 _+_
_+_ : ℕ → ℕ → ℕ
0 + n = n
suc m + n = suc (m + n)
data Vec (A : Set) : ℕ → Set where
[] : Vec A 0
_∷_ : ∀ {n} → A → Vec A n → Vec A (suc n)
_++_ : ∀ {A n m} → Vec A n → Vec A m → Vec A (n + m)
[] ++ ys = ys
(x ∷ xs) ++ ys = x ∷ (xs ++ ys)
infix 4 _≡_
data _≡_ {A : Set} (x : A) : A → Set where
refl : x ≡ x
subst : {A : Set} → (P : A → Set) → ∀{x y} → x ≡ y → P x → P y
subst P refl p = p
cong : {A B : Set} (f : A → B) → {x y : A} → x ≡ y → f x ≡ f y
cong f refl = refl
vec : ∀ {A} (k : ℕ) → Set
vec {A} k = Vec A k
plus_zero : {n : ℕ} → n + 0 ≡ n
plus_zero {zero} = refl
plus_zero {suc n} = cong suc plus_zero
plus_suc : {n : ℕ} → n + (suc 0) ≡ suc n
plus_suc {zero} = refl
plus_suc {suc n} = cong suc (plus_suc {n})
reverse : ∀ {A n} → Vec A n → Vec A n
reverse [] = []
reverse {A} {suc n} (x ∷ xs) = subst vec (plus_suc {n}) (reverse xs ++ (x ∷ []))
```

## Liquid Haskell

This is an advanced section, knowledge of LiquidHaskell is not typically necessary to write Haskell.

LiquidHaskell is an extension to GHC's typesystem that adds the capactity for refinement types using the annotation syntax. The type signatures of functions can be checked by the external for richer type semantics than default GHC provides, including non-exhaustive patterns and complex arithemtic properties that require external SMT solvers to verify. For instance LiquidHaskell can statically verify that a function that operates over a `Maybe a`

is always given a `Just`

or that an arithmetic functions always yields an Int that is even positive number.

To Install LiquidHaskell in Ubuntu add the following line to your `/etc/sources.list`

:

`deb http://ppa.launchpad.net/hvr/z3/ubuntu trusty main`

And then install the external SMT solver.

```
$ sudo apt-key adv --keyserver keyserver.ubuntu.com --recv-keys F6F88286
$ sudo apt-get install z3
```

Then clone the repo and build it using stack.

```
$ git clone --recursive git@github.com:ucsd-progsys/liquidhaskell.git
$ cd liquidhaskell
$ stack install
```

Ensure that `$HOME/.local/bin`

is on your `$PATH`

.

```
import Prelude hiding (mod, gcd)
{-@ mod :: a:Nat -> b:{v:Nat| 0 < v} -> {v:Nat | v < b} @-}
mod :: Int -> Int -> Int
mod a b
| a < b = a
| otherwise = mod (a - b) b
{-@ gcd :: a:Nat -> b:{v:Nat | v < a} -> Int @-}
gcd :: Int -> Int -> Int
gcd a 0 = a
gcd a b = gcd b (a `mod` b)
```

The module can be run through the solver using the `liquid`

command line tool.

```
$ liquid example.hs
Done solving.
**** DONE: solve **************************************************************
**** DONE: annotate ***********************************************************
**** RESULT: SAFE **************************************************************
```

For more extensive documentation and further use cases see the official documentation:

# Generics

Haskell has several techniques for automatic generation of type classes for a variety of tasks that consist largely of boilerplate code generation such as:

- Pretty Printing
- Equality
- Serialization
- Ordering
- Traversal

These are achieved through several tools and techniques outlined in the next few sections:

- Typeable / Dynamic
- Scrap Your Boilerplate
- GHC.Generics
- generics-sop

## Typeable

The `Typeable`

class be used to create runtime type information for arbitrary types.

`typeOf :: Typeable a => a -> TypeRep`

```
{-# LANGUAGE DeriveDataTypeable #-}
import Data.Typeable
data Animal = Cat | Dog deriving Typeable
data Zoo a = Zoo [a] deriving Typeable
equal :: (Typeable a, Typeable b) => a -> b -> Bool
equal a b = typeOf a == typeOf b
example1 :: TypeRep
example1 = typeOf Cat
-- Animal
example2 :: TypeRep
example2 = typeOf (Zoo [Cat, Dog])
-- Zoo Animal
example3 :: TypeRep
example3 = typeOf ((1, 6.636e-34, "foo") :: (Int, Double, String))
-- (Int,Double,[Char])
example4 :: Bool
example4 = equal False ()
-- False
```

Using the Typeable instance allows us to write down a type safe cast function which can safely use `unsafeCast`

and provide a proof that the resulting type matches the input.

```
cast :: (Typeable a, Typeable b) => a -> Maybe b
cast x
| typeOf x == typeOf ret = Just ret
| otherwise = Nothing
where
ret = unsafeCast x
```

Of historical note is that writing our own Typeable classes is currently possible of GHC 7.6 but allows us to introduce dangerous behavior that can cause crashes, and shouldn't be done except by GHC itself. As of 7.8 GHC forbids hand-written Typeable instances. As of 7.10 `-XAutoDeriveDataTypeable`

is enabled by default.

See: Typeable and Data in Haskell

## Dynamic

Since we have a way of querying runtime type information we can use this machinery to implement a `Dynamic`

type. This allows us to box up any monotype into a uniform type that can be passed to any function taking a Dynamic type which can then unpack the underlying value in a type-safe way.

```
toDyn :: Typeable a => a -> Dynamic
fromDyn :: Typeable a => Dynamic -> a -> a
fromDynamic :: Typeable a => Dynamic -> Maybe a
cast :: (Typeable a, Typeable b) => a -> Maybe b
```

```
import Data.Dynamic
import Data.Maybe
dynamicBox :: Dynamic
dynamicBox = toDyn (6.62 :: Double)
example1 :: Maybe Int
example1 = fromDynamic dynamicBox
-- Nothing
example2 :: Maybe Double
example2 = fromDynamic dynamicBox
-- Just 6.62
example3 :: Int
example3 = fromDyn dynamicBox 0
-- 0
example4 :: Double
example4 = fromDyn dynamicBox 0.0
-- 6.62
```

In GHC 7.8 the Typeable class is poly-kinded so polymorphic functions can be applied over functions and higher kinded types.

Use of Dynamic is somewhat rare, except in odd cases that have to deal with foreign memory and FFI interfaces. Using it for business logic is considered a code smell. Consider a more idiomatic solution.

## Data

Just as Typeable lets us create runtime type information, the Data class allows us to reflect information about the structure of datatypes to runtime as needed.

```
class Typeable a => Data a where
gfoldl :: (forall d b. Data d => c (d -> b) -> d -> c b)
-> (forall g. g -> c g)
-> a
-> c a
gunfold :: (forall b r. Data b => c (b -> r) -> c r)
-> (forall r. r -> c r)
-> Constr
-> c a
toConstr :: a -> Constr
dataTypeOf :: a -> DataType
gmapQl :: (r -> r' -> r) -> r -> (forall d. Data d => d -> r') -> a -> r
```

The types for `gfoldl`

and `gunfold`

are a little intimidating ( and depend on `RankNTypes`

), the best way to understand is to look at some examples. First the most trivial case a simple sum type `Animal`

would produce the following code:

`data Animal = Cat | Dog deriving Typeable`

```
instance Data Animal where
gfoldl k z Cat = z Cat
gfoldl k z Dog = z Dog
gunfold k z c
= case constrIndex c of
1 -> z Cat
2 -> z Dog
toConstr Cat = cCat
toConstr Dog = cDog
dataTypeOf _ = tAnimal
tAnimal :: DataType
tAnimal = mkDataType "Main.Animal" [cCat, cDog]
cCat :: Constr
cCat = mkConstr tAnimal "Cat" [] Prefix
cDog :: Constr
cDog = mkConstr tAnimal "Dog" [] Prefix
```

For a type with non-empty containers we get something a little more interesting. Consider the list type:

```
instance Data a => Data [a] where
gfoldl _ z [] = z []
gfoldl k z (x:xs) = z (:) `k` x `k` xs
toConstr [] = nilConstr
toConstr (_:_) = consConstr
gunfold k z c
= case constrIndex c of
1 -> z []
2 -> k (k (z (:)))
dataTypeOf _ = listDataType
nilConstr :: Constr
nilConstr = mkConstr listDataType "[]" [] Prefix
consConstr :: Constr
consConstr = mkConstr listDataType "(:)" [] Infix
listDataType :: DataType
listDataType = mkDataType "Prelude.[]" [nilConstr,consConstr]
```

Looking at `gfoldl`

we see the Data has an implementation of a function for us to walk an applicative over the elements of the constructor by applying a function `k`

over each element and applying `z`

at the spine. For example look at the instance for a 2-tuple as well:

```
instance (Data a, Data b) => Data (a,b) where
gfoldl k z (a,b) = z (,) `k` a `k` b
toConstr (_,_) = tuple2Constr
gunfold k z c
= case constrIndex c of
1 -> k (k (z (,)))
dataTypeOf _ = tuple2DataType
tuple2Constr :: Constr
tuple2Constr = mkConstr tuple2DataType "(,)" [] Infix
tuple2DataType :: DataType
tuple2DataType = mkDataType "Prelude.(,)" [tuple2Constr]
```

This is pretty neat, now within the same typeclass we have a generic way to introspect any `Data`

instance and write logic that depends on the structure and types of its subterms. We can now write a function which allows us to traverse an arbitrary instance of Data and twiddle values based on pattern matching on the runtime types. So let's write down a function `over`

which increments a `Value`

type for both for n-tuples and lists.

```
{-# LANGUAGE DeriveDataTypeable #-}
import Data.Data
import Control.Monad.Identity
import Control.Applicative
data Animal = Cat | Dog deriving (Data, Typeable)
newtype Val = Val Int deriving (Show, Data, Typeable)
incr :: Typeable a => a -> a
incr = maybe id id (cast f)
where f (Val x) = Val (x * 100)
over :: Data a => a -> a
over x = runIdentity $ gfoldl cont base (incr x)
where
cont k d = k <*> (pure $ over d)
base = pure
example1 :: Constr
example1 = toConstr Dog
-- Dog
example2 :: DataType
example2 = dataTypeOf Cat
-- DataType {tycon = "Main.Animal", datarep = AlgRep [Cat,Dog]}
example3 :: [Val]
example3 = over [Val 1, Val 2, Val 3]
-- [Val 100,Val 200,Val 300]
example4 :: (Val, Val, Val)
example4 = over (Val 1, Val 2, Val 3)
-- (Val 100,Val 200,Val 300)
```

We can also write generic operations, for example to count the number of parameters in a data type.

```
numHoles :: Data a => a -> Int
numHoles = gmapQl (+) 0 (const 1)
example1 :: Int
example1 = numHoles (1,2,3,4,5,6,7)
-- 7
example2 :: Int
example2 = numHoles (Just 3)
-- 1
```

## Syb

Using the interface provided by the Data we can retrieve the information we need to, at runtime, inspect the types of expressions and rewrite them, collect terms, and find subterms matching specific predicates.

```
everywhere :: (forall a. Data a => a -> a) -> forall a. Data a => a -> a
everywhereM :: Monad m => GenericM m -> GenericM m
somewhere :: MonadPlus m => GenericM m -> GenericM m
listify :: Typeable r => (r -> Bool) -> GenericQ [r]
everything :: (r -> r -> r) -> GenericQ r -> GenericQ r
```

For example consider we have some custom collection of datatypes for which we want to write generic transformations that transform numerical subexpressions according to set of rewrite rules. We can use `syb`

to write the transformation rules quite succinctly.

```
{-# LANGUAGE DeriveDataTypeable #-}
import Data.Data
import Data.Typeable
import Data.Generics.Schemes
import Data.Generics.Aliases (mkT)
data MyTuple a = MyTuple a Float
deriving (Data, Typeable, Show)
exampleT :: Data a => MyTuple a -> MyTuple a
exampleT = everywhere (mkT go1) . everywhere (mkT go2)
where
go1 :: Int -> Int
go1 x = succ x
go2 :: Float -> Float
go2 x = succ x
findFloat :: Data x => x -> Maybe Float
findFloat = gfindtype
main :: IO ()
main = do
let term = MyTuple (MyTuple (1 :: Int) 2.0) 3.0
print (exampleT term)
print (gsize term)
print (findFloat term)
print (listify ((>0) :: (Int -> Bool)) term)
```

## Generic

The most modern method of doing generic programming uses type families to achieve a better method of deriving the structural properties of arbitrary type classes. Generic implements a typeclass with an associated type `Rep`

( Representation ) together with a pair of functions that form a 2-sided inverse ( isomorphism ) for converting to and from the associated type and the derived type in question.

```
class Generic a where
type Rep a
from :: a -> Rep a
to :: Rep a -> a
class Datatype d where
datatypeName :: t d f a -> String
moduleName :: t d f a -> String
class Constructor c where
conName :: t c f a -> String
```

GHC.Generics defines a set of named types for modeling the various structural properties of types in available in Haskell.

```
-- | Sums: encode choice between constructors
infixr 5 :+:
data (:+:) f g p = L1 (f p) | R1 (g p)
-- | Products: encode multiple arguments to constructors
infixr 6 :*:
data (:*:) f g p = f p :*: g p
-- | Tag for M1: datatype
data D
-- | Tag for M1: constructor
data C
-- | Constants, additional parameters and recursion of kind *
newtype K1 i c p = K1 { unK1 :: c }
-- | Meta-information (constructor names, etc.)
newtype M1 i c f p = M1 { unM1 :: f p }
-- | Type synonym for encoding meta-information for datatypes
type D1 = M1 D
-- | Type synonym for encoding meta-information for constructors
type C1 = M1 C
```

Using the deriving mechanics GHC can generate this Generic instance for us mechanically, if we were to write it by hand for a simple type it might look like this:

```
{-# LANGUAGE TypeOperators #-}
{-# LANGUAGE TypeFamilies #-}
import GHC.Generics
data Animal
= Dog
| Cat
instance Generic Animal where
type Rep Animal = D1 T_Animal ((C1 C_Dog U1) :+: (C1 C_Cat U1))
from Dog = M1 (L1 (M1 U1))
from Cat = M1 (R1 (M1 U1))
to (M1 (L1 (M1 U1))) = Dog
to (M1 (R1 (M1 U1))) = Cat
data T_Animal
data C_Dog
data C_Cat
instance Datatype T_Animal where
datatypeName _ = "Animal"
moduleName _ = "Main"
instance Constructor C_Dog where
conName _ = "Dog"
instance Constructor C_Cat where
conName _ = "Cat"
```

Use `kind!`

in GHCi we can look at the type family `Rep`

associated with a Generic instance.

```
λ: :kind! Rep Animal
Rep Animal :: * -> *
= M1 D T_Animal (M1 C C_Dog U1 :+: M1 C C_Cat U1)
λ: :kind! Rep ()
Rep () :: * -> *
= M1 D GHC.Generics.D1() (M1 C GHC.Generics.C1_0() U1)
λ: :kind! Rep [()]
Rep [()] :: * -> *
= M1
D
GHC.Generics.D1[]
(M1 C GHC.Generics.C1_0[] U1
:+: M1
C
GHC.Generics.C1_1[]
(M1 S NoSelector (K1 R ()) :*: M1 S NoSelector (K1 R [()])))
```

Now the clever bit, instead writing our generic function over the datatype we instead write it over the Rep and then reify the result using `from`

. So for an equivalent version of Haskell's default `Eq`

that instead uses generic deriving we could write:

```
class GEq' f where
geq' :: f a -> f a -> Bool
instance GEq' U1 where
geq' _ _ = True
instance (GEq c) => GEq' (K1 i c) where
geq' (K1 a) (K1 b) = geq a b
instance (GEq' a) => GEq' (M1 i c a) where
geq' (M1 a) (M1 b) = geq' a b
-- Equality for sums.
instance (GEq' a, GEq' b) => GEq' (a :+: b) where
geq' (L1 a) (L1 b) = geq' a b
geq' (R1 a) (R1 b) = geq' a b
geq' _ _ = False
-- Equality for products.
instance (GEq' a, GEq' b) => GEq' (a :*: b) where
geq' (a1 :*: b1) (a2 :*: b2) = geq' a1 a2 && geq' b1 b2
```

To accommodate the two methods of writing classes (generic-deriving or custom implementations) we can use the `DefaultSignatures`

extension to allow the user to leave typeclass functions blank and defer to Generic or to define their own.

```
{-# LANGUAGE DefaultSignatures #-}
class GEq a where
geq :: a -> a -> Bool
default geq :: (Generic a, GEq' (Rep a)) => a -> a -> Bool
geq x y = geq' (from x) (from y)
```

Now anyone using our library need only derive Generic and create an empty instance of our typeclass instance without writing any boilerplate for `GEq`

.

Here is a complete example for deriving equality generics:

```
{-# LANGUAGE TypeOperators #-}
{-# LANGUAGE FlexibleContexts #-}
{-# LANGUAGE DefaultSignatures #-}
import GHC.Generics
-- Auxiliary class
class GEq' f where
geq' :: f a -> f a -> Bool
instance GEq' U1 where
geq' _ _ = True
instance (GEq c) => GEq' (K1 i c) where
geq' (K1 a) (K1 b) = geq a b
instance (GEq' a) => GEq' (M1 i c a) where
geq' (M1 a) (M1 b) = geq' a b
instance (GEq' a, GEq' b) => GEq' (a :+: b) where
geq' (L1 a) (L1 b) = geq' a b
geq' (R1 a) (R1 b) = geq' a b
geq' _ _ = False
instance (GEq' a, GEq' b) => GEq' (a :*: b) where
geq' (a1 :*: b1) (a2 :*: b2) = geq' a1 a2 && geq' b1 b2
--
class GEq a where
geq :: a -> a -> Bool
default geq :: (Generic a, GEq' (Rep a)) => a -> a -> Bool
geq x y = geq' (from x) (from y)
-- Base equalities
instance GEq Char where geq = (==)
instance GEq Int where geq = (==)
instance GEq Float where geq = (==)
-- Equalities derived from structure of (:+:) and (:*:)
instance GEq a => GEq (Maybe a)
instance (GEq a, GEq b) => GEq (a,b)
main :: IO ()
main = do
print $ geq 2 (3 :: Int)
print $ geq 'a' 'b'
print $ geq (Just 'a') (Just 'a')
print $ geq ('a','b') ('a', 'b')
```

See:

- Cooking Classes with Datatype Generic Programming
- Datatype-generic Programming in Haskell
- generic-deriving

## Generic Deriving

Using Generics many common libraries provide a mechanisms to derive common typeclass instances. Some real world examples:

The hashable library allows us to derive hashing functions.

```
{-# LANGUAGE DeriveGeneric #-}
import GHC.Generics (Generic)
import Data.Hashable
data Color = Red | Green | Blue deriving (Generic, Show)
instance Hashable Color where
example1 :: Int
example1 = hash Red
-- 839657738087498284
example2 :: Int
example2 = hashWithSalt 0xDEADBEEF Red
-- 62679985974121021
```

The cereal library allows us to automatically derive a binary representation.

```
{-# LANGUAGE DeriveGeneric #-}
import Data.Word
import Data.ByteString
import Data.Serialize
import GHC.Generics
data Val = A [Val] | B [(Val, Val)] | C
deriving (Generic, Show)
instance Serialize Val where
encoded :: ByteString
encoded = encode (A [B [(C, C)]])
-- "\NUL\NUL\NUL\NUL\NUL\NUL\NUL\NUL\SOH\SOH\NUL\NUL\NUL\NUL\NUL\NUL\NUL\SOH\STX\STX"
bytes :: [Word8]
bytes = unpack encoded
-- [0,0,0,0,0,0,0,0,1,1,0,0,0,0,0,0,0,1,2,2]
decoded :: Either String Val
decoded = decode encoded
```

The aeson library allows us to derive JSON representations for JSON instances.

```
{-# LANGUAGE DeriveGeneric #-}
{-# LANGUAGE OverloadedStrings #-}
import Data.Aeson
import GHC.Generics
data Point = Point { _x :: Double, _y :: Double }
deriving (Show, Generic)
instance FromJSON Point
instance ToJSON Point
example1 :: Maybe Point
example1 = decode "{\"x\":3.0,\"y\":-1.0}"
example2 = encode $ Point 123.4 20
```

See: A Generic Deriving Mechanism for Haskell

##### Higher Kinded Generics

Using the same interface GHC.Generics provides a separate typeclass for higher-kinded generics.

```
class Generic1 f where
type Rep1 f :: * -> *
from1 :: f a -> (Rep1 f) a
to1 :: (Rep1 f) a -> f a
```

So for instance `Maybe`

has `Rep1`

of the form:

```
type instance Rep1 Maybe
= D1
GHC.Generics.D1Maybe
(C1 C1_0Maybe U1
:+: C1 C1_1Maybe (S1 NoSelector Par1))
```

## generics-sop

## Uniplate

Uniplate is a generics library for writing traversals and transformation for arbitrary data structures. It is extremely useful for writing AST transformations and rewriting systems.

```
plate :: from -> Type from to
(|*) :: Type (to -> from) to -> to -> Type from to
(|-) :: Type (item -> from) to -> item -> Type from to
descend :: Uniplate on => (on -> on) -> on -> on
transform :: Uniplate on => (on -> on) -> on -> on
rewrite :: Uniplate on => (on -> Maybe on) -> on -> on
```

The `descend`

function will apply a function to each immediate descendant of an expression and then combines them up into the parent expression.

The `transform`

function will perform a single pass bottom-up transformation of all terms in the expression.

The `rewrite`

function will perform an exhaustive transformation of all terms in the expression to fixed point, using Maybe to signify termination.

```
import Data.Generics.Uniplate.Direct
data Expr a
= Fls
| Tru
| Var a
| Not (Expr a)
| And (Expr a) (Expr a)
| Or (Expr a) (Expr a)
deriving (Show, Eq)
instance Uniplate (Expr a) where
uniplate (Not f) = plate Not |* f
uniplate (And f1 f2) = plate And |* f1 |* f2
uniplate (Or f1 f2) = plate Or |* f1 |* f2
uniplate x = plate x
simplify :: Expr a -> Expr a
simplify = transform simp
where
simp (Not (Not f)) = f
simp (Not Fls) = Tru
simp (Not Tru) = Fls
simp x = x
reduce :: Show a => Expr a -> Expr a
reduce = rewrite cnf
where
-- double negation
cnf (Not (Not p)) = Just p
-- de Morgan
cnf (Not (p `Or` q)) = Just $ (Not p) `And` (Not q)
cnf (Not (p `And` q)) = Just $ (Not p) `Or` (Not q)
-- distribute conjunctions
cnf (p `Or` (q `And` r)) = Just $ (p `Or` q) `And` (p `Or` r)
cnf ((p `And` q) `Or` r) = Just $ (p `Or` q) `And` (p `Or` r)
cnf _ = Nothing
example1 :: Expr String
example1 = simplify (Not (Not (Not (Not (Var "a")))))
-- Var "a"
example2 :: [String]
example2 = [a | Var a <- universe ex]
where
ex = Or (And (Var "a") (Var "b")) (Not (And (Var "c") (Var "d")))
-- ["a","b","c","d"]
example3 :: Expr String
example3 = reduce $ ((a `And` b) `Or` (c `And` d)) `Or` e
where
a = Var "a"
b = Var "b"
c = Var "c"
d = Var "d"
e = Var "e"
```

Alternatively Uniplate instances can be derived automatically from instances of Data without the need to explicitly write a Uniplate instance. This approach carries a slight amount of overhead over an explicit hand-written instance.

```
import Data.Data
import Data.Typeable
import Data.Generics.Uniplate.Data
data Expr a
= Fls
| Tru
| Lit a
| Not (Expr a)
| And (Expr a) (Expr a)
| Or (Expr a) (Expr a)
deriving (Data, Typeable, Show, Eq)
```

**Biplate**

Biplates generalize plates where the target type isn't necessarily the same as the source, it uses multiparameter typeclasses to indicate the type sub of the sub-target. The Uniplate functions all have an equivalent generalized biplate form.

```
descendBi :: Biplate from to => (to -> to) -> from -> from
transformBi :: Biplate from to => (to -> to) -> from -> from
rewriteBi :: Biplate from to => (to -> Maybe to) -> from -> from
descendBiM :: (Monad m, Biplate from to) => (to -> m to) -> from -> m from
transformBiM :: (Monad m, Biplate from to) => (to -> m to) -> from -> m from
rewriteBiM :: (Monad m, Biplate from to) => (to -> m (Maybe to)) -> from -> m from
```

```
{-# LANGUAGE MultiParamTypeClasses #-}
{-# LANGUAGE FlexibleContexts #-}
import Data.Generics.Uniplate.Direct
type Name = String
data Expr
= Var Name
| Lam Name Expr
| App Expr Expr
deriving Show
data Stmt
= Decl [Stmt]
| Let Name Expr
deriving Show
instance Uniplate Expr where
uniplate (Var x ) = plate Var |- x
uniplate (App x y) = plate App |* x |* y
uniplate (Lam x y) = plate Lam |- x |* y
instance Biplate Expr Expr where
biplate = plateSelf
instance Uniplate Stmt where
uniplate (Decl x ) = plate Decl ||* x
uniplate (Let x y) = plate Let |- x |- y
instance Biplate Stmt Stmt where
biplate = plateSelf
instance Biplate Stmt Expr where
biplate (Decl x) = plate Decl ||+ x
biplate (Let x y) = plate Let |- x |* y
rename :: Name -> Name -> Expr -> Expr
rename from to = rewrite f
where
f (Var a) | a == from = Just (Var to)
f (Lam a b) | a == from = Just (Lam to b)
f _ = Nothing
s, k, sk :: Expr
s = Lam "x" (Lam "y" (Lam "z" (App (App (Var "x") (Var "z")) (App (Var "y") (Var "z")))))
k = Lam "x" (Lam "y" (Var "x"))
sk = App s k
m :: Stmt
m = descendBi f $ Decl [ (Let "s" s) , Let "k" k , Let "sk" sk ]
where
f = rename "x" "a"
. rename "y" "b"
. rename "z" "c"
```

# Mathematics

## Numeric Tower

Haskell's numeric tower is unusual and the source of some confusion for novices. Haskell is one of the few languages to incorporate statically typed overloaded literals without a mechanism for "coercions" often found in other languages.

To add to the confusion numerical literals in Haskell are desugared into a function from a numeric typeclass which yields a polymorphic value that can be instantiated to any instance of the `Num`

or `Fractional`

typeclass at the call-site, depending on the inferred type.

To use a blunt metaphor, we're effectively placing an object in a hole and the size and shape of the hole defines the object you place there. This is very different than in other languages where a numeric literal like `2.718`

is hard coded in the compiler to be a specific type ( double or something ) and you cast the value at runtime to be something smaller or larger as needed.

```
42 :: Num a => a
fromInteger (42 :: Integer)
2.71 :: Fractional a => a
fromRational (2.71 :: Rational)
```

The numeric typeclass hierarchy is defined as such:

```
class Num a
class (Num a, Ord a) => Real a
class Num a => Fractional a
class (Real a, Enum a) => Integral a
class (Real a, Fractional a) => RealFrac a
class Fractional a => Floating a
class (RealFrac a, Floating a) => RealFloat a
```

Conversions between concrete numeric types ( from : left column, to : top row ) is accomplished with several generic functions.

Double | Float | Int | Word | Integer | Rational | |
---|---|---|---|---|---|---|

Double | id | fromRational | truncate | truncate | truncate | toRational |

Float | fromRational | id | truncate | truncate | truncate | toRational |

Int | fromIntegral | fromIntegral | id | fromIntegral | fromIntegral | fromIntegral |

Word | fromIntegral | fromIntegral | fromIntegral | id | fromIntegral | fromIntegral |

Integer | fromIntegral | fromIntegral | fromIntegral | fromIntegral | id | fromIntegral |

Rational | fromRatoinal | fromRational | truncate | truncate | truncate | id |

## Integer

The `Integer`

type in GHC is implemented by the GMP (`libgmp`

) arbitrary precision arithmetic library. Unlike the `Int`

type the size of Integer values is bounded only by the available memory. Most notably `libgmp`

is one of the few libraries that compiled Haskell binaries are dynamically linked against.

An alternative library `integer-simple`

can be linked in place of libgmp.

See: GHC, primops and exorcising GMP

## Complex

Haskell supports arithmetic with complex numbers via a Complex datatype from the `Data.Complex`

module. The first argument is the real part, while the second is the imaginary part. The type has a single parameter and inherits it's numerical typeclass components (Num, Fractional, Floating) from the type of this paramater.

```
-- 1 + 2i
let complex = 1 :+ 2
```

```
data Complex a = a :+ a
mkPolar :: RealFloat a => a -> a -> Complex a
```

The `Num`

instance for `Complex`

is only defined if parameter of `Complex`

is an instance of `RealFloat`

.

```
λ: 0 :+ 1
0 :+ 1 :: Complex Integer
λ: (0 :+ 1) + (1 :+ 0)
1.0 :+ 1.0 :: Complex Integer
λ: exp (0 :+ 2 * pi)
1.0 :+ (-2.4492935982947064e-16) :: Complex Double
λ: mkPolar 1 (2*pi)
1.0 :+ (-2.4492935982947064e-16) :: Complex Double
λ: let f x n = (cos x :+ sin x)^n
λ: let g x n = cos (n*x) :+ sin (n*x)
```

## Scientific

Scientific provides arbitrary-precision numbers represented using scientific notation. The constructor takes an arbitrarily sized Integer argument for the digits and an Int for the exponent. Alternatively the value can be parsed from a String or coerced from either Double/Float.

```
scientific :: Integer -> Int -> Scientific
fromFloatDigits :: RealFloat a => a -> Scientific
```

```
import Data.Scientific
c, h, g, a, k :: Scientific
c = scientific 299792458 (0) -- Speed of light
h = scientific 662606957 (-42) -- Planck's constant
g = scientific 667384 (-16) -- Gravitational constant
a = scientific 729735257 (-11) -- Fine structure constant
k = scientific 268545200 (-9) -- Khinchin Constant
tau :: Scientific
tau = fromFloatDigits (2*pi)
maxDouble64 :: Double
maxDouble64 = read "1.7976931348623159e308"
-- Infinity
maxScientific :: Scientific
maxScientific = read "1.7976931348623159e308"
-- 1.7976931348623159e308
```

## Statistics

```
import Data.Vector
import Statistics.Sample
import Statistics.Distribution.Normal
import Statistics.Distribution.Poisson
import qualified Statistics.Distribution as S
s1 :: Vector Double
s1 = fromList [1, 2, 3, 4, 5, 6, 7, 8, 9, 10]
s2 :: PoissonDistribution
s2 = poisson 2.5
s3 :: NormalDistribution
s3 = normalDistr mean stdDev
where
mean = 1
stdDev = 1
descriptive = do
print $ range s1
-- 9.0
print $ mean s1
-- 5.5
print $ stdDev s1
-- 3.0276503540974917
print $ variance s1
-- 8.25
print $ harmonicMean s1
-- 3.414171521474055
print $ geometricMean s1
-- 4.5287286881167645
discrete = do
print $ S.cumulative s2 0
-- 8.208499862389884e-2
print $ S.mean s2
-- 2.5
print $ S.variance s2
-- 2.5
print $ S.stdDev s2
-- 1.5811388300841898
continuous = do
print $ S.cumulative s3 0
-- 0.15865525393145707
print $ S.quantile s3 0.5
-- 1.0
print $ S.density s3 0
-- 0.24197072451914334
print $ S.mean s3
-- 1.0
print $ S.variance s3
-- 1.0
print $ S.stdDev s3
-- 1.0
```

## Constructive Reals

Instead of modeling the real numbers on finite precision floating point numbers we alternatively work with `Num`

which internally manipulate the power series expansions for the expressions when performing operations like arithmetic or transcendental functions without losing precision when performing intermediate computations. Then we simply slice off a fixed number of terms and approximate the resulting number to a desired precision. This approach is not without its limitations and caveats ( notably that it may diverge ).

```
exp(x) = 1 + x + 1/2*x^2 + 1/6*x^3 + 1/24*x^4 + 1/120*x^5 ...
sqrt(1+x) = 1 + 1/2*x - 1/8*x^2 + 1/16*x^3 - 5/128*x^4 + 7/256*x^5 ...
atan(x) = x - 1/3*x^3 + 1/5*x^5 - 1/7*x^7 + 1/9*x^9 - 1/11*x^11 ...
pi = 16 * atan (1/5) - 4 * atan (1/239)
```

```
import Data.Number.CReal
-- algebraic
phi :: CReal
phi = (1 + sqrt 5) / 2
-- transcendental
ramanujan :: CReal
ramanujan = exp (pi * sqrt 163)
main :: IO ()
main = do
putStrLn $ showCReal 30 pi
-- 3.141592653589793238462643383279
putStrLn $ showCReal 30 phi
-- 1.618033988749894848204586834366
putStrLn $ showCReal 15 ramanujan
-- 262537412640768743.99999999999925
```

## SAT Solvers

A collection of constraint problems known as satisfiability problems show up in a number of different disciplines from type checking to package management. Simply put a satisfiability problem attempts to find solutions to a statement of conjoined conjunctions and disjunctions in terms of a series of variables. For example:

`(A v ¬B v C) ∧ (B v D v E) ∧ (D v F)`

To use the picosat library to solve this, it can be written as zero-terminated lists of integers and fed to the solver according to a number-to-variable relation:

```
1 -2 3 -- (A v ¬B v C)
2 4 5 -- (B v D v E)
4 6 -- (D v F)
```

```
import Picosat
main :: IO [Int]
main = do
solve [[1, -2, 3], [2,4,5], [4,6]]
-- Solution [1,-2,3,4,5,6]
```

The SAT solver itself can be used to solve satisfiability problems with millions of variables in this form and is finely tuned.

See:

## SMT Solvers

A generalization of the SAT problem to include predicates other theories gives rise to the very sophisticated domain of "Satisfiability Modulo Theory" problems. The existing SMT solvers are very sophisticated projects ( usually bankrolled by large institutions ) and usually have to called out to via foreign function interface or via a common interface called SMT-lib. The two most common of use in Haskell are `cvc4`

from Stanford and `z3`

from Microsoft Research.

The SBV library can abstract over different SMT solvers to allow us to express the problem in an embedded domain language in Haskell and then offload the solving work to the third party library.

As an example, here's how you can solve a simple cryptarithm

`M` `O` `N` `A` `D` |

+ `B` `U` `R` `R` `I` `T` `O` |

= `B` `A` `N` `D` `A` `I` `D` |

using SBV library:

```
import Data.Foldable
import Data.SBV
-- | val [4,2] == 42
val :: [SInteger] -> SInteger
val = foldr1 (\d r -> d + 10*r) . reverse
puzzle :: Symbolic SBool
puzzle = do
ds@[b,u,r,i,t,o,m,n,a,d] <- sequenceA [ sInteger [v] | v <- "buritomnad" ]
constrain $ allDifferent ds
for_ ds $ \d -> constrain $ inRange d (0,9)
pure $ val [b,u,r,r,i,t,o]
+ val [m,o,n,a,d]
.== val [b,a,n,d,a,i,d]
```

Let's look at all possible solutions,

```
λ: allSat puzzle
Solution #1:
b = 4 :: Integer
u = 1 :: Integer
r = 5 :: Integer
i = 9 :: Integer
t = 7 :: Integer
o = 0 :: Integer
m = 8 :: Integer
n = 3 :: Integer
a = 2 :: Integer
d = 6 :: Integer
This is the only solution.
```

See:

## Z3

TODO

See: z3

# Data Structures

## Map

Functionality | Function | Time Complexity |
---|---|---|

Initialization | empty | O(1) |

Size | size | O(1) |

Lookup | lookup | O(log(n)) |

Insertion | insert | O(log(n)) |

Traversal | traverse | O(n) |

A map is an associative array mapping any instance of `Ord`

keys to values of any type.

```
import qualified Data.Map as Map
kv :: Map.Map Integer String
kv = Map.fromList [(1, "a"), (2, "b")]
lkup :: Integer -> String -> String
lkup key def =
case Map.lookup key kv of
Just val -> val
Nothing -> def
```

## Tree

Functionality | Function | Time Complexity |
---|---|---|

Initialization | empty | O(1) |

Size | size | O(1) |

Lookup | lookup | O(log(n)) |

Insertion | insert | O(log(n)) |

Traversal | traverse | O(n) |

```
import Data.Tree
{-
A
/ \
B C
/ \
D E
-}
tree :: Tree String
tree = Node "A" [Node "B" [], Node "C" [Node "D" [], Node "E" []]]
postorder :: Tree a -> [a]
postorder (Node a ts) = elts ++ [a]
where elts = concat (map postorder ts)
preorder :: Tree a -> [a]
preorder (Node a ts) = a : elts
where elts = concat (map preorder ts)
ex1 = drawTree tree
ex2 = drawForest (subForest tree)
ex3 = flatten tree
ex4 = levels tree
ex5 = preorder tree
ex6 = postorder tree
```

## Set

Functionality | Function | Time Complexity |
---|---|---|

Initialization | empty | O(1) |

Size | size | O(1) |

Insertion | insert | O(log(n)) |

Deletion | delete | O(log(n)) |

Traversal | traverse | O(n) |

Membership Test | member | O(log(n)) |

Sets are an unordered data structures allow `Ord`

values of any type and guaranteeing uniqueness with in the structure. They are not identical to the mathematical notion of a Set even though they share the same namesake.

```
import qualified Data.Set as Set
set :: Set.Set Integer
set = Set.fromList [1..1000]
memtest :: Integer -> Bool
memtest elt = Set.member elt set
```

## Vector

Functionality | Function | Time Complexity |
---|---|---|

Initialization | empty | O(1) |

Size | length | O(1) |

Indexing | (!) | O(1) |

Append | append | O(n) |

Traversal | traverse | O(n) |

Vectors are high performance single dimensional arrays that come come in six variants, two for each of the following types of a mutable and an immutable variant.

- Data.Vector
- Data.Vector.Storable
- Data.Vector.Unboxed

The most notable feature of vectors is constant time memory access with (`(!)`

) as well as variety of efficient map, fold and scan operations on top of a fusion framework that generates surprisingly optimal code.

```
fromList :: [a] -> Vector a
toList :: Vector a -> [a]
(!) :: Vector a -> Int -> a
map :: (a -> b) -> Vector a -> Vector b
foldl :: (a -> b -> a) -> a -> Vector b -> a
scanl :: (a -> b -> a) -> a -> Vector b -> Vector a
zipWith :: (a -> b -> c) -> Vector a -> Vector b -> Vector c
iterateN :: Int -> (a -> a) -> a -> Vector a
```

```
import Data.Vector.Unboxed as V
norm :: Vector Double -> Double
norm = sqrt . V.sum . V.map (\x -> x*x)
example1 :: Double
example1 = norm $ V.iterateN 100000000 (+1) 0.0
```

See: Numerical Haskell: A Vector Tutorial

## Mutable Vectors

Functionality | Function | Time Complexity |
---|---|---|

Initialization | empty | O(1) |

Size | length | O(1) |

Indexing | (!) | O(1) |

Append | append | O(n) |

Traversal | traverse | O(n) |

Update | modify | O(1) |

Read | read | O(1) |

Write | write | O(1) |

```
freeze :: MVector (PrimState m) a -> m (Vector a)
thaw :: Vector a -> MVector (PrimState m) a
```

Within the IO monad we can perform arbitrary read and writes on the mutable vector with constant time reads and writes. When needed a static Vector can be created to/from the `MVector`

using the freeze/thaw functions.

```
import GHC.Prim
import Control.Monad
import Control.Monad.ST
import Control.Monad.Primitive
import Data.Vector.Unboxed (freeze)
import Data.Vector.Unboxed.Mutable
import qualified Data.Vector.Unboxed as V
example :: PrimMonad m => m (V.Vector Int)
example = do
v <- new 10
forM_ [0..9] $ \i ->
write v i (2*i)
freeze v
-- vector computation in IO
vecIO :: IO (V.Vector Int)
vecIO = example
-- vector computation in ST
vecST :: ST s (V.Vector Int)
vecST = example
main :: IO ()
main = do
vecIO >>= print
print $ runST vecST
```

The vector library itself normally does bounds checks on index operations to protect against memory corruption. This can be enabled or disabled on the library level by compiling with `boundschecks`

cabal flag.

## Unordered-Containers

Functionality | Function | Time Complexity |
---|---|---|

Initialization | empty | O(1) |

Size | size | O(1) |

Lookup | lookup | O(log(n)) |

Insertion | insert | O(log(n)) |

Traversal | traverse | O(n) |

```
fromList :: (Eq k, Hashable k) => [(k, v)] -> HashMap k v
lookup :: (Eq k, Hashable k) => k -> HashMap k v -> Maybe v
insert :: (Eq k, Hashable k) => k -> v -> HashMap k v -> HashMap k v
```

Both the `HashMap`

and `HashSet`

are purely functional data structures that are drop in replacements for the `containers`

equivalents but with more efficient space and time performance. Additionally all stored elements must have a `Hashable`

instance.

```
import qualified Data.HashSet as S
import qualified Data.HashMap.Lazy as M
example1 :: M.HashMap Int Char
example1 = M.fromList $ zip [1..10] ['a'..]
example2 :: S.HashSet Int
example2 = S.fromList [1..10]
```

See: Announcing Unordered Containers

## Hashtables

Functionality | Function | Time Complexity |
---|---|---|

Initialization | empty | O(1) |

Size | size | O(1) |

Lookup | lookup | O(1) |

Insertion | insert | O(1) amortized |

Traversal | traverse | O(n) |

Hashtables provides hashtables with efficient lookup within the ST or IO monad.

```
import Prelude hiding (lookup)
import Control.Monad.ST
import Data.HashTable.ST.Basic
-- Hashtable parameterized by ST "thread"
type HT s = HashTable s String String
set :: ST s (HT s)
set = do
ht <- new
insert ht "key" "value1"
return ht
get :: HT s -> ST s (Maybe String)
get ht = do
val <- lookup ht "key"
return val
example :: Maybe String
example = runST (set >>= get)
```

```
new :: ST s (HashTable s k v)
insert :: (Eq k, Hashable k) => HashTable s k v -> k -> v -> ST s ()
lookup :: (Eq k, Hashable k) => HashTable s k v -> k -> ST s (Maybe v)
```

## Graphs

The Graph module in the containers library is a somewhat antiquated API for working with directed graphs. A little bit of data wrapping makes it a little more straightforward to use. The library is not necessarily well-suited for large graph-theoretic operations but is perfectly fine for example, to use in a typechecker which need to resolve strongly connected components of the module definition graph.

```
import Data.Tree
import Data.Graph
data Grph node key = Grph
{ _graph :: Graph
, _vertices :: Vertex -> (node, key, [key])
}
fromList :: Ord key => [(node, key, [key])] -> Grph node key
fromList = uncurry Grph . graphFromEdges'
vertexLabels :: Functor f => Grph b t -> (f Vertex) -> f b
vertexLabels g = fmap (vertexLabel g)
vertexLabel :: Grph b t -> Vertex -> b
vertexLabel g = (\(vi, _, _) -> vi) . (_vertices g)
-- Topologically sort graph
topo' :: Grph node key -> [node]
topo' g = vertexLabels g $ topSort (_graph g)
-- Strongly connected components of graph
scc' :: Grph node key -> [[node]]
scc' g = fmap (vertexLabels g . flatten) $ scc (_graph g)
```

So for example we can construct a simple graph:

```
ex1 :: [(String, String, [String])]
ex1 = [
("a","a",["b"]),
("b","b",["c"]),
("c","c",["a"])
]
ts1 :: [String]
ts1 = topo' (fromList ex1)
-- ["a","b","c"]
sc1 :: [[String]]
sc1 = scc' (fromList ex1)
-- [["a","b","c"]]
```

Or with two strongly connected subgraphs:

```
ex2 :: [(String, String, [String])]
ex2 = [
("a","a",["b"]),
("b","b",["c"]),
("c","c",["a"]),
("d","d",["e"]),
("e","e",["f", "e"]),
("f","f",["d", "e"])
]
ts2 :: [String]
ts2 = topo' (fromList ex2)
-- ["d","e","f","a","b","c"]
sc2 :: [[String]]
sc2 = scc' (fromList ex2)
-- [["d","e","f"],["a","b","c"]]
```

See: GraphSCC

## Graph Theory

The `fgl`

library provides a more efficient graph structure and a wide variety of common graph-theoretic operations. For example calculating the dominance frontier of a graph shows up quite frequently in control flow analysis for compiler design.

```
import qualified Data.Graph.Inductive as G
cyc3 :: G.Gr Char String
cyc3 = G.buildGr
[([("ca",3)],1,'a',[("ab",2)]),
([],2,'b',[("bc",3)]),
([],3,'c',[])]
-- Loop query
ex1 :: Bool
ex1 = G.hasLoop x
-- Dominators
ex2 :: [(G.Node, [G.Node])]
ex2 = G.dom x 0
```

```
x :: G.Gr Int ()
x = G.insEdges edges gr
where
gr = G.insNodes nodes G.empty
edges = [(0,1,()), (0,2,()), (2,1,()), (2,3,())]
nodes = zip [0,1 ..] [2,3,4,1]
```

## DList

Functionality | Function | Time Complexity |
---|---|---|

Initialization | empty | O(1) |

Size | size | O(1) |

Lookup | lookup | O(log(n)) |

Insertion | insert | O(log(n)) |

Traversal | traverse | O(n) |

Append | (|>) | O(1) |

Prepend | (<|) | O(1) |

A dlist is a list-like structure that is optimized for O(1) append operations, internally it uses a Church encoding of the list structure. It is specifically suited for operations which are append-only and need only access it when manifesting the entire structure. It is particularly well-suited for use in the Writer monad.

```
import Data.DList
import Control.Monad
import Control.Monad.Writer
logger :: Writer (DList Int) ()
logger = replicateM_ 100000 $ tell (singleton 0)
```

## Sequence

The sequence data structure behaves structurally similar to list but is optimized for append/prepend operations and traversal.

```
import Data.Sequence
a :: Seq Int
a = fromList [1,2,3]
a0 :: Seq Int
a0 = a |> 4
-- [1,2,3,4]
a1 :: Seq Int
a1 = 0 <| a
-- [0,1,2,3]
```

# FFI

This is an advanced section, knowledge of FFI is not typically necessary to write Haskell.

## Pure Functions

Wrapping pure C functions with primitive types is trivial.

```
/* $(CC) -c simple.c -o simple.o */
int example(int a, int b)
{
return a + b;
}
```

```
-- ghc simple.o simple_ffi.hs -o simple_ffi
{-# LANGUAGE ForeignFunctionInterface #-}
import Foreign.C.Types
foreign import ccall safe "example" example
:: CInt -> CInt -> CInt
main = print (example 42 27)
```

## Storable Arrays

There exists a `Storable`

typeclass that can be used to provide low-level access to the memory underlying Haskell values. `Ptr`

objects in Haskell behave much like C pointers although arithmetic with them is in terms of bytes only, not the size of the type associated with the pointer ( this differs from C).

The Prelude defines Storable interfaces for most of the basic types as well as types in the `Foreign.C`

library.

```
class Storable a where
sizeOf :: a -> Int
alignment :: a -> Int
peek :: Ptr a -> IO a
poke :: Ptr a -> a -> IO ()
```

To pass arrays from Haskell to C we can again use Storable Vector and several unsafe operations to grab a foreign pointer to the underlying data that can be handed off to C. Once we're in C land, nothing will protect us from doing evil things to memory!

```
/* $(CC) -c qsort.c -o qsort.o */
void swap(int *a, int *b)
{
int t = *a;
*a = *b;
*b = t;
}
void sort(int *xs, int beg, int end)
{
if (end > beg + 1) {
int piv = xs[beg], l = beg + 1, r = end;
while (l < r) {
if (xs[l] <= piv) {
l++;
} else {
swap(&xs[l], &xs[--r]);
}
}
swap(&xs[--l], &xs[beg]);
sort(xs, beg, l);
sort(xs, r, end);
}
}
```

```
-- ghc qsort.o ffi.hs -o ffi
{-# LANGUAGE ForeignFunctionInterface #-}
import Foreign.Ptr
import Foreign.C.Types
import qualified Data.Vector.Storable as V
import qualified Data.Vector.Storable.Mutable as VM
foreign import ccall safe "sort" qsort
:: Ptr a -> CInt -> CInt -> IO ()
main :: IO ()
main = do
let vs = V.fromList ([1,3,5,2,1,2,5,9,6] :: [CInt])
v <- V.thaw vs
VM.unsafeWith v $ \ptr -> do
qsort ptr 0 9
out <- V.freeze v
print out
```

The names of foreign functions from a C specific header file can be qualified.

```
foreign import ccall unsafe "stdlib.h malloc"
malloc :: CSize -> IO (Ptr a)
```

Prepending the function name with a `&`

allows us to create a reference to the function pointer itself.

```
foreign import ccall unsafe "stdlib.h &malloc"
malloc :: FunPtr a
```

## Function Pointers

Using the above FFI functionality, it's trivial to pass C function pointers into Haskell, but what about the inverse passing a function pointer to a Haskell function into C using `foreign import ccall "wrapper"`

.

```
#include <stdio.h>
void invoke(void (*fn)(int))
{
int n = 42;
printf("Inside of C, now we'll call Haskell.\n");
fn(n);
printf("Back inside of C again.\n");
}
```

```
{-# LANGUAGE ForeignFunctionInterface #-}
import Foreign
import System.IO
import Foreign.C.Types(CInt(..))
foreign import ccall "wrapper"
makeFunPtr :: (CInt -> IO ()) -> IO (FunPtr (CInt -> IO ()))
foreign import ccall "pointer.c invoke"
invoke :: FunPtr (CInt -> IO ()) -> IO ()
fn :: CInt -> IO ()
fn n = do
putStrLn "Hello from Haskell, here's a number passed between runtimes:"
print n
hFlush stdout
main :: IO ()
main = do
fptr <- makeFunPtr fn
invoke fptr
```

Will yield the following output:

```
Inside of C, now we'll call Haskell
Hello from Haskell, here's a number passed between runtimes:
42
Back inside of C again.
```

# Concurrency

The definitive reference on concurrency and parallelism in Haskell is Simon Marlow's text. This will section will just gloss over these topics because they are far better explained in this book.

See: Parallel and Concurrent Programming in Haskell

`forkIO :: IO () -> IO ThreadId`

Haskell threads are extremely cheap to spawn, using only 1.5KB of RAM depending on the platform and are much cheaper than a pthread in C. Calling forkIO 10^{6} times completes just short of a 1s. Additionally, functional purity in Haskell also guarantees that a thread can almost always be terminated even in the middle of a computation without concern.

See: The Scheduler

## Sparks

The most basic "atom" of parallelism in Haskell is a spark. It is a hint to the GHC runtime that a computation can be evaluated to weak head normal form in parallel.

```
rpar :: a -> Eval a
rseq :: Strategy a
rdeepseq :: NFData a => Strategy a
runEval :: Eval a -> a
```

`rpar a`

spins off a separate spark that evolutes a to weak head normal form and places the computation in the spark pool. When the runtime determines that there is an available CPU to evaluate the computation it will evaluate ( *convert* ) the spark. If the main thread of the program is the evaluator for the spark, the spark is said to have *fizzled*. Fizzling is generally bad and indicates that the logic or parallelism strategy is not well suited to the work that is being evaluated.

The spark pool is also limited ( but user-adjustable ) to a default of 8000 (as of GHC 7.8.3 ). Sparks that are created beyond that limit are said to *overflow*.

```
-- Evaluates the arguments to f in parallel before application.
par2 f x y = x `rpar` y `rpar` f x y
```

An argument to `rseq`

forces the evaluation of a spark before evaluation continues.

Action | Description |
---|---|

`Fizzled` |
The resulting value has already been evaluated by the main thread so the spark need not be converted. |

`Dud` |
The expression has already been evaluated, the computed value is returned and the spark is not converted. |

`GC'd` |
The spark is added to the spark pool but the result is not referenced, so it is garbage collected. |

`Overflowed` |
Insufficient space in the spark pool when spawning. |

The parallel runtime is necessary to use sparks, and the resulting program must be compiled with `-threaded`

. Additionally the program itself can be specified to take runtime options with `-rtsopts`

such as the number of cores to use.

```
ghc -threaded -rtsopts program.hs
./program +RTS -s N8 -- use 8 cores
```

The runtime can be asked to dump information about the spark evaluation by passing the `-s`

flag.

```
$ ./spark +RTS -N4 -s
Tot time (elapsed) Avg pause Max pause
Gen 0 5 colls, 5 par 0.02s 0.01s 0.0017s 0.0048s
Gen 1 3 colls, 2 par 0.00s 0.00s 0.0004s 0.0007s
Parallel GC work balance: 1.83% (serial 0%, perfect 100%)
TASKS: 6 (1 bound, 5 peak workers (5 total), using -N4)
SPARKS: 20000 (20000 converted, 0 overflowed, 0 dud, 0 GC'd, 0 fizzled)
```

The parallel computations themselves are sequenced in the `Eval`

monad, whose evaluation with `runEval`

is itself a pure computation.

```
example :: (a -> b) -> a -> a -> (b, b)
example f x y = runEval $ do
a <- rpar $ f x
b <- rpar $ f y
rseq a
rseq b
return (a, b)
```

## Threadscope

Passing the flag `-l`

generates the eventlog which can be rendered with the threadscope library.

```
$ ghc -O2 -threaded -rtsopts -eventlog Example.hs
$ ./program +RTS -N4 -l
$ threadscope Example.eventlog
```

See Simon Marlows's *Parallel and Concurrent Programming in Haskell* for a detailed guide on interpreting and profiling using Threadscope.

See:

## Strategies

```
type Strategy a = a -> Eval a
using :: a -> Strategy a -> a
```

Sparks themselves form the foundation for higher level parallelism constructs known as `strategies`

which adapt spark creation to fit the computation or data structure being evaluated. For instance if we wanted to evaluate both elements of a tuple in parallel we can create a strategy which uses sparks to evaluate both sides of the tuple.

```
import Control.Parallel.Strategies
parPair' :: Strategy (a, b)
parPair' (a, b) = do
a' <- rpar a
b' <- rpar b
return (a', b')
fib :: Int -> Int
fib 0 = 0
fib 1 = 1
fib n = fib (n-1) + fib (n-2)
serial :: (Int, Int)
serial = (fib 30, fib 31)
parallel :: (Int, Int)
parallel = runEval . parPair' $ (fib 30, fib 31)
```

This pattern occurs so frequently the combinator `using`

can be used to write it equivalently in operator-like form that may be more visually appealing to some.

```
using :: a -> Strategy a -> a
x `using` s = runEval (s x)
parallel ::: (Int, Int)
parallel = (fib 30, fib 31) `using` parPair
```

For a less contrived example consider a parallel `parmap`

which maps a pure function over a list of a values in parallel.

```
import Control.Parallel.Strategies
parMap' :: (a -> b) -> [a] -> Eval [b]
parMap' f [] = return []
parMap' f (a:as) = do
b <- rpar (f a)
bs <- parMap' f as
return (b:bs)
result :: [Int]
result = runEval $ parMap' (+1) [1..1000]
```

The functions above are quite useful, but will break down if evaluation of the arguments needs to be parallelized beyond simply weak head normal form. For instance if the arguments to `rpar`

is a nested constructor we'd like to parallelize the entire section of work in evaluated the expression to normal form instead of just the outer layer. As such we'd like to generalize our strategies so the the evaluation strategy for the arguments can be passed as an argument to the strategy.

`Control.Parallel.Strategies`

contains a generalized version of `rpar`

which embeds additional evaluation logic inside the `rpar`

computation in Eval monad.

`rparWith :: Strategy a -> Strategy a`

Using the deepseq library we can now construct a Strategy variant of rseq that evaluates to full normal form.

```
rdeepseq :: NFData a => Strategy a
rdeepseq x = rseq (force x)
```

We now can create a "higher order" strategy that takes two strategies and itself yields a a computation which when evaluated uses the passed strategies in its scheduling.

```
import Control.DeepSeq
import Control.Parallel.Strategies
evalPair :: Strategy a -> Strategy b -> Strategy (a, b)
evalPair sa sb (a, b) = do
a' <- sa a
b' <- sb b
return (a', b')
parPair :: Strategy a -> Strategy b -> Strategy (a, b)
parPair sa sb = evalPair (rparWith sa) (rparWith sb)
fib :: Int -> Int
fib 0 = 0
fib 1 = 1
fib n = fib (n-1) + fib (n-2)
serial :: ([Int], [Int])
serial = (a, b)
where
a = fmap fib [0..30]
b = fmap fib [1..30]
parallel :: ([Int], [Int])
parallel = (a, b) `using` evalPair rdeepseq rdeepseq
where
a = fmap fib [0..30]
b = fmap fib [1..30]
```

These patterns are implemented in the Strategies library along with several other general forms and combinators for combining strategies to fit many different parallel computations.

```
parTraverse :: Traversable t => Strategy a -> Strategy (t a)
dot :: Strategy a -> Strategy a -> Strategy a
($||) :: (a -> b) -> Strategy a -> a -> b
(.||) :: (b -> c) -> Strategy b -> (a -> b) -> a -> c
```

See:

## STM

```
atomically :: STM a -> IO a
orElse :: STM a -> STM a -> STM a
retry :: STM a
newTVar :: a -> STM (TVar a)
newTVarIO :: a -> IO (TVar a)
writeTVar :: TVar a -> a -> STM ()
readTVar :: TVar a -> STM a
modifyTVar :: TVar a -> (a -> a) -> STM ()
modifyTVar' :: TVar a -> (a -> a) -> STM ()
```

Software Transactional Memory is a technique for guaranteeing atomicity of values in parallel computations, such that all contexts view the same data when read and writes are guaranteed never to result in inconsistent states.

The strength of Haskell's purity guarantees that transactions within STM are pure and can always be rolled back if a commit fails.

```
import Control.Monad
import Control.Concurrent
import Control.Concurrent.STM
type Account = TVar Double
transfer :: Account -> Account -> Double -> STM ()
transfer from to amount = do
available <- readTVar from
when (amount > available) retry
modifyTVar from (+ (-amount))
modifyTVar to (+ amount)
-- Threads are scheduled non-deterministically.
actions :: Account -> Account -> [IO ThreadId]
actions a b = map forkIO [
-- transfer to
atomically (transfer a b 10)
, atomically (transfer a b (-20))
, atomically (transfer a b 30)
-- transfer back
, atomically (transfer a b (-30))
, atomically (transfer a b 20)
, atomically (transfer a b (-10))
]
main :: IO ()
main = do
accountA <- atomically $ newTVar 60
accountB <- atomically $ newTVar 0
sequence_ (actions accountA accountB)
balanceA <- atomically $ readTVar accountA
balanceB <- atomically $ readTVar accountB
print $ balanceA == 60
print $ balanceB == 0
```

## Monad Par

Using the Par monad we express our computation as a data flow graph which is scheduled in order of the connections between forked computations which exchange resulting computations with `IVar`

.

```
new :: Par (IVar a)
put :: NFData a => IVar a -> a -> Par ()
get :: IVar a -> Par a
fork :: Par () -> Par ()
spawn :: NFData a => Par a -> Par (IVar a)
```

```
import Control.Monad
import Control.Monad.Par
f, g :: Int -> Int
f x = x + 10
g x = x * 10
-- f x g x
-- \ /
-- a + b
-- / \
-- f (a+b) g (a+b)
-- \ /
-- (d,e)
example1 :: Int -> (Int, Int)
example1 x = runPar $ do
[a,b,c,d,e] <- replicateM 5 new
fork (put a (f x))
fork (put b (g x))
a' <- get a
b' <- get b
fork (put c (a' + b'))
c' <- get c
fork (put d (f c'))
fork (put e (g c'))
d' <- get d
e' <- get e
return (d', e')
example2 :: [Int]
example2 = runPar $ do
xs <- parMap (+1) [1..25]
return xs
-- foldr (+) 0 (map (^2) [1..xs])
example3 :: Int -> Int
example3 n = runPar $ do
let range = (InclusiveRange 1 n)
let mapper x = return (x^2)
let reducer x y = return (x+y)
parMapReduceRangeThresh 10 range mapper reducer 0
```

## async

Async is a higher level set of functions that work on top of Control.Concurrent and STM.

```
async :: IO a -> IO (Async a)
wait :: Async a -> IO a
cancel :: Async a -> IO ()
concurrently :: IO a -> IO b -> IO (a, b)
race :: IO a -> IO b -> IO (Either a b)
```

```
import Control.Monad
import Control.Applicative
import Control.Concurrent
import Control.Concurrent.Async
import Data.Time
timeit :: IO a -> IO (a,Double)
timeit io = do
t0 <- getCurrentTime
a <- io
t1 <- getCurrentTime
return (a, realToFrac (t1 `diffUTCTime` t0))
worker :: Int -> IO Int
worker n = do
-- simulate some work
threadDelay (10^2 * n)
return (n * n)
-- Spawn 2 threads in parallel, halt on both finished.
test1 :: IO (Int, Int)
test1 = do
val1 <- async $ worker 1000
val2 <- async $ worker 2000
(,) <$> wait val1 <*> wait val2
-- Spawn 2 threads in parallel, halt on first finished.
test2 :: IO (Either Int Int)
test2 = do
let val1 = worker 1000
let val2 = worker 2000
race val1 val2
-- Spawn 10000 threads in parallel, halt on all finished.
test3 :: IO [Int]
test3 = mapConcurrently worker [0..10000]
main :: IO ()
main = do
print =<< timeit test1
print =<< timeit test2
print =<< timeit test3
```

# Graphics

## Diagrams

Diagrams is a a parser combinator library for generating vector images to SVG and a variety of other formats.

```
import Diagrams.Prelude
import Diagrams.Backend.SVG.CmdLine
sierpinski :: Int -> Diagram SVG
sierpinski 1 = eqTriangle 1
sierpinski n =
s
===
(s ||| s) # centerX
where
s = sierpinski (n - 1)
example :: Diagram SVG
example = sierpinski 5 # fc black
main :: IO ()
main = defaultMain example
```

`$ runhaskell diagram1.hs -w 256 -h 256 -o diagram1.svg`

See: Diagrams Quick Start Tutorial

# Parsing

## Parsec

For parsing in Haskell it is quite common to use a family of libraries known as *Parser Combinators* which let us write code to generate parsers which themselves looks very similar to the parser grammar itself!

Combinators | |
---|---|

`<|>` |
The choice operator tries to parse the first argument before proceeding to the second. Can be chained sequentially to generate a sequence of options. |

`many` |
Consumes an arbitrary number of patterns matching the given pattern and returns them as a list. |

`many1` |
Like many but requires at least one match. |

`optional` |
Optionally parses a given pattern returning its value as a Maybe. |

`try` |
Backtracking operator will let us parse ambiguous matching expressions and restart with a different pattern. |

There are two styles of writing Parsec, one can choose to write with monads or with applicatives.

```
parseM :: Parser Expr
parseM = do
a <- identifier
char '+'
b <- identifier
return $ Add a b
```

The same code written with applicatives uses the applicative combinators:

```
-- | Sequential application.
(<*>) :: f (a -> b) -> f a -> f b
-- | Sequence actions, discarding the value of the first argument.
(*>) :: f a -> f b -> f b
(*>) = liftA2 (const id)
-- | Sequence actions, discarding the value of the second argument.
(<*) :: f a -> f b -> f a
(<*) = liftA2 const
```

```
parseA :: Parser Expr
parseA = Add <$> identifier <* char '+' <*> identifier
```

Now for instance if we want to parse simple lambda expressions we can encode the parser logic as compositions of these combinators which yield the string parser when evaluated under with the `parse`

.

```
import Text.Parsec
import Text.Parsec.String
data Expr
= Var Char
| Lam Char Expr
| App Expr Expr
deriving Show
lam :: Parser Expr
lam = do
char '\\'
n <- letter
string "->"
e <- expr
return $ Lam n e
app :: Parser Expr
app = do
apps <- many1 term
return $ foldl1 App apps
var :: Parser Expr
var = do
n <- letter
return $ Var n
parens :: Parser Expr -> Parser Expr
parens p = do
char '('
e <- p
char ')'
return e
term :: Parser Expr
term = var <|> parens expr
expr :: Parser Expr
expr = lam <|> app
decl :: Parser Expr
decl = do
e <- expr
eof
return e
test :: IO ()
test = parseTest decl "\\y->y(\\x->x)y"
main :: IO ()
main = test >>= print
```

## Custom Lexer

In our previous example lexing pass was not necessary because each lexeme mapped to a sequential collection of characters in the stream type. If we wanted to extend this parser with a non-trivial set of tokens, then Parsec provides us with a set of functions for defining lexers and integrating these with the parser combinators. The simplest example builds on top of the builtin Parsec language definitions which define a set of most common lexical schemes.

For instance we'll build on top of the empty language grammar on top of the haskellDef grammer that uses the Text token instead of string.

```
{-# LANGUAGE OverloadedStrings #-}
import Text.Parsec
import Text.Parsec.Text
import qualified Text.Parsec.Token as Tok
import qualified Text.Parsec.Language as Lang
import Data.Functor.Identity (Identity)
import qualified Data.Text as T
import qualified Data.Text.IO as TIO
data Expr
= Var T.Text
| App Expr Expr
| Lam T.Text Expr
deriving (Show)
lexer :: Tok.GenTokenParser T.Text () Identity
lexer = Tok.makeTokenParser style
style :: Tok.GenLanguageDef T.Text () Identity
style = Lang.emptyDef
{ Tok.commentStart = "{-"
, Tok.commentEnd = "-}"
, Tok.commentLine = "--"
, Tok.nestedComments = True
, Tok.identStart = letter
, Tok.identLetter = alphaNum <|> oneOf "_'"
, Tok.opStart = Tok.opLetter style
, Tok.opLetter = oneOf ":!#$%&*+./<=>?@\\^|-~"
, Tok.reservedOpNames = []
, Tok.reservedNames = []
, Tok.caseSensitive = True
}
parens :: Parser a -> Parser a
parens = Tok.parens lexer
reservedOp :: T.Text -> Parser ()
reservedOp op = Tok.reservedOp lexer (T.unpack op)
ident :: Parser T.Text
ident = T.pack <$> Tok.identifier lexer
contents :: Parser a -> Parser a
contents p = do
Tok.whiteSpace lexer
r <- p
eof
return r
var :: Parser Expr
var = do
var <- ident
return (Var var )
app :: Parser Expr
app = do
e1 <- expr
e2 <- expr
return (App e1 e2)
fun :: Parser Expr
fun = do
reservedOp "\\"
binder <- ident
reservedOp "."
rhs <- expr
return (Lam binder rhs)
expr :: Parser Expr
expr = do
es <- many1 aexp
return (foldl1 App es)
aexp :: Parser Expr
aexp = fun <|> var <|> (parens expr)
test :: T.Text -> Either ParseError Expr
test = parse (contents expr) "<stdin>"
repl :: IO ()
repl = do
str <- TIO.getLine
print (test str)
repl
main :: IO ()
main = repl
```

See: Text.Parsec.Language

## Simple Parsing

Putting our lexer and parser together we can write down a more robust parser for our little lambda calculus syntax.

```
module Parser (parseExpr) where
import Text.Parsec
import Text.Parsec.String (Parser)
import Text.Parsec.Language (haskellStyle)
import qualified Text.Parsec.Expr as Ex
import qualified Text.Parsec.Token as Tok
type Id = String
data Expr
= Lam Id Expr
| App Expr Expr
| Var Id
| Num Int
| Op Binop Expr Expr
deriving (Show)
data Binop = Add | Sub | Mul deriving Show
lexer :: Tok.TokenParser ()
lexer = Tok.makeTokenParser style
where ops = ["->","\\","+","*","-","="]
style = haskellStyle {Tok.reservedOpNames = ops }
reservedOp :: String -> Parser ()
reservedOp = Tok.reservedOp lexer
identifier :: Parser String
identifier = Tok.identifier lexer
parens :: Parser a -> Parser a
parens = Tok.parens lexer
contents :: Parser a -> Parser a
contents p = do
Tok.whiteSpace lexer
r <- p
eof
return r
natural :: Parser Integer
natural = Tok.natural lexer
variable :: Parser Expr
variable = do
x <- identifier
return (Var x)
number :: Parser Expr
number = do
n <- natural
return (Num (fromIntegral n))
lambda :: Parser Expr
lambda = do
reservedOp "\\"
x <- identifier
reservedOp "->"
e <- expr
return (Lam x e)
aexp :: Parser Expr
aexp = parens expr
<|> variable
<|> number
<|> lambda
term :: Parser Expr
term = Ex.buildExpressionParser table aexp
where infixOp x f = Ex.Infix (reservedOp x >> return f)
table = [[infixOp "*" (Op Mul) Ex.AssocLeft],
[infixOp "+" (Op Add) Ex.AssocLeft]]
expr :: Parser Expr
expr = do
es <- many1 term
return (foldl1 App es)
parseExpr :: String -> Expr
parseExpr input =
case parse (contents expr) "<stdin>" input of
Left err -> error (show err)
Right ast -> ast
main :: IO ()
main = getLine >>= print . parseExpr >> main
```

Trying it out:

```
λ: runhaskell simpleparser.hs
1+2
Op Add (Num 1) (Num 2)
\i -> \x -> x
Lam "i" (Lam "x" (Var "x"))
\s -> \f -> \g -> \x -> f x (g x)
Lam "s" (Lam "f" (Lam "g" (Lam "x" (App (App (Var "f") (Var "x")) (App (Var "g") (Var "x"))))))
```

## Generic Parsing

Previously we defined generic operations for pretty printing and this begs the question of whether we can write a parser on top of Generics. The answer is generally yes, so long as there is a direct mapping between the specific lexemes and sum and products types. Consider the simplest case where we just read off the names of the constructors using the regular Generics machinery and then build a Parsec parser terms of them.

```
{-# LANGUAGE TypeOperators #-}
{-# LANGUAGE DeriveGeneric #-}
{-# LANGUAGE OverloadedStrings #-}
{-# LANGUAGE FlexibleInstances #-}
{-# LANGUAGE FlexibleContexts #-}
{-# LANGUAGE ScopedTypeVariables #-}
import Text.Parsec
import Text.Parsec.Text.Lazy
import Control.Applicative ((<*), (<*>), (<$>))
import GHC.Generics
class GParse f where
gParse :: Parser (f a)
-- Type synonym metadata for constructors
instance (GParse f, Constructor c) => GParse (C1 c f) where
gParse =
let con = conName (undefined :: t c f a) in
(fmap M1 gParse) <* string con
-- Constructor names
instance GParse f => GParse (D1 c f) where
gParse = fmap M1 gParse
-- Sum types
instance (GParse a, GParse b) => GParse (a :+: b) where
gParse = try (fmap L1 gParse <|> fmap R1 gParse)
-- Product types
instance (GParse f, GParse g) => GParse (f :*: g) where
gParse = (:*:) <$> gParse <*> gParse
-- Nullary constructors
instance GParse U1 where
gParse = return U1
data Scientist
= Newton
| Einstein
| Schrodinger
| Feynman
deriving (Show, Generic)
data Musician
= Vivaldi
| Bach
| Mozart
| Beethoven
deriving (Show, Generic)
gparse :: (Generic g, GParse (Rep g)) => Parser g
gparse = fmap to gParse
scientist :: Parser Scientist
scientist = gparse
musician :: Parser Musician
musician = gparse
```

```
λ: parseTest parseMusician "Bach"
Bach
λ: parseTest parseScientist "Feynman"
Feynman
```

With a little more work and an outer wrapper, this example an easily be extended to automate parsing of a simple recursive type.

```
{-# LANGUAGE TypeOperators #-}
{-# LANGUAGE DeriveGeneric #-}
{-# LANGUAGE OverloadedStrings #-}
{-# LANGUAGE FlexibleInstances #-}
{-# LANGUAGE FlexibleContexts #-}
{-# LANGUAGE ScopedTypeVariables #-}
{-# LANGUAGE DeriveAnyClass #-}
{-# LANGUAGE DefaultSignatures #-}
import Control.Applicative ((<*), (*>), (<*>), (<$>), pure)
import GHC.Generics
import Text.Parsec ((<|>), string, try, many1, digit, char, letter, spaces)
import Text.Parsec.Text.Lazy (Parser)
class GParse f where
gParse :: Parser (f a)
-- Types
instance (Parse a) => GParse (K1 R a) where
gParse = fmap K1 parse
-- Selector names
instance (GParse f, Selector s) => GParse (M1 S s f) where
gParse = fmap M1 gParse
-- Type synonym metadata for constructors
instance (GParse f, Constructor c) => GParse (C1 c f) where
gParse =
let con = conName (undefined :: t c f a) in
(spaces >> string con >> spaces) *> fmap M1 gParse
-- Constructor names
instance (Datatype d, GParse f) => GParse (D1 d f) where
gParse = fmap M1 gParse
-- Sum types
instance (GParse a, GParse b) => GParse (a :+: b) where
gParse = try (fmap L1 gParse) <|> try (fmap R1 gParse)
-- Product types
instance (GParse f, GParse g) => GParse (f :*: g) where
gParse = (:*:) <$> try gParse <*> try gParse
-- Nullary constructors
instance GParse U1 where
gParse = return U1
gparse :: (Generic g, GParse (Rep g)) => Parser g
gparse = fmap to gParse
class Parse a where
parse :: Parser a
default parse :: (Generic a, GParse (Rep a)) => Parser a
parse = spaces >> char '(' >> gparse >>= \e -> char ')' >> return e
instance Parse Integer where
parse = rd <$> (plus <|> minus <|> number)
where rd = read :: String -> Integer
plus = char '+' *> number
minus = (:) <$> char '-' <*> number
number = many1 digit
instance Parse String where
parse = many1 letter
type Name = String
data Exp
= Lit Integer
| Var Name
| Plus Exp Exp
| App Exp Exp
| Abs Name Exp deriving (Show, Generic, Parse)
expr :: Parser Exp
expr = parse
```

```
λ: parseTest expr "(App (Plus (Lit 1) (Var n)) (App (Plus (Lit 5) (Lit 5)) (Plus (Lit 6) (Lit 6))))"
App (Plus (Lit 1) (Var "n")) (App (Plus (Lit 5) (Lit 5)) (Plus (Lit 6) (Lit 6)))
```

## Attoparsec

Attoparsec is a parser combinator like Parsec but more suited for bulk parsing of large text and binary files instead of parsing language syntax to ASTs. When written properly Attoparsec parsers can be efficient.

One notable distinction between Parsec and Attoparsec is that backtracking operator (`try`

) is not present and reflects on attoparsec's different underlying parser model.

For a simple little lambda calculus language we can use attoparsec much in the same we used parsec:

```
{-# LANGUAGE OverloadedStrings #-}
{-# OPTIONS_GHC -fno-warn-unused-do-bind #-}
import Control.Applicative
import Data.Attoparsec.Text
import qualified Data.Text as T
import qualified Data.Text.IO as T
import Data.List (foldl1')
data Name
= Gen Int
| Name T.Text
deriving (Eq, Show, Ord)
data Expr
= Var Name
| App Expr Expr
| Lam [Name] Expr
| Lit Int
| Prim PrimOp
deriving (Eq, Show)
data PrimOp
= Add
| Sub
| Mul
| Div
deriving (Eq, Show)
data Defn = Defn Name Expr
deriving (Eq, Show)
name :: Parser Name
name = Name . T.pack <$> many1 letter
num :: Parser Expr
num = Lit <$> signed decimal
var :: Parser Expr
var = Var <$> name
lam :: Parser Expr
lam = do
string "\\"
vars <- many1 (skipSpace *> name)
skipSpace *> string "->"
body <- expr
return (Lam vars body)
eparen :: Parser Expr
eparen = char '(' *> expr <* skipSpace <* char ')'
prim :: Parser Expr
prim = Prim <$> (
char '+' *> return Add
<|> char '-' *> return Sub
<|> char '*' *> return Mul
<|> char '/' *> return Div)
expr :: Parser Expr
expr = foldl1' App <$> many1 (skipSpace *> atom)
atom :: Parser Expr
atom = try lam
<|> eparen
<|> prim
<|> var
<|> num
def :: Parser Defn
def = do
skipSpace
nm <- name
skipSpace *> char '=' *> skipSpace
ex <- expr
skipSpace <* char ';'
return $ Defn nm ex
file :: T.Text -> Either String [Defn]
file = parseOnly (many def <* skipSpace)
parseFile :: FilePath -> IO (Either T.Text [Defn])
parseFile path = do
contents <- T.readFile path
case file contents of
Left a -> return $ Left (T.pack a)
Right b -> return $ Right b
main :: IO (Either T.Text [Defn])
main = parseFile "simple.ml"
```

For an example try the above parser with the following simple lambda expression.

```
f = g (x - 1);
g = f (x + 1);
h = \x y -> (f x) + (g y);
```

Attoparsec adapts very well to binary and network protocol style parsing as well, this is extracted from a small implementation of a distributed consensus network protocol:

```
{-# LANGUAGE OverloadedStrings #-}
import Control.Monad
import Data.Attoparsec
import Data.Attoparsec.Char8 as A
import Data.ByteString.Char8
data Action
= Success
| KeepAlive
| NoResource
| Hangup
| NewLeader
| Election
deriving Show
type Sender = ByteString
type Payload = ByteString
data Message = Message
{ action :: Action
, sender :: Sender
, payload :: Payload
} deriving Show
proto :: Parser Message
proto = do
act <- paction
send <- A.takeTill (== '.')
body <- A.takeTill (A.isSpace)
endOfLine
return $ Message act send body
paction :: Parser Action
paction = do
c <- anyWord8
case c of
1 -> return Success
2 -> return KeepAlive
3 -> return NoResource
4 -> return Hangup
5 -> return NewLeader
6 -> return Election
_ -> mzero
main :: IO ()
main = do
let msgtext = "\x01\x6c\x61\x70\x74\x6f\x70\x2e\x33\x2e\x31\x34\x31\x35\x39\x32\x36\x35\x33\x35\x0A"
let msg = parseOnly proto msgtext
print msg
```

## Optparse Applicative

Optparse-applicative is a combinator library for building command line interfaces that take in various user flags, commmands and switches and map them into Haskell data structures that can handle the input. The main interface is through the applicative functor `Parser`

and various combinators such as `strArgument`

and `flag`

which populate the option parsing table with some monadic action which returns a Haskell value. The resulting sequence of values can be combined applicatively into a larger Config data structure that holds all the given options. The `--help`

header is also automatically generated from the combinators.

```
./optparse
Usage: optparse.hs [filename...] [--quiet] [--cheetah]
Available options:
-h,--help Show this help text
filename... Input files
--quiet Whether to shut up.
--cheetah Perform task quickly.
```

```
import Data.List
import Data.Monoid
import Options.Applicative
data Opts = Opts
{ _files :: [String]
, _quiet :: Bool
, _fast :: Speed
}
data Speed = Slow | Fast
options :: Parser Opts
options = Opts <$> filename <*> quiet <*> fast
where
filename :: Parser [String]
filename = many $ argument str $
metavar "filename..."
<> help "Input files"
fast :: Parser Speed
fast = flag Slow Fast $
long "cheetah"
<> help "Perform task quickly."
quiet :: Parser Bool
quiet = switch $
long "quiet"
<> help "Whether to shut up."
greet :: Opts -> IO ()
greet (Opts files quiet fast) = do
putStrLn "reading these files:"
mapM_ print files
case fast of
Fast -> putStrLn "quickly"
Slow -> putStrLn "slowly"
case quiet of
True -> putStrLn "quietly"
False -> putStrLn "loudly"
opts :: ParserInfo Opts
opts = info (helper <*> options) fullDesc
main :: IO ()
main = execParser opts >>= greet
```

See: Optparse Applicative Tutorial

## Happy & Alex

Happy is a parser generator system for Haskell, similar to the tool `yacc' for C. It works as a preprocessor with it's own syntax that generates a parse table from two specifications, a lexer file and parser file. Happy does not have the same underlying parser implementation as parser combinators and can effectively work with left-recursive grammars without explicit factorization. It can also easily be modified to track position information for tokens and handle offside parsing rules for indentation-sensitive grammars. Happy is used in GHC itself for Haskell's grammar.

- Lexer.x
- Parser.y

Running the standalone commands will generate the Haskell source for the modules.

```
$ alex Lexer.x -o Lexer.hs
$ happy Parser.y -o Parser.hs
```

The generated modules are not human readable generally and unfortunatly error messages are given in the Haskell source, not the Happy source.

#### Lexer

For instance we could define a little toy lexer with a custom set of tokens.

```
{
module Lexer (
Token(..),
scanTokens
) where
import Syntax
}
%wrapper "basic"
$digit = 0-9
$alpha = [a-zA-Z]
$eol = [\n]
tokens :-
-- Whitespace insensitive
$eol ;
$white+ ;
print { \s -> TokenPrint }
$digit+ { \s -> TokenNum (read s) }
\= { \s -> TokenEq }
$alpha [$alpha $digit \_ \']* { \s -> TokenSym s }
{
data Token
= TokenNum Int
| TokenSym String
| TokenPrint
| TokenEq
| TokenEOF
deriving (Eq,Show)
scanTokens = alexScanTokens
}
```

#### Parser

The associated parser is list of a production rules and a monad to running the parser in. Production rules consist of a set of options on the left and generating Haskell expressions on the right with indexed metavariables (`$1`

, `$2`

, ...) mapping to the ordered terms on the left (i.e. in the second term `term`

~ `$1`

, `term`

~ `$2`

).

```
terms
: term { [$1] }
| term terms { $1 : $2 }
```

```
{
{-# LANGUAGE GeneralizedNewtypeDeriving #-}
module Parser (
parseExpr,
) where
import Lexer
import Syntax
import Control.Monad.Except
}
%name expr
%tokentype { Token }
%monad { Except String } { (>>=) } { return }
%error { parseError }
%token
int { TokenNum $$ }
var { TokenSym $$ }
print { TokenPrint }
'=' { TokenEq }
%%
terms
: term { [$1] }
| term terms { $1 : $2 }
term
: var { Var $1 }
| var '=' int { Assign $1 $3 }
| print term { Print $2 }
{
parseError :: [Token] -> Except String a
parseError (l:ls) = throwError (show l)
parseError [] = throwError "Unexpected end of Input"
parseExpr :: String -> Either String [Expr]
parseExpr input =
let tokenStream = scanTokens input in
runExcept (expr tokenStream)
}
```

As a simple input consider the following simple program.

```
x = 4
print x
y = 5
print y
y = 6
print y
```

## Configurator

```
{-# LANGUAGE OverloadedStrings #-}
import Data.Text
import qualified Data.Configurator as C
data Config = Config
{ verbose :: Bool
, loggingLevel :: Int
, logfile :: FilePath
, dbHost :: Text
, dbUser :: Text
, dbDatabase :: Text
, dbpassword :: Maybe Text
} deriving (Eq, Show)
readConfig :: FilePath -> IO Config
readConfig cfgFile = do
cfg <- C.load [C.Required cfgFile]
verbose <- C.require cfg "logging.verbose"
loggingLevel <- C.require cfg "logging.loggingLevel"
logFile <- C.require cfg "logging.logfile"
hostname <- C.require cfg "database.hostname"
username <- C.require cfg "database.username"
database <- C.require cfg "database.database"
password <- C.lookup cfg "database.password"
return $ Config verbose loggingLevel logFile hostname username database password
main :: IO ()
main = do
cfg <-readConfig "example.config"
print cfg
```

```
logging
{
verbose = true
logfile = "/tmp/app.log"
loggingLevel = 3
}
database
{
hostname = "us-east-1.rds.amazonaws.com"
username = "app"
database = "booktown"
password = "hunter2"
}
```

# Streaming

## Lazy IO

The problem with using the usual monadic approach to processing data accumulated through IO is that the Prelude tools require us to manifest large amounts of data in memory all at once before we can even begin computation.

```
mapM :: Monad m => (a -> m b) -> [a] -> m [b]
sequence :: Monad m => [m a] -> m [a]
```

Reading from the file creates a thunk for the string that forced will then read the file. The problem is then that this method ties the ordering of IO effects to evaluation order which is difficult to reason about in the large.

Consider that normally the monad laws ( in the absence of `seq`

) guarantee that these computations should be identical. But using lazy IO we can construct a degenerate case.

```
import System.IO
main :: IO ()
main = do
withFile "foo.txt" ReadMode $ \fd -> do
contents <- hGetContents fd
print contents
-- "foo\n"
contents <- withFile "foo.txt" ReadMode hGetContents
print contents
-- ""
```

So what we need is a system to guarantee deterministic resource handling with constant memory usage. To that end both the Conduits and Pipes libraries solved this problem using different ( though largely equivalent ) approaches.

## Pipes

```
await :: Monad m => Pipe a y m a
yield :: Monad m => a -> Pipe x a m ()
(>->) :: Monad m
=> Pipe a b m r
-> Pipe b c m r
-> Pipe a c m r
runEffect :: Monad m => Effect m r -> m r
toListM :: Monad m => Producer a m () -> m [a]
```

Pipes is a stream processing library with a strong emphasis on the static semantics of composition. The simplest usage is to connect "pipe" functions with a `(>->)`

composition operator, where each component can `await`

and `yield`

to push and pull values along the stream.

```
import Pipes
import Pipes.Prelude as P
import Control.Monad
import Control.Monad.Identity
a :: Producer Int Identity ()
a = forM_ [1..10] yield
b :: Pipe Int Int Identity ()
b = forever $ do
x <- await
yield (x*2)
yield (x*3)
yield (x*4)
c :: Pipe Int Int Identity ()
c = forever $ do
x <- await
if (x `mod` 2) == 0
then yield x
else return ()
result :: [Int]
result = P.toList $ a >-> b >-> c
```

For example we could construct a "FizzBuzz" pipe.

```
{-# LANGUAGE MultiWayIf #-}
import Pipes
import qualified Pipes.Prelude as P
count :: Producer Integer IO ()
count = each [1..100]
fizzbuzz :: Pipe Integer String IO ()
fizzbuzz = do
n <- await
if | n `mod` 15 == 0 -> yield "FizzBuzz"
| n `mod` 5 == 0 -> yield "Fizz"
| n `mod` 3 == 0 -> yield "Buzz"
| otherwise -> return ()
fizzbuzz
main :: IO ()
main = runEffect $ count >-> fizzbuzz >-> P.stdoutLn
```

To continue with the degenerate case we constructed with Lazy IO, consider than we can now compose and sequence deterministic actions over files without having to worry about effect order.

```
import Pipes
import Pipes.Prelude as P
import System.IO
readF :: FilePath -> Producer String IO ()
readF file = do
lift $ putStrLn $ "Opened" ++ file
h <- lift $ openFile file ReadMode
fromHandle h
lift $ putStrLn $ "Closed" ++ file
lift $ hClose h
main :: IO ()
main = runEffect $ readF "foo.txt" >-> P.take 3 >-> stdoutLn
```

This is simple a sampling of the functionality of pipes. The documentation for pipes is extensive and great deal of care has been taken make the library extremely thorough. `pipes`

is a shining example of an accessible yet category theoretic driven design.

See: Pipes Tutorial

## Safe Pipes

`bracket :: MonadSafe m => Base m a -> (a -> Base m b) -> (a -> m c) -> m c`

As a motivating example, ZeroMQ is a network messaging library that abstracts over traditional Unix sockets to a variety of network topologies. Most notably it isn't designed to guarantee any sort of transactional guarantees for delivery or recovery in case of errors so it's necessary to design a layer on top of it to provide the desired behavior at the application layer.

In Haskell we'd like to guarantee that if we're polling on a socket we get messages delivered in a timely fashion or consider the resource in an error state and recover from it. Using `pipes-safe`

we can manage the life cycle of lazy IO resources and can safely handle failures, resource termination and finalization gracefully. In other languages this kind of logic would be smeared across several places, or put in some global context and prone to introduce errors and subtle race conditions. Using pipes we instead get a nice tight abstraction designed exactly to fit this kind of use case.

For instance now we can bracket the ZeroMQ socket creation and finalization within the `SafeT`

monad transformer which guarantees that after successful message delivery we execute the pipes function as expected, or on failure we halt the execution and finalize the socket.

```
import Pipes
import Pipes.Safe
import qualified Pipes.Prelude as P
import System.Timeout (timeout)
import Data.ByteString.Char8
import qualified System.ZMQ as ZMQ
data Opts = Opts
{ _addr :: String -- ^ ZMQ socket address
, _timeout :: Int -- ^ Time in milliseconds for socket timeout
}
recvTimeout :: Opts -> ZMQ.Socket a -> Producer ByteString (SafeT IO) ()
recvTimeout opts sock = do
body <- liftIO $ timeout (_timeout opts) (ZMQ.receive sock [])
case body of
Just msg -> do
liftIO $ ZMQ.send sock msg []
yield msg
recvTimeout opts sock
Nothing -> liftIO $ print "socket timed out"
collect :: ZMQ.Context
-> Opts
-> Producer ByteString (SafeT IO) ()
collect ctx opts = bracket zinit zclose (recvTimeout opts)
where
-- Initialize the socket
zinit = do
liftIO $ print "waiting for messages"
sock <- ZMQ.socket ctx ZMQ.Rep
ZMQ.bind sock (_addr opts)
return sock
-- On timeout or completion guarantee the socket get closed.
zclose sock = do
liftIO $ print "finalizing"
ZMQ.close sock
runZmq :: ZMQ.Context -> Opts -> IO ()
runZmq ctx opts = runSafeT $ runEffect $
collect ctx opts >-> P.take 10 >-> P.print
main :: IO ()
main = do
ctx <- ZMQ.init 1
let opts = Opts {_addr = "tcp://127.0.0.1:8000", _timeout = 1000000 }
runZmq ctx opts
ZMQ.term ctx
```

## Conduits

```
await :: Monad m => ConduitM i o m (Maybe i)
yield :: Monad m => o -> ConduitM i o m ()
($$) :: Monad m => Source m a -> Sink a m b -> m b
(=$) :: Monad m => Conduit a m b -> Sink b m c -> Sink a m c
type Sink i = ConduitM i Void
type Source m o = ConduitM () o m ()
type Conduit i m o = ConduitM i o m ()
```

Conduits are conceptually similar though philosophically different approach to the same problem of constant space deterministic resource handling for IO resources.

The first initial difference is that await function now returns a `Maybe`

which allows different handling of termination. The composition operators are also split into a connecting operator (`$$`

) and a fusing operator (`=$`

) for combining Sources and Sink and a Conduit and a Sink respectively.

```
{-# LANGUAGE MultiWayIf #-}
import Data.Conduit
import Control.Monad.Trans
import qualified Data.Conduit.List as CL
source :: Source IO Int
source = CL.sourceList [1..100]
conduit :: Conduit Int IO String
conduit = do
val <- await
liftIO $ print val
case val of
Nothing -> return ()
Just n -> do
if | n `mod` 15 == 0 -> yield "FizzBuzz"
| n `mod` 5 == 0 -> yield "Fizz"
| n `mod` 3 == 0 -> yield "Buzz"
| otherwise -> return ()
conduit
sink :: Sink String IO ()
sink = CL.mapM_ putStrLn
main :: IO ()
main = source $$ conduit =$ sink
```

See: Conduit Overview

# Cryptography

## cryptonite

## entropy

## memory

## crypto-pubkey

## crypto-api

## x509

## ed25519

# Compression

## lz4

## zlib

# Data Formats

## JSON

Aeson is library for efficient parsing and generating JSON. It is the canonical JSON library for handling JSON.

```
decode :: FromJSON a => ByteString -> Maybe a
encode :: ToJSON a => a -> ByteString
eitherDecode :: FromJSON a => ByteString -> Either String a
fromJSON :: FromJSON a => Value -> Result a
toJSON :: ToJSON a => a -> Value
```

A point of some subtlety to beginners is that the return types for Aeson functions are **polymorphic in their return types** meaning that the resulting type of decode is specified only in the context of your programs use of the decode function. So if you use decode in a point your program and bind it to a value `x`

and then use `x`

as if it were and integer throughout the rest of your program, Aeson will select the typeclass instance which parses the given input string into a Haskell integer.

#### Value

Aeson uses several high performance data structures (Vector, Text, HashMap) by default instead of the naive versions so typically using Aeson will require that us import them and use `OverloadedStrings`

when indexing into objects.

The underlying Aeson structure is called `Value`

and encodes a recursive tree structure that models the semantics of untyped JSON objects by mapping them onto a large sum type which embodies all possible JSON values.

```
type Object = HashMap Text Value
type Array = Vector Value
-- | A JSON value represented as a Haskell value.
data Value
= Object !Object
| Array !Array
| String !Text
| Number !Scientific
| Bool !Bool
| Null
```

For instance the Value expansion of the following JSON blob:

```
{
"a": [1,2,3],
"b": 1
}
```

Is represented in Aeson as the `Value`

:

```
Object
(fromList
[ ( "a"
, Array (fromList [ Number 1.0 , Number 2.0 , Number 3.0 ])
)
, ( "b" , Number 1.0 )
])
```

Let's consider some larger examples, we'll work with this contrived example JSON:

```
{
"id": 1,
"name": "A green door",
"price": 12.50,
"tags": ["home", "green"],
"refs": {
"a": "red",
"b": "blue"
}
}
```

#### Unstructured JSON

In dynamic scripting languages it's common to parse amorphous blobs of JSON without any a priori structure and then handle validation problems by throwing exceptions while traversing it. We can do the same using Aeson and the Maybe monad.

```
{-# LANGUAGE OverloadedStrings #-}
import Data.Text
import Data.Aeson
import Data.Vector
import qualified Data.HashMap.Strict as M
import qualified Data.ByteString.Lazy as BL
-- Pull a key out of an JSON object.
(^?) :: Value -> Text -> Maybe Value
(^?) (Object obj) k = M.lookup k obj
(^?) _ _ = Nothing
-- Pull the ith value out of a JSON list.
ix :: Value -> Int -> Maybe Value
ix (Array arr) i = arr !? i
ix _ _ = Nothing
readJSON str = do
obj <- decode str
price <- obj ^? "price"
refs <- obj ^? "refs"
tags <- obj ^? "tags"
aref <- refs ^? "a"
tag1 <- tags `ix` 0
return (price, aref, tag1)
main :: IO ()
main = do
contents <- BL.readFile "example.json"
print $ readJSON contents
```

#### Structured JSON

This isn't ideal since we've just smeared all the validation logic across our traversal logic instead of separating concerns and handling validation in separate logic. We'd like to describe the structure before-hand and the invalid case separately. Using Generic also allows Haskell to automatically write the serializer and deserializer between our datatype and the JSON string based on the names of record field names.

```
{-# LANGUAGE DeriveGeneric #-}
import Data.Text
import Data.Aeson
import GHC.Generics
import qualified Data.ByteString.Lazy as BL
import Control.Applicative
data Refs = Refs
{ a :: Text
, b :: Text
} deriving (Show,Generic)
data Data = Data
{ id :: Int
, name :: Text
, price :: Int
, tags :: [Text]
, refs :: Refs
} deriving (Show,Generic)
instance FromJSON Data
instance FromJSON Refs
instance ToJSON Data
instance ToJSON Refs
main :: IO ()
main = do
contents <- BL.readFile "example.json"
let Just dat = decode contents
print $ name dat
print $ a (refs dat)
```

Now we get our validated JSON wrapped up into a nicely typed Haskell ADT.

```
Data
{ id = 1
, name = "A green door"
, price = 12
, tags = [ "home" , "green" ]
, refs = Refs { a = "red" , b = "blue" }
}
```

The functions `fromJSON`

and `toJSON`

can be used to convert between this sum type and regular Haskell types with.

`data Result a = Error String | Success a`

```
λ: fromJSON (Bool True) :: Result Bool
Success True
λ: fromJSON (Bool True) :: Result Double
Error "when expecting a Double, encountered Boolean instead"
```

As of 7.10.2 we can use the new -XDeriveAnyClass to automatically derive instances of FromJSON and TOJSON without the need for standalone instance declarations. These are implemented entirely in terms of the default methods which use Generics under the hood.

```
{-# LANGUAGE DeriveGeneric #-}
{-# LANGUAGE DeriveAnyClass #-}
import Data.Text
import Data.Aeson
import GHC.Generics
import qualified Data.ByteString.Lazy as BL
data Refs = Refs
{ a :: Text
, b :: Text
} deriving (Show,Generic,FromJSON,ToJSON)
data Data = Data
{ id :: Int
, name :: Text
, price :: Int
, tags :: [Text]
, refs :: Refs
} deriving (Show,Generic,FromJSON,ToJSON)
main :: IO ()
main = do
contents <- BL.readFile "example.json"
let Just dat = decode contents
print $ name dat
print $ a (refs dat)
BL.putStrLn $ encode dat
```

#### Hand Written Instances

While it's useful to use generics to derive instances, sometimes you actually want more fine grained control over serialization and de serialization. So we fall back on writing ToJSON and FromJSON instances manually. Using FromJSON we can project into hashmap using the `(.:)`

operator to extract keys. If the key fails to exist the parser will abort with a key failure message. The ToJSON instances can never fail and simply require us to pattern match on our custom datatype and generate an appropriate value.

The law that the FromJSON and ToJSON classes should maintain is that `encode . decode`

and `decode . encode`

should map to the same object. Although in practice there many times when we break this rule and especially if the serialize or de serialize is one way.

```
{-# LANGUAGE OverloadedStrings #-}
{-# LANGUAGE ScopedTypeVariables #-}
import Data.Text
import Data.Aeson
import Data.Maybe
import Data.Aeson.Types
import Control.Applicative
import qualified Data.ByteString.Lazy as BL
data Crew = Crew
{ name :: Text
, rank :: Rank
} deriving (Show)
data Rank
= Captain
| Ensign
| Lieutenant
deriving (Show)
-- Custom JSON Deserializer
instance FromJSON Crew where
parseJSON (Object o) = do
_name <- o .: "name"
_rank <- o .: "rank"
pure (Crew _name _rank)
instance FromJSON Rank where
parseJSON (String s) = case s of
"Captain" -> pure Captain
"Ensign" -> pure Ensign
"Lieutenant" -> pure Lieutenant
_ -> typeMismatch "Could not parse Rank" (String s)
parseJSON x = typeMismatch "Expected String" x
-- Custom JSON Serializer
instance ToJSON Crew where
toJSON (Crew name rank) = object [
"name" .= name
, "rank" .= rank
]
instance ToJSON Rank where
toJSON Captain = String "Captain"
toJSON Ensign = String "Ensign"
toJSON Lieutenant = String "Lieutenant"
roundTrips :: Crew -> Bool
roundTrips = isJust . go
where
go :: Crew -> Maybe Crew
go = decode . encode
picard :: Crew
picard = Crew { name = "Jean-Luc Picard", rank = Captain }
main :: IO ()
main = do
contents <- BL.readFile "crew.json"
let (res :: Maybe Crew) = decode contents
print res
print $ roundTrips picard
```

See: Aeson Documentation

## Yaml

Yaml is a textual serialization format similar to JSON. It uses an indentation sensitive structure to encode nested maps of keys and values. The Yaml interface for Haskell is a precise copy of `Data.Aeson`

```
invoice: 34843
date : 2001-01-23
bill:
given : Chris
family : Dumars
address:
lines: |
458 Walkman Dr.
Suite #292
city : Royal Oak
state : MI
postal : 48046
```

```
Object
(fromList
[ ( "invoice" , Number 34843.0 )
, ( "date" , String "2001-01-23" )
, ( "bill-to"
, Object
(fromList
[ ( "address"
, Object
(fromList
[ ( "state" , String "MI" )
, ( "lines" , String "458 Walkman Dr.\nSuite #292\n" )
, ( "city" , String "Royal Oak" )
, ( "postal" , Number 48046.0 )
])
)
, ( "family" , String "Dumars" )
, ( "given" , String "Chris" )
])
)
])
```

To parse this file we use the following datatypes and functions:

```
{-# LANGUAGE DeriveGeneric #-}
{-# LANGUAGE DeriveAnyClass #-}
{-# LANGUAGE ScopedTypeVariables #-}
import Data.Yaml
import Data.Text (Text)
import qualified Data.ByteString as BL
import GHC.Generics
data Invoice = Invoice
{ invoice :: Int
, date :: Text
, bill :: Billing
} deriving (Show,Generic,FromJSON)
data Billing = Billing
{ address :: Address
, family :: Text
, given :: Text
} deriving (Show,Generic,FromJSON)
data Address = Address
{ lines :: Text
, city :: Text
, state :: Text
, postal :: Int
} deriving (Show,Generic,FromJSON)
main :: IO ()
main = do
contents <- BL.readFile "example.yaml"
let (res :: Either String Invoice) = decodeEither contents
case res of
Right val -> print val
Left err -> putStrLn err
```

Which generates:

```
Invoice
{ invoice = 34843
, date = "2001-01-23"
, bill =
Billing
{ address =
Address
{ lines = "458 Walkman Dr.\nSuite #292\n"
, city = "Royal Oak"
, state = "MI"
, postal = 48046
}
, family = "Dumars"
, given = "Chris"
}
}
```

## CSV

Cassava is an efficient CSV parser library. We'll work with this tiny snippet from the iris dataset:

```
sepal_length,sepal_width,petal_length,petal_width,plant_class
5.1,3.5,1.4,0.2,Iris-setosa
5.0,2.0,3.5,1.0,Iris-versicolor
6.3,3.3,6.0,2.5,Iris-virginica
```

#### Unstructured CSV

Just like with Aeson if we really want to work with unstructured data the library accommodates this.

```
import Data.Csv
import Text.Show.Pretty
import qualified Data.Vector as V
import qualified Data.ByteString.Lazy as BL
type ErrorMsg = String
type CsvData = V.Vector (V.Vector BL.ByteString)
example :: FilePath -> IO (Either ErrorMsg CsvData)
example fname = do
contents <- BL.readFile fname
return $ decode NoHeader contents
```

We see we get the nested set of stringy vectors:

```
[ [ "sepal_length"
, "sepal_width"
, "petal_length"
, "petal_width"
, "plant_class"
]
, [ "5.1" , "3.5" , "1.4" , "0.2" , "Iris-setosa" ]
, [ "5.0" , "2.0" , "3.5" , "1.0" , "Iris-versicolor" ]
, [ "6.3" , "3.3" , "6.0" , "2.5" , "Iris-virginica" ]
]
```

#### Structured CSV

Just like with Aeson we can use Generic to automatically write the deserializer between our CSV data and our custom datatype.

```
{-# LANGUAGE OverloadedStrings #-}
{-# LANGUAGE DeriveGeneric #-}
import Data.Csv
import GHC.Generics
import qualified Data.Vector as V
import qualified Data.ByteString.Lazy as BL
data Plant = Plant
{ sepal_length :: Double
, sepal_width :: Double
, petal_length :: Double
, petal_width :: Double
, plant_class :: String
} deriving (Generic, Show)
instance FromNamedRecord Plant
instance ToNamedRecord Plant
type ErrorMsg = String
type CsvData = (Header, V.Vector Plant)
parseCSV :: FilePath -> IO (Either ErrorMsg CsvData)
parseCSV fname = do
contents <- BL.readFile fname
return $ decodeByName contents
main = parseCSV "iris.csv" >>= print
```

And again we get a nice typed ADT as a result.

```
[ Plant
{ sepal_length = 5.1
, sepal_width = 3.5
, petal_length = 1.4
, petal_width = 0.2
, plant_class = "Iris-setosa"
}
, Plant
{ sepal_length = 5.0
, sepal_width = 2.0
, petal_length = 3.5
, petal_width = 1.0
, plant_class = "Iris-versicolor"
}
, Plant
{ sepal_length = 6.3
, sepal_width = 3.3
, petal_length = 6.0
, petal_width = 2.5
, plant_class = "Iris-virginica"
}
]
```

# Network & Web Programming

## HTTP

Haskell has a variety of HTTP request and processing libraries. The simplest and most flexible is the HTTP library.

```
{-# LANGUAGE OverloadedStrings #-}
import Network.HTTP.Types
import Network.HTTP.Client
import Control.Applicative
import Control.Concurrent.Async
type URL = String
get :: Manager -> URL -> IO Int
get m url = do
req <- parseUrl url
statusCode <$> responseStatus <$> httpNoBody req m
single :: IO Int
single = do
withManager defaultManagerSettings $ \m -> do
get m "http://haskell.org"
parallel :: IO [Int]
parallel = do
withManager defaultManagerSettings $ \m -> do
-- Fetch w3.org 10 times concurrently
let urls = replicate 10 "http://www.w3.org"
mapConcurrently (get m) urls
main :: IO ()
main = do
print =<< single
print =<< parallel
```

## Blaze

Blaze is an HTML combinator library that provides that capacity to build composable bits of HTML programmatically. It doesn't string templating libraries like Hastache but instead provides an API for building up HTML documents from logic where the format out of the output is generated procedurally.

For sequencing HTML elements the elements can either be sequenced in a monad or with monoid operations.

```
{-# LANGUAGE OverloadedStrings #-}
module Html where
import Text.Blaze.Html5
import Text.Blaze.Html.Renderer.Text
import qualified Data.Text.Lazy.IO as T
example :: Html
example = do
h1 "First header"
p $ ul $ mconcat [li "First", li "Second"]
main :: IO ()
main = do
T.putStrLn $ renderHtml example
```

For custom datatypes we can implement the `ToMarkup`

class to convert between Haskell data structures and HTML representation.

```
{-# LANGUAGE RecordWildCards #-}
{-# LANGUAGE OverloadedStrings #-}
module Html where
import Text.Blaze.Html5
import Text.Blaze.Html.Renderer.Text
import qualified Data.Text.Lazy as T
import qualified Data.Text.Lazy.IO as T
data Employee = Employee
{ name :: T.Text
, age :: Int
}
instance ToMarkup Employee where
toMarkup Employee {..} = ul $ mconcat
[ li (toHtml name)
, li (toHtml age)
]
fred :: Employee
fred = Employee { name = "Fred", age = 35 }
main :: IO ()
main = do
T.putStrLn $ renderHtml (toHtml fred)
```

## Warp

Warp is a efficient web server, it's the backend request engine behind several of popular Haskell web frameworks. The internals have been finely tuned to utilize Haskell's concurrent runtime and is capable of handling a great deal of concurrent requests.

```
{-# LANGUAGE OverloadedStrings #-}
import Network.Wai
import Network.Wai.Handler.Warp (run)
import Network.HTTP.Types
app :: Application
app req = return $ responseLBS status200 [] "Engage!"
main :: IO ()
main = run 8000 app
```

See: Warp

## Scotty

Continuing with our trek through web libraries, Scotty is a web microframework similar in principle to Flask in Python or Sinatra in Ruby.

```
{-# LANGUAGE OverloadedStrings #-}
import Web.Scotty
import qualified Text.Blaze.Html5 as H
import Text.Blaze.Html5 (toHtml, Html)
import Text.Blaze.Html.Renderer.Text (renderHtml)
greet :: String -> Html
greet user = H.html $ do
H.head $
H.title "Welcome!"
H.body $ do
H.h1 "Greetings!"
H.p ("Hello " >> toHtml user >> "!")
app = do
get "/" $
text "Home Page"
get "/greet/:name" $ do
name <- param "name"
html $ renderHtml (greet name)
main :: IO ()
main = scotty 8000 app
```

Of importance to note is the Blaze library used here overloads do-notation but is not itself a proper monad so the various laws and invariants that normally apply for monads may break down or fail with error terms.

See: Making a Website with Haskell

## Servant

## Hastache

Hastache is string templating based on the "Mustache" style of encoding metavariables with double braces `{{ x }}`

. Hastache supports automatically converting many Haskell types into strings and uses the efficient Text functions for formatting.

The variables loaded into the template are specified in either a function mapping variable names to printable MuType values. For instance using a function.

```
{-# LANGUAGE OverloadedStrings #-}
import Text.Hastache
import Text.Hastache.Context
import qualified Data.Text as T
import qualified Data.Text.Lazy as TL
import qualified Data.Text.Lazy.IO as TL
import Data.Data
template :: FilePath -> MuContext IO -> IO TL.Text
template = hastacheFile defaultConfig
-- Function strContext
context :: String -> MuType IO
context "body" = MuVariable ("Hello World" :: TL.Text)
context "title" = MuVariable ("Haskell is lovely" :: TL.Text)
context _ = MuVariable ()
main :: IO ()
main = do
output <- template "templates/home.html" (mkStrContext context)
TL.putStrLn output
```

Or using Data-Typeable record and `mkGenericContext`

, the Haskell field names are converted into variable names.

```
{-# LANGUAGE OverloadedStrings #-}
{-# LANGUAGE DeriveDataTypeable #-}
import Text.Hastache
import Text.Hastache.Context
import qualified Data.Text.Lazy as TL
import qualified Data.Text.Lazy.IO as TL
import Data.Data
template :: FilePath -> MuContext IO -> IO TL.Text
template = hastacheFile defaultConfig
-- Record context
data TemplateCtx = TemplateCtx
{ body :: TL.Text
, title :: TL.Text
} deriving (Data, Typeable)
main :: IO ()
main = do
let ctx = TemplateCtx { body = "Hello", title = "Haskell" }
output <- template "templates/home.html" (mkGenericContext ctx)
TL.putStrLn output
```

The MuType and MuContext types can be parameterized by any monad or transformer that implements `MonadIO`

, not just IO.

# Databases

## Postgres

Postgres is an object-relational database management system with a rich extension of the SQL standard. Consider the following tables specified in DDL.

```
CREATE TABLE "books" (
"id" integer NOT NULL,
"title" text NOT NULL,
"author_id" integer,
"subject_id" integer,
Constraint "books_id_pkey" Primary Key ("id")
);
CREATE TABLE "authors" (
"id" integer NOT NULL,
"last_name" text,
"first_name" text,
Constraint "authors_pkey" Primary Key ("id")
);
```

The postgresql-simple bindings provide a thin wrapper to various libpq commands to interact a Postgres server. These functions all take a `Connection`

object to the database instance and allow various bytestring queries to be sent and result sets mapped into Haskell datatypes. There are four primary functions for these interactions:

```
query_ :: FromRow r => Connection -> Query -> IO [r]
query :: (ToRow q, FromRow r) => Connection -> Query -> q -> IO [r]
execute :: ToRow q => Connection -> Query -> q -> IO Int64
execute_ :: Connection -> Query -> IO Int64
```

The result of the `query`

function is a list of elements which implement the FromRow typeclass. This can be many things including a single elemment (Only), a list of tuples where each element implements `FromField`

or a custom datatype that itself implements `FromRow`

. Under the hood the database bindings inspects the Postgres `oid`

objects and then attempts to convert them into the Haskell datatype of the field being scrutinised. This can fail at runtime if the types in the database don't align with the expected types in the logic executing the SQL query.

#### Tuples

```
{-# LANGUAGE OverloadedStrings #-}
{-# LANGUAGE ScopedTypeVariables #-}
import qualified Data.Text as T
import qualified Database.PostgreSQL.Simple as SQL
creds :: SQL.ConnectInfo
creds = SQL.defaultConnectInfo
{ SQL.connectUser = "example"
, SQL.connectPassword = "example"
, SQL.connectDatabase = "booktown"
}
selectBooks :: SQL.Connection -> IO [(Int, T.Text, Int)]
selectBooks conn = SQL.query_ conn "select id, title, author_id from books"
main :: IO ()
main = do
conn <- SQL.connect creds
books <- selectBooks conn
print books
```

This yields the result set:

```
[ ( 7808 , "The Shining" , 4156 )
, ( 4513 , "Dune" , 1866 )
, ( 4267 , "2001: A Space Odyssey" , 2001 )
, ( 1608 , "The Cat in the Hat" , 1809 )
, ( 1590 , "Bartholomew and the Oobleck" , 1809 )
, ( 25908 , "Franklin in the Dark" , 15990 )
, ( 1501 , "Goodnight Moon" , 2031 )
, ( 190 , "Little Women" , 16 )
, ( 1234 , "The Velveteen Rabbit" , 25041 )
, ( 2038 , "Dynamic Anatomy" , 1644 )
, ( 156 , "The Tell-Tale Heart" , 115 )
, ( 41473 , "Programming Python" , 7805 )
, ( 41477 , "Learning Python" , 7805 )
, ( 41478 , "Perl Cookbook" , 7806 )
, ( 41472 , "Practical PostgreSQL" , 1212 )
]
```

#### Custom Types

```
{-# LANGUAGE OverloadedStrings #-}
import qualified Data.Text as T
import qualified Database.PostgreSQL.Simple as SQL
import Database.PostgreSQL.Simple.FromRow (FromRow(..), field)
data Book = Book
{ id_ :: Int
, title :: T.Text
, author_id :: Int
} deriving (Show)
instance FromRow Book where
fromRow = Book <$> field <*> field <*> field
creds :: SQL.ConnectInfo
creds = SQL.defaultConnectInfo
{ SQL.connectUser = "example"
, SQL.connectPassword = "example"
, SQL.connectDatabase = "booktown"
}
selectBooks :: SQL.Connection -> IO [Book]
selectBooks conn = SQL.query_ conn "select id, title, author_id from books limit 4"
main :: IO ()
main = do
conn <- SQL.connect creds
books <- selectBooks conn
print books
```

This yields the result set:

```
[ Book { id_ = 7808 , title = "The Shining" , author_id = 4156 }
, Book { id_ = 4513 , title = "Dune" , author_id = 1866 }
, Book { id_ = 4267 , title = "2001: A Space Odyssey" , author_id = 2001 }
, Book { id_ = 1608 , title = "The Cat in the Hat" , author_id = 1809 }
]
```

#### Quasiquoter

As SQL expressions grow in complexity they often span multiple lines and sometimes its useful to just drop down to a quasiquoter to embed the whole query. The quoter here is pure, and just generates the `Query`

object behind as a ByteString.

```
{-# LANGUAGE QuasiQuotes #-}
{-# LANGUAGE OverloadedStrings #-}
{-# LANGUAGE ScopedTypeVariables #-}
import qualified Data.Text as T
import qualified Database.PostgreSQL.Simple as SQL
import Database.PostgreSQL.Simple.SqlQQ (sql)
import Database.PostgreSQL.Simple.FromRow (FromRow(..), field)
data Book = Book
{ id_ :: Int
, title :: T.Text
, first_name :: T.Text
, last_name :: T.Text
} deriving (Show)
instance FromRow Book where
fromRow = Book <$> field <*> field <*> field <*> field
creds :: SQL.ConnectInfo
creds = SQL.defaultConnectInfo
{ SQL.connectUser = "example"
, SQL.connectPassword = "example"
, SQL.connectDatabase = "booktown"
}
selectBooks :: SQL.Query
selectBooks = [sql|
select
books.id,
books.title,
authors.first_name,
authors.last_name
from books
join authors on
authors.id = books.author_id
limit 5
|]
main :: IO ()
main = do
conn <- SQL.connect creds
(books :: [Book]) <- SQL.query_ conn selectBooks
print books
```

This yields the result set:

```
[ Book
{ id_ = 41472
, title = "Practical PostgreSQL"
, first_name = "John"
, last_name = "Worsley"
}
, Book
{ id_ = 25908
, title = "Franklin in the Dark"
, first_name = "Paulette"
, last_name = "Bourgeois"
}
, Book
{ id_ = 1234
, title = "The Velveteen Rabbit"
, first_name = "Margery Williams"
, last_name = "Bianco"
}
, Book
{ id_ = 190
, title = "Little Women"
, first_name = "Louisa May"
, last_name = "Alcott"
}
]
```

## Redis

Redis is an in-memory key-value store with support for a variety of datastructures. The Haskell exposure is exposed in a `Redis`

monad which sequences a set of redis commands taking ByteString arguments and then executes them against a connection object.

```
{-# LANGUAGE OverloadedStrings #-}
import Database.Redis
import Data.ByteString.Char8
session :: Redis (Either Reply (Maybe ByteString))
session = do
set "hello" "haskell"
get "hello"
main :: IO ()
main = do
conn <- connect defaultConnectInfo
res <- runRedis conn session
print res
```

Redis is quite often used as a lightweight pubsub server, and the bindings integrate with the Haskell concurrency primitives so that listeners can be sparked and shared across threads off without blocking the main thread.

```
{-# LANGUAGE OverloadedStrings #-}
import Database.Redis
import Control.Monad
import Control.Monad.Trans
import Data.ByteString.Char8
import Control.Concurrent
subscriber :: Redis ()
subscriber =
pubSub (subscribe ["news"]) $ \msg -> do
print msg
return mempty
publisher :: Redis ()
publisher = forM_ [1..100] $ \n -> publish "news" (pack (show n))
-- connects to localhost:6379
main :: IO ()
main = do
conn1 <- connect defaultConnectInfo
conn2 <- connect defaultConnectInfo
-- Fork off a publisher
forkIO $ runRedis conn1 publisher
-- Subscribe for messages
runRedis conn2 subscriber
```

## Acid State

Acid-state allows us to build a "database" for around our existing Haskell datatypes that guarantees atomic transactions. For example, we can build a simple key-value store wrapped around the Map type.

```
{-# LANGUAGE TypeFamilies #-}
{-# LANGUAGE TemplateHaskell #-}
{-# LANGUAGE DeriveDataTypeable #-}
import Data.Acid
import Data.Typeable
import Data.SafeCopy
import Control.Monad.Reader (ask)
import qualified Data.Map as Map
import qualified Control.Monad.State as S
type Key = String
type Value = String
data Database = Database !(Map.Map Key Value)
deriving (Show, Ord, Eq, Typeable)
$(deriveSafeCopy 0 'base ''Database)
insertKey :: Key -> Value -> Update Database ()
insertKey key value
= do Database m <- S.get
S.put (Database (Map.insert key value m))
lookupKey :: Key -> Query Database (Maybe Value)
lookupKey key
= do Database m <- ask
return (Map.lookup key m)
deleteKey :: Key -> Update Database ()
deleteKey key
= do Database m <- S.get
S.put (Database (Map.delete key m))
allKeys :: Int -> Query Database [(Key, Value)]
allKeys limit
```