# foldl

Composable, streaming, and efficient left folds

 LTS Haskell 18.18: 1.4.12 Stackage Nightly 2021-11-28: 1.4.12 Latest on Hackage: 1.4.12

See all snapshots `foldl` appears in

Maintained by
This version can be pinned in stack with:`foldl-1.4.12@sha256:18cfbea68999463a2e4c01b78621eca70ea7dbff09f588d9187ebfc699a16996,2566`

#### Module documentation for 1.4.12

Depends on 14 packages(full list with versions):
Used by 16 packages in nightly-2021-07-29(full list with versions):

# `foldl`

Use this `foldl` library when you want to compute multiple folds over a collection in one pass over the data without space leaks.

For example, suppose that you want to simultaneously compute the sum of the list and the length of the list. Many Haskell beginners might write something like this:

``````sumAndLength :: Num a => [a] -> (a, Int)
sumAndLength xs = (sum xs, length xs)

``````

However, this solution will leak space because it goes over the list in two passes. If you demand the result of `sum` the Haskell runtime will materialize the entire list. However, the runtime cannot garbage collect the list because the list is still required for the call to `length`.

Usually people work around this by hand-writing a strict left fold that looks something like this:

``````{-# LANGUAGE BangPatterns #-}

import Data.List (foldl')

sumAndLength :: Num a => [a] -> (a, Int)
sumAndLength xs = foldl' step (0, 0) xs
where
step (x, y) n = (x + n, y + 1)
``````

That now goes over the list in one pass, but will still leak space because the tuple is not strict in both fields! You have to define a strict `Pair` type to fix this:

``````{-# LANGUAGE BangPatterns #-}

import Data.List (foldl')

data Pair a b = Pair !a !b

sumAndLength :: Num a => [a] -> (a, Int)
sumAndLength xs = done (foldl' step (Pair 0 0) xs)
where
step (Pair x y) n = Pair (x + n) (y + 1)

done (Pair x y) = (x, y)
``````

However, this is not satisfactory because you have to reimplement the guts of every fold that you care about and also define a custom strict data type for your fold. Hand-writing the step function, accumulator, and strict data type for every fold that you want to use gets tedious fast. For example, implementing something like reservoir sampling over and over is very error prone.

What if you just stored the step function and accumulator for each individual fold and let some high-level library do the combining for you? That’s exactly what this library does! Using this library you can instead write:

``````import qualified Control.Foldl as Fold

sumAndLength :: Num a => [a] -> (a, Int)
sumAndLength xs = Fold.fold ((,) <\$> Fold.sum <*> Fold.length) xs

-- or, more concisely:
sumAndLength = Fold.fold ((,) <\$> Fold.sum <*> Fold.length)
``````

To see how this works, the `Fold.sum` value is just a datatype storing the step function and the starting state (and a final extraction function):

``````sum :: Num a => Fold a a
sum = Fold (+) 0 id
``````

Same thing for the `Fold.length` value:

``````length :: Fold a Int
length = Fold (\n _ -> n + 1) 0 id
``````

… and the `Applicative` operators combine them into a new datatype storing the composite step function and starting state:

``````(,) <\$> Fold.sum <*> Fold.length = Fold step (Pair 0 0) done
where
step (Pair x y) = Pair (x + n) (y + 1)

done (Pair x y) = (x, y)
``````

… and then `fold` just transforms that to a strict left fold:

``````fold (Fold step begin done) = done (foldl' step begin)
``````

Since we preserve the step function and accumulator, we can use the `Fold` type to fold things other than pure collections. For example, we can fold a `Producer` from `pipes` using the same `Fold`:

``````Fold.purely Pipes.Prelude.fold ((,) <\$> sum <*> length)
:: (Monad m, Num a) => Producer a m () -> m (a, Int)
``````

To learn more about this library, read the documentation in the main `Control.Foldl` module.

## Quick start

Install the `stack` tool and then run:

``````\$ stack setup
\$ stack ghci foldl
Prelude> import qualified Control.Foldl as Fold
Prelude Fold> Fold.fold ((,) <\$> Fold.sum <*> Fold.length) [1..1000000]
(500000500000,1000000)
``````

## How to contribute

Contribute a pull request if you have a `Fold` that you believe other people would find useful.

## Development Status

The `foldl` library is pretty stable at this point. I don’t expect there to be breaking changes to the API from this point forward unless people discover new bugs.

## License (BSD 3-clause)

Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:

• Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.

• Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.

• Neither the name of Gabriel Gonzalez nor the names of other contributors may be used to endorse or promote products derived from this software without specific prior written permission.

THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

## Changes

1.4.12

• `Data.Functor.Extend.Extended` instances for `Fold` / `FoldM`
• Remove dependency on `mwc-random`

1.4.11

• Fix doctest failure when built against newer versions of the `hashable` package

1.4.10

• Fix space leaks in `scan` / `scanM`

1.4.9

• Implement `vector` utility more efficiently

1.4.8

• Only depend on `semigroups` for older GHC versions

1.4.7

• Add `foldByKey{,Hash}Map` functions

1.4.6

• Add `nest`/`predropWhile`/`drop`/`dropM`

1.4.5

• Increase upper bound on `containers`
• Add `either`/`eitherM`

1.4.4

• Increase lower bound on `base`
• Change `mean` to be more numerically stable

1.4.3

• Add `Control.Scanl.scanr`
• Increase upper bound on `mwc-random`

1.4.2

• Add `Semigroupoid` instance for `Fold`
• Increase upper bound on `contravariant` and `profunctors`

1.4.1

• Add `Control.Scanl`
• Drop support for GHC 7.8 and older

1.4.0

• BREAKING CHANGE: Change type of `premapM` to accept a monadic function

1.3.7

• Add `groupBy`

1.3.6

• Documentation improvements

1.3.5

• Add `Choice` instance for `Fold`

1.3.4

• Add `prefilter` and `prefilterM`

1.3.3

• Add back the old `vector` as `vectorM`

1.3.2

• Compatibility with `Semigroup` becoming a super-class of `Monoid`
• Fix `asin` for `Fold`

1.3.1

• Fix `asin` for `FoldM`

1.3.0

• BREAKING CHANGE: Change `vector` to be a pure `Fold` (which is faster, too!)

1.2.5

• Add support for folding new containers: `hashSet`, `map`, and `hashMap`
• Add `prescan`/`postscan` which generalize `scan` to `Traversable` types

1.2.4

• Add `lazy` folds for `Text` and `ByteString`
• Documentation fixes and improvements

1.2.3

• Add `lookup`

1.2.2

• Add numerically stable `mean`, `variance`, and `std` folds
• Add `Control.Foldl.{Text,ByteString}.foldM`
• Add `foldOver`/`foldOverM`

1.2.1

• Performance improvements
• Re-export `filtered`

1.2.0

• Breaking change: Fix `handles` to fold things in the correct order (was previously folding things backwards and also leaking space as a result). No change to behavior of `handlesM`, which was folding things in the right order
• Breaking change: Change the `Monoid` used by `Handler`/`HandlerM`
• Add `folded`

1.1.6

• Add `maximumBy` and `minimumBy`

1.1.5

• Increase lower bound on `base` from `< 4` to `< 4.5`

1.1.4

• Increase upper bound on `comonad` from `< 5` to `< 6`

1.1.3

• Increase upper bound on `profunctors` from `< 5.2` to `< 5.3`
• Add `mapM_`, `hoists`, `purely`, and `impurely`

1.1.2

• Add `lastN`, `randomN`, `sink`, and `duplicateM`
• Add `Comonad` instance for `Fold`
• Add `Profunctor` instance for `FoldM`

1.1.1

• Increase upper bound on `vector` from `< 0.11` to `< 0.12`

1.1.0

• Breaking change: Rename `pretraverse`/`pretraverseM` to `handles`/`handlesM`
• Add `Handler`
• Export `EndoM`

1.0.11

• Add `Profunctor` instance for `Fold`

1.0.10

• Add `random` and `_Fold1`

1.0.9

• Increase upper bound on `primitive` from `< 0.6` to `< 0.7`

1.0.8

• Add `revList`

1.0.7

• Add `Num` and `Fractional` instances for `Fold`/`FoldM`
• Add `count` fold for `Text` and `ByteString`

1.0.6

• Add `pretraverse` and `pretraverseM`

1.0.5

• Add `lastDef`

1.0.4

• Increase upper bounds on `transformers` from `< 0.4` to `< 0.6`
• Add `nub`, `eqNub`, and `set`

1.0.3

• Add `scan`, `generalize`, `simplify`, and `premapM`

1.0.2

• Add `list` and `vector` folds
• Add `fold` function for `Text` and `ByteString`

1.0.1

• Add support for `ByteString` and `Text` folds
• Add `Monoid` instance for `Fold`/`FoldM`

1.0.0

• Initial release