foldl
Composable, streaming, and efficient left folds
| LTS Haskell 24.16: | 1.4.18@rev:1 |
| Stackage Nightly 2025-10-24: | 1.4.18@rev:1 |
| Latest on Hackage: | 1.4.18@rev:1 |
foldl-1.4.18@sha256:aa42a0d8db37909c2bf69861447b9f40b4993166f32a981f6eefbac503d151c7,2793Module documentation for 1.4.18
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:
(,) <$> Fold.sum <*> Fold.length = Fold step (Pair 0 0) done
where
step (Pair x y) n = Pair (x + n) (y + 1)
done (Pair x y) = (x, y)
… and then fold just transforms that to a strict left fold:
fold (Fold step begin done) = done (foldl' step begin)
Since we preserve the step function and accumulator, we can use the Fold type to
fold things other than pure collections. For example, we can fold a Producer
from pipes using the same Fold:
Fold.purely Pipes.Prelude.fold ((,) <$> 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)
Copyright (c) 2016 Gabriella Gonzalez All rights reserved.
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 Gabriella 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
1.4.18
- Add
lifts - Add
nestforFold1 - Add
Choice,Closed,Cosieve,Extend,Semigroupoid,Category,Strong,ArrowandArrowChoiceinstances forFold1 - Add
Closedinstance forFold - Define
stimesfromEndoM’s Semigroup instance for 0
1.4.17
- Add Fold1 utilities:
purely,purely_,premap,handles,foldOver,folded1 - Add pattern synonym
Fold1_that makes the initial, step and extraction functions explicit.
1.4.16
1.4.15
- Add
CosieveandCostronginstances
1.4.14
- Add
Control.Foldl.NonEmpty.nonEmpty - Add
Control.Foldl.NonEmpty.toFold - Generalize
fold1to work withFoldable1
1.4.13
- New “Control.Foldl.NonEmpty” module for folding non-empty containers
1.4.12
Data.Functor.Extend.Extendedinstances forFold/FoldM- Remove dependency on
mwc-random
1.4.11
- Fix doctest failure when built against newer versions of the
hashablepackage
1.4.10
- Fix space leaks in
scan/scanM
1.4.9
- Implement
vectorutility more efficiently
1.4.8
- Only depend on
semigroupsfor older GHC versions
1.4.7
- Add
foldByKey{,Hash}Mapfunctions
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
meanto be more numerically stable
1.4.3
- Add
Control.Scanl.scanr - Increase upper bound on
mwc-random
1.4.2
- Add
Semigroupoidinstance forFold - Increase upper bound on
contravariantandprofunctors
1.4.1
- Add
Control.Scanl - Drop support for GHC 7.8 and older
1.4.0
- BREAKING CHANGE: Change type of
premapMto accept a monadic function
1.3.7
- Add
groupBy
1.3.6
- Documentation improvements
1.3.5
- Add
Choiceinstance forFold
1.3.4
- Add
prefilterandprefilterM
1.3.3
- Add back the old
vectorasvectorM
1.3.2
- Compatibility with
Semigroupbecoming a super-class ofMonoid - Fix
asinforFold
1.3.1
- Fix
asinforFoldM
1.3.0
- BREAKING CHANGE: Change
vectorto be a pureFold(which is faster, too!)
1.2.5
- Add support for folding new containers:
hashSet,map, andhashMap - Add
prescan/postscanwhich generalizescantoTraversabletypes
1.2.4
- Add
lazyfolds forTextandByteString - Documentation fixes and improvements
1.2.3
- Add
lookup
1.2.2
- Add numerically stable
mean,variance, andstdfolds - Add
Control.Foldl.{Text,ByteString}.foldM - Add
foldOver/foldOverM
1.2.1
- Performance improvements
- Re-export
filtered
1.2.0
- Breaking change: Fix
handlesto fold things in the correct order (was previously folding things backwards and also leaking space as a result). No change to behavior ofhandlesM, which was folding things in the right order - Breaking change: Change the
Monoidused byHandler/HandlerM - Add
folded
1.1.6
- Add
maximumByandminimumBy
1.1.5
- Increase lower bound on
basefrom< 4to< 4.5
1.1.4
- Increase upper bound on
comonadfrom< 5to< 6
1.1.3
- Increase upper bound on
profunctorsfrom< 5.2to< 5.3 - Add
mapM_,hoists,purely, andimpurely
1.1.2
- Add
lastN,randomN,sink, andduplicateM - Add
Comonadinstance forFold - Add
Profunctorinstance forFoldM
1.1.1
- Increase upper bound on
vectorfrom< 0.11to< 0.12
1.1.0
- Breaking change: Rename
pretraverse/pretraverseMtohandles/handlesM - Add
Handler - Export
EndoM
1.0.11
- Add
Profunctorinstance forFold
1.0.10
- Add
randomand_Fold1
1.0.9
- Increase upper bound on
primitivefrom< 0.6to< 0.7
1.0.8
- Add
revList
1.0.7
- Add
NumandFractionalinstances forFold/FoldM - Add
countfold forTextandByteString
1.0.6
- Add
pretraverseandpretraverseM
1.0.5
- Add
lastDef
1.0.4
- Increase upper bounds on
transformersfrom< 0.4to< 0.6 - Add
nub,eqNub, andset
1.0.3
- Add
scan,generalize,simplify, andpremapM
1.0.2
- Add
listandvectorfolds - Add
foldfunction forTextandByteString
1.0.1
- Add support for
ByteStringandTextfolds - Add
Monoidinstance forFold/FoldM
1.0.0
- Initial release