Foundation

Linux Build Status Windows Build Status Doc Stackage version Hackage version BSD3 Haskell Coverage Status

Documentation:

Goals

  • provide a base like set of modules that provide a consistent set of features and bugfixes across multiple versions of GHC (unlike base).
  • provide a better and more efficient prelude than base's prelude.
  • be self-sufficient: no external dependencies apart from base (or ghc packages).
  • provide better data-types: packed unicode string by default, arrays.
  • Better numerical classes that better represent mathematical things (No more all-in-one Num).

Usage

How to use with the normal Prelude

Add the foundation package in your cabal file, and it's recommended to import Foundation qualified if you're planning to use with the normal Prelude:

For example:

import qualified Foundation as F

It's also recommended if you're going to deal with packages using text, bytestring, vector.. to use the foundation-edge package.

How to use fully without Prelude

Disable the built-in prelude at the top of your file:

{-# LANGUAGE NoImplicitPrelude #-}

Or directly in your project cabal file:

Default-Extensions: NoImplicitPrelude

Then in your modules:

import Foundation

Advanced settings

Please check out the chapter Advanced Usage Options in the documentation.

How to contribute

Contribution guide can be found here.

Profiling

If you want to see the core (simpl step) or the assembly (asm step), then you can build with

stack build --ghc-options -ddump-simpl --ghc-options -ddump-asm

Note that it doesn't actually will create the necessary extra files if the modules doesn't need building.

you can then find your build by-products in:

.stack-work/dist/<architecture>/Cabal-<CabalVersion>/build/

Each modules that get compiled will create an equivalent file in the build directory:

  • ModuleName.dump-simpl
  • ModuleName.dump-asm

For profiling individual programs, the following command is useful:

stack ghc -- -O --make X.hs -prof -auto-all -caf-all -fforce-recomp

Benchmarking

To get the list of benchmark:

stack bench --benchmark-arguments -l

To compare against other libraries, you need to set the bench-all flag

stack bench --flag foundation:bench-all --benchmark-arguments -l

To run a specific or set of benchmarks :

stack bench --flag foundation:bench-all --benchmark-arguments 'types/String/SplitAt/mascii-10/Text'
stack bench --flag foundation:bench-all --benchmark-arguments '-m prefix types/String/SplitAt'
stack bench --flag foundation:bench-all --benchmark-arguments '-m glob types/String/SplitAt'

To register a set of benchmarks:

stack bench --flag foundation:bench-all --benchmark-arguments "--csv $(git describe).csv"

Design

Foundation started on the simple idea of trying to put everything I need in one simple and consistent package. The amazing haskell ecosystem is extremely fragmented and maintained by different people with different goals, free time, and style. The overall scare of not trying to change anything relatively central (base, bytestring, text, vector, ...) for a promise of stability has pushed many people to work on their own thing, leading to unnecessary work duplication and further fragmentation.

Foundation uses and abuses type families.

Code Organisation

Every foundation modules start by Foundation.

  • Foundation is the prelude replacement module.
  • Foundation.Internal contains only compatibilty and re-export from ghc/ghc-prim/base.
  • Foundation.Primitive is where all the lowlevel magic happens: Important types that underpins many others part Pervasive features

Changes

0.0.13

Block: * Optimise fold

UArray: Re-organise type to be more modular for later change Remove the pinned array explicit status in favor of asking the runtime system directly on demand. Optimise fold operations Optimise all&any Optimise isPrefixOf Optimise isSuffixOf Optimise finding byte Add an optimise function to break on line (CRLF & LF) as part of a stream

String: Optimise length Optimise all&any Optimise foldr Remove many unboxed tuples (next, prev, ..) in favor of a strict unpack constructor * Optimise lines using array breakLine

Collection: * add stripPrefix & stripSuffix

Tests: * Improve performance

Misc: Cleanup Offset/Size types with the C boundary Faster Offset/Size convertions Add Base64 support Add LE/BE instance for NormalForm * Add UUID generation and parsing

0.0.12

  • Fix build windows building & time subsystem
  • Add BlockN: Typed-fixed length block of memory
  • Add Base64
  • Add 'or' and 'and'

0.0.11

  • Add Hlint support (configuration file), and default travis job
  • Property report error through with the ASCII, UTF16, UTF32 string decoders
  • Fix issue with OSX < Sierra
  • Improve Parser and fix backtracking issue
  • Strictify UArray to contains a strict ByteArray#
  • Improve any & all for most collection types
  • Improve minimum & maximum for most collection types
  • Add foldl1 & foldr1
  • Add takeWhile & dropWhile
  • Remove foldl
  • Add basic String builder
  • Add String replace function
  • Add conduit sourceList
  • Improve performance of String uncons, unsnoc, filter and fromBytes UTF8 (validate)
  • Improve UArray filter
  • Fix compilation issue on windows with clock_gettime which is not available though all possible compilation mode

0.0.10

  • Cleanup collection APIs that take a lone Int (length, take, drop, splitAt, ..) to take a CountOf
  • Rename Size to CountOf
  • Add basic time functions
  • Add os dependent timing capability
  • Add simple pattern matching for test names with checks.
  • add '--list-tests' for checks
  • Optimise Eq and Ord for blocks and uarray

0.0.9

  • Introduce Block & MutableBlock which represent a chunk of memory without slices and are faster/leaner in many cases.
  • Cleanup String code and some primitives boundaries
  • Fix storable alignment tests
  • Add These data type (either a, b or both)
  • Implement checks command line
  • Improve checks terminal output
  • drop support for GHC 7.4 and GHC 7.6
  • Improve performance of copy out of block and uarray

0.0.8

  • Add MonadReader and MonadState
  • Improve performance of numerical read parsers (integral, double)
  • Improve precision of double read parser
  • Add Textual conduit combinator (fromBytes, toBytes, lines)
  • Add DList
  • Fix building on latest Win32, RHEL 5.8
  • Add NormalForm
  • Export some functions in Internal module to manipulate unboxed mutable array

0.0.7

  • Improve Checks: random seed, new properties and improved printing
  • Add ability to parse Natural, Integer, and Double from String
  • Temporarily remove compilation of experimental network resolution introduced in 0.0.5 for windows building.
  • Cleanup Offset and Size fixing some bug in String module

0.0.6

  • Fix build on Centos 6.x / older linux distribution
  • Improve test checks generators

0.0.5

  • Generalize monadic map (mapM, mapM_)
  • HostName type
  • Network address / name resolution
  • Fix compilation on FreeBSD & OpenBSD
  • Initial re-implementation for property tests and tests orchestration
  • Fix bug in splitElem, and breakElem
  • Improve splitOn to return empty elements
  • Fix API bug for snoc and cons in Chunked UArray
  • Add UUID
  • Check API
  • Fix compilation on !x86

0.0.4

  • Add Conduit for all your streaming needs
  • Expose Sequential from Foundation
  • Export internal withPtr for optimisation
  • Export ifThenElse
  • Use the proper String type for error instead of [Char]
  • Add any and all to Collection
  • Add defaulting to Integer and Double for numerical types
  • Add negation for Double and Float (and their associated C types)
  • Add/Export system bindings (Posix file/memory handling, Linux Inotify)
  • Add Big Endian (BE) / Little Endian (LE) wrapping types
  • Add a way to transform an UArray into Hexadecimal/Base16
  • Add IPv4 and IPv6 types

0.0.3

Monad:

  • Add MonadCatch and MonadThrow classes
  • Add Transformer base class (MonadTrans)
  • Add IdentityT, StateT, ReaderT

Build:

  • Fix build on !x86

0.0.2

Classes:

  • Add Bifunctor
  • Implement Better storable type class (#111)
  • Expose Nthable for GHC >= 7.10 (product type getter)
  • Split basic function from Sequential to Collection
  • show return a Foundation String now instead of [Char]

Numerical: Overhaul of numerical classes (Integral, Rational, Divisible, ..) add IntegralRounding (i.e. rounding from floating types) Expose IEEE manipulation stuff Expose all trigonometry functions in Foundation.Math.Trigonometry * Export Natural (Unsigned Integer)

Collection: Add partition Add isPrefixOf and isSuffixOf Add ArrayBuilder machinery Add String parser Add minimum and maximum to Collection. Export Foldable and Collection in Foundation * add head,last,tail,init

Types: Basic ArrayUArray support (Array of unboxed Array) Add instance for Float and Double for numerical Boxed array: add native slicing in the type add NonEmpty type * Add some Data declaration for based type

Hashing: Hashing: add FNV, SipHash hash functions family Hashable: add support to hash types

Random support: Add support for system entropy Add pseudo random generation capability using a ChaCha core.

0.0.1

  • Initial version
comments powered byDisqus