https://github.com/bartavelle/fastparser

A fast bytestring parser
https://github.com/bartavelle/fastparser

bytestring fast haskell parsing

Last synced: 10 days ago
JSON representation

A fast bytestring parser

• Host: GitHub
• URL: https://github.com/bartavelle/fastparser
• Owner: bartavelle
• License: bsd-3-clause
• Created: 2016-10-11T09:20:23.000Z (about 8 years ago)
• Default Branch: master
• Last Pushed: 2023-06-08T09:47:21.000Z (over 1 year ago)
• Last Synced: 2024-03-15T11:50:00.740Z (8 months ago)
• Topics: bytestring, fast, haskell, parsing
• Language: Haskell
• Size: 35.2 KB
• Stars: 6
• Watchers: 4
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • Changelog: CHANGELOG.md
  • License: LICENSE

Awesome Lists containing this project

README

        
[![Build Status](https://travis-ci.org/bartavelle/fastparser.svg?branch=master)](https://travis-ci.org/bartavelle/fastparser)

A very simple, backtracking, fast parser combinator library.

It is measurably faster than [attoparsec](https://hackage.haskell.org/package/attoparsec) (36% in [this use case](https://hbtvl.banquise.net/posts/2015-12-14-fastParsing03.html)), but only works on strict `ByteString`, lacks many helper functions, and is not resumable.

It also should consume a tiny bit less memory for equivalent operations.

# When NOT to use fastparser

 * When performance is not the **most** pressing concern.

 * When you need to parse anything else but strict `ByteString`.

 * When you need to use a battle-tested library. While very simple, and in constant use by me, this package is still quite experimental.

 * When you need to parse large inputs that are not easily cut into many smaller pieces that can be parsed independently.

# How to use fastparser

`fastparser` works well with small pieces, such as individual log file lines. It is recommended to use it with a coroutine library (such as [conduit](http://hackage.haskell.org/package/conduit) or [pipe](http://hackage.haskell.org/package/pipes)), so that the input could be incrementaly consumed, cut into individual records, all of which would end up parsed independently.

One such setup, with the `conduit` ecosystem, would look like:

    sourceFile "/tmp/foo" .| Data.Conduit.Binary.lines .| CL.map (parseOnly parser) .| ...

Other than that, `fastparser` is fairly close to any other parser combinators library.