Skip to content

GRbit/go-pcre

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

73 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

go-pcre

This package provides Perl-Compatible RegularExpression support in Go using libpcre or libpcre++.

The origin of this package is glenn-brown pcre but across other forks, this one has JIT compilation available, which makes it much faster. You can check out benchmarks game to see the difference.

Go library Time (lower is better)
This one 3.85 seconds
mdellandrea (also glenn-brown fork) 14.48 seconds
standard one 27.87 seconds

As you can see, this library is almost an order of magnitude faster than standard one and 3-4 times faster tahn pcre without JIT compilation.

Interface / API

API this library provides is a plain copy of C library API. Which may look really ugly to Go programmers. At least, it looks ugly to me, I don't think it's very convenient to set binary flags to use your regexp.

I want to refactor this library and make v2 version which will have API more like a standard library. If you are interested in such library, hit the star button. The more stars I see, the closer I am to implementing this idea.

Documentation

Use godoc.

Installation

  1. install libpcre3-dev or libpcre++-dev

  2. go get

sudo apt-get install libpcre3-dev
go get github.com/GRbit/go-pcre/

Usage

Go programs that depend on this package should import this package as follows to allow automatic downloading:

import (
  "github.com/GRbit/go-pcre/"
)

Building your software

Since this package use cgo it will build dynamically linked. If you plan to use this everywhere without libpcre dependency, you should build it statically linked. You can build your software with the following options:

go build -ldflags="-extldflags=-static"

More details on this here

Performance

Brief performance comparison across other Go libraries is in the beginning of the README, but if you are curious what regex library is the fastest here is an exhaustive research of the question: https://zherczeg.github.io/sljit/regex_perf.html

The answer is: it depends. But in most cases, it's RE2 or PCRE-JIT. RE2 tends to utilize multi-core systems better, while PCRE-JIT is better at using one CPU core for almost all use cases.

LICENSE

This is a fork of hobeone pcre, which is fork of mathpl pcre, which is a fork of glenn-brown pcre. The original package hasn't been updated for several years. But it is still being used in some software, despite its lack of JIT compiling, which gives huge speed-up to regexps. If you somehow can send a message to the original project owner, please inform him about this situation. Maybe he would like to transfer control over the repository to a maintainer who will have time to review pull requests.