trailblazing/urlmatch

A fast URL matcher library

URL matcher lib

This is a small and fast C library duplicating the URL matching functionality of Opera. You might use it to implement ad blocking or similar.

Given a list of patterns, such as

*facebook.com/*
http*google-analytics.*
http://foo.bat/this-annoying-image.jpeg

you can then match any connection attempt against the whole list, getting a yes/no answer back.

Motivation

One of the main components of Opera, the filtering system, supported white- and blacklists with wildcards. It was usable for more than just blocking ads, though it handled those well too.

This is one such function that should never be relegated to Javascript (like Adblock browser extensions do). The average page makes close to a hundred connections, the list is traversed on each connection attempt, and common lists reach a few thousand entries.

Turns out there wasn’t any existing standalone pattern matching library. Regex is too slow (or in glibc’s case, taking gigabytes of RAM), and wildcard functionality is essential.

A simple function (like those you can find dozens of in the web) is included for benchmark comparison purposes. Currently this library is ~5x faster vs the simple function. This is not quite fast enough, so future optimizations will be coming.