glondu / ocaml-re

Pure OCaml regular expressions, with support for Perl and POSIX-style strings

Description

Re is a regular expression library for OCaml.

Contact

This library has been written by Jerome Vouillon ([email protected]). It can be downloaded from https://github.com/ocaml/ocaml-re

Bug reports, suggestions and contributions are welcome.

Features

The following styles of regular expressions are supported:

• Perl-style regular expressions (module Re.Perl);
• Posix extended regular expressions (module Re.Posix);
• Emacs-style regular expressions (module Re.Emacs);
• Shell-style file globbing (module Re.Glob).

It is also possible to build regular expressions by combining simpler regular expressions (module Re).

The most notable missing features are back-references and look-ahead/look-behind assertions.

There is also a subset of the PCRE interface available in the Re.Pcre module. This makes it easier to port code from that library to Re with minimal changes.

Performances

The matches are performed by lazily building a DFA (deterministic finite automaton) from the regular expression. As a consequence, matching takes linear time in the length of the matched string.

The compilation of patterns is slower than with libraries using back-tracking, such as PCRE. But, once a large enough part of the DFA is built, matching is extremely fast.

Of course, for some combinations of regular expression and string, the part of the DFA that needs to be build is so large that this point is never reached, and matching will be slow. This is not expected to happen often in practice, and actually a lot of expressions that behaves badly with a backtracking implementation are very efficient with this implementation.

The library is at the moment entirely written in OCaml. As a consequence, regular expression matching is much slower when the library is compiled to bytecode than when it is compiled to native code.

Here are some timing results (Pentium III 500Mhz):

• Scanning a 1Mb string containing only as, except for the last character which is a b, searching for the pattern aa?b (repeated 100 times):

  • RE: 2.6s
  • PCRE: 68s

• Regular expression example from http://www.bagley.org/~doug/shootout/ [1]

  • RE: 0.43s
  • PCRE: 3.68s

  [1] this page is no longer up but is available via the Internet Archive http://web.archive.org/web/20010429190941/http://www.bagley.org/~doug/shootout/bench/regexmatch/

• The large regular expression (about 2000 characters long) that Unison uses with my preference file to decide whether a file should be ignored or not. This expression is matched against a filename about 20000 times.

  • RE: 0.31s
  • PCRE: 3.7s However, RE is only faster than PCRE when there are more than about 300 filenames.