• via Digitalmars-d-bugs (31/31) May 02 2014 https://issues.dlang.org/show_bug.cgi?id=12690

via Digitalmars-d-bugs <digitalmars-d-bugs puremagic.com> writes:

https://issues.dlang.org/show_bug.cgi?id=12690

          Issue ID: 12690
           Summary: std.regex BacktrackingMatcher bmatch is faster than
                    ThompsonMatcher but discouraged
           Product: D
           Version: D2
          Hardware: x86_64
                OS: Linux
            Status: NEW
          Severity: enhancement
          Priority: P1
         Component: Phobos
          Assignee: nobody puremagic.com
          Reporter: yxcvbasdfgqwert02 gmx.de

Created attachment 1348
  --> https://issues.dlang.org/attachment.cgi?id=1348&action=edit
Source code for test case

The discouraged function std.regex.bmatch is significantly faster then
std.regex.match.

The attached test case already shows the performance difference with a simple
expression and a small haystack. In my real-world program, both are way more
complex and it's many expressions against many haystacks. Most expressions
match, only a few of them not. The performance difference is a knock-out
criterion.

Total real-world program running time:

Using bmatch: 5 seconds
Using match: 47 seconds

All benchmarks with dmd 2.065 on Ubuntu 64 bit, compiled with -release.

Unfortunately I'm not allowed to post the source code or haystack of the
real-world application, so the small test case has to be enough.

--