ruby-boost-regex
Wraps the Boost.Regex library to provide Boost regexes. Why? Why not! (Plus, it's worth seeing if it's faster.)
I'm trying to match the API to be at least completely compatible with
normal Regexp
s, and then add Boost features. So far...
Features
- Matches set global special variables like $`, $1-$9, $&, and so on
- Supports
=~
operator,===
operator (forcase...when
) - Supports
#match
method as in normal Regexps, returning the same type of object (MatchData
). - Spiffy monkey patch (see below)
- Ruby 1.9 Support
- Near-perfect drop-in for built-in Regexes.
Problems
- NO Unicode support. Sorry.
Global Variables
If you use Ruby's regexes, you probably use the handy-dandy global variables that
come along with them, especially for quickly-written scripts. I'm talking about
our buddies, $1, $2, $3
, $\
,
$&,
$~`. This gem makes sure to set those variables
so that these global variables still work.
Cool monkey patch
So monkey patching is bad, right? Right. And rubyists monkey patch all the time, which makes us bad people. Well, this module adds a new, alternate regex, but Ruby has a syntax for regexes already:
reg = /\d{3}/
This is nice because we don't have to escape the backslashes, plus it looks really nice with syntax highlighting. With these boost regexes, you'll either have to do
reg = Boost::Regexp.new("\\d{3}")
# or
reg = Boost::Regexp.new(/\d{3}/)
Why do all that typing? We have a literal regex syntax! But it creates normal regexes. So, we have a compromise:
Boost::Regexp.enable_monkey_patch! # only have to do this once
reg = /\d{3}/.boost!
reg.class # ==> Boost::Regexp
Cool, eh?
Of course, Boost gives us lots of crazy flags:
reg = /abc(def)/.boost!(Boost::Regexp::NO_SUBS & Boost::Regexp::IGNORECASE)
reg =~ "zzzABCDEF" # ==> 3
puts $1 # ==> nil
Benchmarks
I adapted the ruby-benchmark-suite's bm_regex_dna benchmark, which is used in the Computer Language Shootout. However, it uses str.scan(), which will fail if confronted with a non-standard Regexp. So I had to write my own, which is less optimized. The benchmark code is free to be scrutinized! It uses the standard benchmark module because lifting the spiffy one from ruby-benchmark-suite proved to be too much of a hassle for now.
Anyway, here's some results. 2.4Ghz Intel Core 2 Duo, 4GB 1067 Mhz DDR3, Macbook Pro. Ruby 1.8.7 (2008-08-11 patchlevel 72) [universal-darwin10.0]. Oniguruma is the Oniguruma gem for 1.8.x ruby, not the one bundled with Ruby 1.9+. I ran this a few times with differing loads, and they always roughly come out to the same result. YMMV.
DNA-Matching (Computer Language Shootout)
=========================================
Rehearsal ------------------------------------------------
Normal regex 17.240000 0.050000 17.290000 ( 17.353051)
Oniguruma 16.300000 0.030000 16.330000 ( 16.384928)
Boost regex 11.400000 0.040000 11.440000 ( 11.489252)
-------------------------------------- total: 45.060000sec
user system total real
Normal regex 17.190000 0.030000 17.220000 ( 17.273140)
Oniguruma 16.220000 0.040000 16.260000 ( 16.325460)
Boost regex 11.330000 0.030000 11.360000 ( 11.402222)
Failing to match a phone number in a big string of text
=======================================================
Rehearsal ------------------------------------------------
Normal regex 0.070000 0.000000 0.070000 ( 0.072128)
Oniguruma 0.040000 0.000000 0.040000 ( 0.043422)
Boost regex 0.040000 0.000000 0.040000 ( 0.034708)
--------------------------------------- total: 0.150000sec
user system total real
Normal regex 0.070000 0.000000 0.070000 ( 0.071984)
Oniguruma 0.040000 0.000000 0.040000 ( 0.044686)
Boost regex 0.030000 0.000000 0.030000 ( 0.036421)
Ruby 1.9 version:
DNA-Matching (Computer Language Shootout)
=========================================
Rehearsal --------------------------------------------------
1.9 Ruby regex 17.440000 0.030000 17.470000 ( 17.523726)
Boost regex 9.750000 0.010000 9.760000 ( 9.757300)
---------------------------------------- total: 27.230000sec
user system total real
1.9 Ruby regex 17.420000 0.020000 17.440000 ( 17.434514)
Boost regex 9.750000 0.010000 9.760000 ( 9.755085)
Failing to match a phone number in a big string of text
=======================================================
Rehearsal --------------------------------------------------
1.9 Ruby regex 0.020000 0.000000 0.020000 ( 0.016165)
Boost regex 0.040000 0.000000 0.040000 ( 0.035585)
----------------------------------------- total: 0.060000sec
user system total real
1.9 Ruby regex 0.010000 0.000000 0.010000 ( 0.016185)
Boost regex 0.040000 0.000000 0.040000 ( 0.035736)
Very interesting - 1.9 seems to have gotten faster at the "match a phone number in a big block of text" benchmark. Boost is even faster, though, and Oniguruma integrated with 1.9 is back at the old 1.9 speed. I'm hoping to add some kind of XML/HTML-matching benchmark (Yes, I know this is a bad idea). If you know how to benchmark regexps, feel free to fork away!
Usage
Install the gem, use as follows:
require 'ruby-boost-regex'
r = Boost::Regexp.new("(\\d{3})-(\\d{3})-(\\d{4})")
r =~ "555-123-4567"
p $1 # ==> "555"
matches = r.match("123-456-7890")
p matches[2] # ==> "456"
Boost::Regex.enable_monkey_patch!
r = /hello|world/i.boost!
r =~ "i'm Mike. Hello!" #==> 10
Installation
You're going to need Boost.Regex. You can find the Boost Project here, with a Getting Started page for installation.
Ruby-boost-regex expects you to have the boost regex library available as "libboost_regex.so" or "libboost_regex.dylib" or whatever your platform specifies - the key point is the name "boost_regex".
gem install ruby-boost-regex
For the Unicode support I'm trying to get in, you'll need IBM's ICU project installed, with the recommended download page you start from being ICU4C 4.4. In addition, your Boost libraries must be configured to use ICU.
Note on Patches/Pull Requests
- Fork the project.
- Make your feature addition or bug fix.
- Add tests for it. This is important so I don't break it in a future version unintentionally.
- Commit, do not mess with rakefile, version, or history. (if you want to have your own version, that is fine but bump version in a commit by itself I can ignore when I pull)
- Send me a pull request. Bonus points for topic branches.
Copyright
Copyright (c) 2010 Michael Edgar. See LICENSE for details.