ICU4R - ICU Unicode bindings for Ruby

ICU4R is an attempt to provide better Unicode support for Ruby, where it lacks for a long time.

Current code is mostly rewritten string.c from Ruby 1.8.3.

ICU4R is Ruby C-extension binding for ICU library and provides following classes and functionality:

  • UString:

    - String-like class with internal UTF16 storage;
    - UCA rules for UString comparisons (<=>, casecmp);
    - encoding(codepage) conversion;
    - Unicode normalization;
    - transliteration, also rule-based;
    
    Bunch of locale-sensitive functions:
    - upcase/downcase;
    - string collation;
    - string search;
    - iterators over text line/word/char/sentence breaks;
    - message formatting (number/currency/string/time);
    - date and number parsing.
    
  • URegexp - unicode regular expressions.

  • UResourceBundle - access to resource bundles, including ICU locale data.

  • UCalendar - date manipulation and timezone info.

  • UConverter - codepage conversions API

  • UCollator - locale-sensitive string comparison

Install and usage

> ruby extconf.rb
> make && make check
> make install

Now, in your scripts just require ‘icu4r’.

To create RDoc, run

> sh tools/doc.sh

Requirements

To build and use ICU4R you will need GCC and ICU v3.4 libraries.

Differences from Ruby String and Regexp classes

UString vs String

  1. UString substring/index methods use UTF16 codeunit indexes, not code points.

  2. UString supports most methods from String class. Missing methods are:

    capitalize, capitalize!, swapcase, swapcase!
    %, center, ljust, rjust
    chomp, chomp!, chop, chop!
    count, delete, delete!, squeeze, squeeze!, tr, tr!, tr_s, tr_s!
    crypt, intern, sum, unpack
    dump, each_byte, each_line
    hex, oct, to_i, to_sym
    reverse, reverse!
    succ, succ!, next, next!, upto
    
  3. Instead of String#% method, UString#format is provided. See FORMATTING for short reference.

  4. UStrings can be created via String.to_u(encoding=‘utf8’) or global u(str,) calls. Note that encoding parameter must be value of String class.

  5. There’s difference between character grapheme, codepoint and codeunit. See UNICODE reports for gory details, but in short: locale dependent notion of character can be presented using more than one codepoint - base letter and combining (accents) (also possible more than one!), and each codepoint can require more than one codeunit to store (for UTF8 codeunit size is 8bit, though some codepoints require up to 4bytes). So, UString has normalization and locale dependent break iterators.

  1. Currently UString doesn’t include Enumerable module.

  2. UString index/[] methods which accept URegexp, throw exception if Regexp passed.

  3. UString#<=>, UString#casecmp use UCA rules.

URegexp

UString uses ICU regexp library. Pattern syntax is described in [./docs/UNICODE_REGEXPS] and ICU docs.

There are some differences between processing in Ruby Regexp and URegexp:

  1. When UString#sub, UString#gsub are called with block, special vars ($~, $&, $1, …) aren’t set, as their values are processed through deep ruby core code. Instead, block receives UMatch object, which is essentially immutable array of matching groups:

    "test".u.gsub(ure("(e)(.)")) do |match| 
       puts match[0]  # => 'es' <--> $&
       puts match[1]  # => 'e'  <--> $1
       puts match[2]  # => 's'  <--> $2
    end
    
  2. In URegexp search pattern backreferences are in form n (1, 2, …), in replacement string - in form $1, $2, …

    NOTE: URegexp considers char to be a digit NOT ONLY ASCII (0x0030-0x0039), but any Unicode char, which has property Decimal digit number (Nd), e.g.:

    a = [?$, 0x1D7D9].pack("U*").u * 2
    puts a.inspect_names
    <U000024>DOLLAR SIGN
    <U01D7D9>MATHEMATICAL DOUBLE-STRUCK DIGIT ONE
    <U000024>DOLLAR SIGN
    <U01D7D9>MATHEMATICAL DOUBLE-STRUCK DIGIT ONE
    puts "abracadabra".u.gsub(/(b)/.U, a)
    abbracadabbra
    
  3. One can create URegexp using global Kernel#ure function, Regexp#U, Regexp#to_u, or from UString using URegexp.new, e.g:

    /pattern/.U =~ "string".u
    
  4. There are differences about Regexp and URegexp multiline matching options:

    t = "text\ntest"
    # ^,$ handling : URegexp multiline <-> Ruby default
    t.u =~ ure('^\w+$', URegexp::MULTILINE)
    => #<UMatch:0xf6f7de04 @ranges=[0..3], @cg=[\u0074\u0065\u0078\u0074]>
    t =~ /^\w+$/
    => 0
    # . matches \n : URegexp DOTALL <-> /m
    t.u =~ ure('.+test', URegexp::DOTALL)
    => #<UMatch:0xf6fa4d88 ...
    t.u =~ /.+test/m
    
  5. UMatch.range(idx) returns range for capturing group idx. This range is in codeunits.

References

  1. ICU Official Homepage ibm.com/software/globalization/icu/

  2. ICU downloads ibm.com/software/globalization/icu/downloads.jsp

  3. ICU Home Page icu.sf.net

  4. Unicode Home Page www.unicode.org

BUGS, DOCS, TODO

The code is slow and inefficient yet, is still highly experimental, so can have many security and memory leaks, bugs, inconsistent documentation, incomplete test suite. Use it at your own risk.

Bug reports and feature requests are welcome :)

Copying

This extension module is copyrighted free software by Nikolai Lugovoi.

You can redistribute it and/or modify it under the terms of MIT License.

Nikolai Lugovoi <[email protected]>