Submitted by neoneye (Wed Jun 02 19:06:48 UTC 2004)
Magic values are in general bad.
Regexp::IGNORECASE = 1
Regexp::EXTENDED = 2 Regexp::MULTILINE = 4</pre></p>
The lang option are abbreviated with:
unicode = 'U'
none = 'N' euc = 'E'</pre></p>
You can easy make ugly things
Regexp.new('pattern', 7, 'U')
Regexp.new('pattern', :ignorecase=>true, :extended=>true, :multiline=>true, :encoding=>:UTF8)
I have some more ideas for useful options: :interpret_warnings_as_errors, :verbose_mismatch, :suppress_warnings.
compatibility can be maintained, by using ducktyping on the options argument.
If its type are Integer, then use old-style. If type are hash then its new-style.
class Regexp2
def initialize(pattern, options={})
p options
end
end
Regexp2.new('abc', :encoding=>:big5, :suppress_warnings=>true)
| Comments | Current voting | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
|
RCRchive copyright © David Alan Black, 2003-2005.
Powered by .
Additional note: IMHO the encoding name should follow some standard naming scheme (encoding names, use String or Symbol) used all over Ruby. (Is there such a thing already?) We might even want some package method that can emit all known encoding names.
Robert Klemme
> compatibility can be maintained, by using ducktyping on the options argument. > If its type are Integer, then use old-style. If type are hash then its new-style.
I think you mean 'by *not* using ducktyping' :-) Anyway, however it's implemented, I'm voting in favor. I don't use this stuff much but I agree about the crypticness of the existing syntax.
-- David Black
I would only suggest that the format be changed to be more like what will be expected of Ruby2's named arguments, rather than a hash. -- Austin Ziegler
Without looking at the code, can't we just "yield self if block_given?". -- Daniel Berger