Re: [Help-smalltalk] [rfc] regex rewrite

[Mike Anderson]

Paolo Bonzini wrote:
> I was about to make the #copyReplacingRegex:with: methods use
> #bindWithArguments: but I decided that this patch was already big enough.
> What do you think about this additional change?

You mean so that you can use backreferences in the replacement string? 
Emphatically yes. I  think that that it's what you want more often than not.

> Now, all that remains before this stuff can be made ``more'' official,
> is writing some SUnit tests...  I'll get round to it, but if somebody
> wants to help, patches are welcome as usual.

I have written some tests as a way of exploring the code, but I have 
this query:

CRegexRegisters >> #matchBeg and String >> #searchRegex: look like a 
failed match would return nil, but in practice, you get:

| mr |
mr := 'abc' searchRegex: '[0-9]+'.
mr isNil printNl. "false"
mr from printNl. "0".
!

Is that intended? My preference is that a failed match should return 
nil. In some ways, though, always returning a RegexResult, with the 
additional methods #ifMatched:ifNotMatched: and #isMatched methods, is 
more elegant.

I also think it would be more useful if =~ returned MatchResult/nil 
instead of true/false. After all, you can always write:

(myString =~ '[A-Z][a-z]*') notNil

Mike