> Ian Jackson writes ("RE: RegexString: was it really meant to
> be that way ?"):
> > I've just tried my version of RegexString with a compiler snapshot
> > obtained by using cvs up -D '25 December 1999' and it doesn't work
> > properly any more. It seems that re_match now gives a REmatch whose
> > first field, Array Int GroupBounds, is always indexed from 0 to some
> > moderately large number, rather than only having the `relevant'
> > elements in it.
> >
> > I shall work around this in my program by matching against
> a:b:_ &c as
> > you suggest, but it does defeat part of the purpose of my patch.
>
> Oh, and substrPS now fails if the supplied bounds are outside the
> string, even if the end is before the beginning. This is a bit
> unhelpful, but either way the behaviour should be documented.
I don't think the behaviour of substrPS has changed, but you're right it
does fail if end < start, or start < 0. Interestingly, it yields the empty
string if end is past the end of the string.
Would anyone object to removing the failure case, so that substrPS gives the
empty string if either:
- start < 0
- end > len
- end < start
?
Cheers,
Simon
