Re: [xsl] Regex groups / was: Re: [xsl] Move leading/trailing spaces outside (XSLT 2.0)

[Abel Braaksma <abel.online@xxxxxxxxx>]

Yves Forkl wrote:
> Though familiar with RegExp functions in various languages, I am not 
> sure how this one works, so I wonder what happens when the condition 
> of the first regex group cannot be satisfied. Supposing that the 
> second matches, will I then have to refer to it as regex-group(1) or 
> still as regex-group(2)? I.e., do the numbers index the groups defined 
> in the regex or the substrings that actually matched one of the groups?

Quite simply (and not so in some other regex dialects) the number in 
your regex-group is guaranteed to match the content of what is matched 
or not matched by the opening parenthesis with the same index. Meaning: 
just count all "(" in your regex from the left, and you know what will 
be in the regex-group(x) where 'x' equals your count.

I.e., text is: "this is a text"
regex: (t(.+)(.*))

will put the 't' in regex-group(1), the 'his is a text' in 
regex-group(2) and nothing in regex-group(3) (because of the rules of 
greedy matching and backtracking, but that is another story). Note that 
this also includes situations where the branches of the match (by '|' 
character) would logically break this count. Not so in XSLT, i.e.:

regex: (t)|(.)

will put all 't' in regex-group(1) and all other chars in regex-group(2).

If you want to know whether a regex-group actually matched anything, you 
can check its string-length. The same rules for regex-group apply to $1, 
$2 etc, but you can only use them inside replace(), not inside 
analyze-string.

-- Abel Braaksma
  http://www.nuntia.nl