Re: two regexp related questions

On 2011-05-19 21:16, Julian Reschke wrote:

On 2011-05-19 20:51, Brandon Ibach wrote:

For 2), if you're using the regex to both validate the input (making
sure it conforms to the required syntax) and parse/extract the
name/value pairs, you might be able to make the job easier by breaking
these two tasks apart. Use the regex as you have it now to validate
the input and then, if it matches, use a shorter regex that matches
just a single name/value pair with analyze-string to do the actual
processing.

-Brandon :)

That's more or less what I do know. But as long as the regex contains a
repeating pattern, <xsl:matching-substring> will only be invoked once,
and the regex-group function will only return the contents for the last
match, right?

I think it depends on the implementation. I couldn't see anything in the
spec about what regex-group(3) of
([a-z]+)=([a-z]+)(;([a-z]+)=([a-z]+))*
should be. In Saxon, it's ';e=f' for your example, but in principle it
could also be ';c=d'.

As Brandon pointed out, using analyze-string with a repeating pattern
that matches the entire string is not the best approach. There are more
natural approaches that work without recursion. I sketched two of them
below.
..

  value = name ( ";" param )*
  name = token
  param = token "=" (token | quoted-string)
  ...