Front page | perl.perl5.porters |
Postings from August 2000
On Sat, Aug 05, 2000 at 01:31:38AM -0700, Jeffrey Friedl wrote:
> Currently, if you want to abort the current branch, you have to resort
> to hyjinks like
> (?(?{ not blah})(?!))
> which is a lot of typing
(?(?{ blah})|(?!))
> (and eyestrain)
Use custom REx engines to substitute something simpler.
> for what may well be a much more common thing than wanting to pass
> values around in default magic variables during a match.
Do not think so.
> (Along the same lines of utlity vs. eyestrain, I'd love to see a cut
> operator, such as \V, which looks like a wedge being driven in at that
> point. Yes, you can do almost everything with (?>...), but this would
> be so much more readable.
... But much less understandable. I still did not see any
explanation what of what \V would do which would not involve backtracking.
> Actually, I'd like to see two types of cut,
> \v would be one on the same level as (?>...), while \V would even cut
> the bump-along the regex engine does if no match can be found at the
> current position. But, I digress....)
I repeat: even I do not understand what the proposed features are
going to do. What about less accustomed programmers?
> Hugo's idea is also good, which is to use something like
> (?{ if (blah) next })
next *what*? We need to label pieces of REx, and this labeling should
be orthogonal to "save the match to a named variable".
> BTW, Ilya, the last time that I really looked at perl's regex internals (or
> new externals) was just when 5.003 was comming out in 1996 or so. I guess
> since then you've done a lot of work, to say the least. Peaking at this
> stuff now leaves me in more awe than ever.
No wonder. Say, the first-char-cognizance code is just a proof of
concept, not something which is for stay. Nowadays I have no idea why
it works at all. The logic for or-ing/and-ing pieces should be
substituted by append-ing/and-ing (it may be that only the names
should be changed). The runtime start-position sniffer was created by
patching, it should be rethought too.
> When I found the super-linear optimization stuff, I
> just shook my head at how it was just so far beyond me.
No wonder. It is coded wrong.
Ilya