bug in perl 5.005_61 #699
Comments
From swiftkid@bigfoot.comCreated by swiftkid@bigfoot.comUnder Perl 5.005_61, this line produces an error: Unmatched regexp [] ... So I reinstalled Perl 5.005_3, and it works fine with it! Something is wrong in the regexp lib. Perl Info |
From @RandalSchwartz
Faisal> Under Perl 5.005_61, this line produces an error: Faisal> Not that your bug report isn't perhaps valid, you should also know I won't quote the FAQ here. Please read it. -- |
From [Unknown Contact. See original ticket]
The '-' inside [...] is considered as a range. Try '\-' instead. Good luck, |
From [Unknown Contact. See original ticket]Konovalov, Vadim writes:
After \w?! It should not. I wonder whether this is POSIX-classes Ilya |
From @jhiIlya Zakharevich writes:
In 62-to-be I get /^[\w-.]+\@[\w-]+\.[\w-.]+$/: invalid [] range in regexp at -e line 1. which is what I would have expected. A range like \w-. makes no sense. /^[\w.]+\@\w+\.[\w.]+$/ I plead guilty of putting in a stricter check for bogus ranges. [ 3926] By: jhi on 1999/08/05 17:25:19 If anybody can explain to me what [\w-.] would mean, I will rethink. -- |
From @jhiOops, I take my bitching about a range ending into "nothing" back: perlre of both 5.005_04 and 5.005_03: If you want "-" itself to be a member of a class, put it But the other issue remains: Note that \w matches a single alphanumeric character, not a whole word. To -- |
From [Unknown Contact. See original ticket]On Tue, Oct 12, 1999 at 11:12:54AM +0300, Jarkko Hietaniemi wrote:
This is still ambiguous: \w-. may be either \w\-., or a syntax error. Ilya |
From @jhi
Hmmm. On the other hand, it seems that constructs like [\w-.] have Either we take out my patch and fix the documentation to comply, -- |
From [Unknown Contact. See original ticket]: > > match a word you'd need to say \w+. You may use \w, \W, \s, \S, \d That is what I said in my initial message.... it worked in Perl 5.005_3. I vote it to be automatically escaped.... so \w-. whould still be possible. I wonder if my vote matters! :-) |
From @jhiFaisal Nasim writes:
But you still didn't read the documentation :-) -- |
From [Unknown Contact. See original ticket]Ilya Zakharevich <ilya@math.ohio-state.edu> wrote
Since it currently works, this is a bug in the documentation. :-) The documentation should be fixed rather than breaking currently But if we were designing this from scratch, I'd vote for an error. Hmm.. After writing the above, I actually tried some experiments. [\w-.] works (i.e. treats - as escaped) So I also recommend fixing [.-\w] , for consistency. Mike Guy |
From @jhiM.J.T. Guy writes:
It seems that at least my patch consistently made all the above illegal. -- |
From @tamiasOn Tue, Oct 12, 1999 at 02:49:38PM +0100, M.J.T. Guy wrote:
I strongly disagree. Since this is documented not to work, it is an error This is a clear case of programs taking advantage of an obscure bug in Ronald |
From @jhiRonald J Kimball writes:
I can fix it both ways, let me know when the hydra called p5p has -- |
From @jhiJarkko Hietaniemi writes:
Errr, "either way", methinks... -- |
From @jhiRonald J Kimball writes:
As I just happened to finish testing the patch to return to the -- |
From [Unknown Contact. See original ticket]: I strongly disagree. Since this is documented not to work, it is an error And can't you consider it as a feature? if "." is followed by a multiple pattern match regexp, its considered How can I contribute to Perl? (is there any age requirement? :P) |
From @TimToadyjhi@iki.fi writes: Nobody's mentioned the option of issuing a warning. I think I'd prefer Larry |
From @tamiasOn Tue, Oct 12, 1999 at 09:39:27PM +0500, Faisal Nasim wrote:
If you want the character class [\w\-.], I think it's just as easy to write The documentation makes it clear that, just like in regex syntaxes of most Ronald |
From [Unknown Contact. See original ticket]: Nobody's mentioned the option of issuing a warning. I think I'd prefer How about issuing a warning? :-) Why are there qq/q/qx? When you could simply: print "blah is \"cool\" and foobar is \"not\"cool"; ? |
From [Unknown Contact. See original ticket]: Why are there qq/q/qx? When you could simply: and don't forget qw. |
From [Unknown Contact. See original ticket]Larry Wall <larry@wall.org> writes:
But it is "obvious" that \w-\d is a "range" of alphabetics and '_' ;-) -- |
From [Unknown Contact. See original ticket]At 06:05 PM 10/12/99 +0100, Nick Ing-Simmons wrote:
Hey, yeah! Set operations in character classes! Unicode has the set Dan ----------------------------------------"it's like this"------------------- |
From @jhiDan Sugalski writes:
Yes, it has. <...> Ummm, it seems the "+" is kind of taken... :-) -- |
From @jhiLarry Wall writes:
Aargh, a third option! :-) But an option I find awfully tempting... -- |
From [Unknown Contact. See original ticket]At 09:07 PM 10/12/99 +0300, Jarkko Hietaniemi wrote:
I was thinking of the horseshoes (or whatever they're called. Union and Dan ----------------------------------------"it's like this"------------------- |
From @jhiNow 62-to-be is rigged to accept dashes as literals in "false ranges" /akjhfd[xa-\dz]/: false [] range "a-\d" in regexp at -e line 1. The perldiag entry says: (W) A character class range must start and end at a literal character, not -- |
From @jhiWhile staring at regclass() I noticed a possible additional source On the other hand, in 5.005_5X Ilya's patch that complains about Should I add a similar warning for character classes, -- |
From [Unknown Contact. See original ticket]On Fri, Oct 15, 1999 at 12:33:40AM +0300, Jarkko Hietaniemi wrote:
Thanks, yes. I missed this case... Ilya |
From @jhi
Done. Now e.g. [a\zb] (with -w) triggers such a warning. -- |
Migrated from rt.perl.org#1601 (status was 'resolved')
Searchable as RT1601$
The text was updated successfully, but these errors were encountered: