Anonymous Monk has asked for the wisdom of the Perl Monks concerning the following question:

Might someone explain this little oddity. 
$_ = "is this odd or not?";
@a = /(.*? )/gs;  # note the ? is followed by a space
print "@a";

[download]

It only gives one match here (5.6.1, linux). With only /g it matches multiple times. It also matches multiple times with /gsi, but not with combinations of /gs and other modifiers. It does multiple matches if you replace the trailing space in the re with a \s though.

Replies are listed 'Best First'.
Re: regexp /gs oddity or bug
by TheHobbit (Pilgrim) on Apr 22, 2002 at 10:30 UTC

    Hi,
    I ran your code with a use re "debug"; at the beginning and here is what I got: 

    Compiling REx `(.*? )'
size 10 first at 5
   1: OPEN1(3)
   3:   MINMOD(4)
   4:   STAR(6)
   5:     SANY(0)
   6:   EXACT < >(8)
   8: CLOSE1(10)
  10: END(0)
floating ` ' at 0..2147483647 (checking floating) anchored(SBOL) impli
+cit minlen 1 
Guessing start of match, REx `(.*? )' against `is this odd or not?'...
Found floating substr ` ' at offset 2...
Guessed: match at offset 0
Matching REx `(.*? )' against `is this odd or not?'
  Setting an EVAL scope, savestack=6
   0 <> <is this odd >    |  1:  OPEN1
   0 <> <is this odd >    |  3:  MINMOD
   0 <> <is this odd >    |  4:  STAR
  Setting an EVAL scope, savestack=6
                           SANY can match 2 times out of 2...
   2 <is> < this odd >    |  6:    EXACT < >
   3 <is > <this odd >    |  8:    CLOSE1
   3 <is > <this odd >    | 10:    END
Match successful!
Guessing start of match, REx `(.*? )' against `this odd or not?'...
Not at start...
Match rejected by optimizer
Freeing REx: `(.*? )'

    [download]

    Looks like it actualy find other matches, but they where refused by the optimizer. The hell if I know why.

    By the way, I'm using v5.6.1 on a Debian woody system.

    Cheers
    Leo TheHobbit
[reply]
[d/l]
Re: regexp /gs oddity or bug
by Joost (Canon) on Apr 22, 2002 at 09:58 UTC
    Looks like a bug to me, also perl 5.7.3 on linux (latest beta i've got installed here) WILL match multiple times on your code.

    Joost.

[reply]
Re: regexp /gs oddity or bug
by bjelli (Pilgrim) on Apr 22, 2002 at 09:59 UTC
[reply]
Re: regexp /gs oddity or bug
by argathin (Acolyte) on Apr 22, 2002 at 10:06 UTC
    Just as another datapoint: Perl 5.6.1 on Solaris ("sun4-solaris") gives the same result as described in your post.
[reply]
Re: regexp /gs oddity or bug
by belden (Friar) on Apr 22, 2002 at 16:45 UTC
    5.6.1 on alpha is buggy (scalar @a == 1)
    5.6.0 on alpha is ok (scalar @a == 4)

    blyman
    setenv EXINIT 'set noai ts=2'

[reply]
[d/l]
[select]
Re: regexp /gs oddity or bug
by navels (Initiate) on Apr 22, 2002 at 19:17 UTC
    FYI, cygwin's 5.6.1 perl does not have this bug. Lee
[reply]
Re: regexp /gs oddity or bug
by straywalrus (Friar) on Apr 22, 2002 at 23:39 UTC
    Just another data point: I run perl 5.6.0 on SlakWare 7.1.0 and the output I got from the script, 
    $_ = "is this odd or not?";
@a = /(.*? )/gs;
print @a;

    [download]
    was: "is this odd or". nothing was beyond the or
[reply]
[d/l]

Back to Seekers of Perl Wisdom