[sf-perl] RE oddity

[Joseph Brenner]

Rich Morin <rdm at cfcl.com> wrote:

> I'm a big fan of extended regular expressions, but I just
> wrote one that didn't work as I expected.  This code:
> 
>   $line =~ s|[\000-\010                   # nul-bs
>               \012-\037                   # nl-us
>               \177-\377']                 # del-... and '
>             ||gx;                         # Punt weird characters.
> 
> produced the nastygram:
> 
>   Invalid [] range "l-b" in regex;
>   marked by <-- HERE in m/[\000-\010              # nul-b <-- HERE s
>                       \012-\037                   # nl-us
>                       \177-\377']                 # del-...
>                     / at /home/rdm/bin/log_load.pl line 206.
> 
> but this code:
> 
>   $line =~ s|[\000-\010\012-\037\177-\377']||g;
> 
> sails right through.  Is this a bug or a (mis-)feature?

man perlre:

    The "/x" modifier itself needs a little more explanation.  It tells the
    regular expression parser to ignore whitespace that is neither back-
    slashed nor within a character class.  

It's documented.  It's a feature. 

The gotcha I usually get stung on is assuming that /x does something to 
the right hand side of a s///x:

   s{ ^ (.*?)    # capture first word to $1
         /s      # seperated by a space
        (.*?) $  # capture second word to $2
     }{$1 $2}x

That'll remove a space from between two items and then put it right back
again.  (Of course, if you were trying to convert tabs to spaces, then 
this could be useful).