Regex delimiters: '/' vs '|'

[tel2]

Hi Monks,

Why does this:
  perl -e '$a="<abc|def>";$a =~ s/<(.+?)(\|.+?)>/-$1-/;print $a'
give this output:
  -abc-
which looks right to me, but this:
  perl -e '$a="<abc|def>";$a =~ s|<(.+?)(\|.+?)>|-$1-|;print $a'
(which is the same thing with '|' regex pattern delimiters), give this output:
  -a-

I was wanting to use the regex substitution pattern delimiter '|' instead of '/' because my read data had '/' in the text, so '|' meant I didn't have to escape the '/'s.

Perl version: 5.10.1.

Thanks.
tel2

[tye]

I was wanting to use the regex substitution pattern delimiter '|' instead of '/' because my read data had '/' in the text, so '|' meant I didn't have to escape the '/'s.

You don't have to escape delimiters in the read data. You only have to escape delimiters in the regex (or replacement). And the regex you show contains no '/' characters but it does contain a '|' character, so '|' is a strange delimiter choice.

 $ perl -MO=Deparse -e 'm/\|/'
/\|/;

 $ perl -MO=Deparse -e 'm|\||'
/|/;

The above shows how you have changed the meaning of your regex by changing the delimiter character. To include a '|' in a regex delimited by '|'s, you have to type '\|'. The '\' escapes the delimiter and leaves an unescaped '|' for the regex. Worse, you can't get the original regex easily:

 $ perl -MO=Deparse -e 'm|\||x'
/|/x;

 $ perl -MO=Deparse -e 'm|\\||x'
/\\/ | 'x';

 $ perl -MO=Deparse -e 'm|\\\||x'
/\\|/x;

None of those give a regex of /\|/x. You have to go for something different having the same meaning:

 $ perl -MO=Deparse -e 'm|[\|]|'
/[|]/;

You have to use '\' to escape the delimiter and you have to use '\' to escape the '|' regex metacharacter. But Perl provides no way for you to use '\' to both escape the delimiter nature of '|' and also escape the regex metacharacter nature of '|'.

[tel2]

Thanks tye.

Sorry I wasn't clear. My original regex did include '/'s in both sides of it. My simplified regex (which I posted) did not.

I take your points and thanks for your efforts.

Thanks.
tel2

[Anonymous Monk]

What does Basic debugging checklist say about it? ( O=Deparse and re=debug )

You should stick to m// or s/// or m{} or s{}{} because
while you can use s\\\ or s@@@ or s### as delimiters they're the worst option available
while you can use s\\\ or s@@@ or s### as delimiters they're the worst option available
while you can use s\\\ or s@@@ or s### as delimiters they're the worst option available
Perl::Critic::Policy::RegularExpressions::ProhibitUnusualDelimiters - Use only // or {} to delimit regexps.
Perl::Critic::Policy::ValuesAndExpressions::ProhibitQuotesAsQuotelikeOperatorDelimiters - Don't use quotes (', ", `) as delimiters for the quote-like operators.

[tel2]

OK - thanks Anonymous Monk!