Detecting if a string is a regular expression

[smontsar]

Is there some quick and dirty way to detect is a string is a pure string, or if it is a regular expression.

Given:

$a="a.*b";
$b="acdb";

I am looking for a technique or method such that (in pseudocode):

methodIsItRegEx($a) == TRUE
methodIsItRegEx)$b) == FALSE

[MidLifeXis]

Wouldn't $b also be a regular expression, and $a also be a valid string? In other words, what are your criteria?

[davido]

What problem are you trying to solve?

[LanX]

> Is there some quick and dirty way to detect is a string is a pure string, or if it is a regular expression.

no, IMHO any $string can be used as a valid regex like in m/$string/ w/o syntax error.

but if you need a regex datatype to distinguish $scalars use qr// to generate them.

  DB<106> $s='abc'
 => "abc"
 
  DB<107> $r=qr/abc/
 => qr/abc/
 
  DB<108> ref $r eq 'Regexp'
 => 1
 
  DB<109> ref $s eq 'Regexp'
 => ""
  
  DB<110> 'xabcx'=~m/$r/
 => 1
 
  DB<111> 'xabcx'=~m/$s/
 => 1

[AnomalousMonk]

... any $string can be used as a valid regex like in m/$string/ ...

... and you don't even need the  m// or  qr// incantation. The  =~ binding operator can do the trick:

>perl -wMstrict -le
"my $s = 'abc';
 ;;
 print 'match' if 'xabcx' =~ $s;
 print 'match' if 'xabcx' =~ 'abc';
"
match
match

[LanX]

yeah I noticed, but normally I prefer the more obvious syntax. (the m is superfluous too =)

PS: I know someone will show us now benchmarks demonstrating that at least 3 and half processor cycles are waisted this way ... ;)

[BrowserUk]

To test if a string contains regex metachars, quotemeta it and compare it to the original:

print "$_: does", ($_ eq quotemeta() ? 'not' : '' ), 'contain metachar
+s' 
    for 'a.c', '[abc]', 'abcd';;
a.c: does  contain metachars
[abc]: does  contain metachars
abcd: does not contain metachars

Of course, that doesn't test if the metachars make sense as a regex.

[LanX]

for completeness:

> Of course, that doesn't test if the metachars make sense as a regex

let's Perl parse Perl =)

$ perl -ple ' $_ = eval "sub { /$_/ }" ? "OK\n" : "$@" '
a
OK

a(
Unmatched ( in regex; marked by <-- HERE in m/a( <-- HERE / at (eval 2
+) line 1, <> line 2.

a[
Unmatched [ in regex; marked by <-- HERE in m/a[ <-- HERE / at (eval 3
+) line 1, <> line 3.

a()
OK

UPDATE

unfortunately not always correct:

$ perl -ple ' $_ = eval "sub { /$_/ }" ? "OK\n" : "$@" '
a[$]
OK

but

$ perl -e ' "abc" =~ /a[$]/ '
Unmatched [ in regex; marked by <-- HERE in m/a[ <-- HERE 5.010000/ at
+ -e line 1.

UPDATE

interesting, this happens because it's a run-time error ... (why?)

$ perl -e 'sub {"aaa" =~ /a[$]/}'
$ perl -e '"aaa" =~ /a[$]/'
Unmatched [ in regex; marked by <-- HERE in m/a[ <-- HERE 5.010000/ at
+ -e line 1.
$ perl -ce '"aaa" =~ /a[$]/'
-e syntax OK

looks like a parser problem for me!

UPDATE

variable interpolation is part of the problem, a[$] is a valid pattern, as long as it's not interpolated:

perl -E 'say q(a$) =~ q(a[$]) '
1

UPDATE

this is very reliable

perl -ple ' $_ = eval "sub { 'a' =~ q\0$_\0 }" ? "OK\n" : "$@" '

to avoid problems with \0-delim consider using here-docs instead, or even a pack/unpack combination

[Kenosis]

You could create a regex that checks the string for the characters you want to allow, e.g.:

sub isStrOK {
    return $_[0] =~ /\A[\s\w]+\z/;
}

The above's not intended to be exhaustive for allowable characters, but only an example.