how to interpolate string containing capture-group parentheses as regex in Raku?

Question

I want to match against a programmatically-constructed regex, containing a number of (.*) capture groups. I have this regex as a string, say

my $rx = "(.*)a(.*)b(.*)"

I would like to interpolate that string as a regex and match for it. The docs tell me <$rx> should do the trick (i.e. interpolate that string as a regex), but it doesn't. Compare the output of a match (in the perl6 REPL):

> 'xaybz' ~~ rx/<$rx>/
｢xaybz｣

vs the expected/desired output, setting apart the capture groups:

> 'xaybz' ~~ rx/(.*)a(.*)b(.*)/
｢xaybz｣
 0 => ｢x｣
 1 => ｢y｣
 2 => ｢z｣

Comments

One unappealing way I can do this is to EVAL my regex match (also in the REPL):

> use MONKEY; EVAL "'xaybz' ~~ rx/$rx/";
｢xaybz｣
 0 => ｢x｣
 1 => ｢y｣
 2 => ｢z｣

So while this does give me a solution, I'm sure there's a string-interpolation trick I'm missing that would obviate the need to rely on EVAL..

jjmerelo · Accepted Answer · 2020-10-27T09:31:51.570

10

The result of doing the match is being matched when going outside the regex. This will work:

my $rx = '(.*)a(.*)b(.*)';
'xaybz' ~~ rx/$<result>=<$rx>/;
say $<result>;
# OUTPUT: «｢xaybz｣␤ 0 => ｢x｣␤ 1 => ｢y｣␤ 2 => ｢z｣␤»

Since, by assigning to a Match variable, you're accessing the raw Match, which you can then print. The problem is that <$rx> is, actually, a Match, not a String. So what you're doing is a Regex that matches a Match. Possibly the Match is stringified, and then matched. Which is the closest I can be to explaining the result

edited Oct 27 '20 at 09:31

answered Oct 19 '20 at 06:36

jjmerelo

22,578
8
40
86

1

Thank you very much, but I don't think I really understand what happened there. Could you please point me to where this is documented? – grobber Oct 19 '20 at 07:33
What confuses me is that "stringification" does not seem to be what's happening. Note the quotation marks `｢｣` around the result I get. If I stringify the match with `$/.Str` those are gone. So my getting `｢xaybz｣` cannot be due to "stringification", as that's still a Match object. The problem seems to be that the capturing parentheses are ignored in my initial attempt, and I don't see why that is. – grobber Oct 19 '20 at 08:40
@grobber that's right. It's simply being converted into a different Match object. I'll edit a bit to try and explain what's going on. Essentially, `<$rx>` is a Match, and you're matching to a Match, which is stringified... – jjmerelo Oct 19 '20 at 11:31

score 9 · Answer 2 · answered Oct 19 '20 at 12:14

9

The problem is that things in <…> don't capture in general.

'xaybz' ~~ / <:Ll> <:Ll> <:Ll> /
# ｢xay｣

They do capture if the first thing after < is an alphabetic.

my regex foo { (.*)a(.*)b(.*) }

'xaybz' ~~ / <foo> /;
# ｢xaybza｣
#  foo => ｢xaybza｣
#   0 => ｢x｣
#   1 => ｢y｣
#   2 => ｢za｣

That also applies if you use <a=…>

'xaybz' ~~ / <rx=$rx> /;
# ｢xaybza｣
#  rx => ｢xaybza｣
#   0 => ｢x｣
#   1 => ｢y｣
#   2 => ｢za｣

Of course you can assign it on the outside as well.

'xaybz' ~~ / $<rx> = <$rx> /;
# ｢xaybza｣
#  rx => ｢xaybza｣
#   0 => ｢x｣
#   1 => ｢y｣
#   2 => ｢za｣

'xaybz' ~~ / $0 = <$rx> /;
# ｢xaybza｣
#  0 => ｢xaybza｣
#   0 => ｢x｣
#   1 => ｢y｣
#   2 => ｢za｣

Note that <…> is a sub-match, so the $0,$1,$2 from the $rx will never be on the top-level.

answered Oct 19 '20 at 12:14

Brad Gilbert

33,846
11
78
129

Thanks for that explanation. The bit about non-alphabetic `<>` in particular was illuminating. Note that the issue propagates: if I first do `my $rx='(\d)'; my regex R { <$rx> }; '12' ~~ //` it still doesn't capture. – grobber Oct 19 '20 at 15:14
3

grobber: that's because you defined `R` as `<$rx>`. If you do `regex R { }` then it will work. One of the reasons for requiring the naming is that positional captures are counted at a compile time. If you do `regex R { <$rx> (.) }` it will be impossible to know what the number for `(.)` should be. Named captures don't suffer from the same limitation. – user0721090601 Oct 19 '20 at 15:51
2

"Note that the issue propagates". I'm struck by your choice of the word *issue*. The fact that only `<...>` assertions that start with an alphabetic character capture is a deliberate feature. In your example... ah, @user0721090601 has covered that. :) – raiph Oct 19 '20 at 15:52
1

@raiph: only an "issue" from my previously-uninformed perspective. Thank you both (yourself and @user0721090601): the motivation for requiring named `<>` is clear now. – grobber Oct 19 '20 at 18:04

score 1 · Answer 3 · answered Oct 19 '20 at 13:02

You could do the following to expose the inner regex result to an outside variable:

my $rx = "(.*)a(.*)b(.*)";
my $result;

'xaybz' ~~ / $<result>=<$rx> {$result = $<result>}/;

say $result;

# OUTPUT:

# ｢xaybz｣
# 0 => ｢x｣
# 1 => ｢y｣
# 2 => ｢z｣

how to interpolate string containing capture-group parentheses as regex in Raku?

3 Answers3