I'm not sure that is the best way to look at ｢<before>｣ and ｢<after>｣.
> 'abcd123abcd' ~~ / <?before <digit>> .+ <?after <digit>> /
｢123｣
In the above code ｢<?before <digit>>｣ makes sure that the first thing that
｢.+｣ matches is a ｢<digit>｣
And ｢<?after <digit>>｣ makes sure that the last thing ｢.+｣ matches is also
a ｢<digit>｣
The ｢<?before <digit>>｣ is written in front of the ｢.+｣ so it starts at
that position
It does the thing that ｢<digit>｣ would normally do.
' a b c d 1 2 3 a b c d '
' _ _ _ _^1^_ _ _ _ _ _ '
The thing is, ｢<before>｣ resets the position to what it was immediately
before the successful ｢<digit>｣ match.
' a b c d 1 2 3 a b c d '
' _ _ _ _^_ _ _ _ _ _ _ '
The ｢.+｣ then tries to grab everything
' a b c d 1 2 3 a b c d '
' _ _ _ _^1 2 3 a b c d^'
Then ｢<?after <digit>>｣ gets to tell it that it can't do that.
The reason is that ｢<after>｣ looks backwards from the current position. The
current position is at the very end.
It obviously isn't a ｢<digit>｣, so ｢.+｣ has to keep giving up characters
until its last value is a ｢<digit>｣.
' a b c d 1 2 3 a b c d '
' _ _ _ _^1 2 3^_ _ _ _ '
---
You can use ｢<after>｣ to check that is at the beginning.
'abc' ~~ / <!after .> b /
Nil
The reason is that if the current position is anywhere other than the
beginning ｢.｣ would match.
Since we used ｢!｣ that won't fly.
｢<!before .>｣ should probably also prevent the position from being at the
end.
It does work if you write it differently
'abc' ~~ / b <!before( /./ )> /
Nil
Note that ｢<before>｣ and ｢<after>｣ are really just function calls.
It does seem like there could be a bug here.
---
All of that said, I don't think it is useful to tell new Raku programmers
that you can use those features that way.
It make them think that these two regexes are doing something similar.
/ ^ ... /
/ <!after .> ... /
They match the same three characters, but for entirely different reasons.
The ｢^｣ version is basically the same as:
/ <?{ $/.pos == 0 }> ... /
While the other one is something like:
/ <!{ try $/.orig.substr( $/.pos - 1, 1 ) ~~ /./ }> ... /
(The ｢try｣ is needed because ｢.substr( -1 )｣ is a Failure.)
So then these:
/ ... $ /
/ ... <!before .>
Would be
/ ... <?{ $/.pos == $/.orig.chars }> /
/ ... <!{ try $/.orig.substr( $/.pos, 1 ) ~~ /./ }> /
---
What I think is happening is that the ｢<!after .>｣ works because the
｢.substr( -1, 1)｣ creates a Failure.
The thing is that ｢'abc'.substr( 3, 1 )｣ doesn't create a Failure, it just
gives you an empty Str.
(The second argument is the maximum number of characters to return.)
On Mon, May 25, 2020 at 4:10 PM Joseph Brenner <[email protected]> wrote:
> Given this string:
> my $str = "Romp romp ROMP";
>
> We can match just the first or last by using the usual pinning
> features, '^' or '$':
>
> say $str ~~ m:i:g/^romp/; ## (｢Romp｣)
> say $str ~~ m:i:g/romp$/; ## (｢ROMP｣)
>
> Moritz Lenz (Section 3.8 of 'Parsing', p32) makes the point you
> can use 'after' to do something like '^' pinning:
>
> say $str ~~ m:i:g/ <!after .> romp /; ## (｢Romp｣)
>
> That makes sense: the BOL is "not after any character"
> So: I wondered if there was a way to use 'before' to do
> something like '$' pinning:
>
> say $str ~~ m:i:g/ romp <!before .> /; ## (｢Romp｣ ｢romp｣)
>
> That was unexpected: it filters out the one I was trying to
> match for, though the logic seemed reasonable: the EOL is "not
> before any character".
>
> What if we flip this and do a positive before match?
>
> say $str ~~ m:i:g/ romp <?before .> /; ## (｢Romp｣ ｢romp｣)
>
> That does exactly the same thing, but here the logic makes
> sense to me: the first two are "before some character",
> but the last one isn't.
>
