Look behind and global

gingaloon has asked for the wisdom of the Perl Monks concerning the following question:

Dear All, I wish to insert a new line every two digits. Thought I'd use look-behind to do this:

my $x=12345;
$x=~s/(?<=\d{2})/\012/g;
print $x,"\n";
[download]

I get:

I know I can do this other ways, but the result has left me confused as to what is going on with the look behind with the global modifier. Surely at position 3 the look behind 'sees' 12\n? No? Any explanation would be appreciated many thanks james

Comment on Look behind and global Download Code

Replies are listed 'Best First'.
Re: Look behind and global by hv (Prior) on Mar 14, 2005 at 11:55 UTC
When performing `s///g` substitutions, the regexp engine does all the matching on the original unaltered string, and makes a copy on which to perform all the substitutions. So in this case, all the matches are looking at "12345", and at each position from offset 2 onwards it successfully finds that it is "just after two digits". In this case, the simplest way to achieve the desired effect is to have the digits be part of the match rather than perceived only as part of a zero-width assertion: `$x =~ s/(\d{2})/$1\n/g;` [download] Hugo	[reply] [d/l] [select]
Re: Look behind and global by maa (Pilgrim) on Mar 14, 2005 at 11:51 UTC
Hi gingaloon if you have 12345 then the sequences of two digits are 12 23 34 45... HTH - Mark Update: my (guessed) explanation as to why you don't get what you want is that you're not modifying the pattern "in-place", so to speak, so the regexp engine doesn't take account of the changes it makes. This is one use for backreferences (\1 \2 \3) Update 2: why not just `s/(\d\d)/$1\n/g`?	[reply] [d/l]
Re: Look behind and global by thinker (Parson) on Mar 14, 2005 at 12:03 UTC
In the spirit of TIMTOWTDI, you could do this `print map "$_\n", $x =~ /..?/g ; 12 34 5` [download] cheers thinker	[reply] [d/l]
Re: Look behind and global by Tanktalus (Canon) on Mar 14, 2005 at 15:31 UTC
The only nice thing about zero-width assertions is that you don't have to capture anything. So, in that spirit, and in the spirit of TMTOWTDI, and in the spirit that this is what you were asking about (whether it's the best way to do it or not): `$ perl -e '$x=12345; $x=~s/(?<=\G\d{2})/\012/g; print $x,$/' 12 34 5` [download] As you can see - I just added the \G to the assertion. Of course, if you start using anything other than numbers in there, I'm sure this could get all mucked up: `$ perl -e '$x="123d45"; $x=~s/(?<=\G\d{2})/\012/g; print $x,$/' 12 3d45` [download]	[reply] [d/l] [select]


Pathologically Eclectic Rubbish Lister
	PerlMonks