Regex in Ruby question

[J. Cooper]

Another newbie question:

I've just started learning about regex "officially" (reading the
O'reilly book) and was trying out some of the examples (porting them to
Ruby from Perl). From what I can tell, Ruby's regex engine doesn't
support "lookbehind", and as I had only a smattering of knowledge about
regex before and no prior experience with lookaround, I was curious if
this was significant.

i.e. is there any problem that can't be solved without lookbehind?

(Note that I am not facing such a problem; I'm just curious if
lookbehind is just an optional feature that makes certain problems
easier, as opposed to being an essential thing.)

Thanks!

 

[ThoML]

From what I can tell, Ruby's regex engine doesn't

support "lookbehind"

Ruby 1.9 has look-behind:

(?<=subexp) look-behind
(?<!subexp) negative look-behind

With ruby 1.8, you can install Oniguruma.

Regards,
Thomas.

 

[Sebastian Hungerecker]

i.e. is there any problem that can't be solved without lookbehind?

Yes, there are. For example: you want to match any occurence of "bar" except
if it is preceeded by "foo". I.e. you'd want to match "blabar" or "oofbar",
but not "foobar". You can't do that without negative lookbehind.
It might be interesting to note though, that any such problem could also not
be solved by a regular grammar, so "regular" expressions that need lookbehind
aren't, as such, regular anymore.

HTH,
Sebastian

 

[ThoML]

Yes, there are. For example: you want to match any occurence of "bar" except

if it is preceeded by "foo". I.e. you'd want to match "blabar" or "oofbar",
but not "foobar".

I think it's important to state that the look-behind matches with zero
width, i.e. the match isn't included in the match.

If it's okay to include the prefix in the match (e.g., in a gsub, the
prefix could then be referenced as a group), this could also be
achieved without lookbehind:

require 'strscan'
# 0 1 2 3
# 0123456789012345678901234567890123
s = StringScanner.new('blabar oofbar foobar ofobar offbar')
# ^ ^ ^ ^
until s.eos?
m = s.scan_until(/([^o]|[^o]o|[^f]oo)(bar)/)
p s.pos
end

# =>
6
13
27
34

pos 20 is missing.

There are of course situations when this isn't possible.

Regards,
Thomas.

 

[Mark Bush]

There are of course situations when this isn't possible.

require 'strscan'
# 0 1 2 3
# 0123456789012345678901234567890123456789
s = StringScanner.new('obar blabar oofbar foobar ofobar offbar')
# ^ ^ ^ ^
until s.eos?
m = s.scan_until(/([^o]|[^o]o|[^f]oo)(bar)/)
p s.pos
end

# =>
11
18
32
39

:-(
However, with:
m = s.scan_until(/(([^o]|^)|([^o]|^)o|([^f]|^)oo)(bar)/)

=>
4
11
18
32
39

 

[ThoML]

However, with:

m = s.scan_until(/(([^o]|^)|([^o]|^)o|([^f]|^)oo)(bar)/)

Oh well. It's probably a good thing we have look-behind now.

Regards,
Thomas.