Hi
I'm trying to split a path (like an URL, but not one) into bits on the '/' character.
Problem is, I want to be able to be able to escape '/' in the names of bits by doubling
the character to '//' eg
'foo/bar // baz/qux'
=> ["foo", "bar / baz", "qux"]
In a perlish regex I could use zero-width assertions either side thus
string.split(/(?<!\/)\/(?!\/)/)
but there's no lookbehind in Ruby - I wonder if someone could suggest a neat
ruby alternative.
thanks
alex
[snip]
but there's no lookbehind in Ruby - I wonder if someone could suggest a neat
ruby alternative.
with oniguruma this works:
'foo/bar // baz/qux'.split(/(?<!\/)\/(?!\/)/)
# => ["foo", "bar // baz", "qux"]
I am not sure exactly how you want it splitted.
How about this one?
'a/b//c///d////e/////f'.split(/\/(?!\/)/)
# => ["a", "b/", "c//", "d///", "e////", "f"]
On 10/2/05, Alex Fenton <alex@deleteme.pressure.to> wrote:
--
Simon Strandgaard
Don't know if your case is loose enough to allow for a hack like this, but maybe it will give you ideas:
irb(main):001:0> "foo/bar // baz/qux".gsub("//", "\0").split("/").map { |e| e.gsub("\0", "/") }
=> ["foo", "bar / baz", "qux"]
If that doesn't work, it's probably time to break out StringScanner...
James Edward Gray II
On Oct 2, 2005, at 2:06 PM, Alex Fenton wrote:
Hi
I'm trying to split a path (like an URL, but not one) into bits on the '/' character.
Problem is, I want to be able to be able to escape '/' in the names of bits by doubling
the character to '//' eg'foo/bar // baz/qux'
=> ["foo", "bar / baz", "qux"]In a perlish regex I could use zero-width assertions either side thus
string.split(/(?<!\/)\/(?!\/)/)but there's no lookbehind in Ruby - I wonder if someone could suggest a neat
ruby alternative.
Lookbehind is what you want; until we get it in core ruby, I've provided an example of another way to get the above. I've used commas instead of / just to help make the regexp easier to understand - substitute with "\/" as appropriate:
str = 'foo,bar , baz,qux,jorb,jing,blat'
out =
str.scan( /(.+?[^,],{2}*)(?:,(?!,)|$)/ ){ |a,b|
out << a.gsub( ',', ',' )
}
p out
#=> ["foo", "bar , baz", "qux,", "jorb", "jing,blat"]
On Oct 2, 2005, at 1:06 PM, Alex Fenton wrote:
I'm trying to split a path (like an URL, but not one) into bits on the '/' character.
Problem is, I want to be able to be able to escape '/' in the names of bits by doubling
the character to '//' eg'foo/bar // baz/qux'
=> ["foo", "bar / baz", "qux"]
Thanks all for your suggestions. I had forgotten scan was handy in situations like
this - I ended up going with a solution similar to Gavin's (again using commas
to avoid toothpick-itis)
str.scan /(?:[^,]|,)+/
Also helped me see that there are some tricky edge-cases where the separator
character is at the beginning or end of an element - I'm going to prohibit this in
the appplication.
cheers
alex
Gavin Kistner wrote:
On Oct 2, 2005, at 1:06 PM, Alex Fenton wrote:
I'm trying to split a path (like an URL, but not one) into bits on the '/' character.
Problem is, I want to be able to be able to escape '/' in the names of bits by doubling
the character to '//' eg'foo/bar // baz/qux'
=> ["foo", "bar / baz", "qux"]Lookbehind is what you want; until we get it in core ruby, I've provided an example of another way to get the above. I've used commas instead of / just to help make the regexp easier to understand - substitute with "\/" as appropriate:
str = 'foo,bar , baz,qux,jorb,jing,blat'
out =
str.scan( /(.+?[^,],{2}*)(?:,(?!,)|$)/ ){ |a,b|
out << a.gsub( ',', ',' )
}
p out
#=> ["foo", "bar , baz", "qux,", "jorb", "jing,blat"]
Whenever I find myself about to do something like the above, I say to myself:
"Hey, buddy, pre-allocating an array and shoving stuff onto it in a block is neat as an exercise of the closure, but you should be using something like #map."
Unfortunately, it would appear that #scan doesn't automagically map the returned value from each iteration to an array. Man, wouldn't that be nice?
Following is my hackish attempt to make a String#scan_and_map function that does the above.
A few questions for the gurus:
a) Is there a better way to deal with bol? with StringScanner? (Boy, it'd be nice if there was a Regexp#uses_bol_at_start_of_match? method.)
b) Is there a clean way to tell the 'arity' of a regexp (how many captures it has, at max)? (Boy, it'd be nice if there was a Regexp#arity method.)
c) Without knowing the arity, is there a clean/fast way to gather all the 1..n submatches held in StringScanner? (Boy, it'd be nice if StringScanner gave you access to an array of subcaptures as a single property. And if it set the $1..$9 vars.)
require 'strscan'
class String
def scan_and_map( regexp )
# A naive check for beginning of line
use_bol = regexp.inspect =~ /\/(?:\((?:\?:)?)*\^/
# A naive check for sub-expression groups
# Will fail for unescaped ( inside , for example
use_groups = regexp.inspect =~ /(\^|[^\\])\\{2}*\(/
results =
ss = StringScanner.new( self )
while !ss.eos?
ss.scan_until( regexp ) unless ss.match?( regexp )
if use_bol and not ss.bol?
ss.pos += 1
else
result = ss.scan( regexp )
if use_groups
result = (1..9).to_a.map{ |i| ss[i] }
end
results << yield( result )
end
end
results
end
end
str = 'foo,bar , baz,qux,jorb,jing,blat'
p str.scan_and_map( /(.+?[^,],{2}*)(?:,(?!,)|$)/ ){ |saved,others|
saved
}
#=> ["foo", "bar , baz", "qux,", "jorb", "jing,blat"]
On Oct 3, 2005, at 7:01 AM, Gavin Kistner wrote:
str = 'foo,bar , baz,qux,jorb,jing,blat'
out =
str.scan( /(.+?[^,],{2}*)(?:,(?!,)|$)/ ){ |a,b|
out << a.gsub( ',', ',' )
}
p out
#=> ["foo", "bar , baz", "qux,", "jorb", "jing,blat"]
Gavin Kistner wrote:
On Oct 2, 2005, at 1:06 PM, Alex Fenton wrote:
> I'm trying to split a path (like an URL, but not one) into bits on
> the '/' character.
> Problem is, I want to be able to be able to escape '/' in the names
> of bits by doubling
> the character to '//' eg
>
> 'foo/bar // baz/qux'
> => ["foo", "bar / baz", "qux"]Lookbehind is what you want; until we get it in core ruby, I've
provided an example of another way to get the above. I've used commas
instead of / just to help make the regexp easier to understand -
substitute with "\/" as appropriate:str = 'foo,bar , baz,qux,jorb,jing,blat'
out =
str.scan( /(.+?[^,],{2}*)(?:,(?!,)|$)/ ){ |a,b|
out << a.gsub( ',', ',' )
}
p out
#=> ["foo", "bar , baz", "qux,", "jorb", "jing,blat"]
str = 'foo,bar , baz,qux,jorb,jing,blat'
p str.scan( /(.+?[^,],{2}*)(?:,(?!,)|$)/ ).map{|x|
x.first.gsub( /,/, "," ) }
You can avoid toothpick-ities by using a custom regexp delimiter. E.g.
irb(main):001:0> "this/is/a/path".split(%r{/})
=> ["this", "is", "a", "path"]
instead of
irb(main):002:0> "this/is/a/path".split(/\//)
=> ["this", "is", "a", "path"]
You can even do
irb(main):003:0> "this/is/a/path".split(%r / )
=> ["this", "is", "a", "path"]
or even if there are no spaces in the regexp
irb(main):004:0> "this/is/a/path".split %r /
=> ["this", "is", "a", "path"]
(but that is dangerous because you need a trailing space and was done
more for fun)
regards,
Brian
On 03/10/05, Alex Fenton <alex@deleteme.pressure.to> wrote:
Thanks all for your suggestions. I had forgotten scan was handy in
situations like
this - I ended up going with a solution similar to Gavin's (again using
commas
to avoid toothpick-itis)str.scan /(?:[^,]|,)+/
Also helped me see that there are some tricky edge-cases where the separator
character is at the beginning or end of an element - I'm going to
prohibit this in
the appplication.cheers
alexGavin Kistner wrote:
> On Oct 2, 2005, at 1:06 PM, Alex Fenton wrote:
>
>> I'm trying to split a path (like an URL, but not one) into bits on
>> the '/' character.
>> Problem is, I want to be able to be able to escape '/' in the names
>> of bits by doubling
>> the character to '//' eg
>>
>> 'foo/bar // baz/qux'
>> => ["foo", "bar / baz", "qux"]
>
>
> Lookbehind is what you want; until we get it in core ruby, I've
> provided an example of another way to get the above. I've used commas
> instead of / just to help make the regexp easier to understand -
> substitute with "\/" as appropriate:
>
> str = 'foo,bar , baz,qux,jorb,jing,blat'
> out =
> str.scan( /(.+?[^,],{2}*)(?:,(?!,)|$)/ ){ |a,b|
> out << a.gsub( ',', ',' )
> }
> p out
> #=> ["foo", "bar , baz", "qux,", "jorb", "jing,blat"]
--
http://ruby.brian-schroeder.de/
Stringed instrument chords: http://chordlist.brian-schroeder.de/
Following is my hackish attempt to make a String#scan_and_map
function that does the above.
OK, that was dumb. On the way to work I realized that the above can far
more correctly be implemented simply as:
class String
def scan_and_map( regexp )
results =
scan( regexp ){ |args|
results << yield( args )
}
results
end
end
str = 'foo,bar , baz,qux,jorb,jing,blat'
p str.scan_and_map( /(.+?[^,],{2}*)(?:,(?!,)|$)/ ){ |saved,others|
saved
}
#=> ["foo", "bar , baz", "qux,", "jorb", "jing,blat"]
"Phrogz" <gavin@refinery.com> writes:
Following is my hackish attempt to make a String#scan_and_map
function that does the above.OK, that was dumb. On the way to work I realized that the above can far
more correctly be implemented simply as:class String
def scan_and_map( regexp )
results =
scan( regexp ){ |args|
results << yield( args )
}
results
end
endstr = 'foo,bar , baz,qux,jorb,jing,blat'
p str.scan_and_map( /(.+?[^,],{2}*)(?:,(?!,)|$)/ ){ |saved,others|
saved
}
p str.scan( /(.+?[^,],{2}*)(?:,(?!,)|$)/ ).map{ |saved,others|
saved
}
Trivialized?
--
Christian Neukirchen <chneukirchen@gmail.com> http://chneukirchen.org
Oh hell. Well done. 
I didn't realize that it returned something different with no block at
all. Nice work, ye who wrote #scan.