Bug in split with a positive lookahead assertion

[mitchty]

Seems to be introducing extraneous blank elements when used with split.

Example CamelCase=>snake_case convertor on ruby 1.9.3.

    $ ruby -v
    ruby 1.9.3p194 (2012-04-20 revision 35410) [x86_64-darwin11.3.0]
    $ pry
    main> 'FooBar'.split(/(?=[A-Z])/).map{|_| _.downcase}.join('_')
    => "foo_bar"
    main> 'FooBar'.split(/(?=[A-Z])/)
    => ["Foo", "Bar"]

And on a recentish (pulled today) rbx, the downcase on Nil is pointless.

    $ ruby -v
    rubinius 2.0.0dev (1.9.3 e817532e yyyy-mm-dd JI) [x86_64-apple-darwin11.3.0]
    $ rbx -S pry
    main> 'FooBar'.split(/(?=[A-Z])/).map{|_| _.downcase}.join('_')
    NoMethodError: undefined method `downcase' on nil:NilClass.
    from kernel/delta/kernel.rb:81:in `downcase (method_missing)'
    main> 'FooBar'.split(/(?=[A-Z])/)
    => ["Foo", "", "", "Bar"]

Jruby 1.6.7 appears to behave as ruby 1.9.3.

[mitchty]

Actually looking at this after i've had sleep it looks like I accidentally stumbled on two separate issues, one is the split with the regex the second is _ in map.

$ rbx -S pry
main> ["Foo", "", "", "Bar"].map {|_| _.downcase}.join('|')
NoMethodError: undefined method `downcase' on nil:NilClass.
from kernel/delta/kernel.rb:81:in `downcase (method_missing)'
main> ["Foo", "", "", "Bar"].map {|x| x.downcase}.join('|')
=> "foo|||bar"

Should I open a separate issue for the _ in map context or keep it in this issue? To be honest I rarely use _ like this but just happened to type it in pry this one time. 1.9.3-p194 behaves as expected.

[kronos]

I've written a patch that seems to fix a problem with split https://gist.github.com/81e25b635f26a1cb85d2

[ileitch]

@mitchty yes, please open a new issue for the _ issue.
@kronos does this fix an existing spec? If not, you'll need to add a spec along with your patch also.

[IPGlider]

I modified the spec the other day while working on this issue, but did not commit it. Seems that @kronos patch fix the problem.

[kronos]

Problem with "_" isn't fixed but I think it's another issue

[dbussink]

@kronos The issue with _ is indeed another issue, about to push a fix for that.