overriding "y" in kernel causes strange behavior in parslet

[ruby]

overriding "y" in kernel causes strange behavior in parslet

test case: https://gist.github.com/1974317 (sorry, i tried to include it a couple times but didn't know what I was doing)

[tenderlove]

I'm not totally opposed, but y has been defined on kernel since 1.8 (even when requiring syck):

[aaron@higgins rails (master)]$ ruby -v -ryaml -e'p method(:y)'
ruby 1.8.7 (2010-01-10 patchlevel 249) [universal-darwin11.0]
#<Method: Object(Kernel)#y>
[aaron@higgins rails (master)]$

Do you have suggestions for a replacement that would be backwards compatible, yet still provide the y method?

I'm not sure if there's a way to provide the y method without adding it to the Kernel module. There is an easy way to fix it from the Parslet side: the Context class extends from BlankSlate, but in the future it should probably extend BasicObject (this prevents Kernel from being included and fixes the problem). I'm guessing it's using BlankSlate for now for backwards compatibility reasons; BasicObject was only added in 1.9.3.

This only manifests itself in 1.9.3 because Psych became the default yml parser. I'm questioning the usefulness of this method, given that it's added as private and would probably only be used for debugging purposes (though admittedly, I'm new to Psych and haven't gone too deep into anything that uses it). Maybe add a deprecation warning when that method is called, and then removing it in a future release?

[tenderlove]

I'm guessing if you do require "syck" on 1.9.3, you'll get the same issue. I don't think it's unique to just psych.

I think we could push the y method up to Object and everything would still work, but I'd rather just remove the method. The only time I use this method is inside irb, so maybe conditionally loading it if irb is present would be a sufficient compromise.

[kschiess]

I'll fix it on the parslet side as well. This kind of namespace pollution is rather frequent in Ruby libraries and really hinders interop. My fix will be to really define methods instead of using method_missing.

Join the holy crusade: jimweirich/rake#81

;)

I'm fine with whatever works - the irb solution sounds fine, if there's an easy way to detect that. I've already patched the code on my end, so I'm not too worried about it; just wanted to make sure that nobody would run into the same issue.

[kschiess]

The newest release of parslet (1.3.0) contains the fix.