From: "Guy N. Hurst" <gnhurst@...>
Date: 2002-04-21T06:12:21+09:00
Subject: problem with alias

I just discovered a problem with how alias works while
testing some code in irb.

It is common in ruby to alias a method you are about to
override so you can still access the original functionality.
However, if that portion of code is executed more than
once (such as being included or evaluated more than once),
then the original method gets lost.

In short,

alias old_meth meth
alias old_meth meth

will result in meth being lost forever.

Below is an example:

irb(main):001:0> "%s" % 5
"5"
irb(main):002:0> class String       
irb(main):003:1>   alias old_sf %
irb(main):004:1*   def %(other)
irb(main):005:2>     puts "dummy test:"
irb(main):006:2>     old_sf other
irb(main):007:2>   end
irb(main):008:1> end
nil
irb(main):009:0> "%s" % 5
dummy test:
"5"
irb(main):010:0> "%s".old_sf 5
"5"
irb(main):011:0> class String       
irb(main):012:1>   alias old_sf %
irb(main):013:1*   def %(other)
irb(main):014:2>     puts "dummy test:"
irb(main):015:2>     old_sf other
irb(main):016:2>   end
irb(main):017:1> end
nil
irb(main):018:0> "%s" % 5
dummy test:
dummy test:
dummy test:
dummy test:
dummy test:
:
:
(endless)


I just did a search and found some other threads discussing
problems with alias:

http://ruby-talk.com/34131
http://ruby-talk.com/28977

However, I am pointing out something that is slightly different,
since it is not so much an issue of namespace pollution or of
keeping the 'prior' version, as it is keeping the 'original' version!

Perhaps a companion to #send would work, such as #send_original ?
Yet, I suppose someone could override *that* and cause problems.
So it really does point to the possible need for a new keyword, if
not an alteration to the existing functionality of alias. Perhaps
adding a third parameter indicating whether the original or existing
method should be aliased?

alias old_sf % true   # true means use original
alias old_meth meth false  # false means don't use original; this is default
alias old_meth meth   # same as using false for third param

Comments?

Guy N. Hurst