[ruby-talk:16763] uniq and "equivalence"

Array.uniq seems to rely on deleting multiple references to the same object.
Furthermore, this test for sameness seems to be based on object.id rather
than == or ===.  Most of the time this is right and proper, and also 
efficient. I support my belief that this is going on with:

#!/usr/local/bin/ruby -w

class Trapezoidal_fitness

    # Basic trapezoidal fitness funtion class
    # Instance vars are (upper|lower) (left|right) corners
    # of the tapezoid.  The upper points can be co-incident
    # for a triangular fitness function.

    attr_reader :ll, :ul, :ur, :lr

    def initialize(ll, ul, ur, lr)
        @ll, @ul, @ur, @lr = ll, ul, ur, lr
    end 

    def ==(other)
        print "in == self is #{self.inspect}, other is #{other.inspect}\n"
        result = (self.type == other.type) &&
        ( (@ll == other.ll) &&
          (@ul == other.ul) &&
          (@ur == other.ur) &&
          (@lr == other.lr) )
        print "in == result is #{result}\n"
        result
    end

    def |(other)
       [self,other].uniq
    end

end 

if __FILE__ == $0
    require "runit/testcase"
    require 'runit/cui/testrunner'
    require 'runit/testsuite'

    class  Test_Trapezoidal_fitness < RUNIT::TestCase
        def test_fuzzy_or
            first  = Trapezoidal_fitness.new(99.0, 101.0, 101.0, 103.0)
            second = Trapezoidal_fitness.new(99.0, 101.0, 101.0, 103.0)
            assert_equal([first], first | second )
        end
    end

    RUNIT::CUI::TestRunner.run(Test_Trapezoidal_fitness.suite)
end

which gives:

/usr/local/lib/ruby/site_ruby/1.6/runit/assert.rb:146: warning: concat (...) interpreted as method call
/usr/local/lib/ruby/site_ruby/1.6/runit/assert.rb:158: warning: concat (...) interpreted as method call

Test_Trapezoidal_fitness#test_fuzzy_or F.
Time: 0.002461
FAILURES!!!
Test Results:
 Run: 1/1(1 asserts) Failures: 1 Errors: 0
Failures: 1
fit_tmp.rb:42:in `test_fuzzy_or'(Test_Trapezoidal_fitness): expected:<[#<Trapezoidal_fitness:0xd11a0  @lr=103.0, @ur=101.0, @ul=101.0, @ll=99.0>]> but was:<[#<Trapezoidal_fitness:0xd11a0  @lr=103.0, @ur=101.0, @ul=101.0, @ll=99.0>, #<Trapezoidal_fitness:0xd1170  @lr=103.0, @ur=101.0, @ul=101.0, @ll=99.0>]> (RUNIT::AssertionFailedError)
	from fit_tmp.rb:46

So the | operator did not do the uniq as intended, and that can only be
because the objects were deemed to be different.

When this behaviour of uniq is not not what I want, how do I change it?

Possible RCR: uniq and uniq! might take a block specifying
an equality operator, similar to sort's comparison block.  This could
have applications for more complicated ideas of equivalence such as when
"Sasse, Hugh", "Sasse, H. G." and "Hugh Sasse" are to be considered the
same.

	Hugh
	[e-mail]