From: dblack@...
Date: 2003-01-24T12:31:00+09:00
Subject: Re: Bug concerning 1.6.8 and hash equality?

Hi --

On Fri, 24 Jan 2003, E F van de Laar wrote:

> I think I may have stumbled upon a bug in Ruby version 1.6.8. This issue
> arose by a unit test comparing the equality of different types of hashes.
> At first I thought the default behavior of 1.6.8 was correct; hashes
> with equal keys and values are equal when compared with "==" and are not
> equal if one of those hashes has been instantiated with a default value;
> in my case an empty array.
>
> In version 1.6.7 and 1.8.0-pre this behavior differs and both cases
> evaluate as true.  An example should make this all clear. :)
>
> emiel@chyrellos:/usr/local/ruby-1.6.8/bin$ ./irb
> irb(main):001:0> a = Hash.new []
> => {}
> irb(main):002:0> b = {}
> => {}
> irb(main):003:0> a == b
> => false
> irb(main):004:0> p RUBY_VERSION
> "1.6.8"
> => nil
>
> The question is which is the correct behavior? Personally I believe the
> behavior of 1.6.7 is correct; a hash is a hash no matter which default
> value it carries.

It's definitely not a bug; see ChangeLog:

  Wed Aug  7 09:51:54 2002  Yukihiro Matsumoto  <matz@ruby-lang.org>
     * hash.c (rb_hash_equal): should check default values.

I think the reasoning is as follows:

The notion of hash equality is really based in the idea of all the
key/value pairs being the same.  In other words, when this is true:

   a == b

then this should also be true:

   a[v] == b[v]

for every v.

But in 1.6.7, the following situation is possible:

   a == b           # true
   a[v] == b[v]     # false

That means that a == b being true doesn't really tell you anything
useful, because you can't use it to predict what the specific
key-by-key comparisons will give you.


David

-- 
David Alan Black
home: dblack@candle.superlink.net
work: blackdav@shu.edu
Web:  http://pirate.shu.edu/~blackdav