From: William Djaja Tjokroaminata <billtj@...>
Date: 2002-01-25T01:56:24+09:00
Subject: Serious Array Bug in Ruby 1.6.6?

Hi,

I am currently using Ruby 1.6.6 on Linux Mandrake 8.1.  I want to create
an array of hashes, so according to an example in the "Programming
Ruby" book on the new method for class array
(http://www.rubycentral.com/book/ref_c_array.html#Array.new), I wrote

    a = Array.new (2, Hash.new)
    a[0]['key'] = 'value0'
    a[1]['key'] = 'value1'
    puts a[0]['key']
    puts a[1]['key']

Of course (?), this gives

    value1
    value1

On the other hand, the following code works:

    a = Array.new (2)
    for i in 0...a.length
        a[i] = Hash.new
    end
    a[0]['key'] = 'value0'
    a[1]['key'] = 'value1'
    puts a[0]['key']
    puts a[1]['key']

which gives

    value0
    value1

Definitely the book has said "...that is, anInteger references to the same
anObject".  Of course, after reading this, everything is
obvious.  However, it took me several hours to figure this out until I
read the actual description.

At least, probably the book should not give that example.  In the more
general terms, probably Ruby should change the behavior of Array.new (n,
Hash.new) as opposite to, say, Array.new (n, "A").  By having all the
members refer to the same hash object, Ruby has violated the Priciple
of Least Surprise.  Yes, probably it is more consistent in the way Ruby
behaves currently.  However, when a person creates an array of hashes, I
don't think there is anybody who wants all the hashes to refer to the same
thing, because it is useless.

Regards,

Bill