From: Brian Candler <B.Candler@...>
Date: 2003-08-05T23:11:51+09:00
Subject: Re: Problem with ruby-dbi-all-0.0.20 and/or ruby-1.8.0


--zYM0uCDKw75PZbzx
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline

On Tue, Aug 05, 2003 at 10:40:06PM +0900, Brian Candler wrote:
> The following test program works correctly with ruby-dbi-all-0.0.18 under
> ruby-1.6.8, but not ruby-dbi-all-0.0.20 under ruby-1.8.0:

OK, rolling back to dbi-0.0.18 didn't fix this, so I've dug around some more.

The following program demonstrates the problem without any database access:

---- 8< ------------------------------------------------------------------
#!/usr/local/bin/ruby
require 'dbi/row'

res = []
r = DBI::Row.new(["col1","col2"],[nil,nil])

[["one",1],["two",2],["three",3]].each do |x,y|
  r["col1"] = x
  r["col2"] = y
  p r
  res << r.dup
end
p res    # [["three", 3], ["three", 3], ["three", 3]] under ruby-1.8.0
---- 8< ------------------------------------------------------------------

Now, looking at the definition of DBI::Row, it is based on delegation to
Array. So I can boil it down to the following:

---- 8< ------------------------------------------------------------------
#!/usr/local/bin/ruby -w
require 'delegate'

class Foo < DelegateClass(Array)   # distilled from DBI::Row
  def initialize(arr)
    @arr = arr
    super(@arr)
  end
  def clone_with(new_values)
    Foo.new(new_values)
  end
  def clone
    clone_with(@arr.dup)
  end
end

r = Foo.new([nil,nil])
res = []

[["one",1],["two",2],["three",3]].each do |x,y|
  r[0] = x
  r[1] = y
  p r
  res << r.dup
end
p res
---- 8< ------------------------------------------------------------------

$ ruby16 -v bug2.rb
ruby 1.6.8 (2002-12-24) [i386-freebsd4.7]
["one", 1]
["two", 2]
["three", 3]
[["one", 1], ["two", 2], ["three", 3]]           <<< expected
$ ruby18 -v bug2.rb
ruby 1.8.0 (2003-08-04) [i386-freebsd4.7]
["one", 1]
["two", 2]
["three", 3]
[["three", 3], ["three", 3], ["three", 3]]       <<< actual under 1.8.0

So, it looks like DBI is depending on some behaviour of delegate, and/or
'dup' making use of 'clone', which is no longer true in 1.8.0

But I don't know how to fix this; adding my own 'dup' method to class Foo
doesn't seem to help. Anybody got any ideas?

Thanks,

Brian.

--zYM0uCDKw75PZbzx
Content-Type: message/rfc822
Content-Disposition: inline

Date: Tue, 5 Aug 2003 14:39:54 +0100
From: Brian Candler <B.Candler@pobox.com>
To: ruby-talk@ruby-lang.org
Subject: Problem with ruby-dbi-all-0.0.20 and/or ruby-1.8.0
Message-ID: <20030805143954.A32611@linnet.org>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
User-Agent: Mutt/1.2.5i

The following test program works correctly with ruby-dbi-all-0.0.18 under
ruby-1.6.8, but not ruby-dbi-all-0.0.20 under ruby-1.8.0:

------------------------------------------------------------------
#!/usr/local/bin/ruby -w
require 'dbi'
LIMIT = 100

d = DBI.connect('dbi:Mysql:somedatabase','root','')

sql = "select col1,col2 from anytable"
rows = []
d.execute(sql) do |sth|
  loop do
    row = sth.fetch
    p row   # prints the right answer
    break if row.nil?
    rows << row.dup
    break if rows.size >= LIMIT
  end
end

p rows # prints the wrong answer
------------------------------------------------------------------

The problem is that the array 'rows' contains the last row repeated N times,
instead of N different rows. This is despite the use of 'row.dup'.
The object_id's of the elements of row are all different, but the contents
are the same.

I used mysql-ruby-2.4.4 in both cases.

Regards,

Brian.

--zYM0uCDKw75PZbzx--