From: Robert Klemme <shortcutter@...>
Date: 2009-05-12T03:45:01+09:00
Subject: Re: File.basename bug?

On 11.05.2009 20:21, Reid Thompson wrote:
> On Tue, 2009-05-12 at 03:10 +0900, lasitha wrote:
>> On Mon, May 11, 2009 at 9:14 PM, Reid Thompson <reid.thompson@ateb.com> wrote:
>>> How often do people get bitten by this?
>>> When operating on a file in the current directory, './' must be
>>> prepended to a filename for File.basename to work...
>> Hi, i'm either not understanding your issue or not seeing it on my box:
>>
>> $: cd /tmp; touch by.touch
>>
>> $: ruby19 -ve "puts File.basename 'by.touch'"
>> ruby 1.9.1p0 (2009-01-30) [i386-darwin8.11.1]
>> by.touch
>>
>> $: ruby   -ve "puts File.basename 'by.touch'"
>> ruby 1.8.7 (2008-08-11 patchlevel 72) [i686-darwin8]
>> by.touch
>>
>>> $ cat bug.rb
>>> DATAFORMAT_VERSION=1
>>>
>>> def processInputFile(inputFpath)
>>>    puts "inputFpath #{inputFpath}"
>>>    dirpath = File.dirname(inputFpath)
>>>
>>>    outputfile = File.basename(inputFpath)
>>>    outputfile.gsub!("pp_",'')
>>>    outputfile = dirpath + '/' + 'SDF_' + "#{DATAFORMAT_VERSION}" + '_' + outputfile
>>>    puts "outputfile #{outputfile}"
>>>    puts "inputFpath #{inputFpath}"
>>>
>>> end
>>>
>>> processInputFile("pp_a_test_file_name")
>>> rthompso@raker>/home/rthompso
>>> $ ruby bug.rb
>>> inputFpath pp_a_test_file_name
>>> outputfile ./SDF_1_a_test_file_name
>>> inputFpath a_test_file_name
>>> rthompso@raker>/home/rthompso
>>> $ ruby -v
>>> ruby 1.8.7 (2009-04-08 patchlevel 160) [i686-linux]
>>>
>>> rthompso@raker>/home/rthompso
>>> $ vi bug.rb
>>> rthompso@raker>/home/rthompso
>>> $ ruby bug.rb
>>> inputFpath /home/rthompso/pp_a_test_file_name
>>> outputfile /home/rthompso/SDF_1_a_test_file_name
>>> inputFpath /home/rthompso/pp_a_test_file_name
>> Scanned over the above but couldn't spot what you might be on about :)
>> Perhaps you could clarify or wittle the example down a bit further?
>>
>> cheers,
>> lasitha
>>
> 
> example 1 -> processInputFile("pp_a_test_file_name")
> "pp_a_test_file_name" is a string object
> File.basename sets outputfile to the same object as
> "pp_a_test_file_name".
> as a result the gsub! action on outputfile acts on the
> "pp_a_test_file_name" object.
> 
> example 2 -> processInputFile("./pp_a_test_file_name")
> "./pp_a_test_file_name" is a string object
> File.basename sets outputfile to a new object as "pp_a_test_file_name".
> as a result the gsub! action on outputfile acts on the new object,
> without modifying the original "./pp_a_test_file_name" 

Heureka!  It's an aliasing problem which is not there in 1.9:

$ irb19
irb(main):001:0> s = 'foo.txt'
=> "foo.txt"
irb(main):002:0> x = File.basename s
=> "foo.txt"
irb(main):003:0> s.equal? x
=> false
irb(main):004:0> s.object_id
=> 134667930
irb(main):005:0> x.object_id
=> 134637100
irb(main):006:0>

$ irb
irb(main):001:0> s = 'foo'
=> "foo"
irb(main):002:0> x = File.basename s
=> "foo"
irb(main):003:0> s.equal? x
=> true
irb(main):004:0> s.object_id
=> 1073538940
irb(main):005:0> x.object_id
=> 1073538940
irb(main):006:0>

Whether you consider this actually a bug of File.basename - I don't 
know.  It is at least a documentation bug because the docs do not 
mention that fact.

Kind regards

	robert

-- 
remember.guy do |as, often| as.you_can - without end
http://blog.rubybestpractices.com/