From: venkat <venkat@>
Date: 2007-10-05T06:25:02+09:00
Subject: Re: FasterCSV warning: alreadery initialized constant CSV

James Edward Gray II wrote:
>> I wrote a class called "CSVImport" using the default CSV class. Today, 
>> I modified it to work with FasterCSV class by changing the requires as 
>> below:
>>
>> begin
>>   require 'rubygems'
>>   require 'faster_csv'
>>   FasterCSV.build_csv_interface
>> rescue
>>   require 'csv'
>> end
>> # Rest of the code
> 
> You have a bug in that.  A bare rescue will not catch the LoadError we 
> are interested in this case.  Therefore, using just the above code, I 
> don't believe CSV would ever be loaded.  But we know it is loaded because�

I changed that to

    rescue LoadError

> My leading guess is that your code is loading CSV with some code you 
> didn't show me.

I think I found it.

I am using a module called CSVDBUtils:

# start of csvutils.rb
module CSVDBUtils
   #
end

require 'csvexp'
require 'csvimp'

CSVImport = CSVDBUtils::CSVImport
CSVExport = CSVDBUtils::CSVExport

# end of csvutils.rb

The csvdbutils.rb requires 'csvimp' and 'csvexp' and declares, for ease 
of use, CSVImport and CSVExport. If I don't do this, users of CSVImport 
and CSVExport classes have to instatiate with
	
	CSVDBUtils::CSVImport.new
	CSVDBUtils::CSVExport.new
	# any other classes added later.

I think the reason the warning was generated was that csvexp.rb 
requires 'csv'.

I think *FasterCSV#build_csv_interface* should include a check/test if 
it has been done already and not do it if the interface is built. I am 
not sure, how it can be done though.

In an application with different ruby source files using CSV and / 
FasterCSV, it would be difficult to make sure build_csv_interface is 
called only once or that CSV and FasterCSV aren't both used.

I changed the csvutils.rb to following and it doesn't generate any warnings.

# csvutils.rb
begin
   require 'rubygems'
   require 'faster_csv'
   FasterCSV.build_csv_interface
rescue LoadError
   require 'csv'
end
require 'csvexp'
require 'csvimp'
....
....
# rest of the code


Ofcourse, I had to removed requires from both csvimp and csvexp.

Thanks for quick response and I appreciate your help. I am able to 
reduce the time of import of 3000 rows from 90 seconds to less than 9 
seconds using batch import (by submitting in bulk instead of row by row) 
and FasterCSV.

Inorder to reduce the time further, I think I need to be able to collect 
the rows in a single call instead of iterating through rows. Is there a 
fill_array or each_n or similar?

-Venkat