From: James Edward Gray II <james@...>
Date: 2006-04-28T01:42:26+09:00
Subject: Re: Q about the FasterCSV

On Apr 27, 2006, at 10:52 AM, Dave Burt wrote:

> James Edward Gray II wrote:
>> On Apr 27, 2006, at 10:28 AM, Dave Burt wrote:
>>>       :leading_fields =>
>>>         /\A#{Regexp.escape(@col_sep)}+/,      # for empty leading  
>>> fields
>>
>> You should modify the above line too.  It takes both to correctly  
>> parse
>> some lines:
>>
>> /\A\s*#{Regexp.escape(@col_sep)}+/
>
> I looked at this, but I deduced from [1] that a number of fields equal
> to the match size are added, so (I guess) "  ,, foo" would get extra
> leading fields: [nil, nil, nil, nil, "foo"]. So I skipped it. I'm also
> guessing the OP doesn't need it, anyway.
>
>> Looks good to me.  Just don't hold your breath waiting on the  
>> patch...  ;)
>
> Oh, I don't want the patch. It's a terrible idea! "foo, bar, 'baz'"
> aren't CSV, they're CASWSSV (comma and some white-space separated
> values). That's got to be a whole new library :)
>
> Cheers,
> Dave
>
> [1] faster_csv.rb lines 1114..1115:
>       csv = if parse.sub!(@parsers[:leading_fields], "")
>         [nil] * $&.length
>
> P.S.: There's a bug here, and not just here, I think. Maybe
> init_separators should raise an exception if @col_sep.size != 1, or  
> use
> options[:col_sep][0,1]. It currently barfs late and in various
> interesting ways for multi-character values of col_sep.

Good points all around.  Dave knows this code better than I do,  
clearly.  ;)

James Edward Gray II