From: "NAKAMURA, Hiroshi" <nahi@...>
Date: 2003-02-26T11:02:46+09:00
Subject: Re: Regexp help: Parsing a CSV file

Hi, Brian,

Thank you for your comment.

> From: Brian Candler [mailto:B.Candler@pobox.com] 
> Sent: Tuesday, February 25, 2003 11:25 PM

> On Tue, Feb 25, 2003 at 08:01:43PM +0900, NAKAMURA, Hiroshi wrote:
> > > class String
> > >   def csv_to_a
> > >     CSV.parse_line(self)
> > >   end
> > > end
> > > 
> > > class Array
> > >   def to_csv
> > >     CSV.generate_line(self)
> > >   end
> > > end
> > 
> > parse_line and generate_line are methods of CSV2.
> > CSV2 has not yet been released.
> 
> That looks like it will be much nicer, thank you.

Differences are here.
http://www.ruby-lang.org/cgi-bin/cvsweb.cgi/lib/csv/lib/csv.rb?rev=1.17&content-type=text/x-cvsweb-markup
I added one-line parsing/generating methods
(parse_line/generate_line) but made no change for
multi-line parsing/generating.
'line' here is one record in CSV format, not one
line in file.  One CSV line might includes \r or \n
in its fields.

> What happens if the line being parsed is incomplete because 
> it contains a
> quoted newline?

In current API,

class String
  def csv_to_a
    CSV::Reader.parse(self) do |row|
      yield row.to_a
    end
  end
end

$ ruby -rthis_article -e '
  %Q["\n","\n"\r\n"\r","\r"].csv_to_a do |row|
    p row
  end
'
=>
["\n", "\n"]
["\r", "\r"]

> Perhaps it could take a block to fetch any 
> additional lines
> needed, e.g.
> 
>   while line = gets
>     data = CSV.parse_line(line) { gets }
>   end

CSV::Reader.parse(ARGF) { |row| p row.to_a }
...

Oops.  CSV module tries to call ARGF#sysread and crashes...

> In that case it would probably be better to use a block in 
> the first place,
> i.e.
> 
>   while data = CSV.parser { gets }
>     # 'data' is an array containing the cells for one logical row
>   end

Interesting.  As above, current API uses block for
output stream but it might be good to use block for
input stream.  Let me consider.

> The simpler the API the better :-)

Definitely.

Regards,
// NaHi