From: "NAKAMURA, Hiroshi" <nahi@...>
Date: 2004-05-20T12:19:48+09:00
Subject: Re: CSV bug?

Hi,

SER wrote:
> BTW, the csv.rb is really nice.  I wanted to thank you for it.  I'm
> doing a subversive Ruby project at work, and was doing it "the hard
> way" before I discovered csv.rb.  Now the app is smaller and more
> robust.

Great to hear.  Thank you and all users.


I recently added a feature which allows multi-char sequence as a record 
separator and field separator.  And some incompatible changes such as 
CSV::Row -> Array and CSV::Cell -> String.  It passes all tests as 
before so I believe the newer version will work for my "the hard way" 
thing, but API incompatibility will cause problems for many users. 
Sorry for users in advance.

With the new version, you can do;

   0% ruby -rcsv -e '
   CSV::Reader.parse("a||b==c||d", "||", "==") { |r| p r }
   '
   ["a", "b"]
   ["c", "d"]

   0% cat matrix.csv; echo
    | |o
   -+-+-
   x|x|o
   -+-+-
   o| |

   0% ruby -rcsv -e '
   CSV.parse("matrix.csv", "|", "\n-+-+-\n") { |r| p r }
   '
   [" ", " ", "o"]
   ["x", "x", "o"]
   ["o", " ", nil]

And say your boss doesn't like open source because of security;

   0% cat src.rb
   require 'openssl'
   text = ARGV.shift
   key = 'it is my secret key!'
   des = OpenSSL::Cipher::DES.new( 'EDE3', 'CBC' )
   des.decrypt( key )
   result = des.update( text.unpack( 'm' )[0] )
   result += des.final
   p result

   0% ruby src.rb DMZZ+eRpyQbv+nRlr6Y5BQ==
   "hello world"

   0% ruby -rcsv -e '
   CSV.generate("dest.csv", "ae", "oe") { |w|
     CSV.parse("src.rb", " ") { |r| w << r }
   }
   '

   0% cat dest.csv; echo
 
requireae'openssl'oetextae=aeARGV.shiftoekeyae=ae'itaeisaemyaesecretaekey!'oedesae=aeOpenSSL::Cipher::DES.new(ae'EDE3',ae'CBC'ae)oedes.decrypt(aekeyae)oeresultae=aedes.update(aetext.unpack(ae'm'ae)[0]ae)oeresultae+=aedes.finaloepaeresultoe

   0% ruby -rcsv -e '
   CSV.generate("dest.rb", " ") { |w|
     CSV.parse("dest.csv", "ae", "oe") { |r| w << r }
   }
   '

   0% diff src.rb dest.rb

   0%

:-)

The current version is in csv's repository located at
http://www.ruby-lang.org/cgi-bin/cvsweb.cgi/lib/csv/
and I'll merge it to ruby's repository tonight.  Of course at HEAD(1.9). 
  What do you think about the branch for 1.8?  I'm afraid that 1.8 users 
don't want incompatibilities...

Here's the summary of API change (excerpted from README.txt);

[CAUTION] API change: CSV::Row removed.  A row is represented as just an 
Array.  Since CSV::Row was a subclass of Array, it won't hurt almost all 
programs except one which depended CSV::Row#match.

[CAUTION] API change: CSV::Cell removed.  A cell is represented as just 
a String or nil(NULL).  This change will cause widespread destruction.

   CSV.open("foo.csv", "r") do |row|
     row.each do |cell|
       if cell.is_null       # Cell#is_null
         p "(NULL)"
       else
         p cell.data         # Cell#data
       end
     end
   end

must be just;

   CSV.open("foo.csv", "r") do |row|
     row.each do |cell|
       if cell.nil?
         p "(NULL)"
       else
         p cell
       end
     end
   end

[CAUTION] record separator(CR, LF, CR+LF) behavior change: CSV.open,
CSV.parse, and CSV,generate now do not force opened file binmode. 
Formerly it set binmode explicitly.  With CSV.open, binmode of opened 
file depends the given mode parameter "r", "w", "rb", and "wb". 
CSV.parse and CSV.generate open file with "r" and "w".  Setting mode 
properly is user's responsibility now.

Regards,
// NaHi