From: "Y. NOBUOKA" <nobuoka@...>
Date: 2011-01-13T21:55:43+09:00
Subject: Re: Ruby 1.9.2 encoding issue

Please check +File#external_encoding+.

When +File#external_encoding+ is not +nil+, +File#write+ method tries to change
the encoding of the written string into +File#external_encoding+.
If the written data is binary string which contains 0x80 .. 0xFF and
+File#external_encoding+
is "UTF-8", +File#write+ method tries to encode the data, but it can't!!
So an exception will be raised. Please see following example.

 binary_data = [ 0x80, 0x81 ].pack( "C*" )
 File.open( "binary_data", File::CREAT | File::TRUNC | File::WRONLY,
0644 ) do |f|
   p f.external_encoding
     # #<Encoding:UTF-8>
   f.write(binary_data)
     # test.rb:7:in `write': "\x80" from ASCII-8BIT to UTF-8
(Encoding::UndefinedConversionError)
     #         from test.rb:7:in `block in <main>'
     #         from test.rb:7:in `open'
     #         from test.rb:7:in `<main>'
 end

How to avoid this issue?
You must open the file with BINARY MODE. The file object being binary
mode, the +File#write+
method doesn't encode the data automatically.
See below:

 binary_data = [ 0x80, 0x81 ].pack( "C*" )
 File.open( "binary_data", "w+b", 0644 ) do |f|
   p f.external_encoding
     # #<Encoding:ASCII-8BIT>
   f.write(binary_data)
     # OK. NO ERROR!
 end

cf. http://ruby-doc.org/ruby-1.9/classes/IO.html


-- 
nobuoka