[ragel-users] [PATCH] Fix generated code for Ruby 1.9 compatibility.

[Diego Elio Pettenò]

In Ruby 1.9, the String class no longer works as a C-style array of (8-bit)
characters, but supports multiple encoding. While it is obviously a task
for the developer to ensure that the data array passed to the
Ragel-generated code is in a compatible encoding, this also means that the
simple dereference is not going to work:

% ruby18 -e 'puts "foo"[0].class'
Fixnum
% ruby19 -e 'puts "foo"[0].class'
String

This is easily fixed by calling the #ord method on the dereferenced data,
which will provide the ASCII ordinal (or UNICODE codepoint) for the single
character.

The produced code works correctly both on Ruby 1.8 and 1.9.2.
---
 ragel/rubycodegen.cpp |    7 +++++--
 1 files changed, 5 insertions(+), 2 deletions(-)

diff --git a/ragel/rubycodegen.cpp b/ragel/rubycodegen.cpp
index 5117823..f329587 100644
--- a/ragel/rubycodegen.cpp
+++ b/ragel/rubycodegen.cpp
@@ -307,8 +307,11 @@ string RubyCodeGen::GET_KEY()
 		ret << ")";
 	}
 	else {
-		/* Expression for retrieving the key, use simple dereference. */
-		ret << DATA() << "[" << P() << "]";
+		/* Expression for retrieving the key, use dereference
+		 * and read ordinal, for compatibility with Ruby
+		 * 1.9.
+		 */
+		ret << DATA() << "[" << P() << "].ord";
 	}
 	return ret.str();
 }
-- 
1.7.3.3


_______________________________________________
ragel-users mailing list
ragel-users at complang.org
http://www.complang.org/mailman/listinfo/ragel-users