How to do conditional compile for Ruby 1.9 in a C extension?

[Jason Garber]

I'm working on updating RedCloth for Ruby 1.9. Since the output of
#singleton_methods changed from strings to symbols, I think I need a
conditional compile. Here's what I tried:

#if RUBY_VERSION < 0x10900
if (rb_ary_includes(rb_funcall(self, rb_intern("singleton_methods"),
0), btype)) {
#else
if (rb_ary_includes(rb_funcall(self, rb_intern("singleton_methods"),
0), rb_str_intern(btype))) {
#endif

 

[Jason Garber]

Sorry, ruby-forum cut off the rest of what I wrote. The above code
example is doing the first way for both 1.8 and 1.9. What's the proper
way to detect the ruby version or the feature?

Thanks!
Jason Garber

 

[Rick DeNatale]

[Note: parts of this message were removed to make it a legal post.]

I'm working on updating RedCloth for Ruby 1.9. Since the output of
#singleton_methods changed from strings to symbols, I think I need a
conditional compile. Here's what I tried:

#if RUBY_VERSION < 0x10900
if (rb_ary_includes(rb_funcall(self, rb_intern("singleton_methods"),
0), btype)) {
#else
if (rb_ary_includes(rb_funcall(self, rb_intern("singleton_methods"),
0), rb_str_intern(btype))) {
#endif

from version.h in Ruby 1.9
define RUBY_VERSION "1.9.2"
#define RUBY_RELEASE_DATE "2009-02-16"
#define RUBY_PATCHLEVEL -1
#define RUBY_BRANCH_NAME "trunk"

#define RUBY_VERSION_MAJOR 1
#define RUBY_VERSION_MINOR 9
#define RUBY_VERSION_TEENY 1
#define RUBY_RELEASE_YEAR 2009
#define RUBY_RELEASE_MONTH 2
#define RUBY_RELEASE_DAY 16

and for Ruby 1.8.6
#define RUBY_VERSION "1.8.6"
#define RUBY_RELEASE_DATE "2007-11-12"
#define RUBY_VERSION_CODE 186
#define RUBY_RELEASE_CODE 20071112
#define RUBY_PATCHLEVEL 5000

#define RUBY_VERSION_MAJOR 1
#define RUBY_VERSION_MINOR 8
#define RUBY_VERSION_TEENY 6
#define RUBY_RELEASE_YEAR 2007
#define RUBY_RELEASE_MONTH 11
#define RUBY_RELEASE_DAY 12


so I think that you want something like

#if RUBY_VERSION < '1.9.0'
do pre 1.9 stuff
#else
do 1.9 stuff
#end

Matz is careful to use Ruby version numbers which can be compared using
string comparison.

or

#if RUBY_VERSION_MAJOR = 1 && RUBY_VERSION_MINOR == 9
do 1.9 stuff
#else
do 1.8 stuff
#end

or some variant on those.

--
Rick DeNatale

Blog: http://talklikeaduck.denhaven2.com/
Twitter: http://twitter.com/RickDeNatale
WWR: http://www.workingwithrails.com/person/9021-rick-denatale
LinkedIn: http://www.linkedin.com/in/rickdenatale

 

[Jason Garber]

Matz is careful to use Ruby version numbers which can be compared using

string comparison.

Nice! I didn't know that, so where I'd seen string comparisons on the
version number before, I just thought it was foolhardy.

Thanks for the help.

Jason

 

[Jason Garber]

It's not working for me. In my C file I tested the presence of those
variables like so:

#ifdef RUBY_VERSION_MAJOR
printf("RUBY_VERSION_MAJOR defined\n");
#endif
#ifndef RUBY_VERSION_MAJOR
printf("RUBY_VERSION_MAJOR NOT defined ******\n");
#endif

and it outputs RUBY_VERSION_MAJOR NOT defined ******

Why aren't the RUBY_VERSION constants defined?

 

[Rick DeNatale]

[Note: parts of this message were removed to make it a legal post.]

did you include version.h?

It's not working for me. In my C file I tested the presence of those
variables like so:

#ifdef RUBY_VERSION_MAJOR
printf("RUBY_VERSION_MAJOR defined\n");
#endif
#ifndef RUBY_VERSION_MAJOR
printf("RUBY_VERSION_MAJOR NOT defined ******\n");
#endif

and it outputs RUBY_VERSION_MAJOR NOT defined ******

Why aren't the RUBY_VERSION constants defined?

--
Rick DeNatale

Blog: http://talklikeaduck.denhaven2.com/
Twitter: http://twitter.com/RickDeNatale
WWR: http://www.workingwithrails.com/person/9021-rick-denatale
LinkedIn: http://www.linkedin.com/in/rickdenatale

 

[Jason Garber]

version.h couldn't be found. I've read that version.h is deprecated in
Ruby 1.9. There's either got to be another way to detect the change or
a way to work around it. I'd just use map! to convert the strings to
symbols if I knew how to do it in C.

 

[Tim Hunter]

version.h couldn't be found. I've read that version.h is deprecated in
Ruby 1.9. There's either got to be another way to detect the change or
a way to work around it. I'd just use map! to convert the strings to
symbols if I knew how to do it in C.

I'm not saying this is the best way (maybe Nobu can provide a better
answer), but if you can be sure that mkmf.rb is being executed by the
same version of Ruby, then you could cook up a feature test in your
extconf.rb that would create a preprocessor symbol to test. Determine
what singleton_methods is doing, then call $defs.push to set a symbol:

$defs.push("-DNEW_SINGLETON_METHODS")

 

[Nobuyoshi Nakada]

Hi,

At Fri, 20 Feb 2009 02:21:24 +0900,
Jason Garber wrote in [ruby-talk:328775]:

I'm working on updating RedCloth for Ruby 1.9. Since the output of
#singleton_methods changed from strings to symbols, I think I need a
conditional compile. Here's what I tried:

The output? You convert the argument.

#if RUBY_VERSION < 0x10900
if (rb_ary_includes(rb_funcall(self, rb_intern("singleton_methods"),
0), btype)) {
#else
if (rb_ary_includes(rb_funcall(self, rb_intern("singleton_methods"),
0), rb_str_intern(btype))) {
#endif

And #singleton_methods takes a flag, anything not only string
and symbol can be true.

 

[Jason Garber]

Sorry, nobu, I didn't follow you. Try again?

 

[Nobuyoshi Nakada]

Hi,

At Fri, 20 Feb 2009 18:58:55 +0900,
Jason Garber wrote in [ruby-talk:328856]:

Sorry, nobu, I didn't follow you. Try again?

Sorry, I misread the parentheses.

I'm working on updating RedCloth for Ruby 1.9. Since the output of
#singleton_methods changed from strings to symbols, I think I need a
conditional compile. Here's what I tried:

It's a problem indeed.

#if RUBY_VERSION < 0x10900
if (rb_ary_includes(rb_funcall(self, rb_intern("singleton_methods"),
0), btype)) {
#else
if (rb_ary_includes(rb_funcall(self, rb_intern("singleton_methods"),
0), rb_str_intern(btype))) {
#endif

Do you really need to know only if singleton method is defined,
not just self responds to it?

We may have to add a new macro in ruby.h.

 

[Jason Garber]

I ended up sidestepping the whole issue completely and writing a new
accessor, #formatter_methods, that just returned singleton_methods.map!
{|method| method.to_sym }. The reasons for needing #singleton_methods
and not #respond_to? or #method_defined? are hard to explain, but you
can probably understand it fairly quickly by looking at the source:
http://github.com/jgarber/redcloth/commit/c941af12fff3f93154578c160aed0727be98dff0

A macro would've been nice, but it worked out okay for me anyway.
Thanks for your help!

Jason

 

[Nobuyoshi Nakada]

Hi,

At Sat, 21 Feb 2009 04:48:49 +0900,
Jason Garber wrote in [ruby-talk:328928]:

A macro would've been nice, but it worked out okay for me anyway.

Do you have any suggestion for the name? My candidate is
USE_SYMBOL_AS_METHOD_NAME.

 

[Jason Garber]

Hi,


Do you have any suggestion for the name? My candidate is
USE_SYMBOL_AS_METHOD_NAME.

Sounds great to me! Thanks!