HI,
I am writing an extension for Ruby in C++ and as I have more and more
functions, I need to separate them into individual C++ source for
easier maintenance in the long run.
From Programming Ruby, the section on Extending Ruby defines function
as static, if I continue to use the static keyword and the functions
are implemented in separate C++ source file, it would not be visible in
the main ruby method registration file.
How significant is the static keyword when defining functions for use
as Ruby extension?
I had a quick browse of the ruby-netcdf package and noticed that they
do not prefix their function with the static keyword.
Anyone has advice?
Cheers
On 8/14/06, [email protected] [email protected] wrote:
As a general rule, I dislike exposing C++ as a library interface. I
prefer to write a light wrapper in C on top of the C++ functionality-
much easier to link with other code that way because you have much
more control over the names that the linker sees. Then, put the Ruby
wrapper on top of the C wrapper, and all of the Ruby extension
functions can be static, and it will all fit reasonably into one
source file (because the extension functions are only making calls to
another interface).
For an example, look at the extension code in EventMachine (under
branch version_0). The Ruby wrapper is in rubymain.cpp. The C wrapper
is in cmain.cpp. Everything else is C++.
Francis C. wrote:
For an example, look at the extension code in EventMachine (under
branch version_0). The Ruby wrapper is in rubymain.cpp. The C wrapper
is in cmain.cpp. Everything else is C++.
Thanks.
Looking at the code
static VALUE t_add_oneshot_timer (VALUE self, VALUE interval)
{
const char *f = evma_install_oneshot_timer (FIX2INT (interval));
if (!f || !*f)
rb_raise (rb_eRuntimeError, “no timer”);
return rb_str_new2 (f);
}
I have a further question, the parameter I get in my call is an array
e.g.
["pos",[ 0, 0, 0],"dir",[1, 0, 1],"custom",[ 0, 1, 1, 2, 2, 2,
3, 1, 1]]
they are token-value pair
Should I process the array in cmain.cpp or rubymain.cpp
If in cmain.cpp I would need to include “ruby.h” for the definition of
“VALUE” right?
Cheers
On 8/15/06, [email protected] [email protected] wrote:
Should I process the array in cmain.cpp or rubymain.cpp
If in cmain.cpp I would need to include “ruby.h” for the definition of
“VALUE” right?
You should follow whatever conventions make sense for your own code.
For my
code, I would answer your question by saying that all of the Ruby stuff
belongs in the extension wrapper (called rubymain.cpp in my case), so
your
extension code should convert the Ruby array to something directly
accessible from C. And therefore nothing but rubymain.cpp should include
ruby.h. That especially makes sense if you’re writing an extension to
wrap
up already-existing code that knows nothing about Ruby in the first
place.
EventMachine is completely new code, but it was designed to be
completely
independent of Ruby, and if you compile everything except rubymain, you
get
a fully-functional C++ library with a C interface that you can use from
any
language that can dynamically link code. It’s nice to do things in
well-defined layers like that.