From: Charles Mills <cmills@...>
Date: 2004-11-20T05:12:15+09:00
Subject: Re: [ARTICLE] Extending Ruby with C

On Nov 19, 2004, at 11:58 AM, Daniel Berger wrote:

> gabriele renzi <rff_rff@remove-yahoo.it> wrote in message 
> news:<wQknd.40392$Ni.1406459@twister1.libero.it>...
>> An article from Garrett Rooney just appeared at OnLamp.com
>> you can read it at
>> http://www.onlamp.com/pub/a/onlamp/2004/11/18/extending_ruby.html
>> (....)
>
> One thing I wish C extension authors would stop doing is manually
> counting variable length argument lists.  From the article:
>
> static VALUE
> writer_begin_element (int argc, VALUE *argv, VALUE self)
> {
>   genxWriter w;
>   VALUE xmlns, name;
>
>   switch (argc):
>
> ....
>
> If I'm not mistaken, this style of argument checking has already led
> to more than one bug in the Ruby source code itself.  Knock it off and
> use rb_scan_args().  Also, I find that starting all VALUE variables
> with 'rb' leads to clearer code, since it makes it immediately obvious
> which values are Ruby VALUE's versus C variables.

Sounds like a good practice.

> static VALUE
> writer_begin_element (int argc, VALUE *argv, VALUE self)
> {
>   genxWriter w;
>   VALUE rbName, rbXmlns;
>
>   /* One mandatory argument, one optional argument */
>   rb_scan_args(argc,argv,"11",&rbName,&rbXmlns);
>
>   Check_Type(rbName,T_STRING);
>   Check_Type(rbXmlns,T_STRING);
>
> ....

If you use camel case for your variable names this seems OK, but I 
would worry about creating conflicts with existing Ruby method names 
and global variables - since they all start with 'rb'.

>
> This is both shorter, more concise and easier to read all in one shot.
>  It's also a smarter API, since in both cases the name is mandatory,
> while the namespace is optional, afaict.  So, now the front end looks
> like this:
>
> Genx::Writer#begin_element(name,namespace=nil)
>
> Other than a minor modification in GENX4R_ERR, checking against Qnil
> vs 0, this code is identical.  No big, ugly switch.  No need to
> manually raise an ArgumentError if the arg count is wrong.  Nice and
> tidy.
>

I have found that making judicious use of StingValue(), NUM2LONG(), 
rb_Array(), and friends has advantages over using Check_Type() - you 
get better error messages and you add compatibility with other 
extensions/classes that make use "to_str", "to_int", "to_a", ... the 
later being a big plus.

-Charlie