From: Evan Phoenix <evan@...>
Date: 2008-03-16T15:10:45+09:00
Subject: Re: RARRAY_PTR

I second this idea. In fact, Laurent and I have discussed this kind of  
thing before at rubyconf.

I'd love for us to take this even further and have RSTRING_AT(), etc,  
and deprecated RSTRING_PTR, for exactly the same reasons, that expose  
the direct memory as the implementation makes these much more  
difficult for implementors, and more dangerous for users.

- Evan

On Mar 15, 2008, at 11:15 AM, Laurent Sansonetti wrote:

> On Sat, Mar 15, 2008 at 6:42 AM, Yukihiro Matsumoto <matz@ruby-lang.org 
> > wrote:
>> Hi,
>>
>> In message "Re: RARRAY_PTR"
>>
>>   on Sat, 15 Mar 2008 09:40:15 +0900, "Laurent Sansonetti" <laurent.sansonetti@gmail.com 
>> > writes:
>>
>> |Currently, ruby.h defines RARRAY_PTR as a macro to access the  
>> internal
>> |C array of the RArray structure (which represents the Array class).
>> |This macro is used in many places in the Ruby code (and  
>> extensions) to
>> |get an element at a particular index
>> |
>> |  VALUE v = RARRAY_PTR(ary)[42];
>> |
>> |But some people also use it to set values
>> |
>> |  RARRAY_PTR(ary)[0] = Qnil;
>> |
>> |I think this macro is dangerous because it exposes too much  
>> internals.
>> |Also, it is difficult to re-implement RArray using a different  
>> storage
>> |structure and efficiently support this macro at the same time. In
>> |MacRuby I am working on a new implementation of the Array class and
>> |will not be able to support this macro for inline modifications.
>>
>> How about officially declaring RARRAY_PTR() is read-only?  In fact,  
>> it
>> IS read-only.
>>
>
> I have been thinking about this, RARRAY_PTR could return a const
> value. Right now it doesn't seem to.
>
> But there is still the problem of accessing elements in the array.
> People use RARRAY_PTR(ary)[idx] for that. Nobu-san's RARRAY_AT macro
> idea is good because it doesn't expose the underlying implementation
> of the class (and it can also do a boundary check).
>
> If the genuine Ruby in the future wants to re-implement RArray using a
> different storage structure, the interface would already be there.
>
> # If you desire I can work on a patch.
>
> Laurent
>