Re: structure pointers and metatypes using ffi
- From: Daurnimator <quae@xxxxxxxxxxxxxxx>
- To: luajit@xxxxxxxxxxxxx
- Date: Mon, 14 May 2012 22:30:02 +1000
On 14 May 2012 22:05, Mike Pall <mike-1205@xxxxxxxxxx> wrote:
> Daurnimator wrote:
>> I disagree somewhat with this;
>> IMO, pointer comparisons should be handled just like table comparisons:
>> pointer equality being table equality; but when "raw equality" fails:
>> should fall back to metamethods.
>> lua doesn't have a rawequal function for this case....
>
> Ok, so you maybe you get away without rawequal(), if you're lucky.
> But then it gets inconsistent with pointer arithmetic and other
> metamethods. Ok, so one would need to run __add for p1 + 1. But
> that means you cannot _ever_ do plain pointer arithmetic on such a
> pointer without rawadd(), rawsub() and so on. This quickly leads
> to an inflation of raw fallback functions.
>
What if pointer types could have a default metatable?
That is, if all pointer types came with a metatable where __add
implemented point arithmetic;
But a developer could set a new metatable; implementing their own (non
pointer arithmetic) __add metamethod.
We seem to do without a range of raw* functions in lua as it is
(though I admit, I sometimes miss them)
> This is a low-level API -- I don't want this to become as opaque
> as C++.
>
> --Mike
>
It is low level, but needing to wrap pointers in a table or struct or
some such to implement metamethods seems like a waste...
D
Other related posts:
