Re: LUA_USE_APICHECK

[Lorenzo Donati <lorenzodonatibz@...>]

On 26/08/2011 12.54, Axel Kittenberger wrote:
> > The manual doesn't say that non-acceptable indices should return NULL.
>
> In my opinion it does.
>
> "If the value at the given acceptable index is a full userdata,
> returns its block address. If the value is a light userdata, returns
> its pointer. Otherwise, returns NULL"
>
> This english compiles into pseudo-code
>
> if valid(index) and valueat(index) == userdata then return block address
> elseif valid(index) and valueat(index) == light userdata then return pointer
> else return NULL.
>
> So an invalid index returns NULL.
Sadly (for computer manuals writers, not for poets) natural languages 
can be ambiguous, so Josh has a point in saying that valid(index) may be 
a precondition, so the whole lot may also be pseudocoded as:

if valid(index) then
  if valueat(index) == userdata then return block end
  elseif valueat(index) == lightuserdata then return pointer end
  else return NULL end
else
 undefined behaviour
end

From a natural language POV the interpretation may be correct as well, 
especially if you interpret it in the context of the whole manual and 
not in isolation (another "feature" of natural languages - they don't 
have to stick only to locally declared information).

Of course the debate may well indicate that a more clear formulation 
would be helpful, since it is a refman and not a poetry text (to keep 
with the former analogy).

Cheers.
-- Lorenzo