Re: ipairs in Lua 5.3.0 (alpha)

[Jean-Luc Jumpertz <jean-luc@...>]

Le 1 août 2014 à 15:45, Philipp Janda <siffiejoe@gmx.net> a écrit :

> 
> Am 01.08.2014 14:56 schröbte Jean-Luc Jumpertz:
>> Moving this to a new thread to avoid polluting the Lua 5.3.0 (alpha) announcement.
>> 
>> Le 1 août 2014 à 14:41, Philipp Janda <siffiejoe@gmx.net> a écrit :
>> 
>>>>>> And btw testing the  '__len' and '__index'  metamethods in function 'luaB_ipairs' has an almost null performance cost and gives you additional flexibility, so what would it bring to restrict this test to the '__len' metamethod, as Philipp Janda proposed? :-)
>>>>> 
>>>>> It's useless code that causes one extra case of undefined behavior, so why keep it? And what's the additional flexibility?
>>>> 
>>>> The additional flexibility is to allow you to freely define the values you want to return for table[i] and to keep it consistent between direct accesses like << a = my_table[i] >> and for loops like << for i, v in ipairs(my_table) >>, even if you don't redefine __len
>>> 
>>> As I argue above, you can't. `__index` without `__len` is either superfluous ( `__index` never gets called) or highly unpredictable (undefined behavior, could stop in mid-loop in current Lua implementation).
>> 
>> Can you elaborate on this? I really can not get your point here...
> 
> `__index` is only called when you encounter a nil value. If you encounter a nil value, your array has holes. If your array has holes its length is not defined. Without `__len` you can't have both: Either your `__index` function is called or `luaL_len` computes a deterministic length.
> 
>> 
>> It looks to me that  `__index` without `__len` would be appropriate in the case of an indirection-typed data structure, where the value returned by __index is based on the raw value stored in the table and preserves the nil table entries. Which is not an uncommon pattern.
>> 
> 
> The value returned by `__index` can't be based on the raw value, because there is no raw value or else `__index` wouldn't be called in the first place.
> Nevertheless, a data structure with `__index` but without `__len` makes perfect sense -- just not for the current `ipairs` behavior.
> 
> 
> Philipp

Oops, that's right. Working almost exclusively with metatables on userdata with uservalues made me forget this characteristic of the Lua table regarding __index and __newindex. :-)

Jean-Luc