Re: pairs(t, skey) and ipairs(t, skey)

["Claire Lewis" <claire_lewis@...>]

> > > next(t, #t) returns the first key in the hash part (if any).
> > This isn't necessarily true. It is possible to set up a table with
> > multiple holes in the array part such that next( t, #t ) will return
> > an index which still falls in the array.
>
> I wasn't careful in my phrasing of that sentence. Yes, it may return a
> key that follows one of the holes, but this is exactly what I'd
> need/expect in this case.
>
> For example, using this table, {"a", nil, nil, "f", "g", [7] = {}, [9]
> = "i"}, #t gives me 5 (Lua 5.1) and next(t, #t) returns 7.

Just keep in mind this behaviour is pretty much specific to table 
constructors in current implementation, and is (I believe) completely 
undefined.

If you build the same table differently:

t = {}
t[1]='a'
t[2]=nil
t[3]=nil
t[4]='f'
t[5]='g'
t[7]={}
t[9]="i"

=#t
1
=next(t,#t)
5       g