Re: [ANN] Lua 5.2.0 (alpha) now available

[Richard Hundt <richardhundt@...>]

On 12/11/2010 06:40 PM, Florian Weimer wrote:
> * Richard Hundt:
>
> > On 12/11/2010 03:51 PM, Florian Weimer wrote:
> > > * Richard Hundt:
> > >
> > > > My point is that, although there is a way of creating a sparse table
> > > > which works in a numeric for loop which (i.e. the # operator respects
> > > > the number of elements the table was constructed with barring trailing
> > > > nil's),
> > >
> > > This is simply not true, it's an artifact of some Lua implementation,
> > > in some cases.
> >
> > This is Lua 5.2 for which there is a canonical implementation nearing
> > release readiness. Surely conscious effort went into supporting the
> > new table semantics?
>
> I don't think there was a change in 5.2 in this particular area.

Okay, so that brings us to the crux. The question is whether sparse 
tables are an implementation Artifact or a Feature.

If it were a Feature then I'd argue for a programmatic way of creating a 
sparse table (without a metatable), but it seems that it's generally 
considered an Artifact. Perhaps the Lua authors could clarify this?

Thanks for humoring this discussion so far.