Re: Lua and composition

[Francisco Olarte <folarte@...>]

On Thu, Mar 2, 2017 at 7:46 AM, Dirk Laurie <dirk.laurie@gmail.com> wrote:
> Soni has proved to me that I did indeed propose the multi-colon
> version some time ago, so I must confess and ask absolution for
> that sin, but my current preference is also to allow one colon only.

Well, the multi-colon solution seems more 'elegant' in an abstract
kind of way, and nobody forces you to use more than one.

That being said, if I had it I will probably not use it ( falling back
to local t1=a.b.c; local t2=t1.d.e; local t2.g:h(t1, ...), or even
t2.g.h(t2, t1, ....) for clarity, to leave my self something easier to
parse when I'm forced to go back three years later ) except in extreme
or idiomatic cases (like object:behaviour_mixin.method(...), just a
single usage ).

So I would say multi-colon with a trivial extra implementation effort
will just barely pull it's own weight due to the extra documentation
complexity, and would not it with a non trivial one.

Francisco Olarte.