On Thursday, 9 July 2015 at 19:10:39 UTC, Jonathan M Davis wrote:
On Thursday, 9 July 2015 at 18:45:56 UTC, Marc Schütz wrote:
Interestingly, it goes on by saying:
An n-tuple is defined inductively using the construction
of an ordered pair.
Although not stated explicitly, this implies (a kind of) auto
expanding!
=> Fits more than perfectly :-P
No one who has ever seriously used tuples in any programming
language I've ever heard of would expect tuples to auto expand.
Auto expansion makes them _considerably_ less useful. In the
case of TypeTuple/AliasSeq, the situation is a bit different,
because we're not really talking about tuples here. Real tuples
nest, and they don't auto expand.
But my quote above that they _don't_ nest and _do_ auto-expand,
how else could you construct n-tuples from pairs?
TypeTuple/AliasSeq is the _only_ case I've ever seen where
someone tried to claim that something was a tuple when it
didn't nest, or it auto-expanded. Folks have been consistently
confused about the differences between TypeTuple and
std.typecons.Tuple and the fact that TypeTuples auto expand. No
one expects it - because tuples just don't do that in any
actual programming languages. I question the validity of your
interpretation of the theory as well, but even if it's valid,
it doesn't match what's done in actual programming languages.
I personally think this auto-expanding argument is overrated. I
don't have any particular expectation regarding auto-expansion
towards the concept "tuple". Sure, it may be surprising, but that
doesn't stop an auto expanding tuple from being a tuple, just
like Perl arrays being auto-expanding doesn't stop them from
being arrays.
