On Saturday, 20 June 2026 at 14:43:56 UTC, Mindy (0xEAB) wrote:
Those are function calls, hence IFTI[0] applies.
In your original example, however, the template is used as an
alias.
Hmm, I see the difference now, thanks.
Still. Nick suggested a library-side fix above: to separate, on
the template level, the first argument of `format` from the other
arguments. If this separation works, the next logical steps seem
to be:
1. Do this to every library function for better flexibility.
2. Better, on the compiler side, do a lowering of every
multi-argument template into a chain of nested single-argument
templates.
3. Better still, on the compiler side, support something
IFTI-like for aliases.
But it didn't happen. So I guess I'm missing some rationale here.
Perhaps it's something about when each template is instantiated,
and the instantiation cannot be postponed in the general case?..
Just, the original example, where `map!format` fails but
`map!(x=>format(x))` works, still seems like an arbitrary quirk
to me.
Ivan Kazmenko.
