On Wed, 12 Sep 2018 10:05:06 -0700
Ian Lance Taylor <i...@golang.org> wrote:
> I agree that no code is produced by the program, but you're suggesting
> a compile-time decision as to which code should be compiled.
How does it differ from unused code eliding go compiler and linker do
already? Used `for type case` code will compile and run. Unused will be
elided, just a bit earlier.
> but it is something unlike what can be done in Go today.
I thought we're discussing changes. And if we're at it, "Craftsman's
proposal" adds almost **no new[1] concepts** to the language.
Adds no new keywords.
`For type` constraints can be checked with current compiler.
Old Go1 code can use future generic version of current library
without touching a comma: If it calls r := util.Min(a, b) for ints now,
it will be calling generic Min as r := util.Min(a, b) not even knowing
that Min is now string, FooBoos and Mermaids capable.
[1] the actual **single** new thing in the user's code is a generic method
call via package selector:
r := a.util.Min(b) // calls generic Min from util package.
New things in CGG **generic code** are:
- (x type T) declarations (as in team's proposal)
- `for type` contract
- `for type` switch
- a bunch of constrains. But ALL for type constraints are based
on already familiar Go type definitions and casts.
>
> Ian
TC,
--
Wojciech S. Czarnecki
<< ^oo^ >> OHIR-RIPE
--
You received this message because you are subscribed to the Google Groups
"golang-nuts" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to golang-nuts+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
