Imprement tuple inference #520
Conversation
| check_inference( | ||
| r#" | ||
| fn test() { | ||
| let a: (u32, &str) = (1, "a"); |
There was a problem hiding this comment.
the type annotation here kind of makes the type inference for (1, "a") unnecessary, except that we can see that 1 gets inferred to u32. Can you add another line without a type annotation, maybe even just ("a", a) or something, to test a bit more explicitly that tuple expressions get the correct type even when not annotated?
| [60; 68) '(a, "b")': ((u32, &str), [unknown]) | ||
| [61; 62) 'a': (u32, &str) | ||
| [64; 67) '"b"': [unknown] | ||
| [79; 80) 'c': ([unknown], [unknown]) |
There was a problem hiding this comment.
@flodiebold Thanks for your quick review.
I'm not sure inferring result after 10 line is right.
There was a problem hiding this comment.
oh, of course, inferring string literals will only work with #485 :) you could instead add some parameters with declared types and use them instead of literals... so for example
fn test(x: &str, y: isize) {
let a: (u32, &str) = (1, "a");
let b = (a, x);
let c = (y, x);
let d = (c, x);
}should be able to infer everything...
There was a problem hiding this comment.
I see, thanks for your advice.
|
@flodiebold fixed it. |
|
Great! bors r+ |
520: Imprement tuple inference r=flodiebold a=h-michael related #394 I'm sorry I'm late. I try implementing array inference next. Co-authored-by: Hirokazu Hata <[email protected]>
|
|
||
| // we have not infered these case yet. | ||
| let e = (1, "e"); | ||
| let f = (e, "d"); |
There was a problem hiding this comment.
I write these two lines to make spec clearly.
Build succeeded |
related #394
I'm sorry I'm late.
I try implementing array inference next.