Add proposal for "Spread Collections". #48
Conversation
|
Associated feature issue: #47 |
| then the downwards inference context type of a spread element in that list | ||
| is `Iterable<T>`. | ||
|
|
||
| * If a spread element in a list literal has type `Iterable<T>` for some `T`, |
|
|
||
| 1. Evaluate `iterator.current` to a value `newEntry`. | ||
|
|
||
| 1. Call `map[newEntry.key] = value`. |
|
LGTM with comments. |
| This works because the individual elements in `numbers` do happen to have the | ||
| right type even though the list that contains them does not. However, the spread | ||
| object does need to be "spreadable"—it must be some kind of Iterable for a | ||
| list literal or a Map for a map literal. |
There was a problem hiding this comment.
Consider a comment to point out that this also doesn't require that the type of the thing being spread be assignable to the type of the List. e.g. spreading MyList into a List (neither is a subtype of the other).
|
|
||
| If implicit downcasts are disabled, then the "is assignable to" parts here | ||
| become strict subtype checks instead. | ||
|
|
There was a problem hiding this comment.
There is a specified error preventing duplicate keys in map literals when the keys are constant. Should we consider that for nested map literals + spread? Should the following be an error?
var a = {3 : true,
...{3 : false}};
There was a problem hiding this comment.
I thought about duplicate keys, but the static checking for that is even pretty brittle for flat literals since it only works for literal keys (obviously):
var key = "key";
var map = {key: 1, key: 2}; // <-- No error.So I figured the duplicate key checking wasn't very high value and wasn't worth extending into spread. Also, it might get weird if we introduce "conditional spread" features.
There was a problem hiding this comment.
I believe the rule about duplicate keys only apply to constant maps:
It is a compile-time error if two keys of a constant map literal are equal.
so since spreads do not apply to constant maps, that issue is moot.
Here it is!