Dana Jansens
8ea32187d4
Prevent dead code by diagnosing putting the same `impl` in a `match_first` block more than once.
2.2 KiB
Disallow impl in match_first twice
Table of contents
Abstract
Disallow declaring an impl in the same match_first block more than once.
Problem
In
proposal 5337
we have the rule "An impl may appear in at most one match_first block." This
allows an impl to appear in one match_first block more than once, which
creates ambiguity about its ordering.
match_first {
impl forall [T: X] T as Z {}
impl forall [T: Y] T as Z {}
impl forall [T: X] T as Z {}
}
The rules say that the first impl in the block should be selected, so given a
type that implements X & Y, the first impl (with T: X) should be chosen. But
there is also a T: Y impl that wins against the last T: X impl. While the
order of is clear, this leads to more confusing code than is necessary. There is
no need to write an impl twice in a match_first block.
Background
- Proposal 5337
introduced rules for
match_firstblocks.
Proposal
Modify the rule "An impl may appear in at most one match_first block" to instead say "An impl may appear in at most one match_first block, and only once within it".
Details
This will disagnose the case where an impl is written multiple times in a
match_first block, but the later ones would have been ignored. This allows the
user to know about code they've written that isn't doing anything.
Rationale
This advances the Code that is easy to read, understand, and write goal, by avoiding the case where the user can write code that is never used. It may help them catch mistakes such as copy/paste errors.