File size: 1,188 Bytes
c6cb4e0 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 | Fix the following issue:
## Separator DSL methods produce incorrect lookahead / possible tokens
When a rule uses the separator variants of the repetition DSL methods (`MANY_SEP` / `AT_LEAST_ONE_SEP`), the computation of the possible next tokens / lookahead paths gives wrong results.
For example, given a rule along the lines of:
```ts
this.RULE("list", () => {
this.MANY_SEP({
SEP: Comma,
DEF: () => { this.CONSUME(Identifier) }
})
})
```
After consuming an `Identifier`, the set of possible following tokens should include the separator (`Comma`) as one of the options (since the repetition may continue), in addition to whatever can follow the rule. In practice the separator is missing / not produced correctly by the path computation, which then breaks downstream features that rely on it (syntactic content assist, lookahead, etc.).
The non-separator variants (`MANY` / `AT_LEAST_ONE`) behave correctly for the same shape of grammar — the problem is specific to the `*_SEP` DSL methods.
Could the path / lookahead computation be fixed so that separator repetition methods produce the same kind of correct results as their non-separator counterparts? |