Fixes #217
Fixes #217
Conversation
|
I'll take a look at this at some point tomorrow. Thanks for working on this! |
| @@ -66,6 +62,7 @@ module.exports = grammar(C, { | |||
| [$._binary_fold_operator, $._fold_operator], | |||
| [$.expression_statement, $.for_statement], | |||
| [$.init_statement, $.for_statement], | |||
| [$._function_declarator_seq], | |||
There was a problem hiding this comment.
What was the new conflict here?
There was a problem hiding this comment.
repeat($.attribute_specifier) after noexcepts
grammar.js
Outdated
| $.using_declaration, | ||
| $.type_definition, | ||
| $.static_assert_declaration, | ||
| ';', |
There was a problem hiding this comment.
Semicolons can technically appear anywhere a statement is valid. I think you were trying to close a particular issue regarding having an extraneous semicolon (#35?), but I think we should hold off and try to figure out a way to make "empty" statements (a single semicolon) work anywhere a statement could appear (any of which would end with a semicolon), instead of only as a field declaration.
Not sure how terrible that would be for node count though :)
There was a problem hiding this comment.
Hm, maybe something like javascript's automatic semicolons could be used, only adds 7 to state count though
There was a problem hiding this comment.
Not familiar with that, but I think we should make something like an empty_statement: $ => ';' rule and just put that as an option in various places where other statements (including top level) can be. Not the most elegant and I haven't actually tried this so it may not really work in practice but something like that is the best way to make the grammar not choke on extra semicolons.
There was a problem hiding this comment.
https://github.com/tree-sitter/tree-sitter-javascript/blob/master/src/scanner.c, though C++ is a bit different and an empty_statement node would be just fine as well - but I think naming it _empty_statement to hide the rule could be better
|
Hi @amaanq, I am asking this just in case I miss it, are you working on the fix for the precedence issue of parenthesized assignment expressions? |
|
Thanks for the reminder, it slipped my mind completely. I've come up with a solution now but I'd appreciate more testing for potential pitfalls. |
It's great to hear you've got a solution, looking forward to seeing the PR soon. |
Did this to prep for another release, and most of these were easy enough to fix and aren't anything complicated.
The main change that really increased state count a bit is backporting the C updates where I added attribute_specifier to more allowed locations.
Closes #47
Closes #63 (really a C bug fixed in 0.20.4)
Closes #96
Closes #102
Closes #103
Closes #112
Closes #139
Closes #203
The last issue that's a bit difficult is the parenthesized assignment expressions, that's because of fold operators taking precedence over them but I could not figure that out, I'll try a bit more anyways