Add support for optional chaining "?."

[mjambon]

The feature is already being widely used in typescript, in about 10% of the files according to my rough stats. See the related ticket in the tree-sitter-typescript repo: tree-sitter/tree-sitter-typescript#80

I'm thinking of adding the feature to the javascript grammar rather than only adding it to the typescript grammar. Let me know if there's something wrong with the approach. I'll use this reference document: https://tc39.es/ecma262/#sec-optional-chains

[issue-label-bot]

Issue-Label Bot is automatically applying the label feature_request to this issue, with a confidence of 0.98. Please mark this comment with 👍 or 👎 to give our bot feedback!

Links: app homepage, dashboard and code for this bot.

[maxbrunsfeld]

👍 I agree; we should add this to the JavaScript grammar.

[mjambon]

Do you have guidelines for grammar maintenance? Some rules are hidden and some are not. Some rules use the name found in the reference documents (ES6 from 2015 or ES 2021) while some others seem to be workarounds to facilitate parsing or something. For example, the rule augmented_assignment_expression coexists with assignment_expression in grammar.js but only the latter occurs in the ecmascript spec. I'd love to throw away any rule that doesn't appear in the spec and replace it by a literal implementation of the spec... but I also expect this approach to fail. I'd be inclined to leave comments explaining any apparent or actual deviation from the spec but I don't see any. Maybe comments aren't that useful for this type of code, I don't know.

Are there conventions, tools, documents that I should be aware of? Generally speaking, do you have tips on how to modify grammars to as to bring them up to date? Does it work best to proceed incrementally or should I not worry too much, add or modify all the rules I want, and only then start testing?

[maxbrunsfeld]

Keeping a 1:1 correspondence with the language spec isn’t a goal, because that wouldn’t produce a useful CST. There’s some discussion of this in the docs, but we don’t have a formal system that indicates how things should be structured.

In this case, I think we want to add a new expression rule. You might want to run a widely used JS parser to see what name they use for this “foo.?bar” syntax.

[mjambon]

Thank you Max, I'll try something like this.

[maxbrunsfeld]

After scanning the docs on MDN, my suggestion would be to add three new rules:

• optional_member_expression
• optional_call_expression
• optional_subscript_expression

These would be parallel with the existing member/call/subscript expression rules, and they would match foo?.bar, foo?.(bar) and foo?.[bar] respectively.

I haven't looked into it deeply, so I may be missing something, but that might be a good starting point, which is consistent with the existing style.