1297090 - Local variable 'module' treated as keyword

Status: RESOLVED FIXED

(Webtools :: Searchfox, defect)

Description

[Andrew McCreight [:mccr8]]

The function nsScriptLoader::EvaluateScript(nsScriptLoadRequest* aRequest) has a local variable named 'module', which is treated as a keyword (highlighted blue, does not highlight uses when hovering over it with the mouse).

http://searchfox.org/mozilla-central/source/dom/base/nsScriptLoader.cpp#2027

Comment 1

[Bill McCloskey [inactive unless it's an emergency] (:billm)]

I think "module" will be a keyword in C++0something. I guess ideally Searchfox would detect what version of C++ you're compiling with, but it doesn't seem worth it.

Comment 2

[Tooru Fujisawa [:arai]]

According to https://en.cppreference.com/w/cpp/keyword , module is a contextual keyword, which means it's totally fine to be used as variable name even after C++20.

• final
• override
• transaction_safe
• transaction_safe_dynamic
• import
• module

while detecting the context isn't realistic, we could for example do the following:

• make them keyword by default (so, highlight blue)
• if there's analysis record for the token, turn it into symbol syntax

Comment 3

[Tooru Fujisawa [:arai]]

supporting contextual keyword is also beneficial for ECMAScript, given ECMAScript has many kind of contextual keywords, and being able to highlight them based on the syntax (or actually, whether it's detected as variable or not, by indexer) will make the code easier to read.

I have prototype patch, but it conflicts with bug 1781179, and I think it's better fixing the issue based on bug 1781179's change, I'll wait for it.

Comment 4

[Tooru Fujisawa [:arai]]

Attachment: GitHub Pull Request

Comment 5

[Tooru Fujisawa [:arai]]

Rust also has similar concept "weak keywords"
https://doc.rust-lang.org/reference/keywords.html

and bug 1593836 is caused by treating union as strict keyword.

Comment 6

[Tooru Fujisawa [:arai]]

WebIDL's keywords are also very contextual.
https://webidl.spec.whatwg.org/#idl-grammar

There are some case the contextual keyword being the member name, such as includes and any,
where:

• includes is a keyword but is allowed for operation name,
• any is not actually keyword, but has special meaning in the type context

https://webidl.spec.whatwg.org/#prod-OperationName

OperationName ::
    OperationNameKeyword
    identifier

https://webidl.spec.whatwg.org/#prod-SingleType

SingleType ::
    DistinguishableType
    any
    PromiseType

Comment 7

[Tooru Fujisawa [:arai]]

in order to simplify the handling across language, it may be better just making all keywords "contextual",
which means, whenever the word has corresponding analysis record, it becomes an identifier, and otherwise it's kept as keyword.

As long as the indexer correctly handles the keyword vs identifier, the formatter doesn't have to distinguish them

Comment 8

[Tooru Fujisawa [:arai]]

with making everything contextual, this won't conflict with bug 1781179

Comment 9

[Tooru Fujisawa [:arai]]

Attachment: GitHub Pull Request