The [[ExportName]] field of ExportEntry Record can also be null

[seminaoki]

In Table 42: ExportEntry Record Fields, it states that [[ExportName]] filed's value type is String.

But look at the example, [[ExportName]] can also be null, when Export Statement Form is the following:

export * from "mod";

The [[ExportName]] field of ExportEntry Record should be "String | null",
or the [[ExportName]] of export * from "mod"; should not be null.

[jmdyck]

It appears that when an ExportEntry Record has [[ImportName]] = "*", its [[ExportName]] and [[LocalName]] are completely ignored. (ParseModule puts the record in _starExportEntries_, which goes to [[StarExportEntries]], which is then processed in GetExportedNames and ResolveExport.)

So on the one hand, it doesn't matter what [[ExportName]] is set to there. But on the other hand, I agree that the use of the field should conform to its 'type declaration'. The easiest thing would be to change the latter to "String | null". (Personally, I think replacing null with (say) "*ignored*" would be somewhat more informative. Another alternative would be to have 3 'sub-types' of ExportEntry Record, none of which have any null fields.)

[jmdyck]

It appears that when an ExportEntry Record has [[ImportName]] = "*", its [[ExportName]] and [[LocalName]] are completely ignored. (ParseModule puts the record in _starExportEntries_, which goes to [[StarExportEntries]], which is then processed in GetExportedNames and ResolveExport.)

So on the one hand, it doesn't matter what [[ExportName]] is set to there. But on the other hand, I agree that the use of the field should conform to its 'type declaration'. The easiest thing would be to change the latter to "String | null". (Personally, I think replacing null with (say) "*ignored*" would be somewhat more informative. Another alternative would be to have 3 'sub-types' of ExportEntry Record, none of which have any null fields.)

[seminaoki]

I agree that

The easiest thing would be to change the latter to "String | null"

Since other fields of ExportEntry also use "String | null" value type.

[seminaoki]

I agree that

The easiest thing would be to change the latter to "String | null"

Since other fields of ExportEntry also use "String | null" value type.

[devsnek]

wouldn't we say not present or unset here?

[devsnek]

wouldn't we say not present or unset here?

[jmdyck]

wouldn't we say not present or unset here?

The spec doesn't use the word "unset" at all, so that seems unlikely. As for "not present", did you mean like this?

ExportEntry Record { ..., [[ExportName]]: not present, ... }

That would amount to treating not present as a (spec) value, which we don't really do.

[jmdyck]

wouldn't we say not present or unset here?

The spec doesn't use the word "unset" at all, so that seems unlikely. As for "not present", did you mean like this?

ExportEntry Record { ..., [[ExportName]]: not present, ... }

That would amount to treating not present as a (spec) value, which we don't really do.

[ljharb]

It could be ~empty~

[ljharb]

It could be ~empty~

[jmdyck]

It could, but:

• null is clearly the convention for ExportEntry records, and
• you'd still have to change [[ExportEntry]]'s field type.

[jmdyck]

It could, but:

• null is clearly the convention for ExportEntry records, and
• you'd still have to change [[ExportEntry]]'s field type.

[ljharb]

Would you? Empty means there’s nothing there, so the type is still string, it’s just optional.

[ljharb]

Would you? Empty means there’s nothing there, so the type is still string, it’s just optional.

[jmdyck]

I don't think there's a precedent in the spec where something with a declared type of T actually allows T or ~empty~. E.g., the [[Target]] field of completion records is explicitly declared as any ECMAScript string or ~empty~.

(In several programming languages, "optional string" is quite distinct from "string". E.g. Rust, Swift, Kotlin, Scala, C#, F#.)

[jmdyck]

I don't think there's a precedent in the spec where something with a declared type of T actually allows T or ~empty~. E.g., the [[Target]] field of completion records is explicitly declared as any ECMAScript string or ~empty~.

(In several programming languages, "optional string" is quite distinct from "string". E.g. Rust, Swift, Kotlin, Scala, C#, F#.)