JSXIdentifierReference type

[overlookmotel]

Foo in <Foo /> is equivalent to an IdentifierReference. It should have a reference_id, and semantic should resolve its binding.

JSXIdentifier is also used in other positions e.g. as a JSXAttributeName, so would probably need a separate type JSXIdentifierReference for this specific use.

Could also differentiate between <div /> and <Div /> in parser, with JSXIdentifierReference only used for the latter.

[Boshen]

I confirmed with tsc, which uses a plain Identifier:

export type JsxTagNameExpression =
    | Identifier
    | ThisExpression
    | JsxTagNamePropertyAccess
    | JsxNamespacedName
    ;

which means we can remove JsxIdentifier and use IdentifierReference and IdentifierName directly.

pub enum JSXElementName<'a> {
    /// `<Apple />`
    IdentifierReference(Box<'a, IdentifierReference<'a>>),
    /// `<div />
    IdentifierName(Box<'a, IdentifierName<'a>>),
    /// `<Apple:Orange />`
    NamespacedName(Box<'a, JSXNamespacedName<'a>>),
    /// `<Apple.Orange />`
    MemberExpression(Box<'a, JSXMemberExpression<'a>>),
}

but, in order to make it compatible with estree, we need to add at least one new AST node JSXIdentifierReference

:-(

[overlookmotel]

Ditto for:

pub enum JSXMemberExpressionObject<'a> {
    /// `<Apple.Orange />`
    IdentifierReference(Box<'a, IdentifierReference<'a>>),
    /// `<div.weird />`
    IdentifierName(Box<'a, IdentifierName<'a>>),
    /// `<Apple.Orange.Banana />` / `<div.weird.bizarre />`
    MemberExpression(Box<'a, JSXMemberExpression<'a>>),
}

I think we can use IdentifierReference and IdentifierName directly and get ESTree compatibility with a custom Serialize impl.

It's unclear to me how namespaces work. Is Apple in <Apple:Orange /> an IdentifierReference or not? React does not support JSX namespaces, so I think we can ignore this question and leave it as is.

[overlookmotel]

Looks like JSXElementName also needs a ThisExpression variant. <this /> is valid syntax.

[overlookmotel]

Dunqing did bulk of this work in #5223.

Remaining bits to cover: #5352, #5353, #5354.

Closing in favour of those individual issues.