Insonsistent parsing of JsxIdentifier

[ajafff]

TypeScript Version: 3.6.0-dev.20190804

Search Terms:

Code

let a = <\u0061></a>; // works
let ab = <\u0061-b></a-b>; // works
let ac = <a-\u0063></a-c>; // reports mismatching closing tag

Expected behavior:

Consistent parsing of escaped characters in JSX Identifiers.
I don't know the JSX spec, so I cannot tell whether all should be valid or all should be invalid

Actual behavior:

parseJsxIdentifier is only responsible for everything after the first dash. It wasn't changed to support unicode characters. Therefore the regular Identifier parsing allows escaped characters in the first part.

Related Issues:

#32096 forgot to change parseJsxIdentifier. cc @weswigham

[ajafff]

Babel doesn't seem to parse escaped characters in JSX at all.

[weswigham]

JSXIdentifier is still defined in terms of the core IdentifierStart/IdentifierPart, which include Unicode escape sequences, so I suppose they should all work - Babel's wrong here, I suppose.

[RyanCavanaugh]

If the fix for this turns out to be at all nontrivial, this is going to Won't Fix unless there's a compelling scenario here I've overlooked 😅

let a = <\u0061></a>;

Do we know if this is supposed to be legal?

[weswigham]

Do we know if this is supposed to be legal?

With the spec AST as written, yes.

JSXIdentifier is:

IdentifierStart
JSXIdentifier IdentifierPart
JSXIdentifier NO WHITESPACE OR COMMENT -

while ES6+ defined IdentifierStart and IdentifierPart as:

IdentifierStart::
    UnicodeIDStart
    $
    _
    \UnicodeEscapeSequence

IdentifierPart::
    UnicodeIDContinue
    $
    \UnicodeEscapeSequence
    <ZWNJ>
    <ZWJ>

so escape sequences being valid in jsx just falls out from the definitions.