Parser doesn't support whitespace before prolog

[elsassph]

The parser fails completely, and without errors, when there is whitespace before the XML prolog.
I need to preserve positions so I can't trim those spaces; it should be possible to allow them.

{empty lines here}
<?xml version="1.0" encoding="utf-8" ?>
<component>
</component>

[bd82]

As far as I can tell leading whitespace is not allowed before the prolog according to the specifications.

• https://www.w3.org/TR/xml/#sec-well-formed (see Document and Prolog rules).

[elsassph]

That's fair but it's tolerated by many readers/engines consuming XML.

[bd82]

Related issues:

• Parsing fails if there is whitespace before xml prolog libxmljs/libxmljs#72
• https://stackoverflow.com/questions/9350721/whitespace-at-start-of-xml-doc-is-causing-error

I think it would make sense to only show a syntax error in that case but be fault tolerant, meaning
successfully parse the all the document and construct a valid CST.

This may depend on: Chevrotain/chevrotain#646
Although a workaround without a change to Chevrotain may be possible.

[elsassph]

I'd suggest that at least recovering from the error would be nice.

[bd82]

I'd suggest that at least recovering from the error would be nice.

A full productive solution in this library would also have to deal with creating the syntax error and perhaps even making
this grammar variation optional.

However, you may be able to accomplish this yourself by extending the XMLParser class and overriding the document rule to allow
consuming multiple misc rules before the prolog section.

• https://github.com/SAP/xml-tools/blob/master/packages/parser/lib/parser.js#L17-L19

However, Grammar inheritance is a little tricky with Chevrotain, so I'm not sure you can use it without first making some small changes to the XMLParser itself.

• See: https://github.com/SAP/chevrotain/blob/master/examples/parser/inheritance/inheritance.js

Another option could be to create a subclass with a new grammar rule entry point documentWithMiscPrefix

    $.RULE("documentWithMiscPrefix", () => {
      $.MANY(() => {
        $.SUBRULE($.misc);
      });

     // continue parsing using the original logic
     $.SUBRULE($.document);

This may bypass some of the grammar inheritance issues, but I am not 100% sure.

To make this XMLParser easily extensible you would need to enable deferring the performSelfAnalysis call at the end of the constructor.

• e.g by an optional constructor flag argument.

Perhaps this is the most time efficient solution, As I personally don't know when I will get around to implementing a "proper" solution to this issue, however you could pretty easily contribute a PR making the XMLParser more extensible and then
You can implement whichever partial workaround in your own code base.

[elsassph]

Thanks for the pointers. Though I'd love to contribute, this is a bit over our bandwidth right now; this is for another OSS project where switching to this parser is already a major effort 😄 We'll see if it harms our users in practice.