XML / Java: Precise line and character positions whilst parsing tags and attributes?

Question

I’m trying to find a way to precisely determine the line number and character position of both tags and attributes whilst parsing an XML document. I want to do this so that I can report accurately to the author of the XML document (via a web interface) where the document is invalid.

Ultimately I want to set the caret in a to be at the invalid tag or just inside the open quote of the invalid attribute. (I’m not using XML Schema at this point because the exact format of the attributes matters in a way that cannot be validated by schema alone. I may even want report some attributes as being invalid part-way through the attribute’s value. Or similarly, part-way through the text between a start and end tag.)

I’ve tried using SAX (org.xml.sax) and the Locator interface. This works up to a point but isn’t nearly good enough. It will only report the read position after an event; for example, the character immediately after an open tag ends, for startElement(). I can’t just subtract back the length of the tag name because attributes, self-closing tags and/or newlines within the open tag will throw this out. (And Locator provides no information about the position of attributes at all.)

Ideally I was looking to use an event-based approach, as I already have a SAX handler that is building an in-house DOM-like representation or further processing. However, I would be interested in knowing about any DOM or DOM-like library that includes exact position information for the model’s elements.

Has any one solved this issue, or any like it, with the required level of precision?

Answer 1

XML parsers will (and should) smooth over certain things like additional whitespace, so exact mapping back to the character stream is not feasible.

You should rather look into getting a lexer or 'token stream generator' for increased detail, in other words go to the detail level below XML parsers.

There is a few general frameworks for writing lexers in java. This ANTLR 3-based page has a nice overview of lexer vs parser and section one some rudimentory XML Lexer examples.

I'd also like to comment that for a user with a web interface, maybe you should consider a pure client-side (i.e. javascript) solution.