Parsing HTML fragments with libxml SAX

I need to parse HTML fragments, by which I mean the files lack <html>, <head> and <body> elements, otherwise having well-formed XHTML syntax, UTF8 encoding guaranteed. It looks like libxml is ideal for this task, but I have certain constraints which I just don't know how to implement.

htmlSAXParseFile() does its job well enough, but it seems to create DOM itself, inserting body and html elements in the process. I would like to create DOM myself because I may need to skip some elements and modify others on the fly. Is it possible to somehow tell libxml not to create DOM at all and just parse HTML and call my handlers?
If that's not possible for libxml HTML parser, I might as well use xmlSAXUserParseFile() which doesn't seem to create DOM. However, since the files have a structure like <p>...</p><p>...</p>, the parser just spits "Extra content at the end of the document" too early. Is there a way to suppress some parsing errors while still getting notified about them (just because nobody guarantees there will never be other errors in those files)?
There is a whole heck of parsing functions in libxml, some of which accept xmlParserOption as a parameter. Alas, xmlSAXUserParseFile() does not. And those which do all seem to create DOM for some irrelevant API design reasons. Am I missing an obvious candidate?

Oh, and I confess that my reluctance to use libxml's DOM looks like a quirk. I am extremely constrained with RAM, so I desperately need total control over DOM to be able to drop some nodes on low memory conditions and re-read them if necessary.

Thanks in advance.

OK, since nobody has answered the question, I'll try to do it myself.

I wrote all the start/end element handlers and it looks like libxml does not create the DOM any more. At least, the returned document pointer is NULL. It still insists on html and body elements, but I can live with that.

One major problem is that libxml preserves all whitespace nodes, no matter what. So I have to parse text content to eliminate ignorable whitespace. It's ugly, but it works. Should I mention that parsing UTF-8 is the kind of fun you rarely miss?

To be honest, libxml documentation is atrocious. My advice to anybody who ever tries reading the docs: read the source code instead. The code is much more readable and documented.

Thanks for attention.

ansaurus

tags:

views:

answers:

Parsing HTML fragments with libxml SAX

related questions