Parsing with two scanners(ources) as input (?)

["Oliver Bandel" <oliver@first.in-berlin.de>]

Hello,

I want to pasre HTML, and with that I mean I want to parse the  
structure of the tags as well as the contents of the data-elements.

At the moment I'm hacking a special parser for this case,
but it's somehow ugly, because I need to hand-code the state machine  
of the parser, and it somehow becomes ugly.

It would be easier and more elegant, if I could combine the  
HTML-tag-parsing together with the text-parsing on the data-elements.

For HTML-parsing I use Nethtml.
For Text-Scanning I use Pcre.

I want to be able to select certain tags and text that will occur at  
certain positions.

For detecting the found tags I look for Nethtml's
   Element (name, args, subnodes)
and for detecting the data-strings I look into Nethtml's
   Data string
with Pcre.

I would like to find out certain data that occurs after ceratin  
sequences in the tree and then look for certain strings inside that   
Data-strings.

Any idea on how to create the parser?

I thought about somehow wrapping the stuff and give it to ocamlyacc.
Maybe menhir is better for that task?

At the moment I use the Element-match just to call the recursive  
parser on the next doclist.
All my parsing is using   Data-match and looks up for the contents there.
This is, because the information I want to parse out of the document  
is flat text inside that data-string.
But some of that infomation could also be found via Tag-sequences.

So I'm looking for a possibility to combine both kinds of attempts.

How to do it?


   Oliver