Hi John and Malcom,

I know that the HaXml library has a separate error-correcting HTML parser that works around most of the common non-well-formedness bugs in HTML: Text.XML.HaXml.Html.Parse

I believe HXT has a similar parser: Text.XML.HXT.Parser.HtmlParsec

Indeed, some of the similarities suggest this parser was originally lifted directly out of HaXml (as permitted by HaXml's licence), although the two modules have now diverged significantly.

The HTML parser in HXT is based on tagsoup. It's a lazy parser (it does not use parsec) and it tries to parse everything as HTML. But garbage in, garbage out, there is no approach to repair illegal HTML as e.g. the Tidy parsers do. The parser uses tagsoup as a scanner. The table driven approach for inserting missing closing tags is indeed taken from HaXml. Malcom, I hope you don't have a patent on this algorithm. Regards, Uwe

Hi Ivan,

Uwe Schmidt writes:

...

The HTML parser in HXT is based on tagsoup. It's a lazy parser (it does not use parsec) and it tries to parse everything as HTML. But garbage in, garbage out, there is no approach to repair illegal HTML as e.g. the Tidy parsers do. The parser uses tagsoup as a scanner.

So what is parsec used for in HXT then?

for the XML parser. This XML parser also deals with DTDs. This parser only accepts well formed XML, everything else gives an error (not just a warning like HTML parser). tagsoup and the HTML parser do not deal with DTDs, so they can't be used for a full (validating) XML parser. Regards, Uwe