[wwwsearch-general] ClientForm request re ParseErrors

[John J Lee]

On Sun, 9 Jul 2006, bruce wrote:
[...]
> sgmllib.SGMLParseError: expected name token at '<! Others/0/WIN; Too'
>
>
> partial html
> -----------------------------------
> </table>
> <br />
> <FORM NAME='main' METHOD=POST
> Action="/servlets/iclientservlet/a2k_prd/?ICType=Panel&Menu=SA_LEARNER_SERVI
> CES&Market=GBL&PanelGroupName=CLASS_SEARCH"  autocomplete=off>
> <INPUT TYPE=hidden NAME=ICType VALUE=Panel>
> <INPUT TYPE=hidden NAME=ICElementNum VALUE="0">
> <INPUT TYPE=hidden NAME=ICStateNum VALUE="1">
[...]

You don't include the HTML mentioned in the exception message ('<! 
Others/0/WIN; Too') in the part of the HTML that you quote, but that 
snippet is enough to see what's wrong, and lets you find exactly where in 
the HTML the problem lies.  Comments in HTML start with '<!--' and end 
with '-->'.  The comment sgmllib is complaining about is missing the '--'.

You can work around bad HTML using the .set_data() method on response 
objects and the .set_response() method on Browser.  Call the latter before 
you call any other methods that would require parsing the HTML.

r = br.response()
r.set_data(clean_html(br.get_data()))
br.set_response(r)


You must write clean_html yourself (though you may use an external tool to 
do so, of course).

Alternatively, use a more robust parser, e.g.

br = mechanize.Browser(factory=mechanize.RobustFactory())


(you may also integrate another parser of your choice with mechanize, with 
more effort)


John