[Webware-discuss] Re: [python-web-modules] More thoughts on testing

[Chuck E. <ec...@mi...>]

At 11:25 AM 3/20/2001 -0800, Shannon --jj Behrens wrote:
>Chuck,
>
>Do you use the W3C's HTML and CSS validators?  I find
>them to be very useful (I'm actually quite strict
>about DTD compliance).  Also, can you clarify about
>what you're testing for?  If you're simply testing
>that the site is producing "correct" responses, you
>probably only want to set up 1 or 2 screens that
>perform comprehensive diagnostics.  The output from
>such screens don't even have to be in HTML.  They can
>be in whatever format is appropriate for your spider.
>This is a technique that the people over at NaviSite
>use a lot (and I think it's a good one).

I have used the validators a little bit and will be using them more in the 
future since I made an embarrassing news post for Webware that Netscape 
users couldn't even read. :-)

Here is some of what I test for already:

   * no page results in a Python exception
   * Can the user sign in?
   * Can a signed in user visit all his user pages without getting a 
message asking him to sign in again
   * If a non-signed in user visits a user page
     * is he asked to sign in?
     * after doing so, does he get the original page he tried to access

Also, some of my links are programmatically generated. I want to ensure 
they are formed correctly and that when visiting those links I get not just 
any page, but the page I expected. I even want to check that certain links 
are tagged with the right CSS style.

I was also thinking of running the validator on all pages as I go, but 
haven't gotten there yet.

Regarding, your idea of outputting another format of what I would call 
"structured data", that was my original thought. I could output the pages 
as Python dictionaries (containing lists, dictionaries, strings, etc.) 
which I could nab in one eval() and work with.

But then I have a disconnect between what a user would really be served and 
what my test suite is being served. I'm not saying it wouldn't help. It would.

But if I can easily suck in an XHTML file as an easy-to-use Pythonic 
structure then I think I can have my cake and eat it too.


>As an aside, since I do output buffering in Aquarium,
>I've often thought it would be cool if I ran the
>output through HTML Tidy (as a filter) before
>outputing it to the user.  This would have the affect
>of further cleaning up the HTML as well as nicely
>indenting it.  I might or might not have to worry
>about spacing in tables--this can sometimes be a
>problem.  In general, it'd probably be a cool hack.

Sounds like a cool idea.


-Chuck