Re: [OpenSTA-devel] OpenSTA Feature proposal: Alternative for DOM

[Olaf K. <ok...@ab...>]

Daniel Sutcliffe wrote:
 > Olaf Kock wrote:
 >
 >>The issues are regarding addressing certain values on the page. I've
 >>been successful adding "ids" to html pages, e.g. instead of writing
 >>    "You have 3 articles in your cart"
 >>we write
 >>    "You have <span id="number_of_articles">3</span> articles...
 >>
 >>These ids are easily addressed without caring for building a DOM
 >>from possibly broken HTML.
 >
 > So exactly how do/would you address this from SCL?

I currently do address them from Java (HttpUnit/JWebUnit-Tests) like this:
   assertTextInElement("number_of_articles", "3");
(see http://jwebunit.sf.net/ and http://httpunit.sf.net/ for these projects)

One can also retrieve the values with getElement("number_of_articles")
and continue computing things "manually".

I have no preference regarding the addressing from SCL - be it
   LOAD RESPONSE_INFO BODY ON 1 INTO num, WITH "ID/number_of_articles"
or any use of EXTRACT or whatever comes to mind - as long as one can get
rid of html(0)/body(1)/.../table(1)...

 > I think this area is something which does need looking at to provide
 > a more durable solution to writing scripts which don't break so
 > easily when the site changes.  Many people I know will simply use
 > a combination of ~LOCATE and ~EXTRACTs rather than deal with the
 > issues that can occur using the DOM addressing,  but then of course
 > the 64K limit will eventually cause you problems.

I've had some problems with just extracting some text from html pages,
as sometimes they rather look like <td>Articles</td><td>3</td> with any
amount of whitespaces...

But then, this idea comes from functional testing, not load testing. I'd
like to know how frequently other peoples test scripts break upon
changes to the application being tested.

Cheers,
Olaf