[ptt-users] TestGen4Web and HTMLunit: difference in behaviour?

[Dominique de Waleffe]

Continuing my explorations of TM5-B2,

In the script I recorded, tg4w wrote:
<xpath><![CDATA[*/TABLE[1]/TBODY[1]/TR[3]/TD[1]/TABLE[1]/TBODY[1]/TR[1]/TD[1]/TABLE[1]/TBODY[1]/TR[1]/TD[1]/*/A[@CDATA="Select 
Project"]]]></xpath>
as selector. When I run the script in Firefox with recorder, it runs 
successfully, however, when run through htmlunit, I get a null pointer 
exception in the click action ( TestGenForWeb.java:153: prints Elem null)

I suspect the difference of behaviour (in this case) may be linked to 
the treament of white space around the text of interest. Could the 
translation of @cdata=value be more liberal in what it matches as text?
at least, stripping the white space around the contents.

So it looks like the behaviour is different and thus the xpaths recorded 
are not reliable when run under htmlunit. This is annoying...

I then tried to record the xpath directly using xpather (FF extension) 
and plug it into the script definition. This lead to an other type of 
problem: the normalizeXpath() is not very robust when presented an XPATH 
not coming from TG4W (it fails if an element does not have a [, or if 
values are delimited with single quotes).

Examples:
1)
oldXpath 
*/div[@id="PageContainer"]/div[@id="content"]/*/tr[3]/*/td[1]/table/tbody/tr/td/span/a 
newXpath "//div[@id="PageContainer"]/div[@id="content"]//tr[3]//td[1]//////"
Error getting page Expected one of '.', '..', '@', '*', <QName>
org.jaxen.XPathSyntaxException: Expected one of '.', '..', '@', '*', <QName>

2)
if the path contains div[@id='PageContainer'] then the line     	 
sb.append(subPart.substring(subPart.indexOf("="),atIndex+1));
throws an invalid index exception (since we have been looking for ")

Any ideas on how to address this issue globally?

D.

-- 
Dominique de Waleffe