- From: Curt Arnold <carnold@houston.rr.com>
- Date: Sat, 26 May 2001 20:35:34 -0500
- To: <www-dom-ts@w3.org>
- Message-ID: <001301c0e64d$5360b190$ce44a018@curtbnmqewb54k>
Jeroen van Rotterdam wrote: >With the limited knowledge I've got now from the DTD I'm trying to image how a test for >a DOM L3 LS call would look like. > >example: > >void writeNode(in DOMOutputStream destination, in Node node) raises(DOMSystemException); >Would probably something like <DOMWriter.writeNode var="........ I think this is similar to problems that you would also encounter in writing tests to test event dispatching and probably traversals (but I don't know traversals). Definitely I don't think we want our little language to grow to the point that we can extend arbitrary interfaces. I did a few initial tests for event dispatch and the approach that I thought was reasonable was to have a build-in implementation of the listener (or in this case DOMOutputStream) that captured all the events (or output) and allowed you to insert code into the event handler. You could make assertions in the event handler or after event dispatching was complete. I started with some event tests in Java with Junit: http://cvs.sourceforge.net/cgi-bin/viewcvs.cgi/xmlconf/domunit/JUNIT/NET/sourceforge/XMLCONF/DOMUNIT/dom2/events/eventtarget.java?rev=1.1&content-type=text/vnd.viewcvs-markup and then added the minimal language constructs that allows to be represent that test. Which in my schema was the <declare><onEvent/></declare>, the EventMonitor.* elements and the EventMonitor datatype. I would assume a similar approach would work for Traversals and Serialization. Determine what you want to test and prototype a few in the language of your convienience and then add a few constructs to support those types of tests.
Received on Saturday, 26 May 2001 21:32:54 UTC