- From: Curt Arnold <carnold@houston.rr.com>
- Date: Tue, 21 Aug 2001 11:32:38 -0500
- To: <www-dom-ts@w3.org>
Some feedback on the DOM level 1 core testsFreek de Bruijn wrote: >Would it be a good idea to combine multiple tests into one class, instead of using a > separate class for each test? This would reduce the huge number of classes that are > generated at the moment, making the test suite more manageable and improving > performance as well. The original NIST test suite has 15 classes > (one for each interface), which is a lot easier than the 250+ classes we have now. I think that we want to maintain the definition of one test == one XML document since that gives us a revision history of each test instead of a revision history of all tests for a particular interface. In addition, hopefully we will have tests from different sources and trying to have different authors working in the same file is just not a pleasant thing. The one test == one Java file/class flows out of that since it is most natural to do a case by case conversion using XSLT. If it class bloat is really a significant issue, we could address it. But I can't see performance being a crucial factor in running the suite as long as a run is less than a few seconds. >In tests characterdataappenddatanomodificationallowederr.xml, >characterdatadeletedatanomodificationallowederr.xml, >characterdatainsertdatanomodificationallowederr.xml, and >characterdatareplacedatanomodificationallowederr.xml the tagname in the following statement >should probably be "gender": Did a global change of "geneder" to "gender" >For tests characterdataappenddatanomodificationallowederr.xml, >characterdatadeletedatanomodificationallowederr.xml, >characterdatainsertdatanomodificationallowederr.xml, >characterdatareplacedatanomodificationallowederr.xml, and >characterdatasetdatanomodificationallowederr.xml the comment tells that the third gender >element should contain an entity reference, but this is not the case in the older version of the >staff.xml document. I held off on this one till we can get a clarification about the resource files from NIST. >In documentcreateentityreferenceknown.xml, the type of newEntRefNode should probably be >EntityReference (instead of Element): Changed >Several tests seem to test equivalence of two lists where one list contains a String and the other >contains an Integer. Some tests also try to assert null values of nodes with a String value of >"null" in a list, maybe it's better to use assertNull. Since I don't have access to the cvs server >(but who has ;-), could somebody fix these issues? Changed those you mentioned plus one more (nodeclonenodefalse). I did take some editoral liberty with documentcreateprocessinginstruction. There was a discussion on the www-dom-ts mailing list much earlier http://lists.w3.org/Archives/Public/www-dom-ts/2001Apr/0020.html about the inappropriateness of testing createProcessingInstruction with a target ("XML") that is an illegal value per the XML spec. I have changed the target to an innocuous value "TESTPI". Most DOM implementations will create a PI with an "XML" target, but MSXML will reasonably raise an exception. Joe Kesselman's suggestion (http://lists.w3.org/Archives/Public/www-dom-ts/2001Apr/0020.html ) was to change the target and but not to add a negative test for an "XML" target until the Level 3 effort where the appropriate behavior will be prescribed.
Received on Tuesday, 21 August 2001 12:32:42 UTC