Hi Julien, Interesting, I did not realize that you used Jing for RelaxNG. Basically what we are trying to accomplish is supported by (we think) Jing See http://www.xmlhack.com/read.php?item=1675 unfortunately we have done a lot of modeling in W3C Schema...
However, we could get an approximation of "progressive/feasible validation" if we could get the MSV to report errors on all elements. We would ignore all elements that reported errors which were not included in the specific input were validating. What I saw in the MSV documentation that led me to think we could do this is at the end of part 4-1 of the MSV documentation (JARV.html) it says " If you throw an exception from the error handler, that exception will not be catched by the verify method. So the validation is effectively aborted there. If you return from the error handler normally, then MSV will try to recover from the error and find as much errors as possible." I took this to mean that MSV could be configured (through error handling) to catch multiple errors, but I have not had the time to actually try it. OPPS, I wrote schematron-validation and I meant the just validation, sorry :) As for creating schemas on the fly, this could be useful if the partial validation approaches above fail, or are not powerful enough. If we could validate against a generated schema, we could run a transformation against the master schema picking out only the elements that need to be validated and use this as the input to schema input to validation. Though this seems like a much harder approach, I can see how it might be useful. I would imagine that if you allowed annotated validation to be accessed in the same way you do pass/fail validation, the ability to use generated schemas would be a freebee. Let me know if you read the MSV doc the same way I do. Thanks, Peter -----Original Message----- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Julien Mercay Sent: Monday, July 28, 2003 3:30 PM To: OXF Users Subject: Re: Validation Hi Peter, You are right, we are using MSV for W3C Schema validation, and Jing for RelaxNG schemas. MSV and Jing are both fail-fast validator, and only one error can be be reported. This is unfortunate for form validation. I couldn't find anything in MSV documentation about turning off the fast-fail. Please let me know of you can confirm that. MSV doesn't support schematron, which was implemented as an extension at Berkeley. Schematron could be later included in the Validation Processor so it could also be used for XForms validation. > One more thought, with the current XForms validation, is it possible > to generate a schema on the fly, and include a reference to it in the > xform model? Right now it seems that only a schema on the file > systems can be used to do xform validation. No, you can't use a generated schema with XForms validation. Please let me know if it would be useful for you, and we will implement it for the next release. Regards, Julien Peter Charles wrote: > Julien, > > It seems that the xforms input processor is what you are referring to as > "XForms Validation." I assume that processor has a call to the Sun MSV > baked in. I looked for an example of how the xforms-input is used but > could not find any use of the processor in the examples outside the > indirect usage by the web-app controller. > > One thing I noticed about the annotated validation of the xfoms-input > processor is that it only reports the first error. Is this a function > of the xforms-input processor or the MSV(fail-fast)? The MSV > documentation is a bit vague on this but it does suggest that the MSV > can be configured to catch as many errors as possible. > > Since in the context of xforms input there is no chance that of the > instances document having extra elements or being malformed, what is > really being validated is the data type, it would be more useful if > validation reported errors on all elements. This could of course be a > configuration parameter. > > If both the xforms-input processor and the schematron-validation > processor both use the Sun MSV, It would seem that the > schematron-validation processor could be made to take a config input > that would describe how the validation should proceed, i.e. annotated or > non-annotated. > > Peter > > _______________________________________________ oxf-users mailing list [EMAIL PROTECTED] http://mail.orbeon.com/mailman/listinfo/oxf-users _______________________________________________ oxf-users mailing list [EMAIL PROTECTED] http://mail.orbeon.com/mailman/listinfo/oxf-users
