See http://marc.theaimsgroup.com/?l=axis-dev&m=108481109425165&w=2. Got google search working with DynamicInvoker. The result is in a org.w3c.dom.Element object.
-- dims On Sat, 15 May 2004 09:58:28 -0400, Davanum Srinivas <[EMAIL PROTECTED]> wrote: > Tom, > > You will need to traverse the symbol table and register say the > ElementSerializer/ElementDeserializer or your own custom ser/deser > before invoke. It should not be very difficult. > > -- dims > > > > On Sat, 15 May 2004 13:02:37 +0100, Tom Oinn <[EMAIL PROTECTED]> wrote: > > > > Hi all, > > > > We have a workflow system called Taverna (taverna.sf.net) which uses > > axis for the parts it need to call web service based components. I was > > able to use the dynamic invocation example to construct this and to use > > wsdl4j and axis's wsdl parsing to present the user with a view on the > > service that looks like a standard 'chip' type thing, has inputs and > > outputs and a big button marked go (this is per operation, obviously). > > > > This all works very nicely, can consume .net service and others as long > > as they're using simple types, so string, byte[] and arrays of these. As > > it happens this is normally sufficient, in our community > > (bioinformatics) these are far and away the most common services, > > descending as they do from wrapped command line tools - obviously a > > command line tool consumes files so binary or text is cool. > > > > There are a few services out there, however, which are document oriented > > and use complex types. All well and good, but of course our current > > framework panics and so our users can't construct workflows containing > > these types and that's a shame. What I was thinking of doing was > > presenting an operation that consumed complex types as always being > > document oriented, so in the workflow it would have a single input xml > > and a single output xml document. These can then be parsed and created > > by adjacent processing entities in the workflow. > > > > So - how easy is it going to be to modify the existing dynamic > > invocation code to a) notice that there's a complex type and b) switch > > automagically into something that looks like a document style invocation? > > > > Any pointers or experiences would be most helpful. > > > > Cheers, > > > > Tom > > >
