First PR on XercesJ enhancement to add location info to DOM nodes: https://github.com/apache/xerces-j/pull/7 This compiles, but I can't get the unit tests to run, being unfamiliar with edit/compile/test cycle for XercesJ.
On Tue, Feb 18, 2025 at 9:35 PM Mike Beckerle <mbecke...@apache.org> wrote: > Dev team, > > > > This has come up so many times now. In Daffodil, in Internal owl tools > like XSAT2, now in the DFDL coverage tool. > > Frankly anything that processes XML Schemas hits this or other limitations > of this library. > > > > I think we really should fix this the right way. > > > > So it all boils down to two things: > > > > 1. Add a feature - Modify Xerces so that w3c DOM tree nodes, if > downcast to an Impl class or interface, have the location information > populated and available. > 2. Adding enabling of this feature to ws-schema to access to the > underlying DOM object – which would provide access to the location info, as > well as solve all the other limitations mentioned in our Daffodil coding > style wiki article which mentions a number of shortcomings in ws-schema, > not just file/line info. This should be dead-simple. It just needs to > invoke Xerces DOM parser with the new feature option turned on. > > > > This all needs to be optional (default off) since it adds to memory use > and could slow things down marginally. > > > > This would make it possible to eventually strip all the XSD traversal > logic hacking out of Daffodil and have it use ws-schema as the way to > access the XSD-flavor of DFDL, paving the way for alternate non-XSD syntax > variants to fit into the design also, which is a priority feature for DFDL > v2.0 and all the XML-haters in the world. > > > > I forked xerces-j to start investigating this. I already have a fork of > ws-schema. > > > > Collaborators are welcome. > > > > ChatGPT is quite helpful at explaining things about how Xerces J works – > e.g, how to add a new feature flag that works the way the other feature > flags work. > > >
