[xsd-users] Large XSD-schema, speed and identity constraint validation

[Stefan de Konink]

On Thursday, May 14, 2020 2:43:28 PM CEST, Boris Kolpackov wrote:
> Correct. Validation in generated code (also called "perfect parser") works
> well for smaller/simpler schemas (which is the reason why we went this way
> for XSD/e, our mobile/embedded version). But for schemas we are talking
> about (e.g., GML), the size of the generated code becomes impractical
> in many cases.

So XSD/e would be running on the generated C++ code without Xerces? As 
comparison I would still find this an interesting approach.


> Do you perhaps have remote (e.g., http://) schema references in (some
> of) your schemaLocation attributes? That would explain these results
> quite well.

Thanks for this tip. I'll try to run a wireshark session to validate if 
this happens. If this happens, is there any way of registering or caching 
"local" equivalents without changing the XSD?


>> So I am missing the "Key/Value" report but get an ocean of 
>> duplicates where
>> I can't find out the reason.
>
> I haven't looked into this in detail but maybe you can resolve the schema
> names referenced in the error message back to schema locations based on
> the loaded schema grammar.

I have asked a questions in the Xerces C++ group, concerning this issue, 
but I didn't receive any input on it. I am also surprised about the number 
of duplications, either the Java version is not picking it up or the C++ is 
duplicating it.

-- 
Stefan