kxml_compiler

[Frans Englich]

On Wednesday 01 February 2006 15:16, Allen Winter wrote:
> On Tuesday 31 January 2006 17:11, Cornelius Schumacher wrote:
> > On Tuesday 31 January 2006 20:50, Allen Winter wrote:
> > > I was wondering if kxml_compiler (which is currently located in
> > > kdepim/kode/kxml_compiler) should be moved into kdelibs4.  It seems
> > > like a general purpose tool that is not pim specific.
> >
> > Right. But it is far from being generally useful, so it would need a
> > serious amount of work, if it should go to kdelibs. I'm currently not
> > able to make a commitment to work on kxml_compiler, so unless somebody
> > else wants to take over, I would prefer to keep it where it is for now.
>
> Hi Cornelius,
>
> Do you have a list of changes that are needed.  And an estimate
> of how hard the changes are?   Are we talking junior jobs, or expert?
>
> You know that I really want to re-write all the holidays files into a
> RelaxNG schema. And move kdepim/libkholidays into kdelibs.   And I need a
> more robust kode/kxml_compiler to accomplish this.. and then kode will also
> need to be moved into kdelibs.

I'd say a lot(I'm not trying to prick down on someones piece of code here!).

Among XML geeks there is a wide, undisputed consensus that W3C's black sheep, 
also known as W3C XML Schema, is a horror to implement. The interoperability 
problems with schemas are also huge, and that's perhaps why I'm a bit 
paranoid about conformance in this area. W3C's test suite contains 23000+ 
test cases for a reason.

Code is not that interesting. What's interesting is if it can be proved that 
it's actually worth something. For code that will be used by many in general 
ways should be backed up by "This code is intended to do this and here is the 
<insert proof> for that it can do that". For a interoperability thing as WXS, 
I think test results are needed.

I've implemented WXS Part 2(and type hierarchy tools) for KXPath, and the 
implementation is ok. It's currently part of our own 3000+ unit tests, but 
will also be tested by the XQuery test suite. I think that code should be 
re-used and effort saved such that a better result is achieved.

APIs should also be secured with QTestLib tests and code should be 
documented(doxygen), but that's just my two cents.


Cheers,

		Frans