[vtkusers] Experimental enhanced wrappers for VTK

[David Gobbi]

In the new wrappers, there is already full support vtkStdString
support Tcl and Java, and I've even added string regression test for
Tcl.

Unicode was easy to add to Python (a few hours work), and it should be
similarly easy for someone to add it for Tcl and Java.  There are
plenty of people who want it, I'm sure.

Divergence is a concern, but at least all wrappers use the same
front-end, so transferring enhancements from one set of wrappers to
another is that much easier.  And Brad King mentioned a wrapper
utility .cxx that all wrappers use... this would be great, because
right now there is a huge amount of code duplication.

I'll switch over to the branch immediately, and in the future I will
make branches for any other projects that involve multiple classes.

   David


On Wed, May 26, 2010 at 3:58 PM, Marcus D. Hanwell
<marcus.hanwell at kitware.com> wrote:
> Hi David,
> On Wed, May 26, 2010 at 5:46 PM, David Gobbi <david.gobbi at gmail.com> wrote:
>>
>> Thanks for doing this!  I was not looking forward to doing the
>> clean-up work myself... when I started it never occurred to me to
>> create a branch, my plan was to eventually re-base my changes on the
>> VTK master.
>
> Topic branches are much easier for us to work with, and if you avoid merging
> into them/rebasing then the eventual merge into master becomes simpler. Now
> that we have it you could continue your work in that branch, or we could
> cherry pick the commits you make.
>>
>> There are just two minor changes that I can squeeze in over the next
>> two days, and I will email you when they are done.  After that, the
>> sooner that the branch can be merged into master, the better.  In
>> order for the special-type wrapping to be fully realized, most (or
>> all) of the BTX's will have to be removed from the code.  A change
>> that big will probably be easier to do in the master than to merge in
>> from a branch.
>
> A global change like that is definitely something that would be easier done
> in master as you say. We can take care of that once it looks ready to go
> in.
>>
>> I've already tested VTK (with python, tcl, and java) without BTX/ETX.
>> The Tcl and Java wrappers compiled and tested cleanly, and so did
>> Python after I did some small fixes to the way that it does special
>> type wrapping.  It is these python fixes that I need to commit before
>> the merge.
>>
> I think one of the remaining concerns is how much the wrapping might
> diverge, i.e. if Python supports vtkVariant, unicode, vtkStdString etc but
> other languages did not. I am certainly very pleased with the changes in
> Python, and can see how much it will improve the Python wrapping in charts
> and infovis.
>
>>
>> There are other changes that I'm planning for the Python wrappers
>> (enums, etc) but for them I'd prefer to wait until after the merge.
>> I'll put together a wiki page so that they can be reviewed.
>>
> You could publish them in a topic branch for us to review. The main caveats
> remain to avoid rebasing/merging from master. Once we are ready to merge
> into master we can take care of that. Thanks for putting this together, and
> to Keith for preparing this topic branch.
> I am very hopeful that we can merge these changes in soon. We have been
> working away behind the scenes to get this ready to be reviewed and merged
> in.
> Thanks,
> Marcus
>>
>> On Wed, May 26, 2010 at 2:39 PM, Keith Fieldhouse
>> <keith.fieldhouse at kitware.com> wrote:
>> > David
>> > I've been spending some time looking at your Wrapping work.  It appears
>> > that
>> > your git tree was built with a mixture of rebasing and merging which
>> > resulted in a somewhat wonky commit history (duplicated commits
>> > interspersed
>> > with commits coming from VTK master).
>> > In order to get a clean topic branch that will be a little easier for
>> > folks
>> > here to look at (and to eventually merge with master) I spent a  bit of
>> > time
>> > untangling your history (mostly by examining parentage and looking at
>> > commit
>> > subjects).  In the end, by cherry picking the right commits in what I
>> > considered the "right" order I was able to build a clean topic branch
>> > with
>> > no conflicts.  Once I had that, I rebased it against master (as of this
>> > afternoon sometime when I did it).  The result can be found in
>> > git://github.com/keith-fieldhouse/VTK.git
>> > As the topic branch Gobbi/Wrapping.
>> > The result builds and the tests run so I believe that I've built a a
>> > representative branch from your work.
>> > If you plan on doing additional work on this, would it be possible for
>> > you
>> > to grab this topic branch and work on that (without rebasing or merging
>> > for
>> > now)?    That way we'll have a clean topic branch  for an eventual merge
>> > to
>> > "Next" when we switch to branchy development for VTK.  Or if branchy
>> > development comes later, this clean topic branch will be easier to
>> > rebase as
>> > well.
>> > Thanks again for your work on this.
>> > Keith
>> > --
>> > Keith Fieldhouse
>> > R&D Engineer, Kitware Inc.
>> > (518) 836-2190
>> >
>> >
>> >
>> >
>
>