Improving vtk (was RE: [vtk-developers] vtknew tree and testing / examples)

[Prabhu Ramachandran]

hi,

>>>>> "SB" == Sebastien BARRE <sebastien at barre.nom.fr> writes:

    SB> No, but tests help increasing the *coverage* :
    SB> http://public.kitware.com/vtk/quality/MostRecentResults/solarisCoverage/Coverage.html

Oh yes I understand that but to me it seemed like David meant that
examples should be fewer.  I agree that testing code can be separated
but dont agree that we need fewer examples.

    SB> I don't think it makes sense to have a "good example"
    SB> (i.e. not a test) for each class, neither does it make sense
    SB> to cover *all* aspects of a class in a "good example".

A test ideally should test all aspects of a class.  An example neednt
but it helps to have a neat example.  That way you run the example and
know how the class works.

    SB> Hence, ideally there should be more tests than examples, I do
    SB> agree with David, unless I miss something :)

    >> how about readablity for the new user?  Do we want the new
    >> users to see stuff that is speicfic to testing and confuse
    >> them?

    SB> David agrees with separating Testing from Examples, and I
    SB> think too that this is a good idea, because "tests" are not
    SB> written always like "good examples" or "apps", as we try to
    SB> maximize the coverage in "tests" (at the cost of some "task
    SB> logic" sometimes).

I think it is nice to see how you can use a particular thing.  Well,
I'd read the tests anyway so it might not matter.

    >> I am not so sure how folks in Europe/US think about buying a
    >> book that costs $60 in order to understand a library.  I am not
    >> talking of a company or university but of some average user who
    >> wants to do visualization.

    SB> I do not follow you exactly :) OK, I'm not familiar with these
    SB> issues, but is there a lot of independent programmers using
    SB> VTK for fun ? OK VTK is fun, but there are funnier things to

Well, IMHO there _should_ be a lot more folks using VTK for fun.
There arent because (1) there is a learning curve and (2) no free
docs.  (1) and (2) are linked in that you cant go up the learning
curve without spending $$ unless you pester folks on the mailing list
where its unlikely that folks have the time to do hand holding.

    SB> do :) If you are a company or even a small group of
    SB> programmers focusing on VTK, $60 is "peanuts" comparing to the

Ofcourse!  My point is why should only folks with 'focus' in VTK have
to use it?

    SB> other investments. And for a university or a lab, this should
    SB> be no problem.

For universities _abroad_ it isnt a problem.  It doesnt work the same
way here.  Besides, let me say I visit a site on PHP and it catches my
fancy.  Unless I am sure I can learn PHP with some examples that
really get my attention I wont think of buying a book on it.  What I
imagine any person would do is 

  (1) see if the software interests them
  
  (2) look at examples/screenshots/code

  (3) download, test

  (4) learn how to use it.

Now, IMHO there are two important problems in steps (3) and (4).  (3)
because there arent enough binaries.  (4) because there arent good
enough docs.  So unless someone is really dead serious about VTK they
wont cross step (2).  So you basically loose a few potential users
before step (3).

    >> I think VTK is wonderful for doing just that but unfortunately,
    >> its hard to get a feel for it without free docs.  I am sure you
    >> will find more users if there were free docs.

    SB> Let's say that it would be cool to have a free "intermediate"
    SB> doc.  Something between "nothing" and the "user's guide".

Yes, that is exactly what I meant.  So a new user can atleast get
started and then get more details later.

    SB> The best that I could do is :
    SB> http://vtk.barre.nom.fr/links-examples.html Not enough.

    >> One of the greatest things about most free software is they
    >> come with free docs.

    SB> Perl comes with free docs. But Jesus, the "Perl Cookbook"
    SB> (O'Reilly) is an incredibly valuable ressource, and I do not
    SB> regret every buck I put in it.

Yes, I can get some of those books fairly cheaply here too (at about
6-10$).  But I prefer online docs.  One big reason (initially) that I
jumped to Python was that it had much better docs (for me) than Perl.

    >> That said I wouldnt mind helping with writing these docs.  But
    >> as David says we all have limited time and I am really no
    >> expert at VTK.

    SB> A kind of "task-oriented" book would be nice. A joint effort ?
    SB> But I guess someone will have to pay for that.

Why?  Wont the users contribute?  Actually if we get the examples
going along with the VTK tutorial (maybe we could translate the
examples to different languages) this would be a good start.

    SB> I will switch my "VTK links" page to a "open portal structure"
    SB> within the summer. I'll try to put a code/snippet section. I'd
    SB> definitely love to achieve something like
    SB> http://www.zend.com/, which looks like the best portal I've
    SB> seen. A "Code Galley" will be in.

Cool!

    SB> But I'll be honest : if I only I could received as many "Links
    SB> to valuable info" than "How can I do that in VTK ?" emails in
    SB> my mailbox, we would all be happy.

Well, maybe we should mention this on the users list.  Unfortunately,
I am way too busy now and wont be in action for close to two months
from next week.  Hopefully when I get back...

    >> The other thing that would be nice is for more binary
    >> distributions of VTK.  Atleast VTK for different versions of
    >> linux should be available.  Nightly rpms built on RH 5.2 arent
    >> good enough. :(

    SB> VTK will move to Dart this summer (won't it ?). Maybe it can
    SB> help deploy build/test on different platforms ?

What is that?

prabhu