[Xapian-devel] Project for a group of 5

[Olly Betts]

On Tue, Mar 11, 2014 at 08:50:14PM +0530, Siddhant Mutha wrote:
> We would like to do the project on improving Java bindings. It is a GSoC
> project, although I guess we can take it up as such.

The timing is rather problematic - we've had at least four students
already express an interest in that project idea for GSoC, and at least
one has put in a significant amount of research already.  I've no idea
at this stage how many of those would result in proposals actually being
submitted to us, or if any of the proposals would be good enough for us
to want to accept, or how many slots we'll get from Google (I don't even
know how many mentors we'll actually have).

But if you're going to work on this, it doesn't seem sensible for
someone to also be working on it as a GSoC project, so if we're going
to do that we really need to withdraw the idea promptly and let the
students who expressed an interest know and help them find other ideas
to work on.  However, from below it seems you need to get agreement from
your professor, so I'd guess we may not even know for a few more days.

My preference would be to find you something which isn't on the GSoC
list, or at least something which we've not had interest in yet so
we can remove it without messing GSoC students around.

Perhaps implementing bindings for Go (http://golang.org/) would be a
good project for you?  It's an interesting language with some nice
design ideas, and you'd probably find implementing a fresh set of
bindings more interesting than trying to fix up the shortcomings of some
existing bindings.

> But before we go ahead, we would like to have some details as to how much
> work it would be and also how would we evaluate the work done. (compare
> test cases / compare java code with c++ code ?)

For Java, you could also compare how compatible they are with the
hand-wrapped JNI Java bindings, though I think "providing a clean
Java-like API" is more important than such compatibility - there don't
seem to be a lot of existing users of the Java bindings.

> Is there any way in which we can quantify the improvements made to the API?
>
> Lets hope our prof agrees to this project. If not, we'll have to think of
> something else.

You could measure what proportion of C++ API methods can be used from
Java or Go; also what proportion of wrapped methods are exercised by the
testsuite of the Java or Go bindings.

Cheers,
    Olly