[Rivet] HepMC Python module

[Andy Buckley]

Frank Siegert wrote:
> We had the discussion of what to do with the HepMC python module, which we 
> use in both Rivet and AGILe at the moment. The possible solutions we came 
> up with were (please amend if necessary):

> 2. Have it as a standalone package.

I like it! ;)

> 3. Have it in Rivet and AGILe.

Messy!

> I would like to propose a fourth option:
> 
> 4. Move AGILe's HepMC output functionality (it's only a few lines) into 
>    the C++ part, such that we don't need the python HepMC module there. 

This unfortunately blocks the potential use of rivet for running minimal
AGILe generators (a very small addition that could speed up Fortran
generator use a lot). It would also require hard-coding a lot of stuff
into the Rivet C++ interface where it's currently been very nice to have
the flexibility of Python.

> Actually thinking about it now, I don't even see that we couldn't do 
> something similar for Rivet, such that we don't depend on the HepMC python 
> module at all. But I might be stirring Andy's wrath here ...

In principle I don't have a big problem with that idea, but I think that
the module is being used for more than cosmetic things and so it's
better to keep it standalone. I certainly don't want to fork the Python
interface!

We'll look more closely at this when it comes to the 1.1.3 Rivet release
(when rivetgun will disappear), but it's good to have the options
explicitly written down. If the module isn't really giving us anything
necessary (or nice --- being in Python means that the AGILe interface
can be a purely runtime dependency), then we do it all in C++, otherwise
we just add the pyhepmc package to the bootstrap script (and get it into
Genser)

Either way, it definitely needs to be decoupled from mcview --- nice
though that is, it's no longer the important thing in that development tree!

Andy