RFU: py-ipython-0.8.1 - make py-scientific dependency a variant

[parrishmyers@…]

I noticed that when installing py-ipython, it includes py-scientific and subsequently netcdf because of dependencies. As far as I understand, there is no direct dependency on that package... can a variant be made to remove that dependincy for those that don't want to install py-scientific?

[markd@…]

It could be removed as a dependency and only added via a variant made for that. Do you think that behavior is more likely to please users? The package has no maintainer. Any interest in doing that?

[jochen@…]

Physics.PhysicalQuantities and friends in ipython actually need py-scientific, therefore a standard install of py-ipython should include the dependency on py-scientific.

I am not sure a variant that does not use py-scientific makes sense. It could be done, and ipython would also still install, but it might trigger bug-reports that are hard to follow...

[parrishmyers@…]

What confuses me is that iPython is hosted by the Scipy/Numpy folks, and yet there is a ScientificPython dependency in the profile support area... Which package is it a part of? It almost seems that the package could be modified to handle 3 configurations: none, py-scientific, and py-scipy. Although, this sounds like an upstream problem.

As a side note: I have installed iPython without the ScientificPython and it works great (as long as I don't include the physics profile -- which relies on ScientificPython).

[markd@…]

Based on jochen's comments, it sounds like we should leave it alone. It doesn't sound like there is a real advantage to not installing scientific, and potential problems as well. I think this can be closed.

[parrishmyers@…]

I think your right.

[jochen@…]

Closed, as consensus seems to be to leave the port alone.

[anonymous]

Milestone Port Enhancements deleted