Add functionality to remove bins from grids about pineappl

[cschwan]

I'm preparing a test to check this and there seems to be a problem indeed; Grid::delete_bins seems to have no effect.

[cschwan]

Commit 46f3e5f adds a method Grid::delete_bins which does what is needed at a very basic level. We still need:

• Python bindings and
• a CLI subcommand (possibly after merging #95)
• testing and unit tests.

[alecandido]

Python bindings provided in cd7bbaf

It builds, but I had some difficulties to test it. I tried in the interpreter, but it seems not to change the grid...

[cschwan]

Commit fccd890 fixes the problem, and commit 191f62c makes sure it stays fixed. However, the current implementation has the limitation that only bins from the left and/or from the right can be removed. Is that a problem for our purposes?

[cschwan]

Commit 5e629bf fixes yet another bug and commit aeddd4f adds the CLI.

[alecandido]

However, the current implementation has the limitation that only bins from the left and/or from the right can be removed. Is that a problem for our purposes?

I'm not sure to have understood what the limitation consists of...

Can you make an example of something I can't do?

[cschwan]

Say you have 10 bins you can't delete 3 and 4, for instance, because they are neither at the left (first bin would be 0) nor are they at the right (last bin would be 9). You could, however, delete bin 3 and 4 if you'd also delete 0, 1 and 2 at the same time, or alternatively if you'd delete 5, 6, 7, 8 and 9.

[alecandido]

Ok, now I understand, and it might be a limitation in general, but (if I remember correctly) the filters in apfelcomb.db should be all at the beginning.

In any case, we'll raise a dedicated issue if and only if we find something for which is needed. For the time being it is perfect as it is.

(I was confused by the words left and right, since each bin has a left end and a right end, but now I perfectly understand)

[cschwan]

I was bit imprecise, so thanks for asking for a clarification! I agree with your proposal to open a separate Issue when this becomes a problem, and we'll know that as soon as the todo!() macro panics.

[felixhekhorn]

Unfortunately we will hit this case ... since I had already a very preliminary look at the cutting, I realized this already (but thought back then, that in fact you did NOT have this limitation): the case is DYE866P

• it is in the runcard: https://github.com/NNPDF/fktables/blob/7f34897005a90a6bcca38bf45a90b8dfb599dc99/data/runcards/NNPDF40_nnlo_as_01180_1000.yml#L32
• and it has a bin in the middle: https://github.com/NNPDF/fktables/blob/7f34897005a90a6bcca38bf45a90b8dfb599dc99/data/apfelcomb.dat#L200

[cschwan]

OK, no problem, I'll take care of it.

[alecandido]

Oh damn, you're right.

There are also a couple of CMS grids.

[cschwan]

Removing any bins should work since commit 50c12c4.