Fix dis grids

[andreab1997]

These are the new grids fixed after

• the y bug in yadism was solved (which was affecting the pinecards of HERA_CHARM and HERA_BOTTOM ).
• the threshold bug solved after yadism==0.12.3 (used version for previous grids)

This is the reason why all the DIS grids are changed.

CC: @giacomomagni @niclaurenti

[alecandido]

Why all the grids are changing, and not only HERA flavored?

[felixhekhorn]

Why did you change all of them? I guess you wanted also https://github.com/NNPDF/yadism/pull/177 in?

[felixhekhorn]

Actually, looking at the discussion there: we were aware of the problem ... in the end @andreab1997 did you recompute the grids?

[andreab1997]

Why did you change all of them? I guess you wanted also NNPDF/yadism#177 in?

Yes I am sorry, let me edit the description.

[andreab1997]

Actually, looking at the discussion there: we were aware of the problem ... in the end @andreab1997 did you recompute the grids?

Judging from the results I would say no, I did not recompute them.

[felixhekhorn]

Can you please add the comparison to old? (so this script I guess)

[andreab1997]

Can you please add the comparison to old? (so this script I guess)

I added the full comparison of all the sv fktables in b7b797a (for NLO) and in d897e84 (for NNLO)

[alecandido]

Files are exponentially increasing, now we have 1.3k.

Are we really supposed to review? What is the use case for so many files?

[andreab1997]

Files are exponentially increasing, now we have 1.3k.

Are we really supposed to review? What is the use case for so many files?

They are there for reference since we never had such comparisons for scale variations. You can just ignore them if you don't care but @felixhekhorn was explicitly asking for (some) of them so I guess at least him would be interested.

[felixhekhorn]

They are there for reference since we never had such comparisons for scale variations. You can just ignore them if you don't care but @felixhekhorn was explicitly asking for (some) of them so I guess at least him would be interested.

I think they are useful if we want to check a specific dataset and see the difference to old ...

I did look at some (not all - that is impossible) - e.g. I think these are useful:

• NNLO central:
  • charm: https://github.com/NNPDF/theories/blob/d897e8484f8f4aac599698b3d4ede168db585f4c/misc/4001/comparisons_sv/comp_424/HERACOMB_SIGMARED_C.csv up to %
  • other: https://github.com/NNPDF/theories/blob/d897e8484f8f4aac599698b3d4ede168db585f4c/misc/4001/comparisons_sv/comp_424/HERACOMBNCEP920.csv no visible change
• NNLO SV:
  • charm https://github.com/NNPDF/theories/blob/d897e8484f8f4aac599698b3d4ede168db585f4c/misc/4001/comparisons_sv/comp_420/HERACOMB_SIGMARED_C.csv up to 20% and even more
  • other: https://github.com/NNPDF/theories/blob/fix_DIS_grids/misc/4001/comparisons_sv/comp_420/HERACOMBNCEP920.csv up to few percent and sometimes more
• NLO central:
  • charm: https://github.com/NNPDF/theories/blob/d897e8484f8f4aac599698b3d4ede168db585f4c/misc/4005/comparisons_sv/comp_405/HERACOMB_SIGMARED_C.csv same: up to %
  • other: https://github.com/NNPDF/theories/blob/d897e8484f8f4aac599698b3d4ede168db585f4c/misc/4005/comparisons_sv/comp_405/HERACOMBNCEP920.csv same: not visible
• NLO SV:
  • charm https://github.com/NNPDF/theories/blob/d897e8484f8f4aac599698b3d4ede168db585f4c/misc/4005/comparisons_sv/comp_401/HERACOMB_SIGMARED_C.csv yet bigger impact (including change of sign and order of magnitude)
  • other: https://github.com/NNPDF/theories/blob/d897e8484f8f4aac599698b3d4ede168db585f4c/misc/4005/comparisons_sv/comp_401/HERACOMBNCEP920.csv similarly: significantly bigger impact than NNLO

(I hope, I correctly identified the central)

arguably you can drop all the non-DIS stuff - they are useless here (rm -rf misc/400{1,5}/*/*/{ATLAS,CMS}* or something similar might work)

[alecandido]

Yes, but they are all generated files. In Git repos you usually want to store the source, not the generated assets.

In the case of this repo we already accepted to store some generated objects, because they are expensive to compute, and we want to preserve them for debugging. However, all these CSVs have been computed in a shot, so I'd rather store the generating script and the grids, rather than all the comparisons.

[andreab1997]

Yes, but they are all generated files. In Git repos you usually want to store the source, not the generated assets.

In the case of this repo we already accepted to store some generated objects, because they are expensive to compute, and we want to preserve them for debugging. However, all these CSVs have been computed in a shot, so I'd rather store the generating script and the grids, rather than all the comparisons.

We are already storing some comparisons (look at misc/400) and since they are not heavy I am in favor of storing them here.

[andreab1997]

Can we merge this?

[felixhekhorn]

arguably you can drop all the non-DIS stuff - they are useless here (rm -rf misc/400{1,5}/*/*/{ATLAS,CMS}* or something similar might work)

I would still do this before - but can even be skipped if you wish ...

what is theory 4005? (400 = NNLO & 405 = NLO, right?)

[andreab1997]

arguably you can drop all the non-DIS stuff - they are useless here (rm -rf misc/400{1,5}/*/*/{ATLAS,CMS}* or something similar might work)

I would still do this before - but can even be skipped if you wish ...

what is theory 4005? (400 = NNLO & 405 = NLO, right?)

Yes 4005 is NLO but it does not contain the original grids. It contains the grid with the ren_sv order included by pineko

[andreab1997]

arguably you can drop all the non-DIS stuff - they are useless here (rm -rf misc/400{1,5}/*/*/{ATLAS,CMS}* or something similar might work)

I would still do this before - but can even be skipped if you wish ...

what is theory 4005? (400 = NNLO & 405 = NLO, right?)

Actually I would keep also the hadronic: so far we did not have anything to compare with for scale variations. At least, next time we will need to update the grids, we will something to compare to.