Allow categorical variables in smpdf plots

Created by: Zaharid

Fix crash in plot_smpdf when the variable in the color axis is categorical, such as for the W and Z total cross sections. Encode the categorical variables as integers and use an appropriate color map to display them.

The implementation feels a bit untidy, with some variables acting at a distance, but there doesn't seem to be an easy better way considering that we'd like the color bar to be the same for all figures.

Also, the fact that np.unique sorts the output may not be what we want. See https://github.com/numpy/numpy/issues/8621. But don't want to make it more complicated.

requested review from @enocera

Created by: Zaharid

See result here https://vp.nnpdf.science/kqbWu8OWQum14sMYRegLAg==/#atlas_wz_tot_13tev

Created by: RoyStegeman

Review: Commented

np.unique seems fine to me. At the moment I don't imagine the order matters very much so as long as np.unique keeps things consistent I couldn't care less.

Did you mean to have a different figure for the Ws and Z?

Also, there are double parentheses due to group_label and the title both adding them.

Created by: scarlehoff

Review: Commented

Can this be merged?

I guess it would be good to add another smpdf test in which the categorical branch is used, maybe using one of the datasets in your report @Zaharid ? ATLAS_WZ_TOT_13TEV for instance

Created by: RoyStegeman

Review: Changes requested

There are different figures for only W and only Z but both figures contain labels for W and Z.

There are double parentheses due to group_label and the title both adding them.

Created by: RoyStegeman

Maybe this can be moved under the else statement

Created by: RoyStegeman

Review: Approved

Was this just waiting on a review or is there another reason why this hasn't been merged?

Could you run black and isort?

Created by: RoyStegeman

Review: Changes requested

Oh right, the tests...

Created by: Zaharid

@scarlehoff @RoyStegeman I don't know what the current test failure with the evolution is, but don't believe it matters for this. Please merge this.

Created by: RoyStegeman

Review: Approved

I'm happy with this PR. I think the error is because our tests are not compatible with the eko version and I believe this is fixed in #1754. @scarlehoff can you confirm?

Merged by: Zaharid at 2023-06-19 14:09:08 UTC

closed

merged manually