Fix struct.detached #161
Fix struct.detached #161
Conversation
src/eko/output/struct.py
Outdated
| del bases[basis] | ||
| if basis in bases: | ||
| bases[f"_{basis}"] = bases[basis] | ||
| del bases[basis] | ||
| bases["pids"] = np.array(br.flavor_basis_pids) | ||
| for k in ("xgrid", "_inputgrid", "_targetgrid"): |
There was a problem hiding this comment.
Dropping the former loop, also this loop has to be dropped. Essentially, replace k everywhere with "xgrid"
There was a problem hiding this comment.
This is what I did before but then the tests are really a problem. Of course fixing the tests would be the correct solution but, since at the moment we are not using at all rotations, I believe that in the next future we need to get rid completely of it. This will be the proper change and at that moment we can drop these loops and fix the tests accordingly.
BTW I am saying that we are not using at all rotations because even if I use x_grid_reshape and it changes targetgrid, then load() would completely ignore targetgrid and would set it to xgrid (wrongly because instead the operator is rotated). This is the problem that I have with pineko. So now rotations is not even keeping track of the rotations. It is completely useless.
There was a problem hiding this comment.
BTW I am saying that we are not using at all
rotationsbecause even if I usex_grid_reshapeand it changestargetgrid, thenload()would completely ignoretargetgridand would set it toxgrid
This should not happen: we want to "ignore" targetgrid (i.e. not putting it at all) only in the runcard. The reason why it is failing in Pineko is different: metadata are not automatically updated, and you are not doing it specifically.
Metadata are one thing, and runcards are another. Incidentally, they both have rotations - but we can drop rotations at all from the runcard, since we'll never use it any longer.
EDIT: this is partially the reason why I'm pressing for #160, since at the moment there is a huge clash
|
So I removed the loops as agreed yesterday and started to fix the tests. However, as you can see, this is not as trivial as removing every mention (but the ones in |
| operator["rotations"][k], | ||
| log=operator["configs"]["interpolation_is_log"], | ||
| ) | ||
| if operator["rotations"]["xgrid"] is not None: |
There was a problem hiding this comment.
Actually, I'm not sure any longer with need this if: xgrid should not be allowed to be None.
Isn't it? @felixhekhorn
|
I believe that the tests that we are not passing now can only be fixed after we allow for the |
|
This PR updates I would close without merging. |
|
Agreed - let's close without merging.
you mean close all PRs and change the target branch of the last? |
We want
output.struct.detachedto work even if the eko does not containinputpids,targetpids,inputgridandtargetgrid, as this will be the case from now on. @felixhekhorn @alecandido