Make pineappl evolve support EKO v0.15

[Radonirinaunimi]

As the title says, pineappl evolve does not work with EKO v0.15 because of a change in format. Examples of a grid and an EKO (LHCB_WP_8TEV.*) have been uploaded in the usual place on the server.

[Radonirinaunimi]

The current issue when running the code is that it raises the following error:

Error: no operator for x = 0.00020843298688247308 found

I believe this is because the following:

pineappl/pineappl_cli/src/evolve.rs

Line 376 in 5a68926

x1: metadata.xgrid,

should actually be grid.evolve_info.x1? Is that correct @cschwan? And if so, how should we approach it?

[cschwan]

Depends on what metadata.xgrid actually is. Is it x-grid for the input grids or for the output FK-tables? It must be the grid x-grid. Another possibility: have you re-interpolated the EKO? If you don't do that, x-grids probably don't match.

[Radonirinaunimi]

Depends on what metadata.xgrid actually is. Is it x-grid for the input grids or for the output FK-tables? It must be the grid x-grid.

There is now one xgrid in the metadata and it is for the output FK-tables.

Another possibility: have you re-interpolated the EKO? If you don't do that, x-grids probably don't match.

At which stage/where in the pineappl evolve CLI should the EKO be re-interpolated?

[cschwan]

I always had to load the EKO in Python, reinterpolate it for the given x-grid, store it, and only that operator is usable with the CLI. I'll try find my Pyrhon script later this evening.

[Radonirinaunimi]

Ah, I see! If you could provide me with the Python script, that'd be really great! Then I can re-interpolate the EKO, re-upload it, and add a test.

[cschwan]

I pushed it to master, hopefully it still works. You may want to change the hardcoded paths, though.

[Radonirinaunimi]

I pushed it to master, hopefully it still works. You may want to change the hardcoded paths, though.

Thanks! Unfortunately, the script doesn't work anymore with new EKO versions. I opened an issue here: NNPDF/eko#472

Once a fix is provided, I will fix the script in master. I can write it such it is compatible with all the versions of EKO using https://github.com/vmoens/pyvers.

[cschwan]

Once a fix is provided, I will fix the script in master. I can write it such it is compatible with all the versions of EKO using https://github.com/vmoens/pyvers.

Do we need that backwards compatibility? If not there's no need to make it more complicated.

[Radonirinaunimi]

This is now done! The test is failing because the data cache have not been updated. I would prefer to do so after #344 is merged because that PR also unifies the caching action in one place, otherwise there will be some conflicts. So if you are fine with that PR, I will merge that first.

PS: I have update the Python script in master.

[cschwan]

The only missing bit is an entry in CHANGELOG.md, after that feel free to merge!

[felixhekhorn]

actually: this would hopefully correspond to __data_version__ = 3, rather then the API version v0.15, so in the end I think it is a good name 😇