If a facet is longer than 25 characters, preprocessor output files maybe overwritten

[schlunma]

Describe the bug

Example: Two ICON runs

- {project: ICON, dataset: ICON-XPP, exp: exp: atm_amip_r2b5_hybrid1_tuned_version_1}
- {project: ICON, dataset: ICON-XPP, exp: exp: atm_amip_r2b5_hybrid1_tuned_version_2}

will produce just one preprocessor file:

ICON_ICON-XPP_atm-amip-r2b5-hybrid1-tun_atm-2d-ml_Amon_ps_19790101-19800101.nc

Notice that the exp part is stripped at 25 characters. I am pretty sure this is due to the following line:

ESMValCore/esmvalcore/preprocessor/__init__.py

Line 311 in 4137dab

return re.sub("[^a-zA-Z0-9]+", "-", str(value))[:25]

@bouweandela Any particular reason why you implemented it like that?

[bouweandela]

File names/paths that are extremely long can cause trouble, depending on the file system, that's why I thought there should be some sort of limit. This could be implemeneted in a smarter way of course, e.g. limiting the overall lenght and if exceeded replacing the last few characters with a unique value (e.g. from the tempfile module) or a hash computed from the total filename. Some googling suggested that filenames are typically limited to 255 characters and paths between 1024 on older(?) OSX and 4096 on Linux.

[schlunma]

This could be implemeneted in a smarter way of course, e.g. limiting the overall lenght and if exceeded replacing the last few characters with a unique value (e.g. from the tempfile module) or a hash computed from the total filename.

Both of these solutions sound good to me!

[valeriupredoi]

*affecting mostly German facets 😁 🇩🇪