Downloading / updating G4 data about opengate

[nkrah]

In any case, one could consider deleting old G4data when downloading new data upon the import statement after an update.

[dsarrut]

yes indeed! The way the g4 data is managed is not optimal.
However, a warning: the check is performed during the import, so if the check takes time, it may delay the starting phase of all scripts (currently, the import phase is already "slow", mainly due to itk I think).
Please, go with the PR ;) !

[nkrah]

Agree, but don't think the check takes time. Downloading data is slow, but that only occurs once in a while after an update. I'll implement and check.

[nkrah]

Is there any reason not to get the data from the Geant4 installation which is required anyhow? Does it need to be downloaded separately? By looking up the Geant4 install, one would automatically align G4 and opengate in terms of G4 data.
Opinions?

[tbaudier]

Hi @nkrah

You're right when you compile opengate. But when you install opengate on your compute with pip install opengate , Geant4 is not installed on your computer. You need to download the data.

[nkrah]

I see. But don't you need Geant4 even if you install opengate via pip? Opengate uses Geant4 to run the simulation. Or does pip pull a pre-compiled Geant4?

[tbaudier]

Inside the wheel of opengate_core, you can find the compiled g4 librairies. But we cannot insert the data into the wheel because they are too big. This is why we prefer to download them at first import of opengate_core. It's not perfect because, the data are not removed correctly with pip uninstall opengate

[nkrah]

OK, now I got the full picture.

Is it possible to write a "hook" for the uninstall to remove the data? Possibly also when running an update via pip?

[tbaudier]

I have a look and it seems not, eg:
https://stackoverflow.com/questions/70321823/deleting-associated-data-from-a-pip-package-when-uninstalling

The data has to be removed manually

[nkrah]

I am finalizing a PR to solve this issue.

[tbaudier]

Done with #132

[nkrah]

Sorry, I re-open this because I created another PR with some improvements over the previous one. Thomas was too efficient ... ;-)