Allow '_ensure_save_result' to be skipped with a flag in any place it is called.

[GeraldIr]

Often I find myself being unable to start jobs that have different result nodes which are incompatible with save_result and there has to be some workaround implemented for handling these edge cases.

I'd like to propose an optional flag for methods such as create_job and download which allows ensure save result to be skipped so that no superfluous save_result node gets added.

I can also create a PR for this if a change like this is approved.

[jdries]

Do you have an example of these different result nodes? Maybe the check can also be improved? We are for instance thinking about a save_collection process, that could also have this problem.

…

________________________________ From: Gerald Walter Irsiegler ***@***.***> Sent: Monday, December 4, 2023 11:44 AM To: Open-EO/openeo-python-client ***@***.***> Cc: Subscribed ***@***.***> Subject: [Open-EO/openeo-python-client] Allow '_ensure_save_result' to be skipped with a flag in any place it is called. (Issue #513) Often I find myself being unable to start jobs that have different result nodes which are incompatible with save_result and there has to be some workaround implemented for handling these edge cases. I'd like to propose an optional flag for methods such as create_job and download which allows ensure save result to be skipped so that no superfluous save_result node gets added. I can also create a PR for this if a change like this is approved. — Reply to this email directly, view it on GitHub<#513>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ABNJPSH5Q75CQFPLCIQLMK3YHWSQTAVCNFSM6AAAAABAFYEIOOVHI2DSMVQWIX3LMV43ASLTON2WKOZSGAZDGNJVGAYDCMI>. You are receiving this because you are subscribed to this thread.Message ID: ***@***.***> VITO Disclaimer: http://www.vito.be/e-maildisclaimer

[GeraldIr]

We are currently using save_ml_model which stores results of random forest regression, we debated including this functionality in save_result but decided the outputs were different enough to justify its own process (and this might in the future be the case for more things, like UDFs).

I think for more advanced users having the option would be a benefit, and making it optional would save the users relying on it currently from any trouble.

[soxofaan]

is save_ml_model the only use case you're talking about? Because we can easily whitelist that process alongside save_result in the logic.

[soxofaan]

ah no wait, save_ml_model is already whitelisted in MlModel.create_job() so I'm not sure I fully understand the problem here.

can you provide some examples in code snippets?