RakipInitiative / ModelRepository

Joint project of EFSA, Federal Institute For Risk Assessment, DTU and ANSES to create a online model repository.
GNU General Public License v3.0
2 stars 0 forks source link

Reader issue: model archive from URL is deleted too soon #276

Closed schuelet closed 3 years ago

schuelet commented 3 years ago

If a model is loaded by the Reader using a URL, a temporary file is created storing the archive to create a portObject. Then the archive is deleted. But the Runner requires the archive to execute the model.

I believe, the archive should be stored for longer, and only be deleted on reset.

@ahmadswaid since I think you implemented this download from URL system, I am assigning you to it. You can assign it to me if you don't want to do it, though.

https://app.zenhub.com/files/241333403/d87c0ecc-9bd9-47af-82e2-a384954e3b66/download

schuelet commented 3 years ago

fixed in https://app.zenhub.com/workspace/o/silebat/fsk-lab/issues/804.

Testing required, if issues occur when a workflow is shared across different operating systems.

schuelet commented 3 years ago

@mfilter I was thinking about sharing workflows with this and I don't see an easy way to avoid re- executing the reader. Here is the rundown:

I was experimenting with the knime:// Protocoll in the hope that might be an easy solution. I think, if the user uses the Protokoll (or the download from URL option), then we have the guaranteed model path and could avoid the forced re-execution. When I have time, I will implement it.

mfilter commented 3 years ago

OK, I change the prio to medium. My assumption is, that the URL-based models are downloaded into a folder inside the WF that has the Reader is in, correct?

schuelet commented 3 years ago

yes, the user has the files in the workflow folder and can access them if they don't want to re-download them.

schuelet commented 3 years ago

@ahmadswaid maybe we can rename the downloaded model files at some point so they can be identified by the user.

ahmadswaid commented 3 years ago

@schuelet,
yes, one Idea come to my mind is using the enveiroment manager to resolve the knime:// into a real path just in time. then we don't need to re-run the reader node