search

Open-Systems-Pharmacology / QualificationRunner

Qualification runner in charge of managing a qualification workflow
Other
0 stars 1 forks source link

Check existance of referenced objects (e.g. Data) before simulation #76

Open PavelBal opened 4 years ago

PavelBal commented 4 years ago

Currently, during the qualification run, all snapshots are re-created (including running the simulations) before checking if the objects referenced in the Qualification Plan are present in the objects. This may lead to situations where the qualification run is running for 4 hours and then some plots cannot be created because the referenced data object is not in the snapshots (wrong data or just a typo).

Checking of consistency can be performed before running the qualification plan. As an alternative, a separate function for checking the qualification plan could be implemented.

msevestre commented 4 years ago

Checking of consistency can be performed before running the qualification plan.

How do you do this if you need to recreate the projects to check if objects are available? Each instance is running via a CLI. There is no callback mechanism.

As an alternative, a separate function for checking the qualification plan could be implemented.

What would this function check?

I think there is definitely room for improvement but this is very not easy

PavelBal commented 4 years ago

How do you do this if you need to recreate the projects to check if objects are available? Each instance is running via a CLI. There is no callback mechanism.

Can at least check the snapshot-*.json if the repsective entries are present... All referenced simulations, observed data, even building blocks are listed in the snapshot.

What would this function check?

Whether the entries (i.e. their name) are in the snapshot(s) :)

msevestre commented 4 years ago

The qualification runner does not know anythign about the snapshot. This is all in PKSim Maybe there is a way to say OK, from project XYZ I am using {Individual: "Ind1} and {Compound: "Midazolam}. Check that they are here . We would lokad the snapshot twice but this should not be too bad Now if you are dealing with parameters etc.. suddenly we need to load the PKSIm model twice which can take a lot of time. Maybe good to have.

Again, I think it's a good idea, but not easy

PavelBal commented 4 years ago

OK, from project XYZ I am using {Individual: "Ind1} and {Compound: "Midazolam}. Check that they are here .

This would be a small step but would dramatically help in some cases. I just waited 4 hours for the run to finish just to see that the data set I referenced is not in the project (was just a typo during copy-pasting)...