search

ModellingWebLab / weblab-fc

Functional Curation backend for the Modelling Web Lab
Other
1 stars 0 forks source link

Model-protocol combination testing doesn't pick up all required variables #233

Open MichaelClerx opened 3 years ago

MichaelClerx commented 3 years ago

@jonc125 wrote:

I think the front-end and back-end are now using a subtly different algorithm for checking compatibility, in that the front-end uses the new weblab-fc stuff and the back-end is still pycml. So it may be the former isn’t complaining correctly when original_definition is involved?

Rest of conversation:

gary 12:03 Some WL problem is trying to run simulations with missing metadata instead of marking them as inapplicable. Anyone know what might have changed that?!

gary Today at 12:04 e.g. I just re-ran a grey square for Ten Tusscher 2004 epi with same versions of model and protocol with the recycle button thing, and it turned from grey to red

gary 12:08 On a not un-related note, PyCML seems to have nicer checking for duplicate cmetaid or duplicate metadata than cellmlmanip - this version of a CellML file (correctly!) failed on PyCML/WebLab https://scrambler.cs.ox.ac.uk/experiments/27949/versions/28484/stdout.txt/displayContent but Chaste didn't complain when I tried to use it @maurice might be nice to add tests and throw on this error

michael 1 hour ago No idea! Nothing has changed in that code as far as I know, so maybe a front-end thing?

jonathan 42 minutes ago I’ve not deployed anything new since last year…

jonathan 42 minutes ago Does it happen on an initial run, or only an explicitly requested re-run?

gary 30 minutes ago It happened on an initial run when I uploaded a new version of Tomek model and ran new experiment for that model version, and also on a re-run of a previously-grey model/protocol combo.

gary 30 minutes ago It's conceivable we haven't tried to run a grey one for a while....!

gary 28 minutes ago Not sure whether you've permissions to see this page Jon: https://scrambler.cs.ox.ac.uk/experiments/models/227/versions/*/protocols/55/versions/*

gary 27 minutes ago But it should show some previous grey runs for the model that highlighted this, although when I click on some of those I'm getting a 'Not found' error(!) https://scrambler.cs.ox.ac.uk/experiments/3910/versions/4042

gary 26 minutes ago I can say it was definitely working in May 2020 from a model I did upload at that time!

jonathan 1 minute ago I think the front-end and back-end are now using a subtly different algorithm for checking compatibility, in that the front-end uses the new weblab-fc stuff and the back-end is still pycml. So it may be the former isn’t complaining correctly when original_definition is involved? (edited)

jonathan < 1 minute ago I’m getting the ‘not found’ too, I think because I can’t see some versions of the model (it’s owned by you I believe)

jonathan 2 minutes ago I can see the grey experiments where I can see the model version (e.g. https://scrambler.cs.ox.ac.uk/experiments/18537/versions/18909). So it’s a bug really in the matrix view letting me see the existence of versions I can’t access

MichaelClerx commented 3 years ago

The matrix part is a separate issue: https://github.com/ModellingWebLab/WebLab/issues/333

mirams commented 3 years ago

Huh - whatever it is is more complicated than all the grey squares failing, I just sucessfully got this: https://scrambler.cs.ox.ac.uk/experiments/28013/versions/28551

mirams commented 3 years ago

and this is one that should be grey but is coming up red when trying to run a simulation which uses metadata it doesn't have: https://scrambler.cs.ox.ac.uk/experiments/26583/versions/28615/errors.txt/displayContent

MichaelClerx commented 3 years ago

Other block experiments, e.g. ical block and ik1 block do the same