Review Sprint 33

[tarelli]

@mlolson

• [x] Fullscreen mode for the JS editor not working anymore like it used to, see screenshot
• [x] Top and bottom margins are wrong (Chrome)
• [ ] Start button pixel imperfection
• [x] Is anyone seeing these?
• [x] Simulation Spinner won't go away sometimes after simulation is finished loading. Error shown : Invariant Violation : getDomNode() : A component must be mounted to have a DOM Node.
• [x] Sometimes the custom tab in the load simulation dialog shows all of its content in one line

@jrmartin

• [x] Factor out the MetaType strings into a resource file that can be accessed by any widget (to do by@ jrmartin)
• [x] Disable geppetto and show info if webgl is not enabled
• [x] Collada file is not working anymore https://raw.githubusercontent.com/openworm/org.geppetto.samples/development/Collada/GEPPETTO.xml
• [x] getId and getName are not showing as suggestions in the console
• [x] ModelTree has getId undefined, add QUnit test to check all nodes have id, instancepath and name set in a couple of comprehensive examples (hhcell and purkinje after getModelTree should suffice for now).
• [x] Value should not include the unit purkinje.electrical.ModelTree.biophys.ModelViewParmSubset_30.getValue() => "6.8876 uF_per_cm2"
• [x] Error when invoking print on ModelViewParamSubset_30 purkinje.electrical.ModelTree.biophys.ModelViewParmSubset_30.print() => "TypeError: undefined is not a function" @jrmartin let me know if the problem with the last two belongs to the NeuroML population rather than the framework in that case @adrianq and I will take care of them
• [x] Scatter 3D not working https://raw.githubusercontent.com/openworm/org.geppetto.samples/development/LEMS/HindmarshRose/GEPPETTO.xml
• [x] Stopping and restarting an SPH simulation doesn't work
• [x] Fix issue with sending requests to server everytime user types in url.

[tarelli]

• [ ] This happened twice, @mlolson any idea?

[mlolson]

@tarelli When does it happen?

[tarelli]

@mlolson haven't yet identified "when" but it's the second time I open the console and it's there

[mlolson]

Ok I'm not seeing that but I'm pulling everything down and rebuilding the project

[mlolson]

I'll work on all of these except the first.

[jrmartin]

@mlolson The Webgl failing is already fixed too. I will do the first one.

[jrmartin]

I added another issue to the card, sometimes I noticed that the Loading Spinner won't go away after simulation is loaded.

[tarelli]

@jrmartin when that happens (spinner doesn't go away) do you have the Invariant violation error in the console (see last item)?

[jrmartin]

Yes, that error is displayed in the browser debug console.

[mlolson]

@jrmartin When I try to load the Scatter 3D simulation I get the "Cannot read property '0' of undefined " error. Are you familiar with this error?

[jrmartin]

@tarelli The c302 simulation has the variables named like this c302.electrical.SimulationTree.ADAR.0, using ADAR[0] would make more sense, but I didn't create that script so I'm not sure which one is the correct way

[jrmartin]

@mlolson It's a problem with the variable not existing, you can transfer that issue to me since the comment above relating C302 simulation is related to this one.

[mlolson]

Ok cool, this looked like something you were familiar with

[jrmartin]

@tarelli For the issue of value including unit, it's because Purkinje model tree has dummy data. The model tree is populated with random fields from the .nml, and value happened to be the field "erev" which in model tree is shown as erev="137.5262 mV". Issue should not be issue once model tree is populated accurately.

Print method is missing for model tree nodes, this is client side.

[tarelli]

@jrmartin about the .0 vs [0] you are right it comes from LEMS internal representation (@adrianq see the LEMS file and you will find "ADAL/0/generic_iaf_cell/v" which translates to ADAL.0.generic_iaf_cell.v), in that case the question becomes where does the [0] come from?

[tarelli]

@mlolson this is still happening in my dev env after updating

[mlolson]

@tarelli Yeah this is some weird chrome rendering issue I think. I thought it was due to a 1px solid transparent border on disabled buttons, but that seems to not be the case. Will investigate further.

[jrmartin]

@tarelli What branch did you get the SPH not restartign problem? It's working fine in dev.

[tarelli]

@mlolson when you have a chance can you check if in the dev branch you can stop and restart the SPH simulation (remember to use the sample from the dev repository), in my local dev environment when I restart it after stopping it the simulation doesn't restart. It works fine in @jrmartin's.

[mlolson]

I'm seeing the same issue

[pgleeson]

See mail on main list re having more time for NeuroML/OpenWorm issues next week... and this should go for this too... But for now...

It's important to note that variables in different LEMS models will be accessed/named in different ways, depending on whether the model contains:

1) A top level single Component being run directly (e.g. https://github.com/NeuroML/NML2_LEMS_Examples/blob/master/Figure8_SBML_LEMS.xml) 2) A simple component in a network/population (e.g. https://github.com/NeuroML/NeuroML2/blob/development/LEMSexamples/LEMS_NML2_Ex9_FN.xml) 3) An abstract (no morphology) cell with membrane potential in a network: https://github.com/NeuroML/NeuroML2/blob/development/LEMSexamples/LEMS_NML2_Ex8_AdEx.xml 4) A single segment cell with v (or other variables like bioPhys1/membraneProperties/kChans/kChan/n/q) in a population: https://github.com/NeuroML/NeuroML2/blob/development/LEMSexamples/LEMS_NML2_Ex5_DetCell.xml 5) A membrane potential on a named segment in a multicompartmental cell: path will be something like: CellGroup_1/0/L5PC/123/v. This can't be used by jLEMS (no multicompartmental cell support) but will be needed when exported to NEURON.

These various strings are constrained in part by what jLEMS needs to be given, rather than a sensible way to represent paths to variables, and may change in future. To help this, I've got a class here: https://github.com/NeuroML/org.neuroml.export/blob/development/src/main/java/org/neuroml/export/LEMSQuantityPath.java which reads all the different forms in and gives a class to query what it means. As the string changes this class will remain and be updated to read the quantity from LEMS. Hope this helps.

[jrmartin]

@tarelli @pgleeson I spent some time trying to make the console accept any paths for the variables, but due to Javascript syntax restriction variables can't be named as numbers (just like java). So c302.electrical.SimulationTree.ADAR.0 is not allowed, not even by the browse's javascript console. It needs to start with a letter, $ or underscore. Maybe if the variable path contains a number we can add an underscore prior to it for Javascript to allow, so "c302.electrical.SimulationTree.ADAR.0" would become "c302.electrical.SimulationTree.ADAR._0" .

[tarelli]

@jrmartin ok, I started working on the neuromlpopulation branch where among other things I'm also following through the design of extracting entities from the NeuroML file so this problem will be also taken care of. The new path will be c302.ADAR.electrical.SimulationTree etc, the magic will happen in the ModelInterpreter/Simulator. @adrianq these are good branches also to start extracting more stuff from the NeuroML files, to work with it the branch exists for core/simulation/model.neuroml.

[jrmartin]

As server side performance goes, it's all good now that duplicate readModel() is gone. There's still some lagging client side, but only when loading simulation through LoadingModal, doesn't happen if loading simulation via command prompt or URL. @mlolson Do you know if react components take a while to disappear and freeze the client during that time? It's my first thought as to why this might be happening, that the loading modal takes a couple seconds to disappear after clicking "Load" and the socket can't send the command to load the simulation until this happens.

[mlolson]

Hmm, react components should be very quick to render, but there might be something else going on. I'm seeing some very slow load times on live.geppetto.org, a bunch of server requests are taking >4 seconds

[tarelli]

@jrmartin I just noticed that line 170 of GeppettoMessageInbound is hit every time the user edits the Simulation URL textbox, as you type. It should only be called when switching to the custom tab I think.

[jrmartin]

@mlolson Thanks for the clarification on React components, those slow load time on live should be gone on dev now. @tarelli That might be crowding socket, saw them but they seemed to go instantaneously, maybe that's the issue. Thanks for tip

[jrmartin]

@tarelli Well that was it, should follow occam's razor when debugging more often instead of thinking is something more complicated. Thanks.

[tarelli]

@jrmartin where did the fix for this go?

[jrmartin]

@tarelli Development branch

[tarelli]

@jrmartin of which bundle? I merged frontend dev into adrianq-neuromlwidget but it was still happening

[jrmartin]

@tarelli Frontned, this it the commit that takes care of it https://github.com/openworm/org.geppetto.frontend/commit/abfecad0b7739bc6940a5992c1b00896204d2218

[jrmartin]

@tarelli I'm working off the adrianq-neuromlwidget, seems the changes weren't merged properly

[jrmartin]

@tarelli I merged the adrianq-neurml and development branches into a branch called devneuromlmerged which is committed. There were too many manual changes I had to make, didn't want to risk merging back to adrianq-neuroml branch without knowing that branch full functionality.

[adrianq]

If Scatter3d is working fine maybe we want to add the simulation back to the Load Simulation scroll down menu...

[tarelli]

Yup, I'm doing that