Deterministic loading of modules: is it conceivable?

[chir-set]

Deterministic loading of modules: is it conceivable?

[jamesobutler]

It appears that the CrossSectionAnalysis module defines no other modules as a dependency (https://github.com/vmtk/SlicerExtension-VMTK/blob/1cfe4da260bab49a22ee0a740b87805f348230d8/CrossSectionAnalysis/CrossSectionAnalysis.py#L24) where listed dependent modules are loaded first prior to loading the specified module.

Although SlicerVMTK defines a dependency on ExtraMarkups in its s4ext file, that is only so that ExtraMarkups is automatically installed when installing SlicerVMTK from the extensions manager. It doesn’t specify the relation of modules in terms of loading dependency order.

[chir-set]

It doesn’t specify the relation of modules in terms of loading dependency order.

I added 'ExtraMarkups' as a dependency in the relevant modules and things got worse, additional messages are in stdout:

When loading module "CrossSectionAnalysis" , the dependency "ExtraMarkups" failed to be loaded.
void ctkComboBox::setCurrentUserDataAsString(QString) : No item found with user data string "2"
When loading module "GuidedArterySegmentation" , the dependency "ExtraMarkups" failed to be loaded.
When loading module "StenosisMeasurement3D" , the dependency "ExtraMarkups" failed to be loaded.

All modules work as expected once Slicer is up.

Thank you for your reply @jamesobutler .

[jamesobutler]

From my understanding “ExtraMarkups” is the extension name and not the module names. Where the CrossSectionAnalysis module may have a dependency on the “Shape” or “Label modules. In the module, the dependencies is other modules and not extensions.

[chir-set]

the extension name and not the module names

You are right about this. However, using 'Shape' as module name does not cancel the original annoying stdout messages.

Failed to load vtkSlicerStenosisMeasurement3DModuleLogicPython: No module named vtkSlicerShapeModuleMRMLPython
Failed to load vtkSlicerCrossSectionAnalysisModuleLogicPython: No module named vtkSlicerShapeModuleMRMLPython
Error(s):
CLI executable: /home/user/programs/Slicer/slicer.org/Extensions-32703/SlicerVMTK/lib/Slicer-5.7/qt-loadable-modules/vtkvmtk.py
The process failed to start. Either the invoked program is missing, or you may have insufficient permissions to invoke the program.
Fail to instantiate module "vtkvmtk"
The following modules failed to be instantiated:
vtkvmtk
void ctkComboBox::setCurrentUserDataAsString(QString) : No item found with user data string "2"

I think that the solution would be to load anything close to core, like additional markups, before loading usual extensions. This would require core developments, time, funding... ...

Thank you for your time.

[lassoan]

Failed to load vtkSlicerStenosisMeasurement3DModuleLogicPython: No module named vtkSlicerShapeModuleMRMLPython
Failed to load vtkSlicerCrossSectionAnalysisModuleLogicPython: No module named vtkSlicerShapeModuleMRMLPython

These are due to #7579

CLI executable: /home/user/programs/Slicer/slicer.org/Extensions-32703/SlicerVMTK/lib/Slicer-5.7/qt-loadable-modules/vtkvmtk.py
The process failed to start. Either the invoked program is missing, or you may have insufficient permissions to invoke the program.
Fail to instantiate module "vtkvmtk"
The following modules failed to be instantiated:
vtkvmtk

This harmless error message is well known and it is safe to ignore. We can address it as VTK remote module Python wrapping and packaging is improved.

void ctkComboBox::setCurrentUserDataAsString(QString) : No item found with user data string "2"

You can add a breakpoint into this method to see what is the root cause.

[chir-set]

Thank you @lassoan for addressing this with a PR.

The vtkvmtk message is indeed old and harmless.

The other one (::setCurrentUserDataAsString) appeared during tinkering, and does not appear otherwise. I'll get deeper into it if necessary.

[chir-set]

Since 9c87503, the messages in stdout related to vtkSlicerShapeModuleMRMLPython are indeed gone.

('Fail to instantiate module "vtkvmtk"' still persists and was not the target of the patch.)

Thank you all.

@lassoan @jcfr @jamesobutler

[lassoan]

Great, thank you for testing.