paraview still seems to be needed via matplotlib

[doutriaux1]

Today we started uvcdat before Paraview was built. UV-CDAT didn't start becasue matplotlib needs vtk vistrail module (why?) GUI should be able to start w/o vtk

[benbu]

Is there a stack trace or other error message?

I don't see any mention of vtk in the matplotlib package, don't know how it's getting referenced.

[remram44]

UV-CDAT still exits abrubtly with exit code 255 if the vtk VisTrails module is absent.

[remram44]

The error message doesn't show up because of the way UV-CDAT redirects its output streams; in uvcdat.py, both the print "Uncaught exception" and the following print_exc() don't seem to end up anywhere.

The traceback is:

  File "/Users/remirampin/Documents/workspace/install/vistrails/vistrails/uvcdat.py", line 83, in <module>
    v = gui.application.start_application()
  File "/Users/remirampin/Documents/workspace/install/vistrails/vistrails/gui/application.py", line 680, in start_application
    x = VistrailsApplication.init(optionsDict)
  File "/Users/remirampin/Documents/workspace/install/vistrails/vistrails/gui/application.py", line 160, in init
    self.createWindows()
  File "/Users/remirampin/Documents/workspace/install/vistrails/vistrails/gui/application.py", line 533, in createWindows
    self.uvcdatWindow = UVCDATMainWindow()
  File "/Users/remirampin/Documents/workspace/install/vistrails/vistrails/gui/uvcdat/mainwindow.py", line 72, in __init__
    self.createDockWindows()
  File "/Users/remirampin/Documents/workspace/install/vistrails/vistrails/gui/uvcdat/mainwindow.py", line 138, in createDockWindows
    self.varProp = VariableProperties(self)
  File "/Users/remirampin/Documents/workspace/install/vistrails/vistrails/gui/uvcdat/variable.py", line 116, in __init__
    self.createOpenDAPTab()
  File "/Users/remirampin/Documents/workspace/install/vistrails/vistrails/gui/uvcdat/variable.py", line 293, in createOpenDAPTab
    from packages.vtDV3D.RemoteDataBrowser import RemoteDataBrowser
  File "/Users/remirampin/Documents/workspace/install/vistrails/vistrails/packages/vtDV3D/__init__.py", line 76, in <module>
    from packages.vtDV3D.DV3DCell import *
  File "/Users/remirampin/Documents/workspace/install/vistrails/vistrails/packages/vtDV3D/DV3DCell.py", line 11, in <module>
    from packages.vtk.vtkcell import QVTKWidget, QVTKWidgetToolBar
ImportError: No module named vtk.vtkcell

This shows that the UV-CDAT UI tends to import stuff from packages...

Apart from spreadsheet, uvcdat and uvcdat_cdms:

gui.uvcdat imports vtDV3D
gui.application imports ParaView and VisIt
vistrails_window imports vtDV3D (with try-except)

packages.ParaView imports QVTKWidget from vtk (but does not declare a dependency)
pvclimate imports QVTKWidget from vtk (but declares the dependency incorrectly)
scikit_learn imports vis_analytics (declares the dependency)
uvcdat_cdms imports memoryLogger from vtDV3D (but does not declare a dependency)
vtDV3D imports vtk (declares a dependency)
vtDV3D imports pylab (declares a dependency)

This might be showing an issue with how logging is done as well.

[remram44]

I've started working on fixing these in the branch uvcdat-165-fix-package-dependencies

[remram44]

This is what I have:

uvcdat_cdms imports memoryLogger from vtDV3D.vtUtilities; should this really be inside the vtDV3D package?

[ThomasMaxwell]

The memoryLogger refs are diagnostic code that should probably be removed when Ben and I complete our investigations of uvcdat memory leaks. Also, DV3D should not require matplotlib (the dependency is supposed to be optional).

From: Remi Rampin <notifications@github.commailto:notifications@github.com>
Reply-To: UV-CDAT/uvcdat <reply@reply.github.commailto:reply@reply.github.com>
Date: Thursday, July 11, 2013 4:51 PM
To: UV-CDAT/uvcdat <uvcdat@noreply.github.commailto:uvcdat@noreply.github.com>
Subject: Re: [uvcdat] paraview still seems to be needed via matplotlib (#165)

This is what I have:
[https://github-camo.global.ssl.fastly.net/2f61e7456d89039771f2bc97c591fcd95fdcd31c/687474703a2f2f692e696d6775722e636f6d2f5a4d466633754e2e706e67]https://github-camo.global.ssl.fastly.net/2f61e7456d89039771f2bc97c591fcd95fdcd31c/687474703a2f2f692e696d6775722e636f6d2f5a4d466633754e2e706e67

uvcdat_cdms imports memoryLogger from vtDV3D.vtUtilities; should this really be inside the vtDV3D package?

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/165#issuecomment-20841285.

[remram44]

Also, DV3D should not require matplotlib (the dependency is supposed to be optional).

vtDV3D.SlicePlotModule imports MplFigureCellWidget and MplFigureManager.

The import from init to SlicePlotModule was happening in yet another catch-all block, I replaced this with a test on has_package().

[aashish24]

Please update this branch to current uvcdat-master as it was never tested and got aged.

[remram44]

Hmm, I cannot merge with uvcdat-master as it is not related with the uvcdat-next branch I based this on.

I'm thus rebasing this with:

git checkout -b uvcdat-165-fix-package-dependencies_master
git rebase --onto uvcdat-master uvcdat-next

[remram44]

The problems I see are:

• pvclimate -> spreadsheet: dependency is conditional but pvclimate_pipeline_helper imports it
• VisIt -> spreadsheet: dependency is conditional but visit_pipeline_helper imports it
• vtk -> spreadsheet: dependency is conditional but vtkcell imports it

blue: temporary memoryLogger deps
grey: conditional deps (safe)
red: problem

[doutriaux1]

ok merged this branch into master can we close this, or is there still issues?

[remram44]

There is nothing that should block a release.

I'm not sure about the spreadsheet dependencies in red, I think they should either be removed or appear as a package dependency if they are real. So somebody might want to check that.

[doutriaux1]

ok closing for now.

[aashish24]

@doutriaux1 Lets keep it open until we resolve this.