GitLab

Change-Id: I2acdf1fc9dba68668650269fed5c5032ce1a967b

Fixed PrintSelf

Change-Id: I92824f20a80ef863b00a4c73c170609c3f0f0aec

Change-Id: Ib17e807264726ecccd9be5063787f5deb061ae5f

Change-Id: I2819aeab235183f7be1f8076c77c0f3097f91caa

Change-Id: I9bed54cd4fc422292aef5c2507911dfe74bf9ca1

Change-Id: I32515fdb0f70d5cb98f0330530e22ce7d60e9b5a

Change-Id: I4511b3a558b72500ec8b96be65bc6aad63ed892f

Change-Id: Iefada9a77f7f58030cf6e914a9694612508e4e6f

David Gobbi authored May 24, 2011 and Kitware Robot committed May 24, 2011

853599d2 BUG: fix memory leak in last commit.

Change-Id: I1e7ca79290e285d07776ec82c3d7e1325ed99254

David Gobbi authored May 24, 2011 and Kitware Robot committed May 24, 2011

37c59774 ENH: Customized __dir__ for PyVTKClass to enhance help output.

Change-Id: Id8d93b2ddce77b9de576287f8b14cd3ba80126e4

David Gobbi authored May 24, 2011 and Kitware Robot committed May 24, 2011

3c87e3a7 BUG: Fix typo in last commit.

Change-Id: Ic51ad3d3ff3495afe850c5fa9368a28313eb6974

David Gobbi authored May 24, 2011 and Kitware Robot committed May 24, 2011

1948dda0 ENH: Expand the types for vtkArray::CreateArray.

This adds "signed char", "long long", "float" and "vtkUnicodeString".

Change-Id: Ibb5b7a9203a895ebb91af25799979e3dda8df6c4

Andy Wilson authored May 23, 2011 and Kitware Robot committed May 23, 2011

e0c61277 Replace the vtkMPIController::Initialized static variable

Wow, am I ever embarrassed.  I've been compiling and testing all the
changes I've made today to vtkMPIController using a build tree
that had VTK_USE_PARALLEL on but VTK_USE_MPI off.  So... yeah.  None
of this code was actually being compiled.  No wonder I didn't see the
errors.

Anyway.

There are two different stages in initializing MPI for use in a VTK
application.  The first involves a call to MPI_Init().  Bad things
are known to happen if you call that twice.  I'm guarding it by checking
MPI_Initialized(), wrapped in the static method
vtkMPIController::IsMPIInitialized().  The second stage is to set up
all of the global vtkMPICommunicator variables.  *That* is guarded
by the static variable vtkMPIController::Initialized.  That's
what I've restored in this latest batch of fixes.

An analogous process goes on at tear-down time using IsMPIFinalized()
and checks to vtkMPIController::Initialized again.

It compiles.  It passes Ctest.  I can only hope that this really does
fix the problem.

Andy Wilson authored May 23, 2011 and Kitware Robot committed May 23, 2011

beb7aea0 Restore old API; also guard multiple MPI_Finalize calls

It turns out that a few people actually used the three-argument
Initialize() signature.  I restored it and marked it VTK_LEGACY
to try to nudge them toward the newer, saner version.

At the same time, I modified Finalize() to call MPI_Finalized()
instead of trying to keep track of whether VTK had already been
there.  I left the old API in place, wrapped again in VTK_LEGACY
for the time being.

Marcus D. Hanwell authored May 23, 2011 and Kitware Robot committed May 23, 2011

e1c416c0 ENH: Change seriesIndex to vtkIdType.

This now matches the (recently changed) return type for locating the
point in the plot.

Change-Id: I0d80a37379e6ded62436b051016f5e56f359c6b4

Jeff Baumes authored May 23, 2011 and Kitware Robot committed May 23, 2011

dc0e2c22 Removing unused source files

The class has been moved to Common.

Change-Id: I39336aac49535792d05f2633dedfb4c2f4bfe4a4

Andy Wilson authored May 23, 2011 and Kitware Robot committed May 23, 2011

9c849584 Ask MPI whether it's been initialized rather than trying to track it ourselves

Calling MPI_Init multiple times in an application is an error.  This comes up
a lot when trying to run Python scripts that use VTK MPI components.  A while
back I wrote code to try to trap this by tracking (within VTK) whether we had
already initialized MPI.  This turns out to be brittle.  I'm backing that code
out and replacing it with a call to MPI_Initialized() -- we'll just ask MPI
itself whether it's up and running.

David Gobbi authored May 20, 2011 and Kitware Robot committed May 20, 2011

d04c169f BUG: Small issues with python arg conversion.

Fix automatic arg conversion so that it only occurs with
pass-by-reference if the ref is a const ref.  Also check
base classes of special types when doing conversions.

Change-Id: I885424f7f57c890cc316f002862ed8441a7b03eb

Philippe Pebay authored May 20, 2011 and Kitware Robot committed May 20, 2011

8c4af461 Regressions (and corresponding assessments) will now always be calculated.
48c66859 Shortened variable names for legibility (results are difficult to read)
f2656722 Make sure that Assess is always calculated, even when gen. variance < 0
a6e7c461 Added calculations of residuals from linear regressions (both)
3771fadd Don't issue a Warning in Derive for non-positive generalized variance.
e740b71c Make Assessor be explicit about involved columns when determinant <= 0
4b24116d Clarifying code a bit (assessment is way too complicated)

It is just that NaNs will be issued, should the covariance matrix
not be positive definite.

This is because we now also provide the linear residuals in X and Y,
which will always work given a regression line

Just stick 0s in and store the value of the determinant to be sure.