StatsGenerator computes NaN values for standard deviation in the neighbor distribution if bincount is large enough

[cartercocke]

The StatsGenerator filter computes NaN values for the standard deviation within the neighbor distribution when the number of bins is 14 or more (in the case of equiaxed grains) or 8 or more (in the case of rolled grains). The problem occurs in the following areas where the log of a negative number is evaluated giving a NaN value (or log(0) = -inf).

DREAM3D/Source/Plugins/SyntheticBuilding/SyntheticBuildingFilters/Presets/PrecipitateEquiaxedPreset.cpp

Line 157 in 2f6b5ae

mu = log(14.0 + (2.0 * float(i - middlebin)));

DREAM3D/Source/Plugins/SyntheticBuilding/SyntheticBuildingFilters/Presets/PrimaryRecrystallizedPreset.cpp

Line 155 in 2f6b5ae

mu = log(14.0 + (2.0 * float(i - middlebin)));

DREAM3D/Source/Plugins/SyntheticBuilding/SyntheticBuildingFilters/Presets/PrecipitateEquiaxedPreset.cpp

Line 157 in 2f6b5ae

mu = log(14.0 + (2.0 * float(i - middlebin)));

DREAM3D/Source/Plugins/SyntheticBuilding/SyntheticBuildingFilters/Presets/PrimaryRolledPreset.cpp

Line 154 in 2f6b5ae

mu = log(8.0 + (1.0 * float(i - middlebin)));

DREAM3D/Source/Plugins/SyntheticBuilding/SyntheticBuildingFilters/Presets/PrecipitateRolledPreset.cpp

Line 153 in 2f6b5ae

mu = log(8.0 + (1.0 * float(i - middlebin)));

I'm not sure how this carries forward beyond the Stats Generator filter as the Pack Primary Phases filter still works with these NaN values, but I figured it was worth noting as a potential issue. I believe these neighbor distribution statistics are used here:

DREAM3D/Source/Plugins/SyntheticBuilding/SyntheticBuildingFilters/PackPrimaryPhases.cpp

Line 1376 in 2f6b5ae

VectorOfFloatArray Neighdist = pp->getFeatureSize_Neighbors();