Increase default bin size for raster histograms

Fixes #35465 but it is still and arbitrary value, better approaches would require to calculate other dispersion indexes and they seem impractical (inefficient) in this case. See for example: https://en.wikipedia.org/wiki/Freedman%E2%80%93Diaconis_rule
qgis · Apr 11, 2020 · 98261bc · 98261bc
1 parent 7a9c3c6
commit 98261bc
Show file tree

Hide file tree

Showing 2 changed files with 7 additions and 4 deletions.
diff --git a/src/core/raster/qgscolorrampshader.cpp b/src/core/raster/qgscolorrampshader.cpp
@@ -208,7 +208,7 @@ void QgsColorRampShader::classifyColorRamp( const int classes, const int band, c
 
       double cut1 = std::numeric_limits<double>::quiet_NaN();
       double cut2 = std::numeric_limits<double>::quiet_NaN();
-      int sampleSize = 250000;
+      const int sampleSize = 250000;
 
       // set min and max from histogram, used later to calculate number of decimals to display
       input->cumulativeCut( band, 0.0, 1.0, min, max, extent, sampleSize );

diff --git a/src/core/raster/qgsrasterinterface.cpp b/src/core/raster/qgsrasterinterface.cpp
@@ -347,9 +347,12 @@ void QgsRasterInterface::initHistogram( QgsRasterHistogram &histogram,
     }
     else
     {
-      // There is no best default value, to display something reasonable in histogram chart, binCount should be small, OTOH, to get precise data for cumulative cut, the number should be big. Because it is easier to define fixed lower value for the chart, we calc optimum binCount for higher resolution (to avoid calculating that where histogram() is used. In any any case, it does not make sense to use more than width*height;
-      myBinCount = histogram.width * histogram.height;
-      if ( myBinCount > 1000 )  myBinCount = 1000;
+      // There is no best default value, to display something reasonable in histogram chart,
+      // binCount should be small, OTOH, to get precise data for cumulative cut, the number should be big.
+      // Because it is easier to define fixed lower value for the chart, we calc optimum binCount
+      // for higher resolution (to avoid calculating that where histogram() is used. In any any case,
+      // it does not make sense to use more than width*height;
+      myBinCount = std::min( 2000, histogram.width * histogram.height );
 
       // for Int16/Int32 make sure bin count <= actual range, because there is no sense in having
       // bins at fractional values