Modified GafferImage.Reformat to add support for images with negative…

… or offset formats.

Fixed an issue where the output display window would occasionaly be wrong by a pixel.

include/GafferImage/Reformat.h

@@ -84,7 +84,7 @@ class Reformat : public ImageProcessor
  virtual IECore::ConstFloatVectorDataPtr computeChannelData( const std::string &channelName, const Imath::V2i &tileOrigin, const Gaffer::Context *context, const ImagePlug *parent ) const;
 
  // Computes the output scale factor from the input and output formats.
- Imath::V2f scale() const;
+ Imath::V2d scale() const;
 
  private :
 

python/GafferImageTest/ReformatTest.py

@@ -92,7 +92,36 @@ def testDataWindow( self ) :
  reformat["in"].setInput( read["out"] )
  reformatWindow = reformat["out"]["dataWindow"].getValue()
  self.assertEqual( reformatWindow, IECore.Box2i( IECore.V2i( 45, 37 ), IECore.V2i( 119, 99 ) ) )
+
+ def testNegativeDisplayWindowReformat( self ) :
+
+ read = GafferImage.ImageReader()
+ read["fileName"].setValue( os.path.join( self.path, "checkerWithNegWindows.200x150.exr" ) )
+
+ # Resize the image and check the size of the output data window.
+ reformat = GafferImage.Reformat()
+ f = GafferImage.Format( IECore.Box2i( IECore.V2i( -22, -13 ), IECore.V2i( 33, 58 ) ), 1. )
+ reformat["format"].setValue( f )
+ reformat["in"].setInput( read["out"] )
+ reformat["filter"].setValue( "Bilinear" )
+
+ reformatDataWindow = reformat["out"]["dataWindow"].getValue()
+ reformatFormat = reformat["out"]["format"].getValue()
+
+ # Check the data and display windows.
+ self.assertEqual( reformatDataWindow, IECore.Box2i( IECore.V2i( -66, -83 ), IECore.V2i( 365, 333 ) ) )
+ self.assertEqual( reformatFormat, f )
+
+ # Check the image data.
+ expectedFile = os.path.join( self.path, "checkerWithNegWindowsReformated.exr" )
+ expectedOutput = IECore.Reader.create( expectedFile ).read()
+ expectedOutput.blindData().clear()
+
+ reformatOutput = reformat["out"].image()
+ reformatOutput.blindData().clear()
 
+ self.assertEqual( reformatOutput, expectedOutput )
+
  # Test that when the input and output format are the same that the hash is passed through.
  def testHashPassThrough( self ) :
 

src/GafferImage/Reformat.cpp

@@ -161,12 +161,23 @@ void Reformat::hashChannelDataPlug( const GafferImage::ImagePlug *output, const
 Imath::Box2i Reformat::computeDataWindow( const Gaffer::Context *context, const ImagePlug *parent ) const
 {
  // Work out the scale factor of the output image and scale the input data window.
- Imath::V2f s( scale() );
+ Imath::V2d s( scale() );
  Imath::Box2i inDataWindow( inPlug()->dataWindowPlug()->getValue() );
+
+ Imath::V2d inFormatOffset( inPlug()->formatPlug()->getValue().getDisplayWindow().min );
+ Imath::V2d outFormatOffset( formatPlug()->getValue().getDisplayWindow().min );
+
  Imath::Box2i outDataWindow(
- Imath::V2i( int( floor( (inDataWindow.min.x)*s.x ) ), int( floor( (inDataWindow.min.y)*s.y ) ) ),
- Imath::V2i( int( ceil( (inDataWindow.max.x+1.)*s.x-1. ) ), int( ceil( (inDataWindow.max.y+1.)*s.y-1. ) ) )
+ Imath::V2i(
+ IECore::fastFloatFloor( double( inDataWindow.min.x - inFormatOffset.x ) * s.x + outFormatOffset.x ),
+ IECore::fastFloatFloor( double( inDataWindow.min.y - inFormatOffset.y ) * s.y + outFormatOffset.y )
+ ),
+ Imath::V2i(
+ IECore::fastFloatCeil( double( inDataWindow.max.x - inFormatOffset.x + 1. ) * s.x + outFormatOffset.x - 1. ),
+ IECore::fastFloatCeil( double( inDataWindow.max.y - inFormatOffset.y + 1. ) * s.y + outFormatOffset.y - 1. )
+ )
  );
+
  return outDataWindow;
 }
 
@@ -180,13 +191,13 @@ IECore::ConstStringVectorDataPtr Reformat::computeChannelNames( const Gaffer::Co
  return inPlug()->channelNamesPlug()->getValue();
 }
 
-Imath::V2f Reformat::scale() const
+Imath::V2d Reformat::scale() const
 {
  Format inFormat( inPlug()->formatPlug()->getValue() );
  Format outFormat( formatPlug()->getValue() );
- Imath::V2i inWH = Imath::V2i( inFormat.getDisplayWindow().max ) + Imath::V2i(1);
- Imath::V2i outWH = Imath::V2i( outFormat.getDisplayWindow().max ) + Imath::V2i(1);
- Imath::V2f scale( float( outWH.x ) / ( inWH.x ), float( outWH.y ) / inWH.y );
+ Imath::V2d inWH = Imath::V2d( inFormat.getDisplayWindow().size() ) + Imath::V2d(1.);
+ Imath::V2d outWH = Imath::V2d( outFormat.getDisplayWindow().size() ) + Imath::V2d(1.);
+ Imath::V2d scale( double( outWH.x ) / ( inWH.x ), double( outWH.y ) / inWH.y );
  return scale;
 }
 
@@ -204,19 +215,21 @@ IECore::ConstFloatVectorDataPtr Reformat::computeChannelData( const std::string
  out.resize( ImagePlug::tileSize() * ImagePlug::tileSize() );
 
  // Create some useful variables...
- Imath::Box2i tile( tileOrigin, Imath::V2i( tileOrigin.x + ImagePlug::tileSize() - 1, tileOrigin.y + ImagePlug::tileSize() - 1 ) );
+ Imath::V2i formatOffset( formatPlug()->getValue().getDisplayWindow().min ); 
+ Imath::Box2i tile( tileOrigin-formatOffset, Imath::V2i( tileOrigin.x - formatOffset.x + ImagePlug::tileSize() - 1, tileOrigin.y - formatOffset.y + ImagePlug::tileSize() - 1 ) );
  Imath::V2f scaleFactor( scale() );
 
  // Create our filter.
  FilterPtr f = Filter::create( filterPlug()->getValue(), 1.f / scaleFactor.y );
 
  // If we are filtering with a box filter then just don't bother filtering
- // at all and sample instead...
+ // at all and just integer sample instead...
  if ( static_cast<GafferImage::TypeId>( f->typeId() ) == GafferImage::BoxFilterTypeId )
  {
+ Imath::V2d scaleFactorD( scale() );
  Imath::Box2i sampleBox(
- Imath::V2i( IECore::fastFloatFloor( tile.min.x / scaleFactor.x ), IECore::fastFloatCeil( tile.min.y / scaleFactor.y ) ),
- Imath::V2i( IECore::fastFloatFloor( tile.max.x / scaleFactor.x ), IECore::fastFloatCeil( tile.max.y / scaleFactor.y ) )
+ Imath::V2i( IECore::fastFloatFloor( tile.min.x / scaleFactorD.x ), IECore::fastFloatCeil( tile.min.y / scaleFactorD.y ) ),
+ Imath::V2i( IECore::fastFloatFloor( tile.max.x / scaleFactorD.x ), IECore::fastFloatCeil( tile.max.y / scaleFactorD.y ) )
  );
  Sampler sampler( inPlug(), channelName, sampleBox, f, Sampler::Clamp );
  for ( int y = tile.min.y, ty = 0; y <= tile.max.y; ++y, ++ty )
@@ -246,15 +259,15 @@ IECore::ConstFloatVectorDataPtr Reformat::computeChannelData( const std::string
  Imath::V2i( sampleMaxX + fWidth, sampleMaxY + fHeight )
  );
 
- int sampleBoxWidth = sampleBox.max.x - sampleBox.min.x + 1;
- int sampleBoxHeight = sampleBox.max.y - sampleBox.min.y + 1;
+ int sampleBoxWidth = sampleBox.size().x + 1;
+ int sampleBoxHeight = sampleBox.size().y + 1;
 
  // Create a temporary buffer that we can write the result of the first pass to.
  // We extend the buffer vertically as we will need additional information in the
  // vertical squash (the second pass) to properly convolve the filter.
  float buffer[ ImagePlug::tileSize() * sampleBoxHeight ];
 
- // Create several buffers for each pixel in the output row (or coloum depending on the pass)
+ // Create several buffers for each pixel in the output row (or column depending on the pass)
  // into which we can place the indices for the pixels that are contribute to it's result and
  // their weight.