Fix svg2pdf filter

IPython/nbconvert/preprocessors/convertfigures.py

@@ -47,18 +47,12 b' class ConvertFiguresPreprocessor(Preprocessor):'
         """
 
         # Loop through all of the datatypes of the outputs in the cell.
-        for index, cell_out in enumerate(cell.get('outputs', [])):
-            for data_type, data in list(cell_out.items()):
-                # this must run *before* extract outputs,
-                # so figure_name and filename do not exist
-                self._convert_figure(cell_out, resources, data_type, data)
-        return cell, resources
+        for output in cell.get('outputs', []):
+            if output.output_type in {'execute_result', 'display_data'} \
+                    and self.from_format in output.data \
+                    and self.to_format not in output.data:
 
+                output.data[self.to_format] = self.convert_figure(
+                            self.from_format, output.data[self.from_format])
 
-    def _convert_figure(self, cell_out, resources, data_type, data):
-        """
-        Convert a figure and output the results to the cell output
-        """
-        if not self.to_format in cell_out and data_type == self.from_format:
-            data = self.convert_figure(data_type, data)
-            cell_out[self.to_format] = data
+        return cell, resources

IPython/nbconvert/preprocessors/svg2pdf.py

@@ -61,7 +61,7 b' class SVG2PDFPreprocessor(ConvertFiguresPreprocessor):'
         with TemporaryDirectory() as tmpdir:
             
             #Write fig to temp file
-            input_filename = os.path.join(tmpdir, 'figure.' + data_format)
+            input_filename = os.path.join(tmpdir, 'figure.svg')
             # SVG data is unicode text
             with io.open(input_filename, 'w', encoding='utf8') as f:
                 f.write(data)