Add fallbacks to image conversion

camelot/backends/image_conversion.py

@@ -1,15 +1,41 @@
 # -*- coding: utf-8 -*-
 
+import logging
+
 from .poppler_backend import PopplerBackend
 from .ghostscript_backend import GhostscriptBackend
 
+logger = logging.getLogger("camelot")
 backends = {"poppler": PopplerBackend, "ghostscript": GhostscriptBackend}
 
 
 class ImageConversionBackend(object):
-    def __init__(self, backend="poppler"):
+    def __init__(self, backend="poppler", use_fallback=True):
+        if backend not in backends.keys():
+            raise ValueError(f"Image conversion backend '{backend}' not supported")
+
         self.backend = backend
+        self.use_fallback = use_fallback
+        self.fallbacks = list(filter(lambda x: x != backend, backends.keys()))
 
     def convert(self, pdf_path, png_path):
+        try:
             converter = backends[self.backend]()
             converter.convert(pdf_path, png_path)
+        except Exception as e:
+            logger.info(f"Image conversion backend '{self.backend}' failed with {str(e)}")
+
+            if self.use_fallback:
+                for fallback in self.fallbacks:
+                    logger.info(f"Falling back on '{fallback}'")
+
+                    try:
+                        converter = backends[self.backend]()
+                        converter.convert(pdf_path, png_path)
+                    except Exception as e:
+                        logger.info(f"Image conversion backend '{fallback}' failed with {str(e)}")
+
+                        continue
+                    else:
+                        logger.info(f"Image conversion backend '{fallback}' succeeded")
+                        break

camelot/parsers/lattice.py

@@ -129,6 +129,7 @@ class Lattice(BaseParser):
         self.threshold_constant = threshold_constant
         self.iterations = iterations
         self.resolution = resolution
+        self.backend = ImageConversionBackend()
 
     @staticmethod
     def _reduce_index(t, idx, shift_text):
@@ -208,10 +209,6 @@ class Lattice(BaseParser):
                                 t.cells[i][j].text = t.cells[i - 1][j].text
         return t
 
-    def _generate_image(self):
-        converter = ImageConversionBackend()
-        converter.convert(self.filename, self.imagename)
-
     def _generate_table_bbox(self):
         def scale_areas(areas):
             scaled_areas = []
@@ -391,7 +388,8 @@ class Lattice(BaseParser):
                 )
             return []
 
-        self._generate_image()
+        self.backend.convert(self.filename, self.imagename)
+
         self._generate_table_bbox()
 
         _tables = []