Fix the order of args to Bbox

CHANGELOG.rst

@@ -37,6 +37,9 @@ Fixed
   work even when the input mention is not tabular.
   (`#425 <https://github.com/HazyResearch/fonduer/issues/425>`_)
   (`#426 <https://github.com/HazyResearch/fonduer/pull/426>`_)
+* `@HiromuHota`_: Fix the order of args to Bbox.
+  (`#443 <https://github.com/HazyResearch/fonduer/issues/443>`_)
+  (`#444 <https://github.com/HazyResearch/fonduer/pull/444>`_)
 
 0.8.2_ - 2020-04-28
 -------------------

src/fonduer/parser/visual_linker.py

@@ -163,7 +163,7 @@ def _coordinates_from_HTML(
                             word_id = (page_num, i)
                             pdf_word_list.append((word_id, content))
                             coordinate_map[word_id] = Bbox(
-                                page_num, y_min_line, xmin, y_max_line, xmax,
+                                page_num, y_min_line, y_max_line, xmin, xmax,
                             )
                             block_coordinates[word_id] = (y_min_block, x_min_block)
                             i += 1
@@ -318,7 +318,7 @@ def display_match_counts() -> int:
 
     def _update_coordinates(self) -> Iterator[Sentence]:
         for sentence in self.sentences:
-            (page, top, left, bottom, right) = list(
+            (page, top, bottom, left, right) = list(
                 zip(
                     *[
                         self.coordinate_map[self.links[((sentence.stable_id), i)]]

src/fonduer/utils/visualizer.py

@@ -48,13 +48,13 @@ def display_boxes(
             boxes_by_page: DefaultDict[
                 int, List[Tuple[int, int, int, int]]
             ] = defaultdict(list)
-            for i, (page, top, left, bottom, right) in enumerate(boxes):
+            for i, (page, top, bottom, left, right) in enumerate(boxes):
                 boxes_per_page[page] += 1
-                boxes_by_page[page].append((top, left, bottom, right))
+                boxes_by_page[page].append((top, bottom, left, right))
             for i, page_num in enumerate(boxes_per_page.keys()):
                 img = pdf_to_img(pdf_file, page_num)
                 draw.fill_color = transparent
-                for j, (top, left, bottom, right) in enumerate(boxes_by_page[page_num]):
+                for j, (top, bottom, left, right) in enumerate(boxes_by_page[page_num]):
                     draw.stroke_color = colors[j % 2] if alternate_colors else colors[0]
                     draw.rectangle(left=left, top=top, right=right, bottom=bottom)
                 draw(img)
@@ -103,8 +103,8 @@ def display_words(
                         Bbox(
                             sentence.page[i],
                             sentence.top[i],
-                            sentence.left[i],
                             sentence.bottom[i],
+                            sentence.left[i],
                             sentence.right[i],
                         )
                     )
@@ -117,8 +117,8 @@ def get_box(span: SpanMention) -> Bbox:
     return Bbox(
         min(span.get_attrib_tokens("page")),
         min(span.get_attrib_tokens("top")),
-        min(span.get_attrib_tokens("left")),
         max(span.get_attrib_tokens("bottom")),
+        min(span.get_attrib_tokens("left")),
         max(span.get_attrib_tokens("right")),
     )
 

tests/utils/test_visualizer.py

@@ -9,7 +9,7 @@
 
 def test_visualizer():
     """Unit test of visualizer using the md document."""
-    from fonduer.utils.visualizer import Visualizer  # noqa
+    from fonduer.utils.visualizer import Visualizer, get_box  # noqa
 
     docs_path = "tests/data/html_simple/md.html"
     pdf_path = "tests/data/pdf_simple/md.pdf"
@@ -38,12 +38,21 @@ def test_visualizer():
 
     doc = candidate_extractor_udf.apply(doc, split=0)
 
-    cands = doc.organizations
+    # Take one candidate
+    cand = doc.organizations[0]
 
-    # Test visualizer
     pdf_path = "tests/data/pdf_simple"
     vis = Visualizer(pdf_path)
-    vis.display_candidates([cands[0]])
+
+    # Test bounding boxes
+    boxes = [get_box(mention.context) for mention in cand.get_mentions()]
+    for box in boxes:
+        assert box.top <= box.bottom
+        assert box.left <= box.right
+    assert boxes == [mention.context.get_bbox() for mention in cand.get_mentions()]
+
+    # Test visualizer
+    vis.display_candidates([cand])
 
 
 def test_get_pdf_dim():