docling-ibm-models 3.4.0__py3-none-any.whl → 3.4.2__py3-none-any.whl

docling_ibm_models/reading_order/reading_order_rb.py

@@ -201,8 +201,8 @@ class ReadingOrderPredictor:
                     )
                 ):
 
-                    m1 = re.fullmatch(".+([a-z\,\-])(\s*)", elem.text)
-                    m2 = re.fullmatch("(\s*[a-z])(.+)", sorted_elements[ind_p1].text)
+                    m1 = re.fullmatch(r".+([a-z,\-])(\s*)", elem.text)
+                    m2 = re.fullmatch(r"(\s*[a-z])(.+)", sorted_elements[ind_p1].text)
 
                     if m1 and m2:
                         merges[elem.cid] = [sorted_elements[ind_p1].cid]

docling_ibm_models/tableformer/data_management/matching_post_processor.py

@@ -468,112 +468,77 @@ class MatchingPostProcessor:
         return table_cells
 
     def _align_table_cells_to_pdf(self, table_cells, pdf_cells, matches):
-        r"""
-        USED in 8.a step
-        NOT USED in 6. step
-
-        Align table cell bboxes with good matches
-        to encapsulate matching pdf cells
-
-        Parameters
-        ----------
-        table_cells : list of dict
-            Each value is a dictionary with keys: "cell_id", "row_id", "column_id", "bbox", "label"
-        pdf_cells : list of dict
-            List of PDF cells as defined by Docling
-        matches : dictionary of lists of table_cells
-            A dictionary which is indexed by the pdf_cell_id as key and the value is a list
-            of the table_cells that fall inside that pdf cell
-
-        Returns
-        -------
-        clean_table_cells : list of dict
-            Aligned and cleaned table cells
         """
-        # 6
-        # align table cells with matching pdf cells
-        new_table_cells = []
-
-        for pdf_cell_id in matches:
-            match_list = matches[pdf_cell_id]
-            one_table_cells = []
-            for i in range(len(match_list)):
-                otc = int(match_list[i]["table_cell_id"])
-                if otc not in one_table_cells:
-                    one_table_cells.append(otc)
-
-            # Get bbox of pdf_cell:
-            pdf_cell_bbox = []
-            for pdf_cell in pdf_cells:
-                if pdf_cell["id"] == int(pdf_cell_id):
-                    pdf_cell_bbox = pdf_cell["bbox"]
-
-            # Get bbox of pdf_cell:
-            for table_cell in table_cells:
-                if table_cell["cell_id"] in one_table_cells:
-                    # Align bbox vertically to cover PDF cell
-                    new_bbox = [
-                        pdf_cell_bbox[0],
-                        pdf_cell_bbox[1],
-                        pdf_cell_bbox[2],
-                        pdf_cell_bbox[3],
-                    ]
-                    # We are sure cell is not empty,
-                    # because we assign PDF cell to it
-                    new_table_cell_class = "2"
-
-                    if "cell_class" in table_cell:
-                        new_table_cell_class = table_cell["cell_class"]
-
-                    new_table_cell = {
-                        "bbox": new_bbox,
-                        "cell_id": table_cell["cell_id"],
-                        "column_id": table_cell["column_id"],
-                        "label": table_cell["label"],
-                        "row_id": table_cell["row_id"],
-                        "cell_class": new_table_cell_class,
-                    }
+        Align table cell bboxes with good matches to encapsulate matching pdf cells
+        """
+        pdf_cell_dict = {pdf_cell["id"]: pdf_cell["bbox"] for pdf_cell in pdf_cells}
+        table_cell_dict = {cell["cell_id"]: cell for cell in table_cells}
 
-                    if "colspan_val" in table_cell:
-                        new_table_cell["colspan_val"] = table_cell["colspan_val"]
-                    if "rowspan_val" in table_cell:
-                        new_table_cell["rowspan_val"] = table_cell["rowspan_val"]
-                    new_table_cells.append(new_table_cell)
+        # Track unique cells we're going to add
+        processed_cells = set()
 
-        # Rebuild table_cells list deduplicating repeating cells,
-        # encapsulating all duplicate cells dimensions
+        # First pass - create initial new_table_cells with aligned bboxes
+        new_table_cells = []
 
-        for new_table_cell in new_table_cells:
-            cell_id_to_find = new_table_cell["cell_id"]
+        for pdf_cell_id, match_list in matches.items():
+            # Extract unique table cell ids from match_list
+            table_cell_ids = set(int(match["table_cell_id"]) for match in match_list)
 
-            x1s = []
-            y1s = []
-            x2s = []
-            y2s = []
+            # Get bbox of pdf_cell
+            pdf_cell_bbox = pdf_cell_dict.get(int(pdf_cell_id))
+            if not pdf_cell_bbox:
+                continue
 
-            found = 0
+            # Process each unique table cell
+            for cell_id in table_cell_ids:
+                if cell_id in processed_cells:
+                    continue
+
+                table_cell = table_cell_dict.get(cell_id)
+                if not table_cell:
+                    continue
+
+                # Create new table cell with aligned bbox
+                new_table_cell = table_cell.copy()
+                new_table_cell["bbox"] = list(pdf_cell_bbox)
+
+                # Set cell class
+                if "cell_class" not in new_table_cell:
+                    new_table_cell["cell_class"] = "2"
+
+                new_table_cells.append(new_table_cell)
+                processed_cells.add(cell_id)
+
+        # Second pass - aggregate bboxes for duplicate cells
+        cell_to_bboxes = {}
+        for cell in new_table_cells:
+            cell_id = cell["cell_id"]
+            if cell_id not in cell_to_bboxes:
+                cell_to_bboxes[cell_id] = []
+            cell_to_bboxes[cell_id].append(cell["bbox"])
+
+        # Create final clean table cells
+        clean_table_cells = []
+        processed_ids = set()
+
+        for cell in new_table_cells:
+            cell_id = cell["cell_id"]
+            if cell_id in processed_ids:
+                continue
 
-            for found_cell in new_table_cells:
-                if found_cell["cell_id"] == cell_id_to_find:
-                    found += 1
-                    x1s.append(found_cell["bbox"][0])
-                    y1s.append(found_cell["bbox"][1])
-                    x2s.append(found_cell["bbox"][2])
-                    y2s.append(found_cell["bbox"][3])
+            bboxes = cell_to_bboxes[cell_id]
+            if len(bboxes) > 1:
+                # Merge bboxes
+                x1s = [bbox[0] for bbox in bboxes]
+                y1s = [bbox[1] for bbox in bboxes]
+                x2s = [bbox[2] for bbox in bboxes]
+                y2s = [bbox[3] for bbox in bboxes]
 
-            min_x1 = min(x1s)
-            min_y1 = min(y1s)
-            max_x2 = max(x2s)
-            max_y2 = max(y2s)
+                cell["bbox"] = [min(x1s), min(y1s), max(x2s), max(y2s)]
 
-            if found > 1:
-                new_table_cell["bbox"] = [min_x1, min_y1, max_x2, max_y2]
+            clean_table_cells.append(cell)
+            processed_ids.add(cell_id)
 
-        clean_table_cells = [
-            i
-            for n, i in enumerate(new_table_cells)
-            if i not in new_table_cells[n + 1 :]
-        ]
         return clean_table_cells
 
     def _deduplicate_cells(self, tab_columns, table_cells, iou_matches, ioc_matches):

{docling_ibm_models-3.4.0.dist-info → docling_ibm_models-3.4.2.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: docling-ibm-models
-Version: 3.4.0
+Version: 3.4.2
 Summary: This package contains the AI models used by the Docling PDF conversion package
 License: MIT
 Keywords: docling,convert,document,pdf,layout model,segmentation,table structure,table former

{docling_ibm_models-3.4.0.dist-info → docling_ibm_models-3.4.2.dist-info}/RECORD

@@ -11,12 +11,12 @@ docling_ibm_models/layoutmodel/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5
 docling_ibm_models/layoutmodel/layout_predictor.py,sha256=ArVgs7FBOiu23TC-JoybcaTp7F7a4BgYC8uRVxTgx4E,5681
 docling_ibm_models/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 docling_ibm_models/reading_order/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-docling_ibm_models/reading_order/reading_order_rb.py,sha256=PLp4FXHcKT-SPpOO4EJxv29LL5z9lWH0PForotrGadc,20415
+docling_ibm_models/reading_order/reading_order_rb.py,sha256=Vk3ufc47w2FnVaLI5UGpxoBTZFcpWuIrSAaNGa9c5Rg,20416
 docling_ibm_models/tableformer/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 docling_ibm_models/tableformer/common.py,sha256=2zgGZBFf4fXytEaXrZR2NU6FWdX2kxO0DHlGZmuvpNQ,3230
 docling_ibm_models/tableformer/data_management/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 docling_ibm_models/tableformer/data_management/functional.py,sha256=kJntHEXFz2SP7obEcHyjAqZNZC9qh-U75MwUJALLADI,3143
-docling_ibm_models/tableformer/data_management/matching_post_processor.py,sha256=meSM0jLWNLS8P95QjN6pEp095jFEbKdl9KKfRY1ocy0,58046
+docling_ibm_models/tableformer/data_management/matching_post_processor.py,sha256=_MVbsm0l5aKP3ChvKhXFeZ2Gz_DHGLlyMbqbKTan_MU,56721
 docling_ibm_models/tableformer/data_management/tf_cell_matcher.py,sha256=IdZTaWIRhPpyEwzZgCmviZnYacR6kbcUqBvx7ilmkKY,21250
 docling_ibm_models/tableformer/data_management/tf_predictor.py,sha256=BHd6KdAX0-b9TbX01m0872MO10zWDMValyf4UTIRkAU,39008
 docling_ibm_models/tableformer/data_management/transforms.py,sha256=NNaz_7GI7FCVmu_rJuenqH5VfzRSljJHUHpNQQ8Mq3Q,2983
@@ -34,7 +34,7 @@ docling_ibm_models/tableformer/utils/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeu
 docling_ibm_models/tableformer/utils/app_profiler.py,sha256=Pb7o1zcikKXh7ninaNt4_nVa1xuUrogZxbTr6U6jkEE,8392
 docling_ibm_models/tableformer/utils/mem_monitor.py,sha256=NFZUnrfLThXNZQrm3ESRmPSJmPF2J1z3E2v_72O4dRw,6408
 docling_ibm_models/tableformer/utils/utils.py,sha256=8Bxf1rEn977lFbY9NX0r5xh9PvxIRipQZX_EZW92XfA,10980
-docling_ibm_models-3.4.0.dist-info/LICENSE,sha256=mBb7ErEcM8VS9OhiGHnQ2kk75HwPhr54W1Oiz3965MY,1088
-docling_ibm_models-3.4.0.dist-info/METADATA,sha256=Xmp9j4oesChJr4Zed77u2KjRP5nclZq1glz9Vtyfwog,7434
-docling_ibm_models-3.4.0.dist-info/WHEEL,sha256=Nq82e9rUAnEjt98J6MlVmMCZb-t9cYE2Ir1kpBmnWfs,88
-docling_ibm_models-3.4.0.dist-info/RECORD,,
+docling_ibm_models-3.4.2.dist-info/LICENSE,sha256=mBb7ErEcM8VS9OhiGHnQ2kk75HwPhr54W1Oiz3965MY,1088
+docling_ibm_models-3.4.2.dist-info/METADATA,sha256=AC30CNriUSKcviE24Sn1eIEkrwvXzoM5jiP7ImYR4VU,7434
+docling_ibm_models-3.4.2.dist-info/WHEEL,sha256=Nq82e9rUAnEjt98J6MlVmMCZb-t9cYE2Ir1kpBmnWfs,88
+docling_ibm_models-3.4.2.dist-info/RECORD,,