chore: remove merging logic (#447)

Merging logic has been moved to unstructured for a long time now, this
PR removes the leftover logic here.

<!-- CURSOR_SUMMARY -->
---

> [!NOTE]
> Removes obsolete layout merge logic and related tests, drops unused
MERGED source, and bumps version to 1.0.10 with changelog update.
> 
> - **Core**:
> - Remove `merge_inferred_layout_with_extracted_layout` from
`unstructured_inference/inference/layoutelement.py` and associated
unused imports.
> - **Tests**:
> - Delete `test_merge_inferred_layout_with_extracted_layout` and
related imports in `test_unstructured_inference/test_elements.py`.
> - **Constants**:
> - Drop unused `Source.MERGED` from
`unstructured_inference/constants.py`.
> - **Versioning/Docs**:
>   - Bump `__version__` to `1.0.10` and update `CHANGELOG.md`.
> 
> <sup>Written by [Cursor
Bugbot](https://cursor.com/dashboard?tab=bugbot) for commit
5d7f9a1. This will update automatically
on new commits. Configure
[here](https://cursor.com/dashboard?tab=bugbot).</sup>
<!-- /CURSOR_SUMMARY -->

Author: qued

CHANGELOG.md

@@ -1,3 +1,7 @@
+## 1.0.10
+
+* Remove merging logic that's no longer used
+
 ## 1.0.9
 
 * Make OD model loading thread safe

test_unstructured_inference/test_elements.py

@@ -5,20 +5,15 @@
 import numpy as np
 import pytest
 
-from unstructured_inference.constants import ElementType
 from unstructured_inference.inference import elements
 from unstructured_inference.inference.elements import (
-    ImageTextRegion,
     Rectangle,
-    TextRegion,
     TextRegions,
 )
 from unstructured_inference.inference.layoutelement import (
-    LayoutElement,
     LayoutElements,
     clean_layoutelements,
     clean_layoutelements_for_class,
-    merge_inferred_layout_with_extracted_layout,
     partition_groups_from_regions,
     separate,
 )
@@ -297,40 +292,6 @@ def test_separate(rect1, rect2):
     # assert not rect1.intersects(rect2) #TODO: fix this test
 
 
-def test_merge_inferred_layout_with_extracted_layout():
-    inferred_layout = [
-        LayoutElement.from_coords(453, 322, 1258, 408, text=None, type=ElementType.SECTION_HEADER),
-        LayoutElement.from_coords(387, 477, 1320, 537, text=None, type=ElementType.TEXT),
-    ]
-
-    extracted_layout = [
-        TextRegion.from_coords(438, 318, 1272, 407, text="Example Section Header"),
-        TextRegion.from_coords(377, 469, 1335, 535, text="Example Title"),
-    ]
-
-    extracted_layout_with_full_page_image = [
-        ImageTextRegion.from_coords(0, 0, 1700, 2200, text="Example Section Header"),
-    ]
-
-    merged_layout = merge_inferred_layout_with_extracted_layout(
-        inferred_layout=inferred_layout,
-        extracted_layout=extracted_layout,
-        page_image_size=(1700, 2200),
-    )
-    assert merged_layout[0].type == ElementType.SECTION_HEADER
-    assert merged_layout[0].text == "Example Section Header"
-    assert merged_layout[1].type == ElementType.TEXT
-    assert merged_layout[1].text == "Example Title"
-
-    # case: extracted layout with a full page image
-    merged_layout = merge_inferred_layout_with_extracted_layout(
-        inferred_layout=inferred_layout,
-        extracted_layout=extracted_layout_with_full_page_image,
-        page_image_size=(1700, 2200),
-    )
-    assert merged_layout == inferred_layout
-
-
 def test_clean_layoutelements(test_layoutelements):
     elements = clean_layoutelements(test_layoutelements).as_list()
     assert len(elements) == 2

unstructured_inference/__version__.py

@@ -1 +1 @@
-__version__ = "1.0.9"  # pragma: no cover
+__version__ = "1.0.10"  # pragma: no cover

unstructured_inference/constants.py

@@ -5,7 +5,6 @@ class Source(Enum):
     YOLOX = "yolox"
     DETECTRON2_ONNX = "detectron2_onnx"
     DETECTRON2_LP = "detectron2_lp"
-    MERGED = "merged"
 
 
 class ElementType:

unstructured_inference/inference/layoutelement.py

@@ -1,25 +1,18 @@
 from __future__ import annotations
 
 from dataclasses import dataclass, field
-from typing import Any, Collection, Iterable, List, Optional
+from typing import Any, Iterable, List, Optional
 
 import numpy as np
 from pandas import DataFrame
 from scipy.sparse.csgraph import connected_components
 
 from unstructured_inference.config import inference_config
-from unstructured_inference.constants import (
-    FULL_PAGE_REGION_THRESHOLD,
-    ElementType,
-)
 from unstructured_inference.inference.elements import (
-    ImageTextRegion,
     Rectangle,
     TextRegion,
     TextRegions,
     coords_intersections,
-    grow_region_to_match_region,
-    region_bounding_boxes_are_almost_the_same,
 )
 
 EPSILON_AREA = 1e-7
@@ -221,107 +214,6 @@ def from_region(cls, region: TextRegion):
         return cls(text=text, source=source, type=type, prob=prob, bbox=region.bbox)
 
 
-def merge_inferred_layout_with_extracted_layout(
-    inferred_layout: Collection[LayoutElement],
-    extracted_layout: Collection[TextRegion],
-    page_image_size: tuple,
-    same_region_threshold: float = inference_config.LAYOUT_SAME_REGION_THRESHOLD,
-    subregion_threshold: float = inference_config.LAYOUT_SUBREGION_THRESHOLD,
-) -> List[LayoutElement]:
-    """Merge two layouts to produce a single layout."""
-    extracted_elements_to_add: List[TextRegion] = []
-    inferred_regions_to_remove = []
-    w, h = page_image_size
-    full_page_region = Rectangle(0, 0, w, h)
-    for extracted_region in extracted_layout:
-        extracted_is_image = isinstance(extracted_region, ImageTextRegion)
-        if extracted_is_image:
-            # Skip extracted images for this purpose, we don't have the text from them and they
-            # don't provide good text bounding boxes.
-
-            is_full_page_image = region_bounding_boxes_are_almost_the_same(
-                extracted_region.bbox,
-                full_page_region,
-                FULL_PAGE_REGION_THRESHOLD,
-            )
-
-            if is_full_page_image:
-                continue
-        region_matched = False
-        for inferred_region in inferred_layout:
-
-            if inferred_region.bbox.intersects(extracted_region.bbox):
-                same_bbox = region_bounding_boxes_are_almost_the_same(
-                    inferred_region.bbox,
-                    extracted_region.bbox,
-                    same_region_threshold,
-                )
-                inferred_is_subregion_of_extracted = inferred_region.bbox.is_almost_subregion_of(
-                    extracted_region.bbox,
-                    subregion_threshold=subregion_threshold,
-                )
-                inferred_is_text = inferred_region.type not in (
-                    ElementType.FIGURE,
-                    ElementType.IMAGE,
-                    ElementType.PAGE_BREAK,
-                    ElementType.TABLE,
-                )
-                extracted_is_subregion_of_inferred = extracted_region.bbox.is_almost_subregion_of(
-                    inferred_region.bbox,
-                    subregion_threshold=subregion_threshold,
-                )
-                either_region_is_subregion_of_other = (
-                    inferred_is_subregion_of_extracted or extracted_is_subregion_of_inferred
-                )
-                if same_bbox:
-                    # Looks like these represent the same region
-                    if extracted_is_image:
-                        # keep extracted region, remove inferred region
-                        inferred_regions_to_remove.append(inferred_region)
-                    else:
-                        # keep inferred region, remove extracted region
-                        grow_region_to_match_region(inferred_region.bbox, extracted_region.bbox)
-                        inferred_region.text = extracted_region.text
-                        region_matched = True
-                elif extracted_is_subregion_of_inferred and inferred_is_text:
-                    if extracted_is_image:
-                        # keep both extracted and inferred regions
-                        region_matched = False
-                    else:
-                        # keep inferred region, remove extracted region
-                        grow_region_to_match_region(inferred_region.bbox, extracted_region.bbox)
-                        region_matched = True
-                elif (
-                    either_region_is_subregion_of_other
-                    and inferred_region.type != ElementType.TABLE
-                ):
-                    # keep extracted region, remove inferred region
-                    inferred_regions_to_remove.append(inferred_region)
-        if not region_matched:
-            extracted_elements_to_add.append(extracted_region)
-    # Need to classify the extracted layout elements we're keeping.
-    categorized_extracted_elements_to_add = [
-        LayoutElement(
-            text=el.text,
-            type=(
-                ElementType.IMAGE
-                if isinstance(el, ImageTextRegion)
-                else ElementType.UNCATEGORIZED_TEXT
-            ),
-            source=el.source,
-            bbox=el.bbox,
-        )
-        for el in extracted_elements_to_add
-    ]
-    inferred_regions_to_add = [
-        region for region in inferred_layout if region not in inferred_regions_to_remove
-    ]
-
-    final_layout = categorized_extracted_elements_to_add + inferred_regions_to_add
-
-    return final_layout
-
-
 def separate(region_a: Rectangle, region_b: Rectangle):
     """Reduce leftmost rectangle to don't overlap with the other"""