mirror of
https://github.com/DS4SD/docling.git
synced 2025-12-08 12:48:28 +00:00
fix(docx): slow table parsing (#2553)
* chore(docx): remove unnecessary import Signed-off-by: Cesar Berrospi Ramis <ceb@zurich.ibm.com> * fix(docx): simplify parsing of simple tables Simplify the parsing of tables with just text (no rich cells). Move nested function group_cell_elements out of _handle_tables for readability. Signed-off-by: Cesar Berrospi Ramis <ceb@zurich.ibm.com> * chore(docx): reuse method for finding inline pictures Signed-off-by: Cesar Berrospi Ramis <ceb@zurich.ibm.com> * chore(docx): format strikethrough text Signed-off-by: Cesar Berrospi Ramis <ceb@zurich.ibm.com> * tests(docx): use fixtures to avoid converting same file multiple times Signed-off-by: Cesar Berrospi Ramis <ceb@zurich.ibm.com> * fix(docx): remove unnecessary argument docx_obj in functions Signed-off-by: Cesar Berrospi Ramis <ceb@zurich.ibm.com> * tests(docx): add test for rich table cells Signed-off-by: Cesar Berrospi Ramis <ceb@zurich.ibm.com> * chore(docx): small improvements in backend and its unit tests Signed-off-by: Cesar Berrospi Ramis <ceb@zurich.ibm.com> * chore(docx): parse superscript and subscript formatted text Signed-off-by: Cesar Berrospi Ramis <ceb@zurich.ibm.com> --------- Signed-off-by: Cesar Berrospi Ramis <ceb@zurich.ibm.com>
This commit is contained in:
Cesar Berrospi Ramis
committed by
GitHub
GitHub
parent
0ba8d5d9e3
commit
ef623ffcee
BIN
tests/data/docx/docx_rich_cells.docx
vendored
Normal file
BIN
tests/data/docx/docx_rich_cells.docx
vendored
Normal file
Binary file not shown.
107
tests/data/groundtruth/docling_v2/docx_rich_cells.docx.itxt
vendored
Normal file
107
tests/data/groundtruth/docling_v2/docx_rich_cells.docx.itxt
vendored
Normal file
@@ -0,0 +1,107 @@
|
||||
item-0 at level 0: unspecified: group _root_
|
||||
item-1 at level 1: section: group header-0
|
||||
item-2 at level 2: section: group header-1
|
||||
item-3 at level 3: section_header: Table with rich cells
|
||||
item-4 at level 4: table with [4x2]
|
||||
item-5 at level 5: unspecified: group rich_cell_group_1_0_1
|
||||
item-6 at level 6: text: This is a list:
|
||||
item-7 at level 6: list: group list
|
||||
item-8 at level 7: list_item: A First
|
||||
item-9 at level 7: list_item: A Second
|
||||
item-10 at level 7: list_item: A Third
|
||||
item-11 at level 5: unspecified: group rich_cell_group_1_1_1
|
||||
item-12 at level 6: text: This is a formatted list:
|
||||
item-13 at level 6: list: group list
|
||||
item-14 at level 7: list_item:
|
||||
item-15 at level 8: inline: group group
|
||||
item-16 at level 9: text: B
|
||||
item-17 at level 9: text: First
|
||||
item-18 at level 7: list_item:
|
||||
item-19 at level 8: inline: group group
|
||||
item-20 at level 9: text: B
|
||||
item-21 at level 9: text: Second
|
||||
item-22 at level 7: list_item:
|
||||
item-23 at level 8: inline: group group
|
||||
item-24 at level 9: text: B
|
||||
item-25 at level 9: text: Third
|
||||
item-26 at level 5: unspecified: group rich_cell_group_1_0_2
|
||||
item-27 at level 6: text: First Paragraph
|
||||
|
||||
Second Paragraph
|
||||
item-28 at level 6: text: Third paragraph before a numbered list
|
||||
item-29 at level 6: list: group list
|
||||
item-30 at level 7: list_item: Number one
|
||||
item-31 at level 7: list_item: Number two
|
||||
item-32 at level 7: list_item: Number three
|
||||
item-33 at level 5: unspecified: group rich_cell_group_1_1_2
|
||||
item-34 at level 6: text: This is simple text with
|
||||
item-35 at level 6: text: bold
|
||||
item-36 at level 6: text: ,
|
||||
item-37 at level 6: text: strikethrough
|
||||
item-38 at level 6: text: and
|
||||
item-39 at level 6: text: italic
|
||||
item-40 at level 6: text: formatting with x
|
||||
item-41 at level 6: text: 2
|
||||
item-42 at level 6: text: and H
|
||||
item-43 at level 6: text: 2
|
||||
item-44 at level 6: text: O
|
||||
item-45 at level 5: unspecified: group rich_cell_group_1_0_3
|
||||
item-46 at level 6: text: This is a paragraph
|
||||
item-47 at level 6: text: This is another paragraph
|
||||
item-48 at level 4: inline: group group
|
||||
item-49 at level 4: text:
|
||||
item-50 at level 4: text:
|
||||
item-51 at level 4: text:
|
||||
item-52 at level 4: text:
|
||||
item-53 at level 4: text:
|
||||
item-54 at level 4: text:
|
||||
item-55 at level 3: section_header: Table with nested table
|
||||
item-56 at level 4: text: Before table
|
||||
item-57 at level 4: table with [3x2]
|
||||
item-58 at level 5: unspecified: group rich_cell_group_2_1_1
|
||||
item-59 at level 6: text: Simple cell with
|
||||
item-60 at level 6: text: bold
|
||||
item-61 at level 6: text: and
|
||||
item-62 at level 6: text: italic
|
||||
item-63 at level 6: text: text
|
||||
item-64 at level 5: unspecified: group rich_cell_group_3_0_2
|
||||
item-65 at level 6: table with [2x3]
|
||||
item-66 at level 7: unspecified: group rich_cell_group_3_0_1
|
||||
item-67 at level 8: text: Cell 1
|
||||
item-68 at level 7: unspecified: group rich_cell_group_3_1_1
|
||||
item-69 at level 8: text: Cell 2
|
||||
item-70 at level 7: unspecified: group rich_cell_group_3_2_1
|
||||
item-71 at level 8: text: Cell 3
|
||||
item-72 at level 6: text:
|
||||
item-73 at level 5: unspecified: group rich_cell_group_4_1_2
|
||||
item-74 at level 6: text: Rich cell
|
||||
A nested table
|
||||
item-75 at level 6: table with [2x3]
|
||||
item-76 at level 7: unspecified: group rich_cell_group_4_0_1
|
||||
item-77 at level 8: text: Cell 1
|
||||
item-78 at level 7: unspecified: group rich_cell_group_4_1_1
|
||||
item-79 at level 8: text: Cell 2
|
||||
item-80 at level 7: unspecified: group rich_cell_group_4_2_1
|
||||
item-81 at level 8: text: Cell 3
|
||||
item-82 at level 6: text:
|
||||
item-83 at level 4: inline: group group
|
||||
item-84 at level 4: inline: group group
|
||||
item-85 at level 5: text: After table with
|
||||
item-86 at level 5: text: bold
|
||||
item-87 at level 5: text: ,
|
||||
item-88 at level 5: text: underline
|
||||
item-89 at level 5: text: ,
|
||||
item-90 at level 5: text: strikethrough
|
||||
item-91 at level 5: text: , and
|
||||
item-92 at level 5: text: italic
|
||||
item-93 at level 5: text: formatting
|
||||
item-94 at level 4: text:
|
||||
item-95 at level 3: section_header: Table with pictures
|
||||
item-96 at level 4: text:
|
||||
item-97 at level 4: table with [3x2]
|
||||
item-98 at level 5: unspecified: group rich_cell_group_5_1_1
|
||||
item-99 at level 6: picture
|
||||
item-100 at level 5: unspecified: group rich_cell_group_5_0_2
|
||||
item-101 at level 6: text: Text and picture
|
||||
item-102 at level 6: picture
|
||||
item-103 at level 4: text:
|
||||
2928
tests/data/groundtruth/docling_v2/docx_rich_cells.docx.json
vendored
Normal file
2928
tests/data/groundtruth/docling_v2/docx_rich_cells.docx.json
vendored
Normal file
File diff suppressed because one or more lines are too long
25
tests/data/groundtruth/docling_v2/docx_rich_cells.docx.md
vendored
Normal file
25
tests/data/groundtruth/docling_v2/docx_rich_cells.docx.md
vendored
Normal file
@@ -0,0 +1,25 @@
|
||||
### Table with rich cells
|
||||
|
||||
| Column A | Column B |
|
||||
|------------------------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------------|
|
||||
| This is a list: - A First - A Second - A Third | This is a formatted list: - B **First** - B *Second* - B Third |
|
||||
| First Paragraph Second Paragraph Third paragraph before a numbered list 1. Number one 2. Number two 3. Number three | This is simple text with **bold** , ~~strikethrough~~ and *italic* formatting with x 2 and H 2 O |
|
||||
| This is a paragraph This is another paragraph | |
|
||||
|
||||
### Table with nested table
|
||||
|
||||
Before table
|
||||
|
||||
| Column A | Column B |
|
||||
|----------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------------------------------------------------|
|
||||
| Simple cell upper left | Simple cell with **bold** and *italic* text |
|
||||
| | A | B | C | |----------|--------|------------| | *Cell 1* | Cell 2 | **Cell 3** | | Rich cell A nested table | A | B | C | |------------|--------------|--------| | ~~Cell 1~~ | ***Cell 2*** | Cell 3 | |
|
||||
|
||||
After table with **bold** , underline , ~~strikethrough~~ , and *italic* formatting
|
||||
|
||||
### Table with pictures
|
||||
|
||||
| Column A | Column B |
|
||||
|----------------------------------|----------------|
|
||||
| Only text | <!-- image --> |
|
||||
| Text and picture <!-- image --> | |
|
||||
@@ -1,3 +1,4 @@
|
||||
import logging
|
||||
import os
|
||||
from pathlib import Path
|
||||
|
||||
@@ -18,23 +19,109 @@ from docling.document_converter import DocumentConverter
|
||||
from .test_data_gen_flag import GEN_TEST_DATA
|
||||
from .verify_utils import verify_document, verify_export
|
||||
|
||||
_log = logging.getLogger(__name__)
|
||||
|
||||
GENERATE = GEN_TEST_DATA
|
||||
IS_CI = bool(os.getenv("CI"))
|
||||
|
||||
|
||||
@pytest.fixture(scope="module")
|
||||
def docx_paths() -> list[Path]:
|
||||
# Define the directory you want to search
|
||||
directory = Path("./tests/data/docx/")
|
||||
|
||||
# List all docx files in the directory and its subdirectories
|
||||
docx_files = sorted(directory.rglob("*.docx"))
|
||||
|
||||
return docx_files
|
||||
|
||||
|
||||
def get_converter():
|
||||
converter = DocumentConverter(allowed_formats=[InputFormat.DOCX])
|
||||
|
||||
return converter
|
||||
|
||||
|
||||
@pytest.fixture(scope="module")
|
||||
def documents(docx_paths) -> list[tuple[Path, DoclingDocument]]:
|
||||
documents: list[dict[Path, DoclingDocument]] = []
|
||||
|
||||
converter = get_converter()
|
||||
|
||||
for docx_path in docx_paths:
|
||||
_log.debug(f"converting {docx_path}")
|
||||
|
||||
gt_path = (
|
||||
docx_path.parent.parent / "groundtruth" / "docling_v2" / docx_path.name
|
||||
)
|
||||
|
||||
conv_result: ConversionResult = converter.convert(docx_path)
|
||||
|
||||
doc: DoclingDocument = conv_result.document
|
||||
|
||||
assert doc, f"Failed to convert document from file {gt_path}"
|
||||
documents.append((gt_path, doc))
|
||||
|
||||
return documents
|
||||
|
||||
|
||||
def _test_e2e_docx_conversions_impl(docx_paths: list[tuple[Path, DoclingDocument]]):
|
||||
has_libreoffice = False
|
||||
try:
|
||||
cmd = get_libreoffice_cmd(raise_if_unavailable=True)
|
||||
if cmd is not None:
|
||||
has_libreoffice = True
|
||||
except Exception:
|
||||
pass
|
||||
|
||||
for docx_path, doc in docx_paths:
|
||||
if not IS_CI and not has_libreoffice and docx_path.name == "drawingml.docx":
|
||||
print(f"Skipping {docx_path} because no Libreoffice is installed.")
|
||||
continue
|
||||
|
||||
pred_md: str = doc.export_to_markdown()
|
||||
assert verify_export(pred_md, str(docx_path) + ".md", generate=GENERATE), (
|
||||
f"export to markdown failed on {docx_path}"
|
||||
)
|
||||
|
||||
pred_itxt: str = doc._export_to_indented_text(
|
||||
max_text_len=70, explicit_tables=False
|
||||
)
|
||||
assert verify_export(pred_itxt, str(docx_path) + ".itxt", generate=GENERATE), (
|
||||
f"export to indented-text failed on {docx_path}"
|
||||
)
|
||||
|
||||
assert verify_document(doc, str(docx_path) + ".json", generate=GENERATE), (
|
||||
f"DoclingDocument verification failed on {docx_path}"
|
||||
)
|
||||
|
||||
if docx_path.name == "word_tables.docx":
|
||||
pred_html: str = doc.export_to_html()
|
||||
assert verify_export(
|
||||
pred_text=pred_html,
|
||||
gtfile=str(docx_path) + ".html",
|
||||
generate=GENERATE,
|
||||
), f"export to html failed on {docx_path}"
|
||||
|
||||
|
||||
flaky_file = "textbox.docx"
|
||||
|
||||
|
||||
def test_e2e_docx_conversions(documents):
|
||||
target = [item for item in documents if item[0].name != flaky_file]
|
||||
_test_e2e_docx_conversions_impl(target)
|
||||
|
||||
|
||||
@pytest.mark.xfail(strict=False)
|
||||
def test_textbox_extraction():
|
||||
in_path = Path("tests/data/docx/textbox.docx")
|
||||
in_doc = InputDocument(
|
||||
path_or_stream=in_path,
|
||||
format=InputFormat.DOCX,
|
||||
backend=MsWordDocumentBackend,
|
||||
)
|
||||
backend = MsWordDocumentBackend(
|
||||
in_doc=in_doc,
|
||||
path_or_stream=in_path,
|
||||
)
|
||||
doc = backend.convert()
|
||||
def test_textbox_conversion(documents):
|
||||
target = [item for item in documents if item[0].name == flaky_file]
|
||||
_test_e2e_docx_conversions_impl(target)
|
||||
|
||||
|
||||
@pytest.mark.xfail(strict=False)
|
||||
def test_textbox_extraction(documents):
|
||||
name = "textbox.docx"
|
||||
doc = next(item[1] for item in documents if item[0].name == name)
|
||||
|
||||
# Verify if a particular textbox content is extracted
|
||||
textbox_found = False
|
||||
@@ -44,18 +131,9 @@ def test_textbox_extraction():
|
||||
assert textbox_found
|
||||
|
||||
|
||||
def test_heading_levels():
|
||||
in_path = Path("tests/data/docx/word_sample.docx")
|
||||
in_doc = InputDocument(
|
||||
path_or_stream=in_path,
|
||||
format=InputFormat.DOCX,
|
||||
backend=MsWordDocumentBackend,
|
||||
)
|
||||
backend = MsWordDocumentBackend(
|
||||
in_doc=in_doc,
|
||||
path_or_stream=in_path,
|
||||
)
|
||||
doc = backend.convert()
|
||||
def test_heading_levels(documents):
|
||||
name = "word_sample.docx"
|
||||
doc = next(item[1] for item in documents if item[0].name == name)
|
||||
|
||||
found_lvl_1 = found_lvl_2 = False
|
||||
for item, _ in doc.iterate_items():
|
||||
@@ -69,104 +147,11 @@ def test_heading_levels():
|
||||
assert found_lvl_1 and found_lvl_2
|
||||
|
||||
|
||||
def get_docx_paths():
|
||||
# Define the directory you want to search
|
||||
directory = Path("./tests/data/docx/")
|
||||
def test_text_after_image_anchors(documents):
|
||||
"""Test to analyse whether text gets parsed after image anchors."""
|
||||
|
||||
# List all PDF files in the directory and its subdirectories
|
||||
pdf_files = sorted(directory.rglob("*.docx"))
|
||||
return pdf_files
|
||||
|
||||
|
||||
def get_converter():
|
||||
converter = DocumentConverter(allowed_formats=[InputFormat.DOCX])
|
||||
|
||||
return converter
|
||||
|
||||
|
||||
def _test_e2e_docx_conversions_impl(docx_paths: list[Path]):
|
||||
converter = get_converter()
|
||||
|
||||
has_libreoffice = False
|
||||
try:
|
||||
cmd = get_libreoffice_cmd(raise_if_unavailable=True)
|
||||
if cmd is not None:
|
||||
has_libreoffice = True
|
||||
except Exception:
|
||||
pass
|
||||
|
||||
for docx_path in docx_paths:
|
||||
if (
|
||||
not IS_CI
|
||||
and not has_libreoffice
|
||||
and str(docx_path) in ("tests/data/docx/drawingml.docx",)
|
||||
):
|
||||
print(f"Skipping {docx_path} because no Libreoffice is installed.")
|
||||
continue
|
||||
|
||||
gt_path = (
|
||||
docx_path.parent.parent / "groundtruth" / "docling_v2" / docx_path.name
|
||||
)
|
||||
|
||||
conv_result: ConversionResult = converter.convert(docx_path)
|
||||
|
||||
doc: DoclingDocument = conv_result.document
|
||||
|
||||
pred_md: str = doc.export_to_markdown()
|
||||
assert verify_export(pred_md, str(gt_path) + ".md", generate=GENERATE), (
|
||||
f"export to markdown failed on {docx_path}"
|
||||
)
|
||||
|
||||
pred_itxt: str = doc._export_to_indented_text(
|
||||
max_text_len=70, explicit_tables=False
|
||||
)
|
||||
assert verify_export(pred_itxt, str(gt_path) + ".itxt", generate=GENERATE), (
|
||||
f"export to indented-text failed on {docx_path}"
|
||||
)
|
||||
|
||||
assert verify_document(doc, str(gt_path) + ".json", generate=GENERATE), (
|
||||
f"DoclingDocument verification failed on {docx_path}"
|
||||
)
|
||||
|
||||
if docx_path.name == "word_tables.docx":
|
||||
pred_html: str = doc.export_to_html()
|
||||
assert verify_export(
|
||||
pred_text=pred_html,
|
||||
gtfile=str(gt_path) + ".html",
|
||||
generate=GENERATE,
|
||||
), f"export to html failed on {docx_path}"
|
||||
|
||||
|
||||
flaky_path = Path("tests/data/docx/textbox.docx")
|
||||
|
||||
|
||||
def test_e2e_docx_conversions():
|
||||
_test_e2e_docx_conversions_impl(
|
||||
docx_paths=[path for path in get_docx_paths() if path != flaky_path]
|
||||
)
|
||||
|
||||
|
||||
@pytest.mark.xfail(strict=False)
|
||||
def test_textbox_conversion():
|
||||
_test_e2e_docx_conversions_impl(docx_paths=[flaky_path])
|
||||
|
||||
|
||||
def test_text_after_image_anchors():
|
||||
"""
|
||||
Test to analyse whether text gets parsed after image anchors.
|
||||
"""
|
||||
|
||||
in_path = Path("tests/data/docx/word_image_anchors.docx")
|
||||
in_doc = InputDocument(
|
||||
path_or_stream=in_path,
|
||||
format=InputFormat.DOCX,
|
||||
backend=MsWordDocumentBackend,
|
||||
)
|
||||
backend = MsWordDocumentBackend(
|
||||
in_doc=in_doc,
|
||||
path_or_stream=in_path,
|
||||
)
|
||||
doc = backend.convert()
|
||||
name = "word_image_anchors.docx"
|
||||
doc = next(item[1] for item in documents if item[0].name == name)
|
||||
|
||||
found_text_after_anchor_1 = found_text_after_anchor_2 = (
|
||||
found_text_after_anchor_3
|
||||
@@ -188,3 +173,38 @@ def test_text_after_image_anchors():
|
||||
and found_text_after_anchor_3
|
||||
and found_text_after_anchor_4
|
||||
)
|
||||
|
||||
|
||||
def test_is_rich_table_cell(docx_paths):
|
||||
"""Test the function is_rich_table_cell."""
|
||||
|
||||
name = "docx_rich_cells.docx"
|
||||
path = next(item for item in docx_paths if item.name == name)
|
||||
|
||||
in_doc = InputDocument(
|
||||
path_or_stream=path,
|
||||
format=InputFormat.DOCX,
|
||||
backend=MsWordDocumentBackend,
|
||||
filename=name,
|
||||
)
|
||||
backend = MsWordDocumentBackend(
|
||||
in_doc=in_doc,
|
||||
path_or_stream=path,
|
||||
)
|
||||
|
||||
gt_cells: list[bool] = []
|
||||
# table: Table with rich cells
|
||||
gt_cells.extend([False, False, True, True, True, True, True, False])
|
||||
# table: Table with nested table
|
||||
gt_cells.extend([False, False, False, True, True, True])
|
||||
# table: Table with pictures
|
||||
gt_cells.extend([False, False, False, True, True, False])
|
||||
gt_it = iter(gt_cells)
|
||||
|
||||
for idx_t, table in enumerate(backend.docx_obj.tables):
|
||||
for idx_r, row in enumerate(table.rows):
|
||||
for idx_c, cell in enumerate(row.cells):
|
||||
assert next(gt_it) == backend._is_rich_table_cell(cell), (
|
||||
f"Wrong cell type in table {idx_t}, row {idx_r}, col {idx_c} "
|
||||
f"with text: {cell.text}"
|
||||
)
|
||||
|
||||
Reference in New Issue
Block a user