Docling/tests/test_backend_docling_parse_v2.py
Christoph Auer 5e4944f15f
feat: new experimental docling-parse v2 backend (#131)
---------

Signed-off-by: Christoph Auer <cau@zurich.ibm.com>
Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>
Co-authored-by: Michele Dolfi <dol@zurich.ibm.com>
2024-10-11 15:12:49 +02:00

66 lines
2.0 KiB
Python

from pathlib import Path
import pytest
from docling.backend.docling_parse_v2_backend import (
DoclingParseV2DocumentBackend,
DoclingParseV2PageBackend,
)
from docling.datamodel.base_models import BoundingBox
@pytest.fixture
def test_doc_path():
return Path("./tests/data/2206.01062.pdf")
@pytest.mark.skip
def test_text_cell_counts():
pdf_doc = Path("./tests/data/redp5695.pdf")
doc_backend = DoclingParseV2DocumentBackend(pdf_doc, "123456xyz")
for page_index in range(0, doc_backend.page_count()):
last_cell_count = None
for i in range(10):
page_backend: DoclingParseV2PageBackend = doc_backend.load_page(0)
cells = list(page_backend.get_text_cells())
if last_cell_count is None:
last_cell_count = len(cells)
if len(cells) != last_cell_count:
assert (
False
), "Loading page multiple times yielded non-identical text cell counts"
last_cell_count = len(cells)
def test_get_text_from_rect(test_doc_path):
doc_backend = DoclingParseV2DocumentBackend(test_doc_path, "123456xyz")
page_backend: DoclingParseV2PageBackend = doc_backend.load_page(0)
# Get the title text of the DocLayNet paper
textpiece = page_backend.get_text_in_rect(
bbox=BoundingBox(l=102, t=77, r=511, b=124)
)
ref = "DocLayNet: A Large Human-Annotated Dataset for Document-Layout Analysis"
assert textpiece.strip() == ref
def test_crop_page_image(test_doc_path):
doc_backend = DoclingParseV2DocumentBackend(test_doc_path, "123456xyz")
page_backend: DoclingParseV2PageBackend = doc_backend.load_page(0)
# Crop out "Figure 1" from the DocLayNet paper
im = page_backend.get_page_image(
scale=2, cropbox=BoundingBox(l=317, t=246, r=574, b=527)
)
# im.show()
def test_num_pages(test_doc_path):
doc_backend = DoclingParseV2DocumentBackend(test_doc_path, "123456xyz")
doc_backend.page_count() == 9