amorfo77/paperless-ngx
1
0 Fork 0
mirror of https://github.com/paperless-ngx/paperless-ngx.git synced 2026-10-21 23:38:13 +00:00
Code Issues Projects Releases Wiki Activity

Unify OCR zone extraction for preview and saved templates

Browse Source 
Co-authored-by: OpenAI Codex <codex@openai.com>
This commit is contained in:
shamoonandOpenAI Codex committed 2026-10-20 16:58:10 -07:00
1 parent 257c33453d
commit 6562415894
2 files changed
+202 -57
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

src/documents/tests/test_zone_ocr.py
+171
View File
@@ -7,7 +7,9 @@ from unittest.mock import patch
from django.contrib.auth.models import Group
from django.contrib.auth.models import User
from django.test import SimpleTestCase
from django.test import TestCase
from PIL import Image
from documents.models import CustomField
from documents.models import CustomFieldInstance
@@ -17,8 +19,12 @@ from documents.models import OcrTemplate
from documents.models import OcrTemplateZone
from documents.zone_ocr import _apply_transform
from documents.zone_ocr import _convert_value
from documents.zone_ocr import _crop_zone
from documents.zone_ocr import _detect_mime
from documents.zone_ocr import _extract_zone
from documents.zone_ocr import _process_template
from documents.zone_ocr import _resolve_doc_path
from documents.zone_ocr import extract_zone_preview
from documents.zone_ocr import get_templates_for_document
from documents.zone_ocr import run_zone_extraction
from paperless_testing.dirs import DirectoriesMixin
@@ -26,6 +32,171 @@ from paperless_testing.factories import DocumentFactory
from paperless_testing.permissions import grant_object
class TestZoneExtraction(SimpleTestCase):
def setUp(self):
directory = tempfile.TemporaryDirectory()
self.addCleanup(directory.cleanup)
self.tmp_path = Path(directory.name)
self.page = self.tmp_path / "page.png"
with Image.new("RGB", (200, 100), "white") as image:
image.paste("black", (20, 10, 60, 30))
image.save(self.page)
self.zone = OcrTemplateZone(
name="Account",
x=10,
y=5,
width=20,
height=10,
transform="uppercase",
ocr_language="eng",
)
settings = self.settings(SCRATCH_DIR=self.tmp_path)
settings.enable()
self.addCleanup(settings.disable)
def test_crop_scales_coordinates(self):
with _crop_zone(self.page, self.zone, 100, 50) as cropped:
self.assertEqual(cropped.size, (40, 20))
self.assertEqual(cropped.getextrema(), ((0, 0), (0, 0), (0, 0)))
def test_crop_defaults_to_image_dimensions(self):
with _crop_zone(self.page, self.zone, 0, 0) as cropped:
self.assertEqual(cropped.size, (20, 10))
self.assertEqual(cropped.getpixel((0, 0)), (255, 255, 255))
self.assertEqual(cropped.getpixel((19, 9)), (0, 0, 0))
def test_crop_clamps_to_image(self):
self.zone.x, self.zone.y = 90, 40
with _crop_zone(self.page, self.zone, 100, 50) as cropped:
self.assertEqual(cropped.size, (20, 20))
@patch("documents.zone_ocr.subprocess.run")
def test_tiny_crop_skips_reader(self, run):
self.zone.width = 0
result = _extract_zone(self.page, self.zone, 100, 50, self.tmp_path)
self.assertIsNone(result.raw_text)
self.assertIsNone(result.value)
run.assert_not_called()
@patch("documents.zone_ocr.subprocess.run")
def test_unreadable_image_skips_reader(self, run):
self.page.write_bytes(b"not an image")
result = _extract_zone(self.page, self.zone, 100, 50, self.tmp_path)
self.assertIsNone(result.raw_text)
self.assertIsNone(result.value)
run.assert_not_called()
@patch("documents.zone_ocr.subprocess.run")
def test_text_extraction_retains_raw_text_and_transforms(self, run):
def read_crop(args, **kwargs):
with Image.open(args[1]) as crop:
self.assertEqual(crop.size, (40, 20))
self.assertEqual(crop.getpixel((0, 0)), (0, 0, 0))
return MagicMock(stdout=" account abc\n")
run.side_effect = read_crop
result = _extract_zone(self.page, self.zone, 100, 50, self.tmp_path)
self.assertEqual(result.raw_text, "account abc")
self.assertEqual(result.value, "ACCOUNT ABC")
self.assertEqual(run.call_args.args[0][-4:], ["-l", "eng", "--psm", "6"])
self.assertEqual(run.call_args.kwargs["timeout"], 30)
@patch("documents.zone_ocr.subprocess.run")
def test_no_text_returns_no_value(self, run):
run.return_value.stdout = " \n"
result = _extract_zone(self.page, self.zone, 100, 50, self.tmp_path)
self.assertIsNone(result.raw_text)
self.assertIsNone(result.value)
@patch("documents.zone_ocr.subprocess.run")
def test_date_transform_uses_zone_format(self, run):
self.zone.transform = "date"
self.zone.date_format = "%d/%m/%Y"
run.return_value.stdout = "25/12/2025"
result = _extract_zone(self.page, self.zone, 100, 50, self.tmp_path)
self.assertEqual(result.raw_text, "25/12/2025")
self.assertEqual(result.value, "2025-12-25")
@patch("documents.zone_ocr.subprocess.run")
@patch("zxingcpp.read_barcodes")
def test_barcode_extraction_bypasses_tesseract(self, read_barcodes, run):
self.zone.transform = "qr_code"
def read_crop(image):
self.assertEqual(image.size, (40, 20))
self.assertEqual(image.getpixel((0, 0)), (0, 0, 0))
return [MagicMock(text=" https://example.com ")]
read_barcodes.side_effect = read_crop
result = _extract_zone(self.page, self.zone, 100, 50, self.tmp_path)
self.assertEqual(result.raw_text, " https://example.com ")
self.assertEqual(result.value, "https://example.com")
run.assert_not_called()
with self.assertRaises(ValueError):
read_barcodes.call_args.args[0].getpixel((0, 0))
@patch("zxingcpp.read_barcodes", return_value=[])
def test_missing_barcode_returns_no_value(self, read_barcodes):
self.zone.transform = "qr_code"
result = _extract_zone(self.page, self.zone, 100, 50, self.tmp_path)
self.assertIsNone(result.raw_text)
self.assertIsNone(result.value)
@patch("documents.zone_ocr.subprocess.run")
def test_preview_uses_image_dimensions_without_writing(self, run):
run.return_value.stdout = "account abc"
self.assertEqual(
extract_zone_preview(self.page, self.zone, 0, 0, 1),
{"raw_text": "account abc", "value": "ACCOUNT ABC"},
)
@patch("documents.zone_ocr.subprocess.run")
def test_preview_unavailable_page_skips_reader(self, run):
self.zone.page = 2
self.assertEqual(
extract_zone_preview(self.page, self.zone, 100, 50, 1),
{"raw_text": None, "value": None},
)
run.assert_not_called()
class TestSavedZoneExtraction(DirectoriesMixin, TestCase):
@patch("documents.zone_ocr.subprocess.run")
def test_saved_extraction_matches_preview_and_rejects_invalid_value(self, run):
page = Path(self.dirs.scratch_dir) / "page.png"
with Image.new("RGB", (200, 100), "white") as image:
image.save(page)
document = DocumentFactory(title="Original")
template = OcrTemplate.objects.create(
name="Account",
source_width=100,
source_height=50,
)
zone = OcrTemplateZone.objects.create(
template=template,
name="Title",
target="title",
x=0,
y=0,
width=20,
height=10,
transform="uppercase",
validation_regex="ACCOUNT [A-Z]+",
)
for raw_text, expected_title in [
("account abc", "ACCOUNT ABC"),
("invalid", "ACCOUNT ABC"),
]:
with self.subTest(raw_text=raw_text):
run.return_value.stdout = raw_text
preview = extract_zone_preview(page, zone, 100, 50, 1)
self.assertEqual(preview["raw_text"], raw_text)
self.assertEqual(preview["value"], raw_text.upper())
_process_template(document, page, template, [zone])
document.refresh_from_db()
self.assertEqual(document.title, expected_title)
class TestTemplateVisibility(DirectoriesMixin, TestCase):
def setUp(self):
super().setUp()