From 6f2c9dcf8fad21b3850016248422a830c2956fa3 Mon Sep 17 00:00:00 2001
From: Erwan Rouchet <rouchet@teklia.com>
Date: Fri, 16 Nov 2018 14:32:18 +0000
Subject: [PATCH] Compute page text from lines
---
arkindex/dataimport/tasks.py | 9 +++---
arkindex/dataimport/tests/test_tasks.py | 2 ++
arkindex/documents/models.py | 9 ++++++
arkindex/documents/tests/test_page.py | 42 +++++++++++++++++++++++++
4 files changed, 58 insertions(+), 4 deletions(-)
create mode 100644 arkindex/documents/tests/test_page.py
diff --git a/arkindex/dataimport/tasks.py b/arkindex/dataimport/tasks.py
index 9d4a21a233..8a6257d256 100644
--- a/arkindex/dataimport/tasks.py
+++ b/arkindex/dataimport/tasks.py
@@ -214,10 +214,6 @@ def save_ml_results(self, results, **kwargs):
if result is None:
continue
- page.classification = result['classification']
- page.save()
- self.report_message("Updated ML results for {}".format(page))
-
tr_items = result['zones']
# Parse transcription types
for item in tr_items:
@@ -235,6 +231,11 @@ def save_ml_results(self, results, **kwargs):
index_transcriptions(transcriptions)
self.report_message('Indexed transcriptions for {}'.format(page))
+ page.build_text()
+ page.classification = result['classification']
+ page.save()
+ self.report_message("Updated ML results for {}".format(page))
+
return list(map(str, results.keys()))
diff --git a/arkindex/dataimport/tests/test_tasks.py b/arkindex/dataimport/tests/test_tasks.py
index f32dbcc703..60927626f3 100644
--- a/arkindex/dataimport/tests/test_tasks.py
+++ b/arkindex/dataimport/tests/test_tasks.py
@@ -72,6 +72,7 @@ class TestTasks(FixtureMixin, RedisMockAPITestCase):
'label': 'dog',
'probability': 0.9,
}])
+ self.assertEqual(dog.text, '')
dog_ts = dog.transcriptions.get()
self.assertEqual(dog_ts.type, TranscriptionType.Word)
self.assertEqual(dog_ts.text, 'woof')
@@ -83,6 +84,7 @@ class TestTasks(FixtureMixin, RedisMockAPITestCase):
'label': 'cat',
'probability': 0.8,
}])
+ self.assertEqual(cat.text, 'meow')
cat_ts = cat.transcriptions.get()
self.assertEqual(cat_ts.type, TranscriptionType.Line)
self.assertEqual(cat_ts.text, 'meow')
diff --git a/arkindex/documents/models.py b/arkindex/documents/models.py
index 48937a7061..895c282809 100644
--- a/arkindex/documents/models.py
+++ b/arkindex/documents/models.py
@@ -437,6 +437,15 @@ class Page(Element):
# Wait for result
self.text = task.get()
+ def build_text(self):
+ self.text = '\n'.join(
+ t.text
+ for t in sorted(
+ self.transcriptions.filter(type=TranscriptionType.Line).prefetch_related('zone'),
+ key=lambda t: (t.zone.polygon.center.y, t.zone.polygon.center.x),
+ )
+ )
+
class Act(Element):
"""
diff --git a/arkindex/documents/tests/test_page.py b/arkindex/documents/tests/test_page.py
new file mode 100644
index 0000000000..ebb16a2767
--- /dev/null
+++ b/arkindex/documents/tests/test_page.py
@@ -0,0 +1,42 @@
+from arkindex.project.tests import FixtureTestCase
+from arkindex.project.polygon import Polygon
+from arkindex.documents.models import Page, TranscriptionType
+
+
+class TestPage(FixtureTestCase):
+
+ def test_build_text(self):
+ img = self.imgsrv.images.create(path='build_text_test')
+ page = Page.objects.create(
+ corpus=self.corpus,
+ name='Page text test',
+ zone=img.zones.create(
+ polygon=Polygon.from_coords(0, 0, 1000, 1000),
+ ),
+ )
+ # Two lines next to each other
+ page.transcriptions.create(
+ type=TranscriptionType.Line,
+ zone=img.zones.create(polygon=Polygon.from_coords(0, 0, 500, 100)),
+ text='Ligne 1',
+ )
+ page.transcriptions.create(
+ type=TranscriptionType.Line,
+ zone=img.zones.create(polygon=Polygon.from_coords(600, 0, 100, 100)),
+ text='Ligne 2',
+ )
+ # One line under them
+ page.transcriptions.create(
+ type=TranscriptionType.Line,
+ zone=img.zones.create(polygon=Polygon.from_coords(0, 200, 42, 42)),
+ text='Ligne 3',
+ )
+ # A word that should not appear
+ page.transcriptions.create(
+ type=TranscriptionType.Word,
+ zone=img.zones.create(polygon=Polygon.from_coords(0, 0, 42, 42)),
+ text='Mot',
+ )
+
+ page.build_text()
+ self.assertEqual(page.text, "Ligne 1\nLigne 2\nLigne 3")
--
GitLab