Text orientation in base worker

.isort.cfg

@@ -8,4 +8,4 @@ line_length = 88
 
 default_section=FIRSTPARTY
 known_first_party = arkindex,arkindex_common
-known_third_party =PIL,apistar,gitlab,gnupg,peewee,pytest,requests,setuptools,sh,tenacity,yaml
+known_third_party =PIL,apistar,gitlab,gnupg,peewee,playhouse,pytest,requests,setuptools,sh,tenacity,yaml

arkindex_worker/cache.py

@@ -124,6 +124,7 @@ class CachedTranscription(Model):
     element = ForeignKeyField(CachedElement, backref="transcriptions")
     text = TextField()
     confidence = FloatField()
+    orientation = CharField(max_length=50)
     worker_version_id = UUIDField()
 
     class Meta:

arkindex_worker/worker/transcription.py

 # -*- coding: utf-8 -*-
 
+from enum import Enum
+
 from peewee import IntegrityError
 
 from arkindex_worker import logger
@@ -7,8 +9,17 @@ from arkindex_worker.cache import CachedElement, CachedTranscription
 from arkindex_worker.models import Element
 
 
+class TextOrientation(Enum):
+    HorizontalLeftToRight = "horizontal-lr"
+    HorizontalRightToLeft = "horizontal-rl"
+    VerticalRightToLeft = "vertical-rl"
+    VerticalLeftToRight = "vertical-lr"
+
+
 class TranscriptionMixin(object):
-    def create_transcription(self, element, text, score):
+    def create_transcription(
+        self, element, text, score, orientation=TextOrientation.HorizontalLeftToRight
+    ):
         """
         Create a transcription on the given element through the API.
         """
@@ -18,7 +29,9 @@ class TranscriptionMixin(object):
         assert text and isinstance(
             text, str
         ), "text shouldn't be null and should be of type str"
-
+        assert orientation and isinstance(
+            orientation, TextOrientation
+        ), "orientation shouldn't be null and should be of type TextOrientation"
         assert (
             isinstance(score, float) and 0 <= score <= 1
         ), "score shouldn't be null and should be a float in [0..1] range"
@@ -36,6 +49,7 @@ class TranscriptionMixin(object):
                 "text": text,
                 "worker_version": self.worker_version_id,
                 "score": score,
+                "orientation": orientation.value,
             },
         )
 
@@ -50,6 +64,7 @@ class TranscriptionMixin(object):
                         "element_id": element.id,
                         "text": created["text"],
                         "confidence": created["confidence"],
+                        "orientation": created["orientation"],
                         "worker_version_id": self.worker_version_id,
                     }
                 ]
@@ -70,7 +85,10 @@ class TranscriptionMixin(object):
             transcriptions, list
         ), "transcriptions shouldn't be null and should be of type list"
 
-        for index, transcription in enumerate(transcriptions):
+        # Create shallow copies of every transcription to avoid mutating the original payload
+        transcriptions_payload = list(map(dict, transcriptions))
+
+        for (index, transcription) in enumerate(transcriptions_payload):
             element_id = transcription.get("element_id")
             assert element_id and isinstance(
                 element_id, str
@@ -86,11 +104,20 @@ class TranscriptionMixin(object):
                 score is not None and isinstance(score, float) and 0 <= score <= 1
             ), f"Transcription at index {index} in transcriptions: score shouldn't be null and should be a float in [0..1] range"
 
+            orientation = transcription.get(
+                "orientation", TextOrientation.HorizontalLeftToRight
+            )
+            assert orientation and isinstance(
+                orientation, TextOrientation
+            ), f"Transcription at index {index} in transcriptions: orientation shouldn't be null and should be of type TextOrientation"
+            if orientation:
+                transcription["orientation"] = orientation.value
+
         created_trs = self.request(
             "CreateTranscriptions",
             body={
                 "worker_version": self.worker_version_id,
-                "transcriptions": transcriptions,
+                "transcriptions": transcriptions_payload,
             },
         )["transcriptions"]
 
@@ -106,6 +133,7 @@ class TranscriptionMixin(object):
                         "element_id": created_tr["element_id"],
                         "text": created_tr["text"],
                         "confidence": created_tr["confidence"],
+                        "orientation": created_tr["orientation"],
                         "worker_version_id": self.worker_version_id,
                     }
                     for created_tr in created_trs
@@ -132,7 +160,10 @@ class TranscriptionMixin(object):
             transcriptions, list
         ), "transcriptions shouldn't be null and should be of type list"
 
-        for index, transcription in enumerate(transcriptions):
+        # Create shallow copies of every transcription to avoid mutating the original payload
+        transcriptions_payload = list(map(dict, transcriptions))
+
+        for (index, transcription) in enumerate(transcriptions_payload):
             text = transcription.get("text")
             assert text and isinstance(
                 text, str
@@ -143,6 +174,15 @@ class TranscriptionMixin(object):
                 score is not None and isinstance(score, float) and 0 <= score <= 1
             ), f"Transcription at index {index} in transcriptions: score shouldn't be null and should be a float in [0..1] range"
 
+            orientation = transcription.get(
+                "orientation", TextOrientation.HorizontalLeftToRight
+            )
+            assert orientation and isinstance(
+                orientation, TextOrientation
+            ), f"Transcription at index {index} in transcriptions: orientation shouldn't be null and should be of type TextOrientation"
+            if orientation:
+                transcription["orientation"] = orientation.value
+
             polygon = transcription.get("polygon")
             assert polygon and isinstance(
                 polygon, list
@@ -168,7 +208,7 @@ class TranscriptionMixin(object):
             body={
                 "element_type": sub_element_type,
                 "worker_version": self.worker_version_id,
-                "transcriptions": transcriptions,
+                "transcriptions": transcriptions_payload,
                 "return_elements": True,
             },
         )
@@ -216,6 +256,9 @@ class TranscriptionMixin(object):
                         "element_id": annotation["element_id"],
                         "text": transcription["text"],
                         "confidence": transcription["score"],
+                        "orientation": transcription.get(
+                            "orientation", TextOrientation.HorizontalLeftToRight
+                        ).value,
                         "worker_version_id": self.worker_version_id,
                     }
                 )

tests/conftest.py

@@ -15,6 +15,7 @@ from arkindex.mock import MockApiClient
 from arkindex_worker.cache import MODELS, CachedElement, CachedTranscription
 from arkindex_worker.git import GitHelper, GitlabHelper
 from arkindex_worker.worker import BaseWorker, ElementsWorker
+from arkindex_worker.worker.transcription import TextOrientation
 
 FIXTURES_DIR = Path(__file__).resolve().parent / "data"
 
@@ -381,6 +382,7 @@ def mock_cached_transcriptions():
         element_id=UUID("11111111-1111-1111-1111-111111111111"),
         text="This",
         confidence=0.42,
+        orientation=TextOrientation.HorizontalLeftToRight,
         worker_version_id=UUID("56785678-5678-5678-5678-567856785678"),
     )
     CachedTranscription.create(
@@ -388,6 +390,7 @@ def mock_cached_transcriptions():
         element_id=UUID("22222222-2222-2222-2222-222222222222"),
         text="is",
         confidence=0.42,
+        orientation=TextOrientation.HorizontalLeftToRight,
         worker_version_id=UUID("90129012-9012-9012-9012-901290129012"),
     )
     CachedTranscription.create(
@@ -395,6 +398,7 @@ def mock_cached_transcriptions():
         element_id=UUID("33333333-3333-3333-3333-333333333333"),
         text="a",
         confidence=0.42,
+        orientation=TextOrientation.HorizontalLeftToRight,
         worker_version_id=UUID("90129012-9012-9012-9012-901290129012"),
     )
     CachedTranscription.create(
@@ -402,6 +406,7 @@ def mock_cached_transcriptions():
         element_id=UUID("44444444-4444-4444-4444-444444444444"),
         text="good",
         confidence=0.42,
+        orientation=TextOrientation.HorizontalLeftToRight,
         worker_version_id=UUID("90129012-9012-9012-9012-901290129012"),
     )
     CachedTranscription.create(
@@ -409,6 +414,7 @@ def mock_cached_transcriptions():
         element_id=UUID("55555555-5555-5555-5555-555555555555"),
         text="test",
         confidence=0.42,
+        orientation=TextOrientation.HorizontalLeftToRight,
         worker_version_id=UUID("90129012-9012-9012-9012-901290129012"),
     )
 
@@ -461,6 +467,7 @@ def mock_databases(tmpdir):
             element_id=UUID("42424242-4242-4242-4242-424242424242"),
             text="Hello!",
             confidence=0.42,
+            orientation=TextOrientation.HorizontalLeftToRight,
             worker_version_id=UUID("56785678-5678-5678-5678-567856785678"),
         )
 
@@ -477,6 +484,7 @@ def mock_databases(tmpdir):
             element_id=UUID("42424242-4242-4242-4242-424242424242"),
             text="Hello again neighbor !",
             confidence=0.42,
+            orientation=TextOrientation.HorizontalLeftToRight,
             worker_version_id=UUID("56785678-5678-5678-5678-567856785678"),
         )
 

tests/test_cache.py

@@ -59,7 +59,7 @@ CREATE TABLE "elements" ("id" TEXT NOT NULL PRIMARY KEY, "parent_id" TEXT, "type
 CREATE TABLE "entities" ("id" TEXT NOT NULL PRIMARY KEY, "type" VARCHAR(50) NOT NULL, "name" TEXT NOT NULL, "validated" INTEGER NOT NULL, "metas" text, "worker_version_id" TEXT NOT NULL)
 CREATE TABLE "images" ("id" TEXT NOT NULL PRIMARY KEY, "width" INTEGER NOT NULL, "height" INTEGER NOT NULL, "url" TEXT NOT NULL)
 CREATE TABLE "transcription_entities" ("transcription_id" TEXT NOT NULL, "entity_id" TEXT NOT NULL, "offset" INTEGER NOT NULL CHECK (offset >= 0), "length" INTEGER NOT NULL CHECK (length > 0), "worker_version_id" TEXT NOT NULL, PRIMARY KEY ("transcription_id", "entity_id"), FOREIGN KEY ("transcription_id") REFERENCES "transcriptions" ("id"), FOREIGN KEY ("entity_id") REFERENCES "entities" ("id"))
-CREATE TABLE "transcriptions" ("id" TEXT NOT NULL PRIMARY KEY, "element_id" TEXT NOT NULL, "text" TEXT NOT NULL, "confidence" REAL NOT NULL, "worker_version_id" TEXT NOT NULL, FOREIGN KEY ("element_id") REFERENCES "elements" ("id"))"""
+CREATE TABLE "transcriptions" ("id" TEXT NOT NULL PRIMARY KEY, "element_id" TEXT NOT NULL, "text" TEXT NOT NULL, "confidence" REAL NOT NULL, "orientation" VARCHAR(50) NOT NULL, "worker_version_id" TEXT NOT NULL, FOREIGN KEY ("element_id") REFERENCES "elements" ("id"))"""
 
     actual_schema = "\n".join(
         [

tests/test_elements_worker/test_entities.py

@@ -13,6 +13,7 @@ from arkindex_worker.cache import (
 )
 from arkindex_worker.models import Element
 from arkindex_worker.worker import EntityType
+from arkindex_worker.worker.transcription import TextOrientation
 
 from . import BASE_API_CALLS
 
@@ -465,6 +466,7 @@ def test_create_transcription_entity_with_cache(
         element=UUID("12341234-1234-1234-1234-123412341234"),
         text="Hello, it's me.",
         confidence=0.42,
+        orientation=TextOrientation.HorizontalLeftToRight,
         worker_version_id=UUID("12341234-1234-1234-1234-123412341234"),
     )
     CachedEntity.create(