.gitlab-ci.yml

@@ -11,7 +11,7 @@ include:
 
 # For jobs that run backend scripts directly
 .backend-setup:
-  image: registry.gitlab.com/teklia/arkindex/backend/base:lxml-4.9.2
+  image: registry.gitlab.teklia.com/arkindex/backend/base:gitlab-teklia
 
   cache:
     paths:
@@ -19,8 +19,8 @@ include:
 
   before_script:
     # Custom line to install our own deps from Git using GitLab CI credentials
-    - "pip install -e git+https://gitlab-ci-token:${CI_JOB_TOKEN}@gitlab.com/teklia/arkindex/transkribus#egg=transkribus-client"
-    - "pip install -e git+https://gitlab-ci-token:${CI_JOB_TOKEN}@gitlab.com/teklia/arkindex/license#egg=teklia-license"
+    - "pip install -e git+https://gitlab-ci-token:${CI_JOB_TOKEN}@gitlab.teklia.com/arkindex/transkribus#egg=transkribus-client"
+    - "pip install -e git+https://gitlab-ci-token:${CI_JOB_TOKEN}@gitlab.teklia.com/arkindex/license#egg=teklia-license"
     - pip install -r tests-requirements.txt
     - "echo 'database: {host: postgres, port: 5432}\npublic_hostname: http://ci.arkindex.localhost' > $CONFIG_PATH"
 
@@ -245,7 +245,7 @@ sentry-release:
 
 release-notes:
   stage: deploy
-  image: registry.gitlab.com/teklia/devops:latest
+  image: registry.gitlab.teklia.com/infra/devops:latest
 
   rules:
     - if: '$CI_COMMIT_TAG && $CI_COMMIT_TAG !~ /^base-.*/'
@@ -257,7 +257,7 @@ release-notes:
 
 bump-python-deps:
   stage: deploy
-  image: registry.gitlab.com/teklia/devops:latest
+  image: registry.gitlab.teklia.com/infra/devops:latest
 
   only:
     - schedules

Dockerfile

-FROM registry.gitlab.com/teklia/arkindex/backend/base:lxml-4.9.2 as build
+FROM registry.gitlab.teklia.com/arkindex/backend/base:gitlab-teklia as build
 
 RUN mkdir build
 ADD . build
 RUN cd build && python3 setup.py sdist
 
-FROM registry.gitlab.com/teklia/arkindex/backend/base:lxml-4.9.2
+FROM registry.gitlab.teklia.com/arkindex/backend/base:gitlab-teklia
 ARG TRANSKRIBUS_BRANCH=master
-ARG TRANSKRIBUS_ID=11180199
+ARG TRANSKRIBUS_ID=63
 ARG LICENSE_BRANCH=master
-ARG LICENSE_ID=45943500
-ARG GITLAB_TOKEN="gaFM7LRa9zy9QMowcUhx"
+ARG LICENSE_ID=37
+
+# Auth token expires on 01/07/2024
+ARG GITLAB_TOKEN="glpat-3sBZPFgkZbqJxfSqjcAa"
 
 # Install transkribus-client from private repo
 RUN \
   mkdir /tmp/transkribus && \
-  wget --header "PRIVATE-TOKEN: $GITLAB_TOKEN" https://gitlab.com/api/v4/projects/$TRANSKRIBUS_ID/repository/archive.tar.gz?sha=$TRANSKRIBUS_BRANCH -O /tmp/transkribus/archive.tar.gz && \
+  wget --header "PRIVATE-TOKEN: $GITLAB_TOKEN" https://gitlab.teklia.com/api/v4/projects/$TRANSKRIBUS_ID/repository/archive.tar.gz?sha=$TRANSKRIBUS_BRANCH -O /tmp/transkribus/archive.tar.gz && \
   tar --strip-components=1 -xvf /tmp/transkribus/archive.tar.gz -C /tmp/transkribus && \
   cd /tmp/transkribus && pip install --disable-pip-version-check --no-cache-dir --quiet . && \
   rm -rf /tmp/transkribus
@@ -22,7 +24,7 @@ RUN \
 # Install teklia-license from private repo
 RUN \
   mkdir /tmp/teklia-license && \
-  wget --header "PRIVATE-TOKEN: $GITLAB_TOKEN" https://gitlab.com/api/v4/projects/$LICENSE_ID/repository/archive.tar.gz?sha=$LICENSE_BRANCH -O /tmp/teklia-license.tar.gz && \
+  wget --header "PRIVATE-TOKEN: $GITLAB_TOKEN" https://gitlab.teklia.com/api/v4/projects/$LICENSE_ID/repository/archive.tar.gz?sha=$LICENSE_BRANCH -O /tmp/teklia-license.tar.gz && \
   tar --strip-components=1 -xvf /tmp/teklia-license.tar.gz -C /tmp/teklia-license && \
   cd /tmp/teklia-license && pip install --disable-pip-version-check --no-cache-dir --quiet . && \
   rm -rf /tmp/teklia-license

Dockerfile.binary

-FROM python:3.10-slim AS compilation
+FROM python:3.10-slim-bookworm AS compilation
 
 RUN apt-get update && apt-get install --no-install-recommends -y build-essential wget
 
 RUN pip install nuitka
 
 ARG TRANSKRIBUS_BRANCH=master
-ARG TRANSKRIBUS_ID=11180199
+ARG TRANSKRIBUS_ID=63
 ARG LICENSE_BRANCH=master
-ARG LICENSE_ID=45943500
-ARG GITLAB_TOKEN="gaFM7LRa9zy9QMowcUhx"
+ARG LICENSE_ID=37
+
+# Auth token expires on 01/07/2024
+ARG GITLAB_TOKEN="glpat-3sBZPFgkZbqJxfSqjcAa"
 
 # We build in /usr/share because Django will try to load some files relative to that path
 # once executed in the binary (management commands, ...)
@@ -22,14 +24,14 @@ ADD requirements.txt /tmp/requirements-arkindex.txt
 # Install transkribus-client from private repo
 RUN \
   mkdir /tmp/transkribus && \
-  wget --header "PRIVATE-TOKEN: $GITLAB_TOKEN" https://gitlab.com/api/v4/projects/$TRANSKRIBUS_ID/repository/archive.tar.gz?sha=$TRANSKRIBUS_BRANCH -O /tmp/transkribus.tar.gz && \
+  wget --header "PRIVATE-TOKEN: $GITLAB_TOKEN" https://gitlab.teklia.com/api/v4/projects/$TRANSKRIBUS_ID/repository/archive.tar.gz?sha=$TRANSKRIBUS_BRANCH -O /tmp/transkribus.tar.gz && \
   tar --strip-components=1 -xvf /tmp/transkribus.tar.gz -C /tmp/transkribus && \
   mv /tmp/transkribus/transkribus /usr/share
 
 # Install teklia-license from private repo
 RUN \
   mkdir /tmp/teklia-license && \
-  wget --header "PRIVATE-TOKEN: $GITLAB_TOKEN" https://gitlab.com/api/v4/projects/$LICENSE_ID/repository/archive.tar.gz?sha=$LICENSE_BRANCH -O /tmp/teklia-license.tar.gz && \
+  wget --header "PRIVATE-TOKEN: $GITLAB_TOKEN" https://gitlab.teklia.com/api/v4/projects/$LICENSE_ID/repository/archive.tar.gz?sha=$LICENSE_BRANCH -O /tmp/teklia-license.tar.gz && \
   tar --strip-components=1 -xvf /tmp/teklia-license.tar.gz -C /tmp/teklia-license && \
   mv /tmp/teklia-license/teklia_license /usr/share && \
   cp /tmp/teklia-license/requirements.txt /tmp/requirements-license-arkindex.txt
@@ -61,7 +63,7 @@ RUN python -m nuitka \
       arkindex/manage.py
 
 # Start over from a clean setup
-FROM registry.gitlab.com/teklia/arkindex/backend/base:lxml-4.9.2 as build
+FROM registry.gitlab.teklia.com/arkindex/backend/base:gitlab-teklia as build
 
 # Import files from compilation
 RUN mkdir /usr/share/arkindex

Makefile

 ROOT_DIR:=$(shell dirname $(realpath $(lastword $(MAKEFILE_LIST))))
-IMAGE_TAG=registry.gitlab.com/teklia/arkindex/backend
+IMAGE_TAG=registry.gitlab.teklia.com/arkindex/backend
 
 .PHONY: all release
 
@@ -32,7 +32,7 @@ test-fixtures:
 test-fixtures-run:
 	arkindex/manage.py migrate
 	arkindex/manage.py build_fixtures
-	arkindex/manage.py dumpdata --indent 4 process documents images users auth ponos > arkindex/documents/fixtures/data.json
+	arkindex/manage.py dumpdata --indent 4 process documents images users auth ponos training > arkindex/documents/fixtures/data.json
 
 test-fixtures-restore:
 	# This first renaming ensures that arkindex_tmp_fixtures exists; we don't want to drop arkindex_dev without a backup

README.md

 Backend for Historical Manuscripts Indexing
 ===========================================
 
-[![pipeline status](https://gitlab.com/teklia/arkindex/backend/badges/master/pipeline.svg)](https://gitlab.com/teklia/arkindex/backend/commits/master)
+[![pipeline status](https://gitlab.teklia.com/arkindex/backend/badges/master/pipeline.svg)](https://gitlab.teklia.com/arkindex/backend/commits/master)
 
 ## Requirements
 
-* Clone of the [architecture](https://gitlab.com/teklia/arkindex/architecture)
+* Clone of the [architecture](https://gitlab.teklia.com/arkindex/architecture)
 * Git
 * Make
 * Python 3.6+
@@ -15,13 +15,13 @@ Backend for Historical Manuscripts Indexing
 ## Dev Setup
 
 ```
-git clone git@gitlab.com:arkindex/backend.git
+git clone git@gitlab.teklia.com:arkindex/backend.git
 cd backend
 mkvirtualenv ark -a .
 pip install -e .[test]
 ```
 
-When the [architecture](https://gitlab.com/teklia/arkindex/architecture) is running locally to provide required services:
+When the [architecture](https://gitlab.teklia.com/arkindex/architecture) is running locally to provide required services:
 
 ```
 arkindex/manage.py migrate
@@ -47,7 +47,7 @@ The line that sets the PDF policy is `<policy domain="coder" rights="none" patte
 
 Arkindex uses OAuth to let a user connect their GitLab account(s) and register Git repositories. In local development, you will need to register Arkindex as a GitLab OAuth application for it to work.
 
-Go to GitLab's [Applications settings](https://gitlab.com/profile/applications) and create a new application with the `api` scope and add the following callback URIs:
+Go to GitLab's [Applications settings](https://gitlab.teklia.com/profile/applications) and create a new application with the `api` scope and add the following callback URIs:
 
 ```
 http://127.0.0.1:8000/api/v1/oauth/providers/gitlab/callback/

VERSION

-1.4.2
+1.5.0

arkindex/documents/api/elements.py

@@ -68,7 +68,7 @@ from arkindex.documents.serializers.light import CorpusAllowedMetaDataSerializer
 from arkindex.documents.serializers.ml import ElementTranscriptionSerializer
 from arkindex.images.models import Image
 from arkindex.ponos.utils import is_admin_or_ponos_task
-from arkindex.process.models import WorkerRun, WorkerVersion
+from arkindex.process.models import WorkerConfiguration, WorkerRun, WorkerVersion
 from arkindex.project.fields import Unnest
 from arkindex.project.mixins import ACLMixin, CorpusACLMixin, SelectionMixin
 from arkindex.project.openapi import UUID_OR_FALSE, AutoSchema
@@ -82,6 +82,7 @@ from arkindex.project.triggers import (
     selection_worker_results_delete,
     worker_results_delete,
 )
+from arkindex.training.models import ModelVersion
 from arkindex.users.models import Role
 from arkindex.users.utils import filter_rights
 
@@ -1165,14 +1166,14 @@ class ElementChildren(ElementsListBase):
         # Let `self.get_filters` handle filtering optimisations
         if self.clean_params.get('order', 'position').lower() == 'position':
             # This condition is necessary because when ordering by position ('paths__ordering')
-            # we run into this bug https://gitlab.com/teklia/arkindex/backend/-/issues/769 and unless
+            # we run into this bug https://gitlab.teklia.com/arkindex/backend/-/issues/769 and unless
             # the ordering is also used in the .distinct(), this leads to some results being
             # "cut off" from the response. For example, if you have 5 distinct elements, but two
             # of these elements have multiple paths with different orderings within the same parent,
             # the DISTINCT will work as expected during the COUNT query, and find 5 elements;
             # however when next the elements are selected, the path ordering is used in the
             # SELECT query which triggers the "duplicating" bug
-            # (https://gitlab.com/teklia/arkindex/backend/-/issues/76).This SELECT would return 7 elements
+            # (https://gitlab.teklia.com/arkindex/backend/-/issues/76).This SELECT would return 7 elements
             # (2 duplicates) but the previous result from the COUNT would mean that only 5 of these
             # 7 elements would get returned, and some elements would therefore (randomly) be missing
             # from the results. Adding the 'paths__ordering' to the .distinct() here forces the
@@ -1263,14 +1264,6 @@ class ElementRetrieve(ACLMixin, RetrieveUpdateDestroyAPIView):
     get=extend_schema(
         operation_id='ListElementNeighbors',
         tags=['elements'],
-        parameters=[
-            OpenApiParameter(
-                'n',
-                type={'type': 'integer', 'minimum': 1, 'maximum': 10},
-                description='Number of neighbors to retrieve around the element',
-                required=False,
-            )
-        ],
     )
 )
 class ElementNeighbors(ACLMixin, ListAPIView):
@@ -1284,16 +1277,9 @@ class ElementNeighbors(ACLMixin, ListAPIView):
     queryset = Element.objects.none()
 
     def get_queryset(self):
-        n = self.request.query_params.get('n', 1)
-        try:
-            n = int(n)
-        except (TypeError, ValueError):
-            raise ValidationError({'n': 'Should be an integer between 1 and 10'})
-        if not 1 <= n <= 10:
-            raise ValidationError({'n': 'Should be an integer between 1 and 10'})
 
         element = get_object_or_404(
-            Element.objects.select_related('corpus'),
+            Element.objects.select_related('corpus').only('id', 'corpus__public'),
             id=self.kwargs['pk']
         )
 
@@ -1301,7 +1287,7 @@ class ElementNeighbors(ACLMixin, ListAPIView):
         if not self.has_access(element.corpus, Role.Guest.value):
             raise PermissionDenied(detail='You do not have a read access to this element.')
 
-        return Element.objects.get_neighbors(element, n)
+        return Element.objects.get_neighbors(element)
 
 
 @extend_schema(tags=['elements'], request=None)
@@ -2194,6 +2180,21 @@ class CorpusSelectionDestroy(CorpusACLMixin, SelectionMixin, DestroyAPIView):
                 description='Only delete Worker Results on selected elements in this corpus. '
                             'Cannot be used together with `element_id`.',
             ),
+            OpenApiParameter(
+                'model_version_id',
+                type=UUID,
+                required=False,
+                description='Only delete Worker Results produced by a specific model version.',
+            ),
+            OpenApiParameter(
+                'configuration_id',
+                type=UUID_OR_FALSE,
+                required=False,
+                description=dedent("""
+                    Only delete Worker Results produced by a specific worker configuration.
+                    If set to false, only delete results that use no specific configuration.
+                """)
+            ),
         ],
         tags=['ml'],
     )
@@ -2246,6 +2247,38 @@ class WorkerResultsDestroy(CorpusACLMixin, DestroyAPIView):
                 except WorkerVersion.DoesNotExist:
                     errors['worker_version_id'].append('This worker version does not exist.')
 
+        model_version = None
+        if 'model_version_id' in self.request.query_params:
+            try:
+                model_version_id = UUID(self.request.query_params['model_version_id'])
+            except (TypeError, ValueError):
+                errors['model_version_id'].append('Invalid UUID.')
+            else:
+                try:
+                    model_version = ModelVersion.objects.get(id=model_version_id)
+                except ModelVersion.DoesNotExist:
+                    errors['model_version_id'].append('This model version does not exist.')
+
+        configuration = None
+        if 'configuration_id' in self.request.query_params:
+            conf_id = self.request.query_params['configuration_id']
+            if conf_id.lower() in ('false', '0'):
+                configuration = False
+            else:
+                try:
+                    conf_id = UUID(conf_id)
+                except (TypeError, ValueError):
+                    errors['configuration_id'].append(
+                        'Invalid UUID. You can set "false" to exclude results with a configuration.'
+                    )
+                else:
+                    try:
+                        configuration = WorkerConfiguration.objects.get(id=conf_id)
+                    except WorkerConfiguration.DoesNotExist:
+                        errors['configuration_id'].append(
+                            'This worker configuration does not exist.'
+                        )
+
         if errors:
             raise ValidationError(errors)
 
@@ -2253,6 +2286,8 @@ class WorkerResultsDestroy(CorpusACLMixin, DestroyAPIView):
             selection_worker_results_delete(
                 corpus=corpus,
                 version=worker_version,
+                model_version=model_version,
+                configuration=configuration,
                 user_id=self.request.user.id,
             )
         else:
@@ -2260,6 +2295,8 @@ class WorkerResultsDestroy(CorpusACLMixin, DestroyAPIView):
                 corpus_id=corpus.id,
                 version=worker_version,
                 element_id=element_id,
+                model_version=model_version,
+                configuration=configuration,
                 user_id=self.request.user.id,
             )
 

arkindex/documents/deletion.py

@@ -70,6 +70,35 @@ def delete_element(element_id: UUID) -> None:
         """, {'id': element_id})
         logger.info(f"Deleted {cursor.rowcount} usage from process as element")
 
+        # Set folders references on training processes to None
+        cursor.execute("""
+        UPDATE process_process
+        SET train_folder_id = NULL
+        WHERE train_folder_id = %(id)s
+        OR train_folder_id IN (
+            SELECT element_id FROM documents_elementpath WHERE path && ARRAY[%(id)s]
+        )
+        """, {'id': element_id})
+        logger.info(f"Deleted {cursor.rowcount} usage from process as train folder")
+        cursor.execute("""
+        UPDATE process_process
+        SET validation_folder_id = NULL
+        WHERE validation_folder_id = %(id)s
+        OR validation_folder_id IN (
+            SELECT element_id FROM documents_elementpath WHERE path && ARRAY[%(id)s]
+        )
+        """, {'id': element_id})
+        logger.info(f"Deleted {cursor.rowcount} usage from process as validation folder")
+        cursor.execute("""
+        UPDATE process_process
+        SET test_folder_id = NULL
+        WHERE test_folder_id = %(id)s
+        OR test_folder_id IN (
+            SELECT element_id FROM documents_elementpath WHERE path && ARRAY[%(id)s]
+        )
+        """, {'id': element_id})
+        logger.info(f"Deleted {cursor.rowcount} usage from process as test folder")
+
         # Remove user selections
         cursor.execute("""
         DELETE FROM documents_selection selection

arkindex/documents/export/structure.sql

 PRAGMA foreign_keys = ON;
 
-CREATE TABLE export_version AS SELECT 6 AS version;
+CREATE TABLE export_version AS SELECT 7 AS version;
 
 CREATE TABLE image_server (
-    id VARCHAR(37) NOT NULL,
+    id INTEGER NOT NULL,
     display_name VARCHAR(250) NOT NULL,
     url TEXT NOT NULL,
     max_width INTEGER,
@@ -18,7 +18,7 @@ CREATE TABLE image (
     url TEXT NOT NULL,
     width INTEGER NOT NULL,
     height INTEGER NOT NULL,
-    server_id VARCHAR(37) NOT NULL,
+    server_id INTEGER NOT NULL,
     PRIMARY KEY (id),
     FOREIGN KEY (server_id) REFERENCES image_server (id) ON DELETE CASCADE,
     UNIQUE (url),
@@ -146,7 +146,7 @@ CREATE TABLE entity_type (
 );
 
 CREATE TABLE transcription_entity (
-    id VARCHAR(37) NOT NULL,
+    id INTEGER NOT NULL,
     transcription_id VARCHAR(37) NOT NULL,
     entity_id VARCHAR(37) NOT NULL,
     offset INTEGER NOT NULL,

arkindex/documents/management/commands/bootstrap.py

@@ -29,7 +29,7 @@ PONOS_FARM_NAME = 'Bootstrap farm'
 PONOS_FARM_SEED = "b12868101dab84984481741663d809d2393784894d6e807ceee0bd95051bf971"
 IMPORT_WORKER_VERSION_ID = 'f2bb8dd7-55e9-49ae-9bd9-b1d2e5d491b9'
 IMPORT_WORKER_SLUG = 'file_import'
-IMPORT_WORKER_REPO = 'https://gitlab.com/teklia/arkindex/tasks'
+IMPORT_WORKER_REPO = 'https://gitlab.teklia.com/arkindex/tasks'
 IMPORT_WORKER_REVISION_MESSAGE = 'File import worker bootstrap'
 IMPORT_WORKER_REVISION_AUTHOR = 'Dev Bootstrap'
 ADMIN_API_TOKEN = "deadbeefTestToken"

arkindex/documents/management/commands/build_fixtures.py

@@ -17,16 +17,11 @@ from arkindex.process.models import (
     WorkerVersion,
     WorkerVersionGPUUsage,
     WorkerVersionState,
-    Workflow,
 )
+from arkindex.project.tools import fake_now
 from arkindex.users.models import Group, Right, Role, User
 
 
-def fake_now():
-    # Use the same creation date for all created objects
-    return datetime(2020, 2, 2, 1, 23, 45, 678000, tzinfo=timezone.utc)
-
-
 def square_polygon(x1, x2):
     return LinearRing((x1, x1), (x2, x1), (x2, x2), (x1, x2), (x1, x1))
 
@@ -65,11 +60,11 @@ class Command(BaseCommand):
         group = Group.objects.create(name="User group", public=False)
         group.add_member(user=user, level=Role.Admin.value)
         group.add_member(
-            user=User.objects.create(email='user2@user.fr', display_name='Test user write'),
+            user=User.objects.create(email='user2@user.fr', display_name='Test user write', verified_email=True),
             level=Role.Contributor.value
         )
         group.add_member(
-            user=User.objects.create(email='user3@user.fr', display_name='Test user read'),
+            user=User.objects.create(email='user3@user.fr', display_name='Test user read', verified_email=True),
             level=Role.Guest.value
         )
 
@@ -109,15 +104,26 @@ class Command(BaseCommand):
         )
 
         # Create worker types
-        dla_worker_type = WorkerType.objects.create(slug="dla")
-        recognizer_worker_type = WorkerType.objects.create(slug="recognizer")
-        gpu_worker_type = WorkerType.objects.create(slug="worker")
+        dla_worker_type = WorkerType.objects.create(slug="dla", display_name="Document Layout Analysis")
+        recognizer_worker_type = WorkerType.objects.create(slug="recognizer", display_name="Recognizer")
+        gpu_worker_type = WorkerType.objects.create(slug="worker", display_name="Worker requiring a GPU")
         import_worker_type = WorkerType.objects.create(slug="import", display_name="Import")
 
         # Create a fake docker build with a docker image task
         farm = Farm.objects.create(name="Wheat farm")
-        workflow = Workflow.objects.create(farm=farm)
-        build_task = workflow.tasks.create(run=0, depth=0, slug='docker_build', state=State.Completed)
+        build_process = Process.objects.create(
+            farm=farm,
+            creator=superuser,
+            mode=ProcessMode.Repository,
+        )
+        build_task = build_process.tasks.create(
+            run=0,
+            depth=0,
+            slug='docker_build',
+            state=State.Completed,
+            # Use an expiry very far away so that task is never expired
+            expiry=datetime(2100, 12, 31, 23, 59, 59, 999999, timezone.utc),
+        )
         docker_image = build_task.artifacts.create(size=42_000, path='/path/to/docker_build')
 
         # Create some workers for the repository with their available version
@@ -273,6 +279,10 @@ class Command(BaseCommand):
             display_name='Word',
         )
 
+        # Create 2 datasets
+        corpus.datasets.create(name='First Dataset', description='dataset number one', creator=user)
+        corpus.datasets.create(name='Second Dataset', description='dataset number two', creator=user)
+
         # Create 2 volumes
         vol1 = Element.objects.create(
             corpus=corpus,

arkindex/documents/management/commands/load_export.py

@@ -41,7 +41,7 @@ from arkindex.process.models import (
 from arkindex.training.models import Model
 from arkindex.users.models import Role, User
 
-EXPORT_VERSION = 6
+EXPORT_VERSION = 7
 
 TABLE_NAMES = {
     'export_version',

arkindex/documents/management/commands/reorder_paths.py

@@ -90,7 +90,7 @@ class Command(BaseCommand):
                 ROW_NUMBER() OVER (
                     PARTITION BY parent_id
                     ORDER BY ordering, name, element_id
-                ) AS ordering
+                ) - 1 AS ordering
             FROM (
                 SELECT
                     DISTINCT ON (element_id, path[array_length(path, 1)])

arkindex/documents/migrations/0003_auto_20200218_1118.py

-# Generated by Django 2.2.9 on 2020-02-18 11:18
-
-from django.db import migrations, models
-
-
-class Migration(migrations.Migration):
-
-    dependencies = [
-        ('documents', '0002_initial'),
-    ]
-
-    operations = [
-        migrations.AlterField(
-            model_name='mlclass',
-            name='name',
-            field=models.CharField(max_length=1024),
-        ),
-    ]