Auto-generate label shortkeys on corpus import

app/server/models.py

@@ -142,7 +142,7 @@ class Label(models.Model):
         ('shift', 'shift'),
         ('ctrl shift', 'ctrl shift')
     )
-    SUFFIX_KEYS = (
+    SUFFIX_KEYS = tuple(
         (c, c) for c in string.ascii_lowercase
     )
 

app/server/serializers.py

@@ -34,12 +34,17 @@ class LabelSerializer(serializers.ModelSerializer):
             raise ValidationError('Shortcut key may not have a suffix key.')
 
         # Don't allow to save same shortcut key when prefix_key is null.
-        context = self.context['request'].parser_context
-        project_id = context['kwargs'].get('project_id')
-        if Label.objects.filter(suffix_key=suffix_key,
-                                prefix_key__isnull=True,
-                                project=project_id).exists():
-            raise ValidationError('Duplicate key.')
+        try:
+            context = self.context['request'].parser_context
+            project_id = context['kwargs']['project_id']
+        except (AttributeError, KeyError):
+            pass  # unit tests don't always have the correct context set up
+        else:
+            if Label.objects.filter(suffix_key=suffix_key,
+                                    prefix_key__isnull=True,
+                                    project=project_id).exists():
+                raise ValidationError('Duplicate key.')
+
         return super().validate(attrs)
 
     class Meta:

app/server/tests/data/classification.jsonl

@@ -1,3 +1,4 @@
 {"text": "example", "labels": ["positive"], "meta": {"wikiPageID": 1}}
 {"text": "example", "labels": ["positive", "negative"], "meta": {"wikiPageID": 2}}
 {"text": "example", "labels": ["negative"], "meta": {"wikiPageID": 3}}
+{"text": "example", "labels": ["neutral"], "meta": {"wikiPageID": 4}}

app/server/tests/test_api.py

@@ -747,8 +747,9 @@ class TestUploader(APITestCase):
                                 expected_status=status.HTTP_201_CREATED)
 
         self.label_test_helper(self.classification_labels_url, expected_labels=[
-            {'text': 'positive'},
-            {'text': 'negative'},
+            {'text': 'positive', 'suffix_key': 'p', 'prefix_key': None},
+            {'text': 'negative', 'suffix_key': 'n', 'prefix_key': None},
+            {'text': 'neutral', 'suffix_key': 'n', 'prefix_key': 'ctrl'},
         ])
 
     def test_can_upload_labeling_jsonl(self):
@@ -758,9 +759,9 @@ class TestUploader(APITestCase):
                                 expected_status=status.HTTP_201_CREATED)
 
         self.label_test_helper(self.labeling_labels_url, expected_labels=[
-            {'text': 'LOC'},
-            {'text': 'ORG'},
-            {'text': 'PER'},
+            {'text': 'LOC', 'suffix_key': 'l', 'prefix_key': None},
+            {'text': 'ORG', 'suffix_key': 'o', 'prefix_key': None},
+            {'text': 'PER', 'suffix_key': 'p', 'prefix_key': None},
         ])
 
     def test_can_upload_seq2seq_jsonl(self):

app/server/utils.py

@@ -74,16 +74,60 @@ class BaseStorage(object):
         """
         return [label for label in labels if label not in created]
 
-    def to_serializer_format(self, labels):
-        """Exclude created labels.
+    @classmethod
+    def to_serializer_format(cls, labels, created):
+        """Convert a label to model dictionary.
+
+        Also assigns shortkeys for each label that don't clash with existing
+        label shortkeys.
 
         Example:
             >>> labels = ["positive"]
-            >>> self.to_serializer_format(labels)
-            [{"text": "negative"}]
-        ```
+            >>> created = {}
+            >>> BaseStorage.to_serializer_format(labels, created)
+            [{"text": "positive", "suffix_key": "p", "prefix_key": None}]
+        """
+        existing_shortkeys = {(label.suffix_key, label.prefix_key)
+                              for label in created.values()}
+
+        serializer_labels = []
+
+        for label in sorted(labels):
+            serializer_label = {'text': label}
+
+            shortkey = cls.get_shortkey(label, existing_shortkeys)
+            if shortkey:
+                serializer_label['suffix_key'] = shortkey[0]
+                serializer_label['prefix_key'] = shortkey[1]
+                existing_shortkeys.add(shortkey)
+
+            serializer_labels.append(serializer_label)
+
+        return serializer_labels
+
+    @classmethod
+    def get_shortkey(cls, label, existing_shortkeys):
+        """Find the first non existing shortkey for the label.
+
+        Example without existing shortkey:
+            >>> BaseStorage.get_shortkey("positive", set())
+            ("p", None)
+
+        Example with existing shortkey:
+            >>> BaseStorage.get_shortkey("positive", {("p", None)})
+            ("p", "ctrl")
         """
-        return [{'text': label} for label in labels]
+        model_prefix_keys = [key for (key, _) in Label.PREFIX_KEYS]
+        prefix_keys = [None] + model_prefix_keys
+
+        model_suffix_keys = {key for (key, _) in Label.SUFFIX_KEYS}
+        suffix_keys = [key for key in label.lower() if key in model_suffix_keys]
+
+        for shortkey in itertools.product(suffix_keys, prefix_keys):
+            if shortkey not in existing_shortkeys:
+                return shortkey
+
+        return None
 
     def update_saved_labels(self, saved, new):
         """Update saved labels.
@@ -120,7 +164,7 @@ class ClassificationStorage(BaseStorage):
             labels = self.extract_label(data)
             unique_labels = self.extract_unique_labels(labels)
             unique_labels = self.exclude_created_labels(unique_labels, saved_labels)
-            unique_labels = self.to_serializer_format(unique_labels)
+            unique_labels = self.to_serializer_format(unique_labels, saved_labels)
             new_labels = self.save_label(unique_labels)
             saved_labels = self.update_saved_labels(saved_labels, new_labels)
             annotations = self.make_annotations(docs, labels, saved_labels)
@@ -170,7 +214,7 @@ class SequenceLabelingStorage(BaseStorage):
             labels = self.extract_label(data)
             unique_labels = self.extract_unique_labels(labels)
             unique_labels = self.exclude_created_labels(unique_labels, saved_labels)
-            unique_labels = self.to_serializer_format(unique_labels)
+            unique_labels = self.to_serializer_format(unique_labels, saved_labels)
             new_labels = self.save_label(unique_labels)
             saved_labels = self.update_saved_labels(saved_labels, new_labels)
             annotations = self.make_annotations(docs, labels, saved_labels)