Merge branch 'releases/0.x' into bugfixes/resolve-1551

environment_dev.yml

@@ -32,7 +32,7 @@ dependencies:
       - pre-commit==2.15.0
       # extra test dependencies
       - cleanlab
-      - datasets>1.17.0,<2.3.0 # TODO: Remove this when 2.3.0 compatibility is resolved
+      - datasets>1.17.0,<2.3.0  # TODO: push_to_hub fails up to 2.3.2, check patches when they come out eventually
       - huggingface_hub != 0.5.0 # some backward comp. problems introduced in 0.5.0
       - https://github.com/explosion/spacy-models/releases/download/en_core_web_sm-3.1.0/en_core_web_sm-3.1.0.tar.gz
       - flair==0.10

frontend/components/text2text/results/RecordText2Text.vue

@@ -139,6 +139,7 @@ export default {
     &:hover {
       ::v-deep .button-primary--outline {
         opacity: 1 !important;
+        pointer-events: all;
         transition: opacity 0.5s ease-in-out 0.2s !important;
       }
     }

frontend/plugins/highlight-search.js

@@ -53,7 +53,7 @@ export default (context, inject) => {
     );
     text = htmlText(text);
     sortedKeywords.forEach((keyword) => {
-      const regex = new RegExp(`\\b${keyword}\\b`, "gmi");
+      const regex = new RegExp(`([^a-zA-ZÀ-ÿ\u00f1\u00d1]|^)${keyword}`, "gmi");
       text = text.replace(regex, (match) => htmlHighlightText(match));
     });
 
@@ -62,7 +62,7 @@ export default (context, inject) => {
 
   const keywordsSpans = function (text, keywords) {
     return (keywords || []).flatMap((keyword) => {
-      const regex = new RegExp(`\\b${keyword}\\b`, "gmi");
+      const regex = new RegExp(`([^a-zA-ZÀ-ÿ\u00f1\u00d1]|^)${keyword}`, "gmi");
       return [...text.matchAll(regex)].map((match) => {
         return {
           start: match.index,

src/rubrix/client/datasets.py

@@ -14,7 +14,7 @@
 #  limitations under the License.
 import functools
 import logging
-from typing import Dict, List, Optional, Tuple, Type, Union
+from typing import Any, Dict, List, Optional, Tuple, Type, Union
 
 import pandas as pd
 from pkg_resources import parse_version
@@ -516,6 +516,11 @@ def _from_datasets(
                 row["inputs"] = {
                     key: val for key, val in row["inputs"].items() if val is not None
                 }
+            if row.get("annotation") is not None:
+                row["annotation"] = cls._parse_annotation_field2(
+                    row["annotation"], dataset.features["annotation"]
+                )
+
             if row.get("prediction"):
                 row["prediction"] = (
                     [
@@ -544,6 +549,40 @@ def _from_datasets(
             records.append(TextClassificationRecord.parse_obj(row))
         return cls(records)
 
+    @staticmethod
+    def _parse_annotation_field2(
+        annotation: Union[str, List[str], int, List[int]],
+        feature: Optional[Any],
+    ) -> Optional[Union[str, List[str], int, List[int]]]:
+        """Helper function to parse the annotation field.
+
+        Args:
+            annotation: The value from the annotation column.
+            feature: The feature of the annotation column to optionally convert ints to strs.
+
+        Returns:
+            The input value for the annotation field.
+        """
+        import datasets
+
+        # extract ClassLabel feature
+        if isinstance(feature, list):
+            feature = feature[0]
+        if isinstance(feature, datasets.Sequence):
+            feature = feature.feature
+        if not isinstance(feature, datasets.ClassLabel):
+            feature = None
+
+        if feature is None:
+            return annotation
+
+        try:
+            return feature.int2str(annotation)
+        # integers don't have to map to the names ...
+        # it seems that sometimes -1 is used to denote "no label"
+        except ValueError:
+            return None
+
     @classmethod
     def _parse_inputs_field(
         cls, dataset: "datasets.Dataset", fields: Optional[Union[str, List[str]]]