Implement Dataset add_item

src/datasets/arrow_dataset.py

@@ -2368,6 +2368,10 @@ def add_elasticsearch_index(
             )
         return self
 
+    def add_item(self, item):
+        table = pa.Table.from_pydict({k: [v] for k, v in item.items()})
+        self._data = pa.concat_tables([self._data, table]) if self._data.shape != (0, 0) else table
+
 
 def concatenate_datasets(
     dsets: List[Dataset],

tests/test_arrow_dataset.py

@@ -1666,3 +1666,11 @@ def test_dataset_from_file(in_memory, dataset, arrow_file):
     assert dataset_from_file.features == dataset.features
     assert dataset_from_file.cache_files == ([{"filename": filename}] if not in_memory else [])
     assert increased_allocated_memory == in_memory
+
+
+def test_dataset_add_item():
+    item = {"input_ids": np.array([4, 4, 2])}
+    ds = Dataset(pa.Table.from_pydict({}))
+    ds.add_item(item)
+    assert ds.data.shape == (1, 1)
+    assert ds.data.column_names == ["input_ids"]