Merge remote-tracking branch 'origin/main' into tswast-polars-session-names

Author: tswast

bigframes/operations/blob.py

@@ -228,9 +228,14 @@ def display(
         df._set_internal_query_job(query_job)
 
         def display_single_url(
-            read_url: str, content_type: Union[str, pd._libs.missing.NAType]
+            read_url: Union[str, pd._libs.missing.NAType],
+            content_type: Union[str, pd._libs.missing.NAType],
         ):
-            if content_type is pd.NA:  # display as raw data or error
+            if pd.isna(read_url):
+                ipy_display.display("<NA>")
+                return
+
+            if pd.isna(content_type):  # display as raw data or error
                 response = requests.get(read_url)
                 ipy_display.display(response.content)
                 return

noxfile.py

@@ -124,6 +124,10 @@
 # Sessions are executed in the order so putting the smaller sessions
 # ahead to fail fast at presubmit running.
 nox.options.sessions = [
+    # Include unit_noextras to ensure at least some unit tests contribute to
+    # coverage.
+    # TODO(tswast): Consider removing this when unit_noextras and cover is run
+    # from GitHub actions.
     "unit_noextras",
     "system-3.9",  # No extras.
     f"system-{LATEST_FULLY_SUPPORTED_PYTHON}",  # All extras.

tests/system/large/blob/test_function.py

@@ -12,6 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+import logging
 import os
 import traceback
 from typing import Generator
@@ -434,6 +435,15 @@ def test_blob_transcribe(
         actual_text = actual[0]["content"]
     else:
         actual_text = actual[0]
+
+    if pd.isna(actual_text) or actual_text == "":
+        # Ensure the tests are robust to flakes in the model, which isn't
+        # particularly useful information for the bigframes team.
+        logging.warning(
+            f"blob_transcribe() model {model_name} verbose={verbose} failure"
+        )
+        return
+
     actual_len = len(actual_text)
 
     relative_length_tolerance = 0.2

tests/system/small/blob/test_io.py

@@ -12,6 +12,9 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+from unittest import mock
+
+import IPython.display
 import pandas as pd
 
 import bigframes
@@ -92,3 +95,33 @@ def test_blob_create_read_gbq_object_table(
     pd.testing.assert_frame_equal(
         pd_blob_df, expected_df, check_dtype=False, check_index_type=False
     )
+
+
+def test_display_images(monkeypatch, images_mm_df: bpd.DataFrame):
+    mock_display = mock.Mock()
+    monkeypatch.setattr(IPython.display, "display", mock_display)
+
+    images_mm_df["blob_col"].blob.display()
+
+    for call in mock_display.call_args_list:
+        args, _ = call
+        arg = args[0]
+        assert isinstance(arg, IPython.display.Image)
+
+
+def test_display_nulls(
+    monkeypatch,
+    bq_connection: str,
+    session: bigframes.Session,
+):
+    uri_series = bpd.Series([None, None, None], dtype="string", session=session)
+    blob_series = uri_series.str.to_blob(connection=bq_connection)
+    mock_display = mock.Mock()
+    monkeypatch.setattr(IPython.display, "display", mock_display)
+
+    blob_series.blob.display()
+
+    for call in mock_display.call_args_list:
+        args, _ = call
+        arg = args[0]
+        assert arg == "<NA>"