fix: error for object dtype on read_pandas (#570)

Author: chelsea-lin

bigframes/session/__init__.py

@@ -1050,7 +1050,12 @@ def _read_pandas(
  inline_df = self._read_pandas_inline(pandas_dataframe)
  if inline_df is not None:
  return inline_df
- return self._read_pandas_load_job(pandas_dataframe, api_name)
+ try:
+ return self._read_pandas_load_job(pandas_dataframe, api_name)
+ except pa.ArrowInvalid as e:
+ raise pa.ArrowInvalid(
+ f"Could not convert with a BigQuery type: `{e}`. "
+ ) from e
 
  def _read_pandas_inline(
  self, pandas_dataframe: pandas.DataFrame
@@ -1064,6 +1069,10 @@ def _read_pandas_inline(
  inline_df = dataframe.DataFrame(
  blocks.Block.from_local(pandas_dataframe, self)
  )
+ except pa.ArrowInvalid as e:
+ raise pa.ArrowInvalid(
+ f"Could not convert with a BigQuery type: `{e}`. "
+ ) from e
  except ValueError: # Thrown by ibis for some unhandled types
  return None
  except pa.ArrowTypeError: # Thrown by arrow for types without mapping (geo).

tests/system/small/test_session.py

@@ -24,6 +24,7 @@
 import google.cloud.bigquery as bigquery
 import numpy as np
 import pandas as pd
+import pyarrow as pa
 import pytest
 
 import bigframes
@@ -436,6 +437,11 @@ def test_read_pandas_index(session):
  pd.testing.assert_index_equal(bf_idx.to_pandas(), pd_idx)
 
 
+def test_read_pandas_w_unsupported_mixed_dtype(session):
+ with pytest.raises(pa.ArrowInvalid, match="Could not convert"):
+ session.read_pandas(pd.DataFrame({"a": [1, "hello"]}))
+
+
 def test_read_pandas_inline_respects_location():
  options = bigframes.BigQueryOptions(location="europe-west1")
  session = bigframes.Session(options)