feat: add write_engine parameter to read_FORMATNAME methods to control how data is written to BigQuery

bigframes/dtypes.py

@@ -371,12 +371,19 @@ def arrow_dtype_to_bigframes_dtype(arrow_dtype: pa.DataType) -> Dtype:
         return pd.ArrowDtype(arrow_dtype)
     if pa.types.is_struct(arrow_dtype):
         return pd.ArrowDtype(arrow_dtype)
+
+    # BigFrames doesn't distinguish between string and large_string because the
+    # largest string (2 GB) is already larger than the largest BigQuery row.
+    if pa.types.is_string(arrow_dtype) or pa.types.is_large_string(arrow_dtype):
+        return STRING_DTYPE
+
     if arrow_dtype == pa.null():
         return DEFAULT_DTYPE
-    else:
-        raise ValueError(
-            f"Unexpected Arrow data type {arrow_dtype}. {constants.FEEDBACK_LINK}"
-        )
+
+    # No other types matched.
+    raise ValueError(
+        f"Unexpected Arrow data type {arrow_dtype}. {constants.FEEDBACK_LINK}"
+    )
 
 
 _BIGFRAMES_TO_ARROW = {

bigframes/pandas/io/api.py

@@ -30,6 +30,7 @@
     Union,
 )
 
+import bigframes_vendored.constants as constants
 import bigframes_vendored.pandas.io.gbq as vendored_pandas_gbq
 from google.cloud import bigquery
 import numpy
@@ -103,6 +104,7 @@ def read_csv(
         Literal["c", "python", "pyarrow", "python-fwf", "bigquery"]
     ] = None,
     encoding: Optional[str] = None,
+    write_engine: constants.WriteEngineType = "default",
     **kwargs,
 ) -> bigframes.dataframe.DataFrame:
     return global_session.with_default_session(
@@ -116,6 +118,7 @@ def read_csv(
         dtype=dtype,
         engine=engine,
         encoding=encoding,
+        write_engine=write_engine,
         **kwargs,
     )
 
@@ -133,6 +136,7 @@ def read_json(
     encoding: Optional[str] = None,
     lines: bool = False,
     engine: Literal["ujson", "pyarrow", "bigquery"] = "ujson",
+    write_engine: constants.WriteEngineType = "default",
     **kwargs,
 ) -> bigframes.dataframe.DataFrame:
     return global_session.with_default_session(
@@ -143,6 +147,7 @@ def read_json(
         encoding=encoding,
         lines=lines,
         engine=engine,
+        write_engine=write_engine,
         **kwargs,
     )
 
@@ -243,24 +248,41 @@ def read_gbq_table(
 
 
 @typing.overload
-def read_pandas(pandas_dataframe: pandas.DataFrame) -> bigframes.dataframe.DataFrame:
+def read_pandas(
+    pandas_dataframe: pandas.DataFrame,
+    *,
+    write_engine: constants.WriteEngineType = "default",
+) -> bigframes.dataframe.DataFrame:
     ...
 
 
 @typing.overload
-def read_pandas(pandas_dataframe: pandas.Series) -> bigframes.series.Series:
+def read_pandas(
+    pandas_dataframe: pandas.Series,
+    *,
+    write_engine: constants.WriteEngineType = "default",
+) -> bigframes.series.Series:
     ...
 
 
 @typing.overload
-def read_pandas(pandas_dataframe: pandas.Index) -> bigframes.core.indexes.Index:
+def read_pandas(
+    pandas_dataframe: pandas.Index,
+    *,
+    write_engine: constants.WriteEngineType = "default",
+) -> bigframes.core.indexes.Index:
     ...
 
 
-def read_pandas(pandas_dataframe: Union[pandas.DataFrame, pandas.Series, pandas.Index]):
+def read_pandas(
+    pandas_dataframe: Union[pandas.DataFrame, pandas.Series, pandas.Index],
+    *,
+    write_engine: constants.WriteEngineType = "default",
+):
     return global_session.with_default_session(
         bigframes.session.Session.read_pandas,
         pandas_dataframe,
+        write_engine=write_engine,
     )
 
 
@@ -271,25 +293,32 @@ def read_pickle(
     filepath_or_buffer: FilePath | ReadPickleBuffer,
     compression: CompressionOptions = "infer",
     storage_options: StorageOptions = None,
+    *,
+    write_engine: constants.WriteEngineType = "default",
 ):
     return global_session.with_default_session(
         bigframes.session.Session.read_pickle,
         filepath_or_buffer=filepath_or_buffer,
         compression=compression,
         storage_options=storage_options,
+        write_engine=write_engine,
     )
 
 
 read_pickle.__doc__ = inspect.getdoc(bigframes.session.Session.read_pickle)
 
 
 def read_parquet(
-    path: str | IO["bytes"], *, engine: str = "auto"
+    path: str | IO["bytes"],
+    *,
+    engine: str = "auto",
+    write_engine: constants.WriteEngineType = "default",
 ) -> bigframes.dataframe.DataFrame:
     return global_session.with_default_session(
         bigframes.session.Session.read_parquet,
         path,
         engine=engine,
+        write_engine=write_engine,
     )