ValueError: Must pass schema, or at least one RecordBatch in BigQuery

[grieve54706]

We found the error when we query the table with INTERVAL or JSON column and the table is empty.

if record_batches and bqstorage_client is not None:
    return pyarrow.Table.from_batches(record_batches)
else:
    # No records (not record_batches), use schema based on BigQuery schema
    # **or**
    # we used the REST API (bqstorage_client is None),
    # which doesn't add arrow extension metadata, so we let
    # `bq_to_arrow_schema` do it.
    arrow_schema = _pandas_helpers.bq_to_arrow_schema(self._schema)
    return pyarrow.Table.from_batches(record_batches, schema=arrow_schema)

https://github.com/googleapis/python-bigquery/blob/main/google/cloud/bigquery/table.py#L1965-L1974

When the record_batches is None, the BigQuery client will get the arrow_schema. But the types INTERVAL and JSON are not in the mapping.

_BQ_TO_ARROW_SCALARS = {
    "BOOL": pyarrow.bool_,
    "BOOLEAN": pyarrow.bool_,
    "BYTES": pyarrow.binary,
    "DATE": pyarrow.date32,
    "DATETIME": pyarrow_datetime,
    "FLOAT": pyarrow.float64,
    "FLOAT64": pyarrow.float64,
    "GEOGRAPHY": pyarrow.string,
    "INT64": pyarrow.int64,
    "INTEGER": pyarrow.int64,
    "NUMERIC": pyarrow_numeric,
    "STRING": pyarrow.string,
    "TIME": pyarrow_time,
    "TIMESTAMP": pyarrow_timestamp,
}

https://github.com/googleapis/python-bigquery/blob/106161180ead01aca1ead909cf06ca559f68666d/google/cloud/bigquery/_pyarrow_helpers.py#L56-L71

They don't want to handle these. We can see these pending or closed issues and PRs.
googleapis/python-bigquery#1580
googleapis/python-bigquery#1832
googleapis/python-bigquery#826