def dataframe_to_parquet(dataframe,
bq_schema,
filepath,
parquet_compression="SNAPPY"):
"""Write dataframe as a Parquet file, according to the desired BQ schema.
This function requires the :mod:`pyarrow` package. Arrow is used as an
intermediate format.
Args:
dataframe (pandas.DataFrame):
DataFrame to convert to Parquet file.
bq_schema (Sequence[Union[ \
:class:`~google.cloud.bigquery.schema.SchemaField`, \
Mapping[str, Any] \
]]):
Desired BigQuery schema. Number of columns must match number of
columns in the DataFrame.
filepath (str):
Path to write Parquet file to.
parquet_compression (Optional[str]):
The compression codec to use by the the ``pyarrow.parquet.write_table``
serializing method. Defaults to "SNAPPY".
https://arrow.apache.org/docs/python/generated/pyarrow.parquet.write_table.html#pyarrow-parquet-write-table
"""
if pyarrow is None:
raise ValueError("pyarrow is required for BigQuery schema conversion.")
bq_schema = schema._to_schema_fields(bq_schema)
arrow_table = dataframe_to_arrow(dataframe, bq_schema)
pyarrow.parquet.write_table(arrow_table,
filepath,
compression=parquet_compression)
def dataframe_to_arrow(dataframe, bq_schema):
"""Convert pandas dataframe to Arrow table, using BigQuery schema.
Args:
dataframe (pandas.DataFrame):
DataFrame to convert to Arrow table.
bq_schema (Sequence[Union[ \
:class:`~google.cloud.bigquery.schema.SchemaField`, \
Mapping[str, Any] \
]]):
Desired BigQuery schema. The number of columns must match the
number of columns in the DataFrame.
Returns:
pyarrow.Table:
Table containing dataframe data, with schema derived from
BigQuery schema.
"""
column_names = set(dataframe.columns)
column_and_index_names = set(
name for name, _ in list_columns_and_indexes(dataframe)
)
bq_schema = schema._to_schema_fields(bq_schema)
bq_field_names = set(field.name for field in bq_schema)
extra_fields = bq_field_names - column_and_index_names
if extra_fields:
raise ValueError(
u"bq_schema contains fields not present in dataframe: {}".format(
extra_fields
)
)
# It's okay for indexes to be missing from bq_schema, but it's not okay to
# be missing columns.
missing_fields = column_names - bq_field_names
if missing_fields:
raise ValueError(
u"bq_schema is missing fields from dataframe: {}".format(missing_fields)
)
arrow_arrays = []
arrow_names = []
arrow_fields = []
for bq_field in bq_schema:
arrow_fields.append(bq_to_arrow_field(bq_field))
arrow_names.append(bq_field.name)
arrow_arrays.append(
bq_to_arrow_array(get_column_or_index(dataframe, bq_field.name), bq_field)
)
if all((field is not None for field in arrow_fields)):
return pyarrow.Table.from_arrays(
arrow_arrays, schema=pyarrow.schema(arrow_fields)
)
return pyarrow.Table.from_arrays(arrow_arrays, names=arrow_names)
def schema(self, value):
if value is None:
self._del_sub_prop("schema")
return
value = _to_schema_fields(value)
_helpers._set_sub_prop(
self._properties,
["load", "schema", "fields"],
[field.to_api_repr() for field in value],
)
def dataframe_to_parquet(
dataframe,
bq_schema,
filepath,
parquet_compression="SNAPPY",
parquet_use_compliant_nested_type=True,
):
"""Write dataframe as a Parquet file, according to the desired BQ schema.
This function requires the :mod:`pyarrow` package. Arrow is used as an
intermediate format.
Args:
dataframe (pandas.DataFrame):
DataFrame to convert to Parquet file.
bq_schema (Sequence[Union[ \
:class:`~google.cloud.bigquery.schema.SchemaField`, \
Mapping[str, Any] \
]]):
Desired BigQuery schema. Number of columns must match number of
columns in the DataFrame.
filepath (str):
Path to write Parquet file to.
parquet_compression (Optional[str]):
The compression codec to use by the the ``pyarrow.parquet.write_table``
serializing method. Defaults to "SNAPPY".
https://arrow.apache.org/docs/python/generated/pyarrow.parquet.write_table.html#pyarrow-parquet-write-table
parquet_use_compliant_nested_type (bool):
Whether the ``pyarrow.parquet.write_table`` serializing method should write
compliant Parquet nested type (lists). Defaults to ``True``.
https://github.com/apache/parquet-format/blob/master/LogicalTypes.md#nested-types
https://arrow.apache.org/docs/python/generated/pyarrow.parquet.write_table.html#pyarrow-parquet-write-table
This argument is ignored for ``pyarrow`` versions earlier than ``4.0.0``.
"""
pyarrow = _helpers.PYARROW_VERSIONS.try_import(raise_if_error=True)
import pyarrow.parquet
kwargs = ({
"use_compliant_nested_type": parquet_use_compliant_nested_type
} if _helpers.PYARROW_VERSIONS.use_compliant_nested_type else {})
bq_schema = schema._to_schema_fields(bq_schema)
arrow_table = dataframe_to_arrow(dataframe, bq_schema)
pyarrow.parquet.write_table(
arrow_table,
filepath,
compression=parquet_compression,
**kwargs,
)
def download_arrow_row_iterator(pages, bq_schema):
"""Use HTTP JSON RowIterator to construct an iterable of RecordBatches.
Args:
pages (Iterator[:class:`google.api_core.page_iterator.Page`]):
An iterator over the result pages.
bq_schema (Sequence[Union[ \
:class:`~google.cloud.bigquery.schema.SchemaField`, \
Mapping[str, Any] \
]]):
A decription of the fields in result pages.
Yields:
:class:`pyarrow.RecordBatch`
The next page of records as a ``pyarrow`` record batch.
"""
bq_schema = schema._to_schema_fields(bq_schema)
column_names = bq_to_arrow_schema(bq_schema) or [field.name for field in bq_schema]
arrow_types = [bq_to_arrow_data_type(field) for field in bq_schema]
for page in pages:
yield _row_iterator_page_to_arrow(page, column_names, arrow_types)
def download_dataframe_row_iterator(pages, bq_schema, dtypes):
"""Use HTTP JSON RowIterator to construct a DataFrame.
Args:
pages (Iterator[:class:`google.api_core.page_iterator.Page`]):
An iterator over the result pages.
bq_schema (Sequence[Union[ \
:class:`~google.cloud.bigquery.schema.SchemaField`, \
Mapping[str, Any] \
]]):
A decription of the fields in result pages.
dtypes(Mapping[str, numpy.dtype]):
The types of columns in result data to hint construction of the
resulting DataFrame. Not all column types have to be specified.
Yields:
:class:`pandas.DataFrame`
The next page of records as a ``pandas.DataFrame`` record batch.
"""
bq_schema = schema._to_schema_fields(bq_schema)
column_names = [field.name for field in bq_schema]
for page in pages:
yield _row_iterator_page_to_dataframe(page, column_names, dtypes)
def _rows_from_json(values, schema):
"""Convert JSON row data to rows with appropriate types.
Args:
values (Sequence[Dict]): The list of responses (JSON rows) to convert.
schema (Sequence[Union[ \
:class:`~google.cloud.bigquery.schema.SchemaField`, \
Mapping[str, Any] \
]]):
The table's schema. If any item is a mapping, its content must be
compatible with
:meth:`~google.cloud.bigquery.schema.SchemaField.from_api_repr`.
Returns:
List[:class:`~google.cloud.bigquery.Row`]
"""
from google.cloud.bigquery import Row
from google.cloud.bigquery.schema import _to_schema_fields
schema = _to_schema_fields(schema)
field_to_index = _field_to_index_mapping(schema)
return [Row(_row_tuple_from_json(r, schema), field_to_index) for r in values]
def _row_tuple_from_json(row, schema):
"""Convert JSON row data to row with appropriate types.
Note: ``row['f']`` and ``schema`` are presumed to be of the same length.
Args:
row (Dict): A JSON response row to be converted.
schema (Sequence[Union[ \
:class:`~google.cloud.bigquery.schema.SchemaField`, \
Mapping[str, Any] \
]]): Specification of the field types in ``row``.
Returns:
Tuple: A tuple of data converted to native types.
"""
from google.cloud.bigquery.schema import _to_schema_fields
schema = _to_schema_fields(schema)
row_data = []
for field, cell in zip(schema, row["f"]):
row_data.append(_field_from_json(cell["v"], field))
return tuple(row_data)
def dataframe_to_bq_schema(dataframe, bq_schema):
"""Convert a pandas DataFrame schema to a BigQuery schema.
Args:
dataframe (pandas.DataFrame):
DataFrame for which the client determines the BigQuery schema.
bq_schema (Sequence[Union[ \
:class:`~google.cloud.bigquery.schema.SchemaField`, \
Mapping[str, Any] \
]]):
A BigQuery schema. Use this argument to override the autodetected
type for some or all of the DataFrame columns.
Returns:
Optional[Sequence[google.cloud.bigquery.schema.SchemaField]]:
The automatically determined schema. Returns None if the type of
any column cannot be determined.
"""
if bq_schema:
bq_schema = schema._to_schema_fields(bq_schema)
bq_schema_index = {field.name: field for field in bq_schema}
bq_schema_unused = set(bq_schema_index.keys())
else:
bq_schema_index = {}
bq_schema_unused = set()
bq_schema_out = []
unknown_type_fields = []
for column, dtype in list_columns_and_indexes(dataframe):
# Use provided type from schema, if present.
bq_field = bq_schema_index.get(column)
if bq_field:
bq_schema_out.append(bq_field)
bq_schema_unused.discard(bq_field.name)
continue
# Otherwise, try to automatically determine the type based on the
# pandas dtype.
bq_type = _PANDAS_DTYPE_TO_BQ.get(dtype.name)
bq_field = schema.SchemaField(column, bq_type)
bq_schema_out.append(bq_field)
if bq_field.field_type is None:
unknown_type_fields.append(bq_field)
# Catch any schema mismatch. The developer explicitly asked to serialize a
# column, but it was not found.
if bq_schema_unused:
raise ValueError(
u"bq_schema contains fields not present in dataframe: {}".format(
bq_schema_unused))
# If schema detection was not successful for all columns, also try with
# pyarrow, if available.
if unknown_type_fields:
if not pyarrow:
msg = u"Could not determine the type of columns: {}".format(
", ".join(field.name for field in unknown_type_fields))
warnings.warn(msg)
return None # We cannot detect the schema in full.
# The augment_schema() helper itself will also issue unknown type
# warnings if detection still fails for any of the fields.
bq_schema_out = augment_schema(dataframe, bq_schema_out)
return tuple(bq_schema_out) if bq_schema_out else None
def _call_fut(schema):
from google.cloud.bigquery.schema import _to_schema_fields
return _to_schema_fields(schema)
