maxdebayser commented on code in PR #7831: URL: https://github.com/apache/iceberg/pull/7831#discussion_r1258846345
########## python/pyiceberg/utils/file_stats.py: ########## @@ -0,0 +1,333 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. + +import struct +from typing import ( + Any, + Dict, + List, + Union, +) + +import pyarrow.lib +import pyarrow.parquet as pq + +from pyiceberg.manifest import DataFile, FileFormat +from pyiceberg.schema import Schema, SchemaVisitor, visit +from pyiceberg.types import ( + IcebergType, + ListType, + MapType, + NestedField, + PrimitiveType, + StructType, +) + +BOUND_TRUNCATED_LENGHT = 16 + +# Serialization rules: https://iceberg.apache.org/spec/#binary-single-value-serialization +# +# Type Binary serialization +# boolean 0x00 for false, non-zero byte for true +# int Stored as 4-byte little-endian +# long Stored as 8-byte little-endian +# float Stored as 4-byte little-endian +# double Stored as 8-byte little-endian +# date Stores days from the 1970-01-01 in an 4-byte little-endian int +# time Stores microseconds from midnight in an 8-byte little-endian long +# timestamp without zone Stores microseconds from 1970-01-01 00:00:00.000000 in an 8-byte little-endian long +# timestamp with zone Stores microseconds from 1970-01-01 00:00:00.000000 UTC in an 8-byte little-endian long +# string UTF-8 bytes (without length) +# uuid 16-byte big-endian value, see example in Appendix B +# fixed(L) Binary value +# binary Binary value (without length) +# + + +def bool_to_avro(value: bool) -> bytes: + return struct.pack("?", value) + + +def int32_to_avro(value: int) -> bytes: + return struct.pack("<i", value) + + +def int64_to_avro(value: int) -> bytes: + return struct.pack("<q", value) + + +def float_to_avro(value: float) -> bytes: + return struct.pack("<f", value) + + +def double_to_avro(value: float) -> bytes: + return struct.pack("<d", value) + + +def bytes_to_avro(value: Union[bytes, str]) -> bytes: + if type(value) == str: + return value.encode() + else: + assert isinstance(value, bytes) # appeases mypy + return value + + +class StatsAggregator: + def __init__(self, type_string: str): + self.current_min: Any = None + self.current_max: Any = None + self.serialize: Any = None + + if type_string == "BOOLEAN": + self.serialize = bool_to_avro + elif type_string == "INT32": + self.serialize = int32_to_avro + elif type_string == "INT64": + self.serialize = int64_to_avro + elif type_string == "INT96": + raise NotImplementedError("Statistics not implemented for INT96 physical type") + elif type_string == "FLOAT": + self.serialize = float_to_avro + elif type_string == "DOUBLE": + self.serialize = double_to_avro + elif type_string == "BYTE_ARRAY": + self.serialize = bytes_to_avro + elif type_string == "FIXED_LEN_BYTE_ARRAY": + self.serialize = bytes_to_avro + else: + raise AssertionError(f"Unknown physical type {type_string}") + + def add_min(self, val: bytes) -> None: + if not self.current_min: + self.current_min = val + elif val < self.current_min: + self.current_min = val + + def add_max(self, val: bytes) -> None: + if not self.current_max: + self.current_max = val + elif self.current_max < val: + self.current_max = val + + def get_min(self) -> bytes: + return self.serialize(self.current_min)[:BOUND_TRUNCATED_LENGHT] + + def get_max(self) -> bytes: + return self.serialize(self.current_max)[:BOUND_TRUNCATED_LENGHT] + + +def fill_parquet_file_metadata( + df: DataFile, metadata: pq.FileMetaData, col_path_2_iceberg_id: Dict[str, int], file_size: int +) -> None: + """ + Computes and fills the following fields of the DataFile object. + + - file_format + - record_count + - file_size_in_bytes + - column_sizes + - value_counts + - null_value_counts + - nan_value_counts + - lower_bounds + - upper_bounds + - split_offsets + + Args: + df (DataFile): A DataFile object representing the Parquet file for which metadata is to be filled. + metadata (pyarrow.parquet.FileMetaData): A pyarrow metadata object. + col_path_2_iceberg_id: A mapping of column paths as in the `path_in_schema` attribute of the colum + metadata to iceberg schema IDs. For scalar columns this will be the column name. For complex types + it could be something like `my_map.key_value.value` + file_size (int): The total compressed file size cannot be retrieved from the metadata and hence has to + be passed here. Depending on the kind of file system and pyarrow library call used, different + ways to obtain this value might be appropriate. + """ + col_index_2_id = {} + + col_names = set(metadata.schema.names) + + first_group = metadata.row_group(0) + + for c in range(metadata.num_columns): + column = first_group.column(c) + col_path = column.path_in_schema + + if col_path in col_path_2_iceberg_id: + col_index_2_id[c] = col_path_2_iceberg_id[col_path] + else: + raise AssertionError(f"Column path {col_path} couldn't be mapped to an iceberg ID") + + column_sizes: Dict[int, int] = {} + value_counts: Dict[int, int] = {} + split_offsets: List[int] = [] + + null_value_counts: Dict[int, int] = {} + nan_value_counts: Dict[int, int] = {} + + col_aggs = {} + + for r in range(metadata.num_row_groups): + # References: + # https://github.com/apache/iceberg/blob/fc381a81a1fdb8f51a0637ca27cd30673bd7aad3/parquet/src/main/java/org/apache/iceberg/parquet/ParquetUtil.java#L232 + # https://github.com/apache/parquet-mr/blob/ac29db4611f86a07cc6877b416aa4b183e09b353/parquet-hadoop/src/main/java/org/apache/parquet/hadoop/metadata/ColumnChunkMetaData.java#L184 + + row_group = metadata.row_group(r) + + data_offset = row_group.column(0).data_page_offset + dictionary_offset = row_group.column(0).dictionary_page_offset + + if row_group.column(0).has_dictionary_page and dictionary_offset < data_offset: + split_offsets.append(dictionary_offset) + else: + split_offsets.append(data_offset) + + for c in range(metadata.num_columns): + col_id = col_index_2_id[c] + + column = row_group.column(c) + + column_sizes[col_id] = column_sizes.get(col_id, 0) + column.total_compressed_size + value_counts[col_id] = value_counts.get(col_id, 0) + column.num_values + + if column.is_stats_set: + try: + statistics = column.statistics + + null_value_counts[col_id] = null_value_counts.get(col_id, 0) + statistics.null_count + + if column.path_in_schema in col_names: + # Iceberg seems to only have statistics for scalar columns + + if col_id not in col_aggs: + col_aggs[col_id] = StatsAggregator(statistics.physical_type) + + col_aggs[col_id].add_min(statistics.min) Review Comment: It's because there are 3 different concerns here: - dealing with the parquet type that leaks through the arrow API: https://github.com/apache/arrow/blob/d676078c13a02ad920eeea2acd5fa517f14526e2/cpp/src/parquet/parquet.thrift#L34 - dealing with with the metrics mode (full or truncate) - actually computing min and max. I think these should stay out of the inner loop to keep it readable. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
