| 123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236 |
- """encoding.py - methods for reading parquet encoded data blocks."""
- from __future__ import absolute_import
- from __future__ import division
- from __future__ import print_function
- from __future__ import unicode_literals
- import array
- import io
- import logging
- import math
- import os
- import struct
- import sys
- import thriftpy
- THRIFT_FILE = os.path.join(os.path.dirname(__file__), "parquet.thrift")
- parquet_thrift = thriftpy.load(THRIFT_FILE, module_name=str("parquet_thrift")) # pylint: disable=invalid-name
- logger = logging.getLogger("parquet") # pylint: disable=invalid-name
- PY3 = sys.version_info.major > 2
- ARRAY_BYTE_STR = u'B' if PY3 else b'B'
- def read_plain_boolean(file_obj, count):
- """Read `count` booleans using the plain encoding."""
- # for bit packed, the count is stored shifted up. But we want to pass in a count,
- # so we shift up.
- # bit width is 1 for a single-bit boolean.
- return read_bitpacked(file_obj, count << 1, 1, logger.isEnabledFor(logging.DEBUG))
- def read_plain_int32(file_obj, count):
- """Read `count` 32-bit ints using the plain encoding."""
- length = 4 * count
- data = file_obj.read(length)
- if len(data) != length:
- raise EOFError("Expected {0} bytes but got {1} bytes".format(length, len(data)))
- res = struct.unpack(b"<{0}i".format(count).encode("utf-8"), data)
- return res
- def read_plain_int64(file_obj, count):
- """Read `count` 64-bit ints using the plain encoding."""
- return struct.unpack(b"<{0}q".format(count).encode("utf-8"), file_obj.read(8 * count))
- def read_plain_int96(file_obj, count):
- """Read `count` 96-bit ints using the plain encoding."""
- items = struct.unpack(b"<qi" * count, file_obj.read(12) * count)
- args = [iter(items)] * 2
- return [q << 32 | i for (q, i) in zip(*args)]
- def read_plain_float(file_obj, count):
- """Read `count` 32-bit floats using the plain encoding."""
- return struct.unpack(b"<{0}f".format(count).encode("utf-8"), file_obj.read(4 * count))
- def read_plain_double(file_obj, count):
- """Read `count` 64-bit float (double) using the plain encoding."""
- return struct.unpack(b"<{0}d".format(count).encode("utf-8"), file_obj.read(8 * count))
- def read_plain_byte_array(file_obj, count):
- """Read `count` byte arrays using the plain encoding."""
- return [file_obj.read(struct.unpack(b"<i", file_obj.read(4))[0]) for i in range(count)]
- def read_plain_byte_array_fixed(file_obj, fixed_length):
- """Read a byte array of the given fixed_length."""
- return file_obj.read(fixed_length)
- DECODE_PLAIN = {
- parquet_thrift.Type.BOOLEAN: read_plain_boolean,
- parquet_thrift.Type.INT32: read_plain_int32,
- parquet_thrift.Type.INT64: read_plain_int64,
- parquet_thrift.Type.INT96: read_plain_int96,
- parquet_thrift.Type.FLOAT: read_plain_float,
- parquet_thrift.Type.DOUBLE: read_plain_double,
- parquet_thrift.Type.BYTE_ARRAY: read_plain_byte_array,
- parquet_thrift.Type.FIXED_LEN_BYTE_ARRAY: read_plain_byte_array_fixed
- }
- def read_plain(file_obj, type_, count):
- """Read `count` items `type` from the fo using the plain encoding."""
- if count == 0:
- return []
- conv = DECODE_PLAIN[type_]
- return conv(file_obj, count)
- def read_unsigned_var_int(file_obj):
- """Read a value using the unsigned, variable int encoding."""
- result = 0
- shift = 0
- while True:
- byte = struct.unpack(b"<B", file_obj.read(1))[0]
- result |= ((byte & 0x7F) << shift)
- if (byte & 0x80) == 0:
- break
- shift += 7
- return result
- def read_rle(file_obj, header, bit_width, debug_logging):
- """Read a run-length encoded run from the given fo with the given header and bit_width.
- The count is determined from the header and the width is used to grab the
- value that's repeated. Yields the value repeated count times.
- """
- count = header >> 1
- zero_data = b"\x00\x00\x00\x00"
- width = (bit_width + 7) // 8
- data = file_obj.read(width)
- data = data + zero_data[len(data):]
- value = struct.unpack(b"<i", data)[0]
- if debug_logging:
- logger.debug("Read RLE group with value %s of byte-width %s and count %s",
- value, width, count)
- for _ in range(count):
- yield value
- def width_from_max_int(value):
- """Convert the value specified to a bit_width."""
- return int(math.ceil(math.log(value + 1, 2)))
- def _mask_for_bits(i):
- """Generate a mask to grab `i` bits from an int value."""
- return (1 << i) - 1
- def read_bitpacked(file_obj, header, width, debug_logging):
- """Read a bitpacked run of the rle/bitpack hybrid.
- Supports width >8 (crossing bytes).
- """
- num_groups = header >> 1
- count = num_groups * 8
- byte_count = (width * count) // 8
- if debug_logging:
- logger.debug("Reading a bit-packed run with: %s groups, count %s, bytes %s",
- num_groups, count, byte_count)
- raw_bytes = array.array(ARRAY_BYTE_STR, file_obj.read(byte_count)).tolist()
- current_byte = 0
- data = raw_bytes[current_byte]
- mask = _mask_for_bits(width)
- bits_wnd_l = 8
- bits_wnd_r = 0
- res = []
- total = len(raw_bytes) * 8
- while total >= width:
- # NOTE zero-padding could produce extra zero-values
- if debug_logging:
- logger.debug(" read bitpacked: width=%s window=(%s %s) b=%s,"
- " current_byte=%s",
- width, bits_wnd_l, bits_wnd_r, bin(data), current_byte)
- if bits_wnd_r >= 8:
- bits_wnd_r -= 8
- bits_wnd_l -= 8
- data >>= 8
- elif bits_wnd_l - bits_wnd_r >= width:
- res.append((data >> bits_wnd_r) & mask)
- total -= width
- bits_wnd_r += width
- if debug_logging:
- logger.debug(" read bitpackage: added: %s", res[-1])
- elif current_byte + 1 < len(raw_bytes):
- current_byte += 1
- data |= (raw_bytes[current_byte] << bits_wnd_l)
- bits_wnd_l += 8
- return res
- def read_bitpacked_deprecated(file_obj, byte_count, count, width, debug_logging):
- """Read `count` values from `fo` using the deprecated bitpacking encoding."""
- raw_bytes = array.array(ARRAY_BYTE_STR, file_obj.read(byte_count)).tolist()
- mask = _mask_for_bits(width)
- index = 0
- res = []
- word = 0
- bits_in_word = 0
- while len(res) < count and index <= len(raw_bytes):
- if debug_logging:
- logger.debug("index = %d", index)
- logger.debug("bits in word = %d", bits_in_word)
- logger.debug("word = %s", bin(word))
- if bits_in_word >= width:
- # how many bits over the value is stored
- offset = (bits_in_word - width)
- # figure out the value
- value = (word & (mask << offset)) >> offset
- if debug_logging:
- logger.debug("offset = %d", offset)
- logger.debug("value = %d (%s)", value, bin(value))
- res.append(value)
- bits_in_word -= width
- else:
- word = (word << 8) | raw_bytes[index]
- index += 1
- bits_in_word += 8
- return res
- def read_rle_bit_packed_hybrid(file_obj, width, length=None):
- """Read values from `fo` using the rel/bit-packed hybrid encoding.
- If length is not specified, then a 32-bit int is read first to grab the
- length of the encoded data.
- """
- debug_logging = logger.isEnabledFor(logging.DEBUG)
- io_obj = file_obj
- if length is None:
- length = read_plain_int32(file_obj, 1)[0]
- raw_bytes = file_obj.read(length)
- if raw_bytes == b'':
- return None
- io_obj = io.BytesIO(raw_bytes)
- res = []
- while io_obj.tell() < length:
- header = read_unsigned_var_int(io_obj)
- if header & 1 == 0:
- res += read_rle(io_obj, header, width, debug_logging)
- else:
- res += read_bitpacked(io_obj, header, width, debug_logging)
- return res
|