py-libp2p/libp2p/utils.py
NIC619 879f193aa1
Handle errors from
- `read_delim`
    - `read_varint_prefixed_bytes`
    - `decode_uvarint_from_stream`
2019-09-15 16:58:08 +08:00

103 lines
2.9 KiB
Python

import itertools
import math
from libp2p.exceptions import ParseError
from libp2p.io.abc import Reader
from .io.utils import read_exactly
# Unsigned LEB128(varint codec)
# Reference: https://github.com/ethereum/py-wasm/blob/master/wasm/parsers/leb128.py
LOW_MASK = 2 ** 7 - 1
HIGH_MASK = 2 ** 7
# The maximum shift width for a 64 bit integer. We shouldn't have to decode
# integers larger than this.
SHIFT_64_BIT_MAX = int(math.ceil(64 / 7)) * 7
def encode_uvarint(number: int) -> bytes:
"""Pack `number` into varint bytes"""
buf = b""
while True:
towrite = number & 0x7F
number >>= 7
if number:
buf += bytes((towrite | 0x80,))
else:
buf += bytes((towrite,))
break
return buf
async def decode_uvarint_from_stream(reader: Reader) -> int:
"""
https://en.wikipedia.org/wiki/LEB128
"""
res = 0
for shift in itertools.count(0, 7):
if shift > SHIFT_64_BIT_MAX:
raise ParseError("TODO: better exception msg: Integer is too large...")
byte = await reader.read(1)
try:
value = byte[0]
except IndexError:
raise ParseError(
"Unexpected end of stream while parsing LEB128 encoded integer"
)
res += (value & LOW_MASK) << shift
if not value & HIGH_MASK:
break
return res
def encode_varint_prefixed(msg_bytes: bytes) -> bytes:
varint_len = encode_uvarint(len(msg_bytes))
return varint_len + msg_bytes
async def read_varint_prefixed_bytes(reader: Reader) -> bytes:
len_msg = await decode_uvarint_from_stream(reader)
data = await read_exactly(reader, len_msg)
return data
# Delimited read/write, used by multistream-select.
# Reference: https://github.com/gogo/protobuf/blob/07eab6a8298cf32fac45cceaac59424f98421bbc/io/varint.go#L109-L126 # noqa: E501
def encode_delim(msg: bytes) -> bytes:
delimited_msg = msg + b"\n"
return encode_varint_prefixed(delimited_msg)
async def read_delim(reader: Reader) -> bytes:
msg_bytes = await read_varint_prefixed_bytes(reader)
# TODO: Investigate if it is possible to have empty `msg_bytes`
if len(msg_bytes) != 0 and msg_bytes[-1:] != b"\n":
raise ValueError(f'msg_bytes is not delimited by b"\\n": msg_bytes={msg_bytes}')
return msg_bytes[:-1]
SIZE_LEN_BYTES = 4
# Fixed-prefixed read/write, used by "/plaintext/2.0.0".
# Reference: https://github.com/libp2p/go-msgio/blob/d5bbf59d3c4240266b1d2e5df9dc993454c42011/num.go#L11-L33 # noqa: E501 # noqa: E501
def encode_fixedint_prefixed(msg_bytes: bytes) -> bytes:
len_prefix = len(msg_bytes).to_bytes(SIZE_LEN_BYTES, "big")
return len_prefix + msg_bytes
async def read_fixedint_prefixed(reader: Reader) -> bytes:
len_bytes = await reader.read(SIZE_LEN_BYTES)
len_int = int.from_bytes(len_bytes, "big")
return await reader.read(len_int)