mirror of
https://github.com/davidhalter/typeshed.git
synced 2025-12-08 04:54:47 +08:00
Re-organize directory structure (#4971)
See discussion in #2491 Co-authored-by: Ivan Levkivskyi <ilevkivskyi@dropbox.com>
This commit is contained in:
164
stdlib/lzma.pyi
Normal file
164
stdlib/lzma.pyi
Normal file
@@ -0,0 +1,164 @@
|
||||
import io
|
||||
from _typeshed import AnyPath, ReadableBuffer
|
||||
from typing import IO, Any, Mapping, Optional, Sequence, TextIO, TypeVar, Union, overload
|
||||
from typing_extensions import Literal
|
||||
|
||||
_OpenBinaryWritingMode = Literal["w", "wb", "x", "xb", "a", "ab"]
|
||||
_OpenTextWritingMode = Literal["wt", "xt", "at"]
|
||||
|
||||
_PathOrFile = Union[AnyPath, IO[bytes]]
|
||||
|
||||
_FilterChain = Sequence[Mapping[str, Any]]
|
||||
_T = TypeVar("_T")
|
||||
|
||||
FORMAT_AUTO: int
|
||||
FORMAT_XZ: int
|
||||
FORMAT_ALONE: int
|
||||
FORMAT_RAW: int
|
||||
CHECK_NONE: int
|
||||
CHECK_CRC32: int
|
||||
CHECK_CRC64: int
|
||||
CHECK_SHA256: int
|
||||
CHECK_ID_MAX: int
|
||||
CHECK_UNKNOWN: int
|
||||
FILTER_LZMA1: int
|
||||
FILTER_LZMA2: int
|
||||
FILTER_DELTA: int
|
||||
FILTER_X86: int
|
||||
FILTER_IA64: int
|
||||
FILTER_ARM: int
|
||||
FILTER_ARMTHUMB: int
|
||||
FILTER_SPARC: int
|
||||
FILTER_POWERPC: int
|
||||
MF_HC3: int
|
||||
MF_HC4: int
|
||||
MF_BT2: int
|
||||
MF_BT3: int
|
||||
MF_BT4: int
|
||||
MODE_FAST: int
|
||||
MODE_NORMAL: int
|
||||
PRESET_DEFAULT: int
|
||||
PRESET_EXTREME: int
|
||||
|
||||
# from _lzma.c
|
||||
class LZMADecompressor(object):
|
||||
def __init__(
|
||||
self, format: Optional[int] = ..., memlimit: Optional[int] = ..., filters: Optional[_FilterChain] = ...
|
||||
) -> None: ...
|
||||
def decompress(self, data: bytes, max_length: int = ...) -> bytes: ...
|
||||
@property
|
||||
def check(self) -> int: ...
|
||||
@property
|
||||
def eof(self) -> bool: ...
|
||||
@property
|
||||
def unused_data(self) -> bytes: ...
|
||||
@property
|
||||
def needs_input(self) -> bool: ...
|
||||
|
||||
# from _lzma.c
|
||||
class LZMACompressor(object):
|
||||
def __init__(
|
||||
self, format: Optional[int] = ..., check: int = ..., preset: Optional[int] = ..., filters: Optional[_FilterChain] = ...
|
||||
) -> None: ...
|
||||
def compress(self, data: bytes) -> bytes: ...
|
||||
def flush(self) -> bytes: ...
|
||||
|
||||
class LZMAError(Exception): ...
|
||||
|
||||
class LZMAFile(io.BufferedIOBase, IO[bytes]):
|
||||
def __init__(
|
||||
self,
|
||||
filename: Optional[_PathOrFile] = ...,
|
||||
mode: str = ...,
|
||||
*,
|
||||
format: Optional[int] = ...,
|
||||
check: int = ...,
|
||||
preset: Optional[int] = ...,
|
||||
filters: Optional[_FilterChain] = ...,
|
||||
) -> None: ...
|
||||
def __enter__(self: _T) -> _T: ...
|
||||
def close(self) -> None: ...
|
||||
@property
|
||||
def closed(self) -> bool: ...
|
||||
def fileno(self) -> int: ...
|
||||
def seekable(self) -> bool: ...
|
||||
def readable(self) -> bool: ...
|
||||
def writable(self) -> bool: ...
|
||||
def peek(self, size: int = ...) -> bytes: ...
|
||||
def read(self, size: Optional[int] = ...) -> bytes: ...
|
||||
def read1(self, size: int = ...) -> bytes: ...
|
||||
def readline(self, size: Optional[int] = ...) -> bytes: ...
|
||||
def write(self, data: ReadableBuffer) -> int: ...
|
||||
def seek(self, offset: int, whence: int = ...) -> int: ...
|
||||
def tell(self) -> int: ...
|
||||
|
||||
@overload
|
||||
def open(
|
||||
filename: _PathOrFile,
|
||||
mode: Literal["r", "rb"] = ...,
|
||||
*,
|
||||
format: Optional[int] = ...,
|
||||
check: Literal[-1] = ...,
|
||||
preset: None = ...,
|
||||
filters: Optional[_FilterChain] = ...,
|
||||
encoding: None = ...,
|
||||
errors: None = ...,
|
||||
newline: None = ...,
|
||||
) -> LZMAFile: ...
|
||||
@overload
|
||||
def open(
|
||||
filename: _PathOrFile,
|
||||
mode: _OpenBinaryWritingMode,
|
||||
*,
|
||||
format: Optional[int] = ...,
|
||||
check: int = ...,
|
||||
preset: Optional[int] = ...,
|
||||
filters: Optional[_FilterChain] = ...,
|
||||
encoding: None = ...,
|
||||
errors: None = ...,
|
||||
newline: None = ...,
|
||||
) -> LZMAFile: ...
|
||||
@overload
|
||||
def open(
|
||||
filename: AnyPath,
|
||||
mode: Literal["rt"],
|
||||
*,
|
||||
format: Optional[int] = ...,
|
||||
check: Literal[-1] = ...,
|
||||
preset: None = ...,
|
||||
filters: Optional[_FilterChain] = ...,
|
||||
encoding: Optional[str] = ...,
|
||||
errors: Optional[str] = ...,
|
||||
newline: Optional[str] = ...,
|
||||
) -> TextIO: ...
|
||||
@overload
|
||||
def open(
|
||||
filename: AnyPath,
|
||||
mode: _OpenTextWritingMode,
|
||||
*,
|
||||
format: Optional[int] = ...,
|
||||
check: int = ...,
|
||||
preset: Optional[int] = ...,
|
||||
filters: Optional[_FilterChain] = ...,
|
||||
encoding: Optional[str] = ...,
|
||||
errors: Optional[str] = ...,
|
||||
newline: Optional[str] = ...,
|
||||
) -> TextIO: ...
|
||||
@overload
|
||||
def open(
|
||||
filename: _PathOrFile,
|
||||
mode: str,
|
||||
*,
|
||||
format: Optional[int] = ...,
|
||||
check: int = ...,
|
||||
preset: Optional[int] = ...,
|
||||
filters: Optional[_FilterChain] = ...,
|
||||
encoding: Optional[str] = ...,
|
||||
errors: Optional[str] = ...,
|
||||
newline: Optional[str] = ...,
|
||||
) -> Union[LZMAFile, TextIO]: ...
|
||||
def compress(
|
||||
data: bytes, format: int = ..., check: int = ..., preset: Optional[int] = ..., filters: Optional[_FilterChain] = ...
|
||||
) -> bytes: ...
|
||||
def decompress(data: bytes, format: int = ..., memlimit: Optional[int] = ..., filters: Optional[_FilterChain] = ...) -> bytes: ...
|
||||
def is_check_supported(check: int) -> bool: ...
|
||||
Reference in New Issue
Block a user