Mini Shell
import inspect
import logging
import operator
import re
import string
import sys
import typing
import typing as t
from datetime import date
from datetime import datetime
from datetime import timezone
from itertools import chain
from weakref import WeakKeyDictionary
if t.TYPE_CHECKING:
from _typeshed.wsgi import StartResponse
from _typeshed.wsgi import WSGIApplication
from _typeshed.wsgi import WSGIEnvironment
from .wrappers.request import Request # noqa: F401
_logger: t.Optional[logging.Logger] = None
_signature_cache = WeakKeyDictionary() # type: ignore
_epoch_ord = date(1970, 1, 1).toordinal()
_legal_cookie_chars = frozenset(
c.encode("ascii")
for c in f"{string.ascii_letters}{string.digits}/=!#$%&'*+-.^_`|~:"
)
_cookie_quoting_map = {b",": b"\\054", b";": b"\\073", b'"': b'\\"', b"\\": b"\\\\"}
for _i in chain(range(32), range(127, 256)):
_cookie_quoting_map[_i.to_bytes(1, sys.byteorder)] = f"\\{_i:03o}".encode("latin1")
_octal_re = re.compile(rb"\\[0-3][0-7][0-7]")
_quote_re = re.compile(rb"[\\].")
_legal_cookie_chars_re = rb"[\w\d!#%&\'~_`><@,:/\$\*\+\-\.\^\|\)\(\?\}\{\=]"
_cookie_re = re.compile(
rb"""
(?P<key>[^=;]+)
(?:\s*=\s*
(?P<val>
"(?:[^\\"]|\\.)*" |
(?:.*?)
)
)?
\s*;
""",
flags=re.VERBOSE,
)
class _Missing:
def __repr__(self) -> str:
return "no value"
def __reduce__(self) -> str:
return "_missing"
_missing = _Missing()
@typing.overload
def _make_encode_wrapper(reference: str) -> t.Callable[[str], str]:
...
@typing.overload
def _make_encode_wrapper(reference: bytes) -> t.Callable[[str], bytes]:
...
def _make_encode_wrapper(reference: t.AnyStr) -> t.Callable[[str], t.AnyStr]:
"""Create a function that will be called with a string argument. If
the reference is bytes, values will be encoded to bytes.
"""
if isinstance(reference, str):
return lambda x: x
return operator.methodcaller("encode", "latin1")
def _check_str_tuple(value: t.Tuple[t.AnyStr, ...]) -> None:
"""Ensure tuple items are all strings or all bytes."""
if not value:
return
item_type = str if isinstance(value[0], str) else bytes
if any(not isinstance(item, item_type) for item in value):
raise TypeError(f"Cannot mix str and bytes arguments (got {value!r})")
_default_encoding = sys.getdefaultencoding()
def _to_bytes(
x: t.Union[str, bytes], charset: str = _default_encoding, errors: str = "strict"
) -> bytes:
if x is None or isinstance(x, bytes):
return x
if isinstance(x, (bytearray, memoryview)):
return bytes(x)
if isinstance(x, str):
return x.encode(charset, errors)
raise TypeError("Expected bytes")
@typing.overload
def _to_str( # type: ignore
x: None,
charset: t.Optional[str] = ...,
errors: str = ...,
allow_none_charset: bool = ...,
) -> None:
...
@typing.overload
def _to_str(
x: t.Any,
charset: t.Optional[str] = ...,
errors: str = ...,
allow_none_charset: bool = ...,
) -> str:
...
def _to_str(
x: t.Optional[t.Any],
charset: t.Optional[str] = _default_encoding,
errors: str = "strict",
allow_none_charset: bool = False,
) -> t.Optional[t.Union[str, bytes]]:
if x is None or isinstance(x, str):
return x
if not isinstance(x, (bytes, bytearray)):
return str(x)
if charset is None:
if allow_none_charset:
return x
return x.decode(charset, errors) # type: ignore
def _wsgi_decoding_dance(
s: str, charset: str = "utf-8", errors: str = "replace"
) -> str:
return s.encode("latin1").decode(charset, errors)
def _wsgi_encoding_dance(
s: str, charset: str = "utf-8", errors: str = "replace"
) -> str:
if isinstance(s, bytes):
return s.decode("latin1", errors)
return s.encode(charset).decode("latin1", errors)
def _get_environ(obj: t.Union["WSGIEnvironment", "Request"]) -> "WSGIEnvironment":
env = getattr(obj, "environ", obj)
assert isinstance(
env, dict
), f"{type(obj).__name__!r} is not a WSGI environment (has to be a dict)"
return env
def _has_level_handler(logger: logging.Logger) -> bool:
"""Check if there is a handler in the logging chain that will handle
the given logger's effective level.
"""
level = logger.getEffectiveLevel()
current = logger
while current:
if any(handler.level <= level for handler in current.handlers):
return True
if not current.propagate:
break
current = current.parent # type: ignore
return False
class _ColorStreamHandler(logging.StreamHandler):
"""On Windows, wrap stream with Colorama for ANSI style support."""
def __init__(self) -> None:
try:
import colorama
except ImportError:
stream = None
else:
stream = colorama.AnsiToWin32(sys.stderr)
super().__init__(stream)
def _log(type: str, message: str, *args: t.Any, **kwargs: t.Any) -> None:
"""Log a message to the 'werkzeug' logger.
The logger is created the first time it is needed. If there is no
level set, it is set to :data:`logging.INFO`. If there is no handler
for the logger's effective level, a :class:`logging.StreamHandler`
is added.
"""
global _logger
if _logger is None:
_logger = logging.getLogger("werkzeug")
if _logger.level == logging.NOTSET:
_logger.setLevel(logging.INFO)
if not _has_level_handler(_logger):
_logger.addHandler(_ColorStreamHandler())
getattr(_logger, type)(message.rstrip(), *args, **kwargs)
def _parse_signature(func): # type: ignore
"""Return a signature object for the function.
.. deprecated:: 2.0
Will be removed in Werkzeug 2.1 along with ``utils.bind`` and
``validate_arguments``.
"""
# if we have a cached validator for this function, return it
parse = _signature_cache.get(func)
if parse is not None:
return parse
# inspect the function signature and collect all the information
tup = inspect.getfullargspec(func)
positional, vararg_var, kwarg_var, defaults = tup[:4]
defaults = defaults or ()
arg_count = len(positional)
arguments = []
for idx, name in enumerate(positional):
if isinstance(name, list):
raise TypeError(
"cannot parse functions that unpack tuples in the function signature"
)
try:
default = defaults[idx - arg_count]
except IndexError:
param = (name, False, None)
else:
param = (name, True, default)
arguments.append(param)
arguments = tuple(arguments)
def parse(args, kwargs): # type: ignore
new_args = []
missing = []
extra = {}
# consume as many arguments as positional as possible
for idx, (name, has_default, default) in enumerate(arguments):
try:
new_args.append(args[idx])
except IndexError:
try:
new_args.append(kwargs.pop(name))
except KeyError:
if has_default:
new_args.append(default)
else:
missing.append(name)
else:
if name in kwargs:
extra[name] = kwargs.pop(name)
# handle extra arguments
extra_positional = args[arg_count:]
if vararg_var is not None:
new_args.extend(extra_positional)
extra_positional = ()
if kwargs and kwarg_var is None:
extra.update(kwargs)
kwargs = {}
return (
new_args,
kwargs,
missing,
extra,
extra_positional,
arguments,
vararg_var,
kwarg_var,
)
_signature_cache[func] = parse
return parse
@typing.overload
def _dt_as_utc(dt: None) -> None:
...
@typing.overload
def _dt_as_utc(dt: datetime) -> datetime:
...
def _dt_as_utc(dt: t.Optional[datetime]) -> t.Optional[datetime]:
if dt is None:
return dt
if dt.tzinfo is None:
return dt.replace(tzinfo=timezone.utc)
elif dt.tzinfo != timezone.utc:
return dt.astimezone(timezone.utc)
return dt
_TAccessorValue = t.TypeVar("_TAccessorValue")
class _DictAccessorProperty(t.Generic[_TAccessorValue]):
"""Baseclass for `environ_property` and `header_property`."""
read_only = False
def __init__(
self,
name: str,
default: t.Optional[_TAccessorValue] = None,
load_func: t.Optional[t.Callable[[str], _TAccessorValue]] = None,
dump_func: t.Optional[t.Callable[[_TAccessorValue], str]] = None,
read_only: t.Optional[bool] = None,
doc: t.Optional[str] = None,
) -> None:
self.name = name
self.default = default
self.load_func = load_func
self.dump_func = dump_func
if read_only is not None:
self.read_only = read_only
self.__doc__ = doc
def lookup(self, instance: t.Any) -> t.MutableMapping[str, t.Any]:
raise NotImplementedError
@typing.overload
def __get__(
self, instance: None, owner: type
) -> "_DictAccessorProperty[_TAccessorValue]":
...
@typing.overload
def __get__(self, instance: t.Any, owner: type) -> _TAccessorValue:
...
def __get__(
self, instance: t.Optional[t.Any], owner: type
) -> t.Union[_TAccessorValue, "_DictAccessorProperty[_TAccessorValue]"]:
if instance is None:
return self
storage = self.lookup(instance)
if self.name not in storage:
return self.default # type: ignore
value = storage[self.name]
if self.load_func is not None:
try:
return self.load_func(value)
except (ValueError, TypeError):
return self.default # type: ignore
return value # type: ignore
def __set__(self, instance: t.Any, value: _TAccessorValue) -> None:
if self.read_only:
raise AttributeError("read only property")
if self.dump_func is not None:
self.lookup(instance)[self.name] = self.dump_func(value)
else:
self.lookup(instance)[self.name] = value
def __delete__(self, instance: t.Any) -> None:
if self.read_only:
raise AttributeError("read only property")
self.lookup(instance).pop(self.name, None)
def __repr__(self) -> str:
return f"<{type(self).__name__} {self.name}>"
def _cookie_quote(b: bytes) -> bytes:
buf = bytearray()
all_legal = True
_lookup = _cookie_quoting_map.get
_push = buf.extend
for char_int in b:
char = char_int.to_bytes(1, sys.byteorder)
if char not in _legal_cookie_chars:
all_legal = False
char = _lookup(char, char)
_push(char)
if all_legal:
return bytes(buf)
return bytes(b'"' + buf + b'"')
def _cookie_unquote(b: bytes) -> bytes:
if len(b) < 2:
return b
if b[:1] != b'"' or b[-1:] != b'"':
return b
b = b[1:-1]
i = 0
n = len(b)
rv = bytearray()
_push = rv.extend
while 0 <= i < n:
o_match = _octal_re.search(b, i)
q_match = _quote_re.search(b, i)
if not o_match and not q_match:
rv.extend(b[i:])
break
j = k = -1
if o_match:
j = o_match.start(0)
if q_match:
k = q_match.start(0)
if q_match and (not o_match or k < j):
_push(b[i:k])
_push(b[k + 1 : k + 2])
i = k + 2
else:
_push(b[i:j])
rv.append(int(b[j + 1 : j + 4], 8))
i = j + 4
return bytes(rv)
def _cookie_parse_impl(b: bytes) -> t.Iterator[t.Tuple[bytes, bytes]]:
"""Lowlevel cookie parsing facility that operates on bytes."""
i = 0
n = len(b)
while i < n:
match = _cookie_re.search(b + b";", i)
if not match:
break
key = match.group("key").strip()
value = match.group("val") or b""
i = match.end(0)
yield key, _cookie_unquote(value)
def _encode_idna(domain: str) -> bytes:
# If we're given bytes, make sure they fit into ASCII
if isinstance(domain, bytes):
domain.decode("ascii")
return domain
# Otherwise check if it's already ascii, then return
try:
return domain.encode("ascii")
except UnicodeError:
pass
# Otherwise encode each part separately
return b".".join(p.encode("idna") for p in domain.split("."))
def _decode_idna(domain: t.Union[str, bytes]) -> str:
# If the input is a string try to encode it to ascii to do the idna
# decoding. If that fails because of a unicode error, then we
# already have a decoded idna domain.
if isinstance(domain, str):
try:
domain = domain.encode("ascii")
except UnicodeError:
return domain # type: ignore
# Decode each part separately. If a part fails, try to decode it
# with ascii and silently ignore errors. This makes sense because
# the idna codec does not have error handling.
def decode_part(part: bytes) -> str:
try:
return part.decode("idna")
except UnicodeError:
return part.decode("ascii", "ignore")
return ".".join(decode_part(p) for p in domain.split(b"."))
@typing.overload
def _make_cookie_domain(domain: None) -> None:
...
@typing.overload
def _make_cookie_domain(domain: str) -> bytes:
...
def _make_cookie_domain(domain: t.Optional[str]) -> t.Optional[bytes]:
if domain is None:
return None
domain = _encode_idna(domain)
if b":" in domain:
domain = domain.split(b":", 1)[0]
if b"." in domain:
return domain
raise ValueError(
"Setting 'domain' for a cookie on a server running locally (ex: "
"localhost) is not supported by complying browsers. You should "
"have something like: '127.0.0.1 localhost dev.localhost' on "
"your hosts file and then point your server to run on "
"'dev.localhost' and also set 'domain' for 'dev.localhost'"
)
def _easteregg(app: t.Optional["WSGIApplication"] = None) -> "WSGIApplication":
"""Like the name says. But who knows how it works?"""
def bzzzzzzz(gyver: bytes) -> str:
import base64
import zlib
return zlib.decompress(base64.b64decode(gyver)).decode("ascii")
gyver = "\n".join(
[
x + (77 - len(x)) * " "
for x in bzzzzzzz(
b"""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"""
).splitlines()
]
)
def easteregged(
environ: "WSGIEnvironment", start_response: "StartResponse"
) -> t.Iterable[bytes]:
def injecting_start_response(
status: str, headers: t.List[t.Tuple[str, str]], exc_info: t.Any = None
) -> t.Callable[[bytes], t.Any]:
headers.append(("X-Powered-By", "Werkzeug"))
return start_response(status, headers, exc_info)
if app is not None and environ.get("QUERY_STRING") != "macgybarchakku":
return app(environ, injecting_start_response)
injecting_start_response("200 OK", [("Content-Type", "text/html")])
return [
f"""\
<!DOCTYPE html>
<html>
<head>
<title>About Werkzeug</title>
<style type="text/css">
body {{ font: 15px Georgia, serif; text-align: center; }}
a {{ color: #333; text-decoration: none; }}
h1 {{ font-size: 30px; margin: 20px 0 10px 0; }}
p {{ margin: 0 0 30px 0; }}
pre {{ font: 11px 'Consolas', 'Monaco', monospace; line-height: 0.95; }}
</style>
</head>
<body>
<h1><a href="http://werkzeug.pocoo.org/">Werkzeug</a></h1>
<p>the Swiss Army knife of Python web development.</p>
<pre>{gyver}\n\n\n</pre>
</body>
</html>""".encode(
"latin1"
)
]
return easteregged
Zerion Mini Shell 1.0