"""
The MIT License (MIT)
Copyright (c) 2015-2021 Rapptz
Copyright (c) 2021-present Pycord Development
Permission is hereby granted, free of charge, to any person obtaining a
copy of this software and associated documentation files (the "Software"),
to deal in the Software without restriction, including without limitation
the rights to use, copy, modify, merge, publish, distribute, sublicense,
and/or sell copies of the Software, and to permit persons to whom the
Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in
all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
DEALINGS IN THE SOFTWARE.
"""
from __future__ import annotations
import asyncio
import time
from collections import deque
from typing import TYPE_CHECKING, Any, Callable, Deque, TypeVar
import discord.abc
from discord.enums import Enum
from ...abc import PrivateChannel
from .errors import MaxConcurrencyReached
if TYPE_CHECKING:
from ...message import Message
__all__ = (
"BucketType",
"Cooldown",
"CooldownMapping",
"DynamicCooldownMapping",
"MaxConcurrency",
)
C = TypeVar("C", bound="CooldownMapping")
MC = TypeVar("MC", bound="MaxConcurrency")
[docs]
class BucketType(Enum):
default = 0
user = 1
guild = 2
channel = 3
member = 4
category = 5
role = 6
def get_key(self, msg: Message) -> Any:
if self is BucketType.user:
return msg.author.id
elif self is BucketType.guild:
return (msg.guild or msg.author).id
elif self is BucketType.channel:
return msg.channel.id
elif self is BucketType.member:
return (msg.guild and msg.guild.id), msg.author.id
elif self is BucketType.category:
return (
msg.channel.category.id
if isinstance(msg.channel, discord.abc.GuildChannel)
and msg.channel.category
else msg.channel.id
)
elif self is BucketType.role:
# we return the channel id of a private-channel as there are only roles in guilds
# and that yields the same result as for a guild with only the @everyone role
# NOTE: PrivateChannel doesn't actually have an id attribute, but we assume we are
# receiving a DMChannel or GroupChannel which inherit from PrivateChannel and do
return (msg.channel if isinstance(msg.channel, PrivateChannel) else msg.author.top_role).id # type: ignore
def __call__(self, msg: Message) -> Any:
return self.get_key(msg)
[docs]
class Cooldown:
"""Represents a cooldown for a command.
Attributes
----------
rate: :class:`int`
The total number of tokens available per :attr:`per` seconds.
per: :class:`float`
The length of the cooldown period in seconds.
"""
__slots__ = ("rate", "per", "_window", "_tokens", "_last")
def __init__(self, rate: float, per: float) -> None:
self.rate: int = int(rate)
self.per: float = float(per)
self._window: float = 0.0
self._tokens: int = self.rate
self._last: float = 0.0
[docs]
def get_tokens(self, current: float | None = None) -> int:
"""Returns the number of available tokens before rate limiting is applied.
Parameters
----------
current: Optional[:class:`float`]
The time in seconds since Unix epoch to calculate tokens at.
If not supplied then :func:`time.time()` is used.
Returns
-------
:class:`int`
The number of tokens available before the cooldown is to be applied.
"""
if not current:
current = time.time()
tokens = self._tokens
if current > self._window + self.per:
tokens = self.rate
return tokens
[docs]
def get_retry_after(self, current: float | None = None) -> float:
"""Returns the time in seconds until the cooldown will be reset.
Parameters
----------
current: Optional[:class:`float`]
The current time in seconds since Unix epoch.
If not supplied, then :func:`time.time()` is used.
Returns
-------
:class:`float`
The number of seconds to wait before this cooldown will be reset.
"""
current = current or time.time()
tokens = self.get_tokens(current)
if tokens == 0:
return self.per - (current - self._window)
return 0.0
[docs]
def update_rate_limit(self, current: float | None = None) -> float | None:
"""Updates the cooldown rate limit.
Parameters
----------
current: Optional[:class:`float`]
The time in seconds since Unix epoch to update the rate limit at.
If not supplied, then :func:`time.time()` is used.
Returns
-------
Optional[:class:`float`]
The retry-after time in seconds if rate limited.
"""
current = current or time.time()
self._last = current
self._tokens = self.get_tokens(current)
# first token used means that we start a new rate limit window
if self._tokens == self.rate:
self._window = current
# check if we are rate limited
if self._tokens == 0:
return self.per - (current - self._window)
# we're not so decrement our tokens
self._tokens -= 1
[docs]
def reset(self) -> None:
"""Reset the cooldown to its initial state."""
self._tokens = self.rate
self._last = 0.0
[docs]
def copy(self) -> Cooldown:
"""Creates a copy of this cooldown.
Returns
-------
:class:`Cooldown`
A new instance of this cooldown.
"""
return Cooldown(self.rate, self.per)
def __repr__(self) -> str:
return (
f"<Cooldown rate: {self.rate} per: {self.per} window:"
f" {self._window} tokens: {self._tokens}>"
)
class CooldownMapping:
def __init__(
self,
original: Cooldown | None,
type: Callable[[Message], Any],
) -> None:
if not callable(type):
raise TypeError("Cooldown type must be a BucketType or callable")
self._cache: dict[Any, Cooldown] = {}
self._cooldown: Cooldown | None = original
self._type: Callable[[Message], Any] = type
def copy(self) -> CooldownMapping:
ret = CooldownMapping(self._cooldown, self._type)
ret._cache = self._cache.copy()
return ret
@property
def valid(self) -> bool:
return self._cooldown is not None
@property
def type(self) -> Callable[[Message], Any]:
return self._type
@classmethod
def from_cooldown(cls: type[C], rate, per, type) -> C:
return cls(Cooldown(rate, per), type)
def _bucket_key(self, msg: Message) -> Any:
return self._type(msg)
def _verify_cache_integrity(self, current: float | None = None) -> None:
# we want to delete all cache objects that haven't been used
# in a cooldown window. e.g. if we have a command that has a
# cooldown of 60s, and it has not been used in 60s then that key should be deleted
current = current or time.time()
dead_keys = [k for k, v in self._cache.items() if current > v._last + v.per]
for k in dead_keys:
del self._cache[k]
def create_bucket(self, message: Message) -> Cooldown:
return self._cooldown.copy() # type: ignore
def get_bucket(self, message: Message, current: float | None = None) -> Cooldown:
if self._type is BucketType.default:
return self._cooldown # type: ignore
self._verify_cache_integrity(current)
key = self._bucket_key(message)
if key not in self._cache:
bucket = self.create_bucket(message)
if bucket is not None:
self._cache[key] = bucket
else:
bucket = self._cache[key]
return bucket
def update_rate_limit(
self, message: Message, current: float | None = None
) -> float | None:
bucket = self.get_bucket(message, current)
return bucket.update_rate_limit(current)
class DynamicCooldownMapping(CooldownMapping):
def __init__(
self, factory: Callable[[Message], Cooldown], type: Callable[[Message], Any]
) -> None:
super().__init__(None, type)
self._factory: Callable[[Message], Cooldown] = factory
def copy(self) -> DynamicCooldownMapping:
ret = DynamicCooldownMapping(self._factory, self._type)
ret._cache = self._cache.copy()
return ret
@property
def valid(self) -> bool:
return True
def create_bucket(self, message: Message) -> Cooldown:
return self._factory(message)
class _Semaphore:
"""This class is a version of a semaphore.
If you're wondering why asyncio.Semaphore isn't being used,
it's because it doesn't expose the internal value. This internal
value is necessary because I need to support both `wait=True` and
`wait=False`.
An asyncio.Queue could have been used to do this as well -- but it is
not as efficient since internally that uses two queues and is a bit
overkill for what is basically a counter.
"""
__slots__ = ("value", "loop", "_waiters")
def __init__(self, number: int) -> None:
self.value: int = number
self.loop: asyncio.AbstractEventLoop = asyncio.get_event_loop()
self._waiters: Deque[asyncio.Future] = deque()
def __repr__(self) -> str:
return f"<_Semaphore value={self.value} waiters={len(self._waiters)}>"
def locked(self) -> bool:
return self.value == 0
def is_active(self) -> bool:
return len(self._waiters) > 0
def wake_up(self) -> None:
while self._waiters:
future = self._waiters.popleft()
if not future.done():
future.set_result(None)
return
async def acquire(self, *, wait: bool = False) -> bool:
if not wait and self.value <= 0:
# signal that we're not acquiring
return False
while self.value <= 0:
future = self.loop.create_future()
self._waiters.append(future)
try:
await future
except:
future.cancel()
if self.value > 0 and not future.cancelled():
self.wake_up()
raise
self.value -= 1
return True
def release(self) -> None:
self.value += 1
self.wake_up()
class MaxConcurrency:
__slots__ = ("number", "per", "wait", "_mapping")
def __init__(self, number: int, *, per: BucketType, wait: bool) -> None:
self._mapping: dict[Any, _Semaphore] = {}
self.per: BucketType = per
self.number: int = number
self.wait: bool = wait
if number <= 0:
raise ValueError("max_concurrency 'number' cannot be less than 1")
if not isinstance(per, BucketType):
raise TypeError(
f"max_concurrency 'per' must be of type BucketType not {type(per)!r}"
)
def copy(self: MC) -> MC:
return self.__class__(self.number, per=self.per, wait=self.wait)
def __repr__(self) -> str:
return (
f"<MaxConcurrency per={self.per!r} number={self.number} wait={self.wait}>"
)
def get_key(self, message: Message) -> Any:
return self.per.get_key(message)
async def acquire(self, message: Message) -> None:
key = self.get_key(message)
try:
sem = self._mapping[key]
except KeyError:
self._mapping[key] = sem = _Semaphore(self.number)
acquired = await sem.acquire(wait=self.wait)
if not acquired:
raise MaxConcurrencyReached(self.number, self.per)
async def release(self, message: Message) -> None:
# Technically there's no reason for this function to be async
# But it might be more useful in the future
key = self.get_key(message)
try:
sem = self._mapping[key]
except KeyError:
# ...? peculiar
return
else:
sem.release()
if sem.value >= self.number and not sem.is_active():
del self._mapping[key]