Migration to discord.py 2.0 (#5600)

* Temporarily set d.py to use latest git revision

* Remove `bot` param to Client.start

* Switch to aware datetimes

A lot of this is removing `.replace(...)` which while not technically
needed, simplifies the code base. There's only a few changes that are
actually necessary here.

* Update to work with new Asset design

* [threads] Update core ModLog API to support threads

- Added proper support for passing `Thread` to `channel`
  when creating/editing case
- Added `parent_channel_id` attribute to Modlog API's Case
    - Added `parent_channel` property that tries to get parent channel
- Updated case's content to show both thread and parent information

* [threads] Disallow usage of threads in some of the commands

- announceset channel
- filter channel clear
- filter channel add
- filter channel remove
- GlobalUniqueObjectFinder converter
    - permissions addglobalrule
    - permissions removeglobalrule
    - permissions removeserverrule
    - Permissions cog does not perform any validation for IDs
      when setting through YAML so that has not been touched
- streamalert twitch/youtube/picarto
- embedset channel
- set ownernotifications adddestination

* [threads] Handle threads in Red's permissions system (Requires)

- Made permissions system apply rules of (only) parent in threads

* [threads] Update embed_requested to support threads

- Threads don't have their own embed settings and inherit from parent

* [threads] Update Red.message_eligible_as_command to support threads

* [threads] Properly handle invocation of [p](un)mutechannel in threads

Usage of a (un)mutechannel will mute/unmute user in the parent channel
if it's invoked in a thread.

* [threads] Update Filter cog to properly handle threads

- `[p]filter channel list` in a threads sends list for parent channel
- Checking for filter hits for a message in a thread checks its parent
  channel's word list. There's no separate word list for threads.

* [threads] Support threads in Audio cog

- Handle threads being notify channels
- Update type hint for `is_query_allowed()`

* [threads] Update type hints and documentation to reflect thread support

- Documented that `{channel}` in CCs might be a thread
- Allowed (documented) usage of threads with `Config.channel()`
    - Separate thread scope is still in the picture though
      if it were to be done, it's going to be in separate in PR
- GuildContext.channel might be Thread

* Use less costy channel check in customcom's on_message_without_command

This isn't needed for d.py 2.0 but whatever...

* Update for in-place edits

* Embed's bool changed behavior, I'm hoping it doesn't affect us

* Address User.permissions_in() removal

* Swap VerificationLevel.extreme with VerificationLevel.highest

* Change to keyword-only parameters

* Change of `Guild.vanity_invite()` return type

* avatar -> display_avatar

* Fix metaclass shenanigans with Converter

* Update Red.add_cog() to be inline with `dpy_commands.Bot.add_cog()`

This means adding `override` keyword-only parameter and causing
small breakage by swapping RuntimeError with discord.ClientException.

* Address all DEP-WARNs

* Remove Context.clean_prefix and use upstream implementation instead

* Remove commands.Literal and use upstream implementation instead

Honestly, this was a rather bad implementation anyway...

Breaking but actually not really - it was provisional.

* Update Command.callback's setter

Support for functools.partial is now built into d.py

* Add new perms in HUMANIZED_PERM mapping (some from d.py 1.7 it seems)

BTW, that should really be in core instead of what we have now...

* Remove the part of do_conversion that has not worked for a long while

* Stop wrapping BadArgument in ConversionFailure

This is breaking but it's best to resolve it like this.

The functionality of ConversionFailure can be replicated with
Context.current_parameter and Context.current_argument.

* Add custom errors for int and float converters

* Remove Command.__call__ as it's now implemented in d.py

* Get rid of _dpy_reimplements

These were reimplemented for the purpose of typing
so it is no longer needed now that d.py is type hinted.

* Add return to Red.remove_cog

* Ensure we don't delete messages that differ only by used sticker

* discord.InvalidArgument->ValueError

* Move from raw <t:...> syntax to discord.utils.format_dt()

* Address AsyncIter removal

* Swap to pos-only for params that are pos-only in upstream

* Update for changes to Command.params

* [threads] Support threads in ignore checks and allow ignoring them

- Updated `[p](un)ignore channel` to accept threads
- Updated `[p]ignore list` to list ignored threads
- Updated logic in `Red.ignored_channel_or_guild()`

Ignores for guild channels now work as follows (only changes for threads):
- if channel is not a thread:
    - check if user has manage channels perm in channel
      and allow command usage if so
    - check if channel is ignored and disallow command usage if so
    - allow command usage if none of the conditions above happened
- if channel is a thread:
    - check if user has manage channels perm in parent channel
      and allow command usage if so
    - check if parent channel is ignored and disallow command usage
      if so
    - check if user has manage thread perm in parent channel
      and allow command usage if so
    - check if thread is ignored and disallow command usage if so
    - allow command usage if none of the conditions above happened

* [partial] Raise TypeError when channel is of PartialMessageable type

- Red.embed_requested
- Red.ignored_channel_or_guild

* [partial] Discard command messages when channel is PartialMessageable

* [threads] Add utilities for checking appropriate perms in both channels & threads

* [threads] Update code to use can_react_in() and @bot_can_react()

* [threads] Update code to use can_send_messages_in

* [threads] Add send_messages_in_threads perm to mute role and overrides

* [threads] Update code to use (bot/user)_can_manage_channel

* [threads] Update [p]diagnoseissues to work with threads

* Type hint fix

* [threads] Patch vendored discord.ext.menus to check proper perms in threads

I guess we've reached time when we have to patch the lib we vendor...

* Make docs generation work with non-final d.py releases

* Update discord.utils.oauth_url() usage

* Swap usage of discord.Embed.Empty/discord.embeds.EmptyEmbed to None

* Update usage of Guild.member_count to work with `None`

* Switch from Guild.vanity_invite() to Guild.vanity_url

* Update startup process to work with d.py's new asynchronous startup

* Use setup_hook() for pre-connect actions

* Update core's add_cog, remove_cog, and load_extension methods

* Update all setup functions to async and add awaits to bot.add_cog calls

* Modernize cogs by using async cog_load and cog_unload

* Address StoreChannel removal

* [partial] Disallow passing PartialMessageable to Case.channel

* [partial] Update cogs and utils to work better with PartialMessageable

- Ignore messages with PartialMessageable channel in CustomCommands cog
- In Filter cog, don't pass channel to modlog.create_case()
  if it's PartialMessageable
- In Trivia cog, only compare channel IDs
- Make `.utils.menus.menu()` work for messages
  with PartialMessageable channel
- Make checks in `.utils.tunnel.Tunnel.communicate()` more rigid

* Add few missing DEP-WARNs
This commit is contained in:
jack1142
2022-04-03 03:21:20 +02:00
committed by GitHub
parent c9a0971945
commit febca8ccbb
104 changed files with 1427 additions and 999 deletions

View File

@@ -285,13 +285,6 @@ class Command(CogCommandMixin, DPYCommand):
(type used will be of the inner type instead)
"""
def __call__(self, *args, **kwargs):
if self.cog:
# We need to inject cog as self here
return self.callback(self.cog, *args, **kwargs)
else:
return self.callback(*args, **kwargs)
def __init__(self, *args, **kwargs):
self.ignore_optional_for_conversion = kwargs.pop("ignore_optional_for_conversion", False)
super().__init__(*args, **kwargs)
@@ -323,60 +316,27 @@ class Command(CogCommandMixin, DPYCommand):
@callback.setter
def callback(self, function):
"""
Below should be mostly the same as discord.py
# Below should be mostly the same as discord.py
#
# Here's the list of cases where the behavior differs:
# - `typing.Optional` behavior is changed
# when `ignore_optional_for_conversion` option is used
super(Command, Command).callback.__set__(self, function)
Currently, we modify behavior for
if not self.ignore_optional_for_conversion:
return
- functools.partial support
- typing.Optional behavior change as an option
"""
self._callback = function
if isinstance(function, functools.partial):
self.module = function.func.__module__
globals_ = function.func.__globals__
else:
self.module = function.__module__
globals_ = function.__globals__
signature = inspect.signature(function)
self.params = signature.parameters.copy()
# PEP-563 allows postponing evaluation of annotations with a __future__
# import. When postponed, Parameter.annotation will be a string and must
# be replaced with the real value for the converters to work later on
_NoneType = type(None)
for key, value in self.params.items():
if isinstance(value.annotation, str):
self.params[key] = value = value.replace(
annotation=eval(value.annotation, globals_)
)
# fail early for when someone passes an unparameterized Greedy type
if value.annotation is Greedy:
raise TypeError("Unparameterized Greedy[...] is disallowed in signature.")
if not self.ignore_optional_for_conversion:
continue # reduces indentation compared to alternative
try:
vtype = value.annotation.__origin__
if vtype is Union:
_NoneType = type if TYPE_CHECKING else type(None)
args = value.annotation.__args__
if _NoneType in args:
args = tuple(a for a in args if a is not _NoneType)
if len(args) == 1:
# can't have a union of 1 or 0 items
# 1 prevents this from becoming 0
# we need to prevent 2 become 1
# (Don't change that to becoming, it's intentional :musical_note:)
self.params[key] = value = value.replace(annotation=args[0])
else:
# and mypy wretches at the correct Union[args]
temp_type = type if TYPE_CHECKING else Union[args]
self.params[key] = value = value.replace(annotation=temp_type)
except AttributeError:
origin = getattr(value.annotation, "__origin__", None)
if origin is not Union:
continue
args = value.annotation.__args__
if _NoneType in args:
args = tuple(a for a in args if a is not _NoneType)
# typing.Union is automatically deduplicated and flattened
# so we don't need to anything else here
self.params[key] = value = value.replace(annotation=Union[args])
@property
def help(self):
@@ -420,6 +380,7 @@ class Command(CogCommandMixin, DPYCommand):
async def can_run(
self,
ctx: "Context",
/,
*,
check_all_parents: bool = False,
change_permission_state: bool = False,
@@ -476,7 +437,7 @@ class Command(CogCommandMixin, DPYCommand):
if not change_permission_state:
ctx.permission_state = original_state
async def prepare(self, ctx):
async def prepare(self, ctx, /):
ctx.command = self
if not self.enabled:
@@ -502,39 +463,6 @@ class Command(CogCommandMixin, DPYCommand):
await self._max_concurrency.release(ctx)
raise
async def do_conversion(
self, ctx: "Context", converter, argument: str, param: inspect.Parameter
):
"""Convert an argument according to its type annotation.
Raises
------
ConversionFailure
If doing the conversion failed.
Returns
-------
Any
The converted argument.
"""
# Let's not worry about all of this junk if it's just a str converter
if converter is str:
return argument
try:
return await super().do_conversion(ctx, converter, argument, param)
except BadArgument as exc:
raise ConversionFailure(converter, argument, param, *exc.args) from exc
except ValueError as exc:
# Some common converters need special treatment...
if converter in (int, float):
message = _('"{argument}" is not a number.').format(argument=argument)
raise ConversionFailure(converter, argument, param, message) from exc
# We should expose anything which might be a bug in the converter
raise exc
async def can_see(self, ctx: "Context"):
"""Check if this command is visible in the given context.
@@ -636,7 +564,7 @@ class Command(CogCommandMixin, DPYCommand):
break
return old_rule, new_rule
def error(self, coro):
def error(self, coro, /):
"""
A decorator that registers a coroutine as a local error handler.
@@ -796,7 +724,7 @@ class Group(GroupMixin, Command, CogGroupMixin, DPYGroup):
self.autohelp = kwargs.pop("autohelp", True)
super().__init__(*args, **kwargs)
async def invoke(self, ctx: "Context"):
async def invoke(self, ctx: "Context", /):
# we skip prepare in some cases to avoid some things
# We still always want this part of the behavior though
ctx.command = self
@@ -971,7 +899,7 @@ class CogMixin(CogGroupMixin, CogCommandMixin):
"""
raise RedUnhandledAPI()
async def can_run(self, ctx: "Context", **kwargs) -> bool:
async def can_run(self, ctx: "Context", /, **kwargs) -> bool:
"""
This really just exists to allow easy use with other methods using can_run
on commands and groups such as help formatters.
@@ -999,7 +927,7 @@ class CogMixin(CogGroupMixin, CogCommandMixin):
return can_run
async def can_see(self, ctx: "Context") -> bool:
async def can_see(self, ctx: "Context", /) -> bool:
"""Check if this cog is visible in the given context.
In short, this will verify whether
@@ -1112,7 +1040,7 @@ class _AlwaysAvailableMixin:
This particular class is not supported for 3rd party use
"""
async def can_run(self, ctx, *args, **kwargs) -> bool:
async def can_run(self, ctx, /, *args, **kwargs) -> bool:
return not ctx.author.bot
can_see = can_run
@@ -1161,7 +1089,7 @@ class _ForgetMeSpecialCommand(_RuleDropper, Command):
We need special can_run behavior here
"""
async def can_run(self, ctx, *args, **kwargs) -> bool:
async def can_run(self, ctx, /, *args, **kwargs) -> bool:
return await ctx.bot._config.datarequests.allow_user_requests()
can_see = can_run