diff options
Diffstat (limited to 'bot/exts/backend')
| -rw-r--r-- | bot/exts/backend/branding/__init__.py | 7 | ||||
| -rw-r--r-- | bot/exts/backend/branding/_cog.py | 653 | ||||
| -rw-r--r-- | bot/exts/backend/branding/_repository.py | 240 | ||||
| -rw-r--r-- | bot/exts/backend/error_handler.py | 67 | ||||
| -rw-r--r-- | bot/exts/backend/logging.py | 2 | ||||
| -rw-r--r-- | bot/exts/backend/sync/_syncers.py | 10 |
6 files changed, 965 insertions, 14 deletions
diff --git a/bot/exts/backend/branding/__init__.py b/bot/exts/backend/branding/__init__.py new file mode 100644 index 000000000..20a747b7f --- /dev/null +++ b/bot/exts/backend/branding/__init__.py @@ -0,0 +1,7 @@ +from bot.bot import Bot +from bot.exts.backend.branding._cog import Branding + + +def setup(bot: Bot) -> None: + """Load Branding cog.""" + bot.add_cog(Branding(bot)) diff --git a/bot/exts/backend/branding/_cog.py b/bot/exts/backend/branding/_cog.py new file mode 100644 index 000000000..47c379a34 --- /dev/null +++ b/bot/exts/backend/branding/_cog.py @@ -0,0 +1,653 @@ +import asyncio +import contextlib +import logging +import random +import typing as t +from datetime import timedelta +from enum import Enum +from operator import attrgetter + +import async_timeout +import discord +from arrow import Arrow +from async_rediscache import RedisCache +from discord.ext import commands, tasks + +from bot.bot import Bot +from bot.constants import Branding as BrandingConfig, Channels, Colours, Guild, MODERATION_ROLES +from bot.decorators import mock_in_debug +from bot.exts.backend.branding._repository import BrandingRepository, Event, RemoteObject + +log = logging.getLogger(__name__) + + +class AssetType(Enum): + """ + Recognised Discord guild asset types. + + The value of each member corresponds exactly to a kwarg that can be passed to `Guild.edit`. + """ + + BANNER = "banner" + ICON = "icon" + + +def compound_hash(objects: t.Iterable[RemoteObject]) -> str: + """ + Join SHA attributes of `objects` into a single string. + + Compound hashes are cached to check for change in any of the member `objects`. + """ + return "-".join(item.sha for item in objects) + + +def make_embed(title: str, description: str, *, success: bool) -> discord.Embed: + """ + Construct simple response embed. + + If `success` is True, use green colour, otherwise red. + + For both `title` and `description`, empty string are valid values ~ fields will be empty. + """ + colour = Colours.soft_green if success else Colours.soft_red + return discord.Embed(title=title[:256], description=description[:2048], colour=colour) + + +def extract_event_duration(event: Event) -> str: + """ + Extract a human-readable, year-agnostic duration string from `event`. + + In the case that `event` is a fallback event, resolves to 'Fallback'. + + For 1-day events, only the single date is shown, instead of a period. + """ + if event.meta.is_fallback: + return "Fallback" + + fmt = "%B %d" # Ex: August 23 + start_date = event.meta.start_date.strftime(fmt) + end_date = event.meta.end_date.strftime(fmt) + + if start_date == end_date: + return start_date + + return f"{start_date} - {end_date}" + + +def extract_event_name(event: Event) -> str: + """ + Extract title-cased event name from the path of `event`. + + An event with a path of 'events/black_history_month' will resolve to 'Black History Month'. + """ + name = event.path.split("/")[-1] # Inner-most directory name. + words = name.split("_") # Words from snake case. + + return " ".join(word.title() for word in words) + + +class Branding(commands.Cog): + """ + Guild branding management. + + Extension responsible for automatic synchronisation of the guild's branding with the branding repository. + Event definitions and assets are automatically discovered and applied as appropriate. + + All state is stored in Redis. The cog should therefore seamlessly transition across restarts and maintain + a consistent icon rotation schedule for events with multiple icon assets. + + By caching hashes of banner & icon assets, we discover changes in currently applied assets and always keep + the latest version applied. + + The command interface allows moderators+ to control the daemon or request asset synchronisation, while + regular users can see information about the current event and the overall event schedule. + """ + + # RedisCache[ + # "daemon_active": bool | If True, daemon starts on start-up. Controlled via commands. + # "event_path": str | Current event's path in the branding repo. + # "event_description": str | Current event's Markdown description. + # "event_duration": str | Current event's human-readable date range. + # "banner_hash": str | SHA of the currently applied banner. + # "icons_hash": str | Compound SHA of all icons in current rotation. + # "last_rotation_timestamp": float | POSIX UTC timestamp. + # ] + cache_information = RedisCache() + + # Icons in current rotation. Keys (str) are download URLs, values (int) track the amount of times each + # icon has been used in the current rotation. + cache_icons = RedisCache() + + # All available event names & durations. Cached by the daemon nightly; read by the calendar command. + cache_events = RedisCache() + + def __init__(self, bot: Bot) -> None: + """Instantiate repository abstraction & allow daemon to start.""" + self.bot = bot + self.repository = BrandingRepository(bot) + + self.bot.loop.create_task(self.maybe_start_daemon()) # Start depending on cache. + + # region: Internal logic & state management + + @mock_in_debug(return_value=True) # Mocked in development environment to prevent API spam. + async def apply_asset(self, asset_type: AssetType, download_url: str) -> bool: + """ + Download asset from `download_url` and apply it to PyDis as `asset_type`. + + Return a boolean indicating whether the application was successful. + """ + log.info(f"Applying '{asset_type.value}' asset to the guild.") + + try: + file = await self.repository.fetch_file(download_url) + except Exception: + log.exception(f"Failed to fetch '{asset_type.value}' asset.") + return False + + await self.bot.wait_until_guild_available() + pydis: discord.Guild = self.bot.get_guild(Guild.id) + + timeout = 10 # Seconds. + try: + with async_timeout.timeout(timeout): # Raise after `timeout` seconds. + await pydis.edit(**{asset_type.value: file}) + except discord.HTTPException: + log.exception("Asset upload to Discord failed.") + return False + except asyncio.TimeoutError: + log.error(f"Asset upload to Discord timed out after {timeout} seconds.") + return False + else: + log.trace("Asset uploaded successfully.") + return True + + async def apply_banner(self, banner: RemoteObject) -> bool: + """ + Apply `banner` to the guild and cache its hash if successful. + + Banners should always be applied via this method to ensure that the last hash is cached. + + Return a boolean indicating whether the application was successful. + """ + success = await self.apply_asset(AssetType.BANNER, banner.download_url) + + if success: + await self.cache_information.set("banner_hash", banner.sha) + + return success + + async def rotate_icons(self) -> bool: + """ + Choose and apply the next-up icon in rotation. + + We keep track of the amount of times each icon has been used. The values in `cache_icons` can be understood + to be iteration IDs. When an icon is chosen & applied, we bump its count, pushing it into the next iteration. + + Once the current iteration (lowest count in the cache) depletes, we move onto the next iteration. + + In the case that there is only 1 icon in the rotation and has already been applied, do nothing. + + Return a boolean indicating whether a new icon was applied successfully. + """ + log.debug("Rotating icons.") + + state = await self.cache_icons.to_dict() + log.trace(f"Total icons in rotation: {len(state)}.") + + if not state: # This would only happen if rotation not initiated, but we can handle gracefully. + log.warning("Attempted icon rotation with an empty icon cache. This indicates wrong logic.") + return False + + if len(state) == 1 and 1 in state.values(): + log.debug("Aborting icon rotation: only 1 icon is available and has already been applied.") + return False + + current_iteration = min(state.values()) # Choose iteration to draw from. + options = [download_url for download_url, times_used in state.items() if times_used == current_iteration] + + log.trace(f"Choosing from {len(options)} icons in iteration {current_iteration}.") + next_icon = random.choice(options) + + success = await self.apply_asset(AssetType.ICON, next_icon) + + if success: + await self.cache_icons.increment(next_icon) # Push the icon into the next iteration. + + timestamp = Arrow.utcnow().timestamp() + await self.cache_information.set("last_rotation_timestamp", timestamp) + + return success + + async def maybe_rotate_icons(self) -> None: + """ + Call `rotate_icons` if the configured amount of time has passed since last rotation. + + We offset the calculated time difference into the future to avoid off-by-a-little-bit errors. Because there + is work to be done before the timestamp is read and written, the next read will likely commence slightly + under 24 hours after the last write. + """ + log.debug("Checking whether it's time for icons to rotate.") + + last_rotation_timestamp = await self.cache_information.get("last_rotation_timestamp") + + if last_rotation_timestamp is None: # Maiden case ~ never rotated. + await self.rotate_icons() + return + + last_rotation = Arrow.utcfromtimestamp(last_rotation_timestamp) + difference = (Arrow.utcnow() - last_rotation) + timedelta(minutes=5) + + log.trace(f"Icons last rotated at {last_rotation} (difference: {difference}).") + + if difference.days >= BrandingConfig.cycle_frequency: + await self.rotate_icons() + + async def initiate_icon_rotation(self, available_icons: t.List[RemoteObject]) -> None: + """ + Set up a new icon rotation. + + This function should be called whenever available icons change. This is generally the case when we enter + a new event, but potentially also when the assets of an on-going event change. In such cases, a reset + of `cache_icons` is necessary, because it contains download URLs which may have gotten stale. + + This function does not upload a new icon! + """ + log.debug("Initiating new icon rotation.") + + await self.cache_icons.clear() + + new_state = {icon.download_url: 0 for icon in available_icons} + await self.cache_icons.update(new_state) + + log.trace(f"Icon rotation initiated for {len(new_state)} icons.") + + await self.cache_information.set("icons_hash", compound_hash(available_icons)) + + async def send_info_embed(self, channel_id: int, *, is_notification: bool) -> None: + """ + Send the currently cached event description to `channel_id`. + + When `is_notification` holds, a short contextual message for the #changelog channel is added. + + We read event information from `cache_information`. The caller is therefore responsible for making + sure that the cache is up-to-date before calling this function. + """ + log.debug(f"Sending event information event to channel: {channel_id} ({is_notification=}).") + + await self.bot.wait_until_guild_available() + channel: t.Optional[discord.TextChannel] = self.bot.get_channel(channel_id) + + if channel is None: + log.warning(f"Cannot send event information: channel {channel_id} not found!") + return + + log.trace(f"Destination channel: #{channel.name}.") + + description = await self.cache_information.get("event_description") + duration = await self.cache_information.get("event_duration") + + if None in (description, duration): + content = None + embed = make_embed("No event in cache", "Is the daemon enabled?", success=False) + + else: + content = "Python Discord is entering a new event!" if is_notification else None + embed = discord.Embed(description=description[:2048], colour=discord.Colour.blurple()) + embed.set_footer(text=duration[:2048]) + + await channel.send(content=content, embed=embed) + + async def enter_event(self, event: Event) -> t.Tuple[bool, bool]: + """ + Apply `event` assets and update information cache. + + We cache `event` information to ensure that we: + * Remember which event we're currently in across restarts + * Provide an on-demand informational embed without re-querying the branding repository + + An event change should always be handled via this function, as it ensures that the cache is populated. + + The #changelog notification is omitted when `event` is fallback, or already applied. + + Return a 2-tuple indicating whether the banner, and the icon, were applied successfully. + """ + log.info(f"Entering event: '{event.path}'.") + + banner_success = await self.apply_banner(event.banner) # Only one asset ~ apply directly. + + await self.initiate_icon_rotation(event.icons) # Prepare a new rotation. + icon_success = await self.rotate_icons() # Apply an icon from the new rotation. + + # This will only be False in the case of a manual same-event re-synchronisation. + event_changed = event.path != await self.cache_information.get("event_path") + + # Cache event identity to avoid re-entry in case of restart. + await self.cache_information.set("event_path", event.path) + + # Cache information shown in the 'about' embed. + await self.populate_cache_event_description(event) + + # Notify guild of new event ~ this reads the information that we cached above. + if event_changed and not event.meta.is_fallback: + await self.send_info_embed(Channels.change_log, is_notification=True) + else: + log.trace("Omitting #changelog notification. Event has not changed, or new event is fallback.") + + return banner_success, icon_success + + async def synchronise(self) -> t.Tuple[bool, bool]: + """ + Fetch the current event and delegate to `enter_event`. + + This is a convenience function to force synchronisation via a command. It should generally only be used + in a recovery scenario. In the usual case, the daemon already has an `Event` instance and can pass it + to `enter_event` directly. + + Return a 2-tuple indicating whether the banner, and the icon, were applied successfully. + """ + log.debug("Synchronise: fetching current event.") + + current_event, available_events = await self.repository.get_current_event() + + await self.populate_cache_events(available_events) + + if current_event is None: + log.error("Failed to fetch event. Cannot synchronise!") + return False, False + + return await self.enter_event(current_event) + + async def populate_cache_events(self, events: t.List[Event]) -> None: + """ + Clear `cache_events` and re-populate with names and durations of `events`. + + For each event, we store its name and duration string. This is the information presented to users in the + calendar command. If a format change is needed, it has to be done here. + + The cache does not store the fallback event, as it is not shown in the calendar. + """ + log.debug("Populating events cache.") + + await self.cache_events.clear() + + no_fallback = [event for event in events if not event.meta.is_fallback] + chronological_events = sorted(no_fallback, key=attrgetter("meta.start_date")) + + log.trace(f"Writing {len(chronological_events)} events (fallback omitted).") + + with contextlib.suppress(ValueError): # Cache raises when updated with an empty dict. + await self.cache_events.update({ + extract_event_name(event): extract_event_duration(event) + for event in chronological_events + }) + + async def populate_cache_event_description(self, event: Event) -> None: + """ + Cache `event` description & duration. + + This should be called when entering a new event, and can be called periodically to ensure that the cache + holds fresh information in the case that the event remains the same, but its description changes. + + The duration is stored formatted for the frontend. It is not intended to be used programmatically. + """ + log.debug("Caching event description & duration.") + + await self.cache_information.set("event_description", event.meta.description) + await self.cache_information.set("event_duration", extract_event_duration(event)) + + # endregion + # region: Daemon + + async def maybe_start_daemon(self) -> None: + """ + Start the daemon depending on cache state. + + The daemon will only start if it has been explicitly enabled via a command. + """ + log.debug("Checking whether daemon should start.") + + should_begin: t.Optional[bool] = await self.cache_information.get("daemon_active") # None if never set! + + if should_begin: + self.daemon_loop.start() + + def cog_unload(self) -> None: + """ + Cancel the daemon in case of cog unload. + + This is **not** done automatically! The daemon otherwise remains active in the background. + """ + log.debug("Cog unload: cancelling daemon.") + + self.daemon_loop.cancel() + + async def daemon_main(self) -> None: + """ + Synchronise guild & caches with branding repository. + + Pull the currently active event from the branding repository and check whether it matches the currently + active event in the cache. If not, apply the new event. + + However, it is also possible that an event's assets change as it's active. To account for such cases, + we check the banner & icons hashes against the currently cached values. If there is a mismatch, each + specific asset is re-applied. + """ + log.info("Daemon main: checking current event.") + + new_event, available_events = await self.repository.get_current_event() + + await self.populate_cache_events(available_events) + + if new_event is None: + log.warning("Daemon main: failed to get current event from branding repository, will do nothing.") + return + + if new_event.path != await self.cache_information.get("event_path"): + log.debug("Daemon main: new event detected!") + await self.enter_event(new_event) + return + + await self.populate_cache_event_description(new_event) # Cache fresh frontend info in case of change. + + log.trace("Daemon main: event has not changed, checking for change in assets.") + + if new_event.banner.sha != await self.cache_information.get("banner_hash"): + log.debug("Daemon main: detected banner change.") + await self.apply_banner(new_event.banner) + + if compound_hash(new_event.icons) != await self.cache_information.get("icons_hash"): + log.debug("Daemon main: detected icon change.") + await self.initiate_icon_rotation(new_event.icons) + await self.rotate_icons() + else: + await self.maybe_rotate_icons() + + @tasks.loop(hours=24) + async def daemon_loop(self) -> None: + """ + Call `daemon_main` every 24 hours. + + The scheduler maintains an exact 24-hour frequency even if this coroutine takes time to complete. If the + coroutine is started at 00:01 and completes at 00:05, it will still be started at 00:01 the next day. + """ + log.trace("Daemon loop: calling daemon main.") + + try: + await self.daemon_main() + except Exception: + log.exception("Daemon loop: failed with an unhandled exception!") + + @daemon_loop.before_loop + async def daemon_before(self) -> None: + """ + Call `daemon_loop` immediately, then block the loop until the next-up UTC midnight. + + The first iteration is invoked directly such that synchronisation happens immediately after daemon start. + We then calculate the time until the next-up midnight and sleep before letting `daemon_loop` begin. + """ + log.trace("Daemon before: performing start-up iteration.") + + await self.daemon_loop() + + log.trace("Daemon before: calculating time to sleep before loop begins.") + now = Arrow.utcnow() + + # The actual midnight moment is offset into the future to prevent issues with imprecise sleep. + tomorrow = now.shift(days=1) + midnight = tomorrow.replace(hour=0, minute=1, second=0, microsecond=0) + + sleep_secs = (midnight - now).total_seconds() + log.trace(f"Daemon before: sleeping {sleep_secs} seconds before next-up midnight: {midnight}.") + + await asyncio.sleep(sleep_secs) + + # endregion + # region: Command interface (branding) + + @commands.group(name="branding") + async def branding_group(self, ctx: commands.Context) -> None: + """Control the branding cog.""" + if not ctx.invoked_subcommand: + await ctx.send_help(ctx.command) + + @branding_group.command(name="about", aliases=("current", "event")) + async def branding_about_cmd(self, ctx: commands.Context) -> None: + """Show the current event's description and duration.""" + await self.send_info_embed(ctx.channel.id, is_notification=False) + + @commands.has_any_role(*MODERATION_ROLES) + @branding_group.command(name="sync") + async def branding_sync_cmd(self, ctx: commands.Context) -> None: + """ + Force branding synchronisation. + + Show which assets have failed to synchronise, if any. + """ + async with ctx.typing(): + banner_success, icon_success = await self.synchronise() + + failed_assets = ", ".join( + name + for name, status in [("banner", banner_success), ("icon", icon_success)] + if status is False + ) + + if failed_assets: + resp = make_embed("Synchronisation unsuccessful", f"Failed to apply: {failed_assets}.", success=False) + resp.set_footer(text="Check log for details.") + else: + resp = make_embed("Synchronisation successful", "Assets have been applied.", success=True) + + await ctx.send(embed=resp) + + # endregion + # region: Command interface (branding calendar) + + @branding_group.group(name="calendar", aliases=("schedule", "events")) + async def branding_calendar_group(self, ctx: commands.Context) -> None: + """ + Show the current event calendar. + + We draw event information from `cache_events` and use each key-value pair to create a field in the response + embed. As such, we do not need to query the API to get event information. The cache is automatically + re-populated by the daemon whenever it makes a request. A moderator+ can also explicitly request a cache + refresh using the 'refresh' subcommand. + + Due to Discord limitations, we only show up to 25 events. This is entirely sufficient at the time of writing. + In the case that we find ourselves with more than 25 events, a warning log will alert core devs. + + In the future, we may be interested in a field-paginating solution. + """ + if ctx.invoked_subcommand: + # If you're wondering why this works: when the 'refresh' subcommand eventually re-invokes + # this group, the attribute will be automatically set to None by the framework. + return + + available_events = await self.cache_events.to_dict() + log.trace(f"Found {len(available_events)} cached events available for calendar view.") + + if not available_events: + resp = make_embed("No events found!", "Cache may be empty, try `branding calendar refresh`.", success=False) + await ctx.send(embed=resp) + return + + embed = discord.Embed(title="Current event calendar", colour=discord.Colour.blurple()) + + # Because Discord embeds can only contain up to 25 fields, we only show the first 25. + first_25 = list(available_events.items())[:25] + + if len(first_25) != len(available_events): # Alert core devs that a paginating solution is now necessary. + log.warning(f"There are {len(available_events)} events, but the calendar view can only display 25.") + + for name, duration in first_25: + embed.add_field(name=name[:256], value=duration[:1024]) + + embed.set_footer(text="Otherwise, the fallback season is used.") + + await ctx.send(embed=embed) + + @commands.has_any_role(*MODERATION_ROLES) + @branding_calendar_group.command(name="refresh") + async def branding_calendar_refresh_cmd(self, ctx: commands.Context) -> None: + """ + Refresh event cache and show current event calendar. + + Supplementary subcommand allowing force-refreshing the event cache. Implemented as a subcommand because + unlike the supergroup, it requires moderator privileges. + """ + log.info("Performing command-requested event cache refresh.") + + async with ctx.typing(): + available_events = await self.repository.get_events() + await self.populate_cache_events(available_events) + + await ctx.invoke(self.branding_calendar_group) + + # endregion + # region: Command interface (branding daemon) + + @commands.has_any_role(*MODERATION_ROLES) + @branding_group.group(name="daemon", aliases=("d",)) + async def branding_daemon_group(self, ctx: commands.Context) -> None: + """Control the branding cog's daemon.""" + if not ctx.invoked_subcommand: + await ctx.send_help(ctx.command) + + @branding_daemon_group.command(name="enable", aliases=("start", "on")) + async def branding_daemon_enable_cmd(self, ctx: commands.Context) -> None: + """Enable the branding daemon.""" + await self.cache_information.set("daemon_active", True) + + if self.daemon_loop.is_running(): + resp = make_embed("Daemon is already enabled!", "", success=False) + else: + self.daemon_loop.start() + resp = make_embed("Daemon enabled!", "It will now automatically awaken on start-up.", success=True) + + await ctx.send(embed=resp) + + @branding_daemon_group.command(name="disable", aliases=("stop", "off")) + async def branding_daemon_disable_cmd(self, ctx: commands.Context) -> None: + """Disable the branding daemon.""" + await self.cache_information.set("daemon_active", False) + + if self.daemon_loop.is_running(): + self.daemon_loop.cancel() + resp = make_embed("Daemon disabled!", "It will not awaken on start-up.", success=True) + else: + resp = make_embed("Daemon is already disabled!", "", success=False) + + await ctx.send(embed=resp) + + @branding_daemon_group.command(name="status") + async def branding_daemon_status_cmd(self, ctx: commands.Context) -> None: + """Check whether the daemon is currently enabled.""" + if self.daemon_loop.is_running(): + resp = make_embed("Daemon is enabled", "Use `branding daemon disable` to stop.", success=True) + else: + resp = make_embed("Daemon is disabled", "Use `branding daemon enable` to start.", success=False) + + await ctx.send(embed=resp) + + # endregion diff --git a/bot/exts/backend/branding/_repository.py b/bot/exts/backend/branding/_repository.py new file mode 100644 index 000000000..7b09d4641 --- /dev/null +++ b/bot/exts/backend/branding/_repository.py @@ -0,0 +1,240 @@ +import logging +import typing as t +from datetime import date, datetime + +import frontmatter + +from bot.bot import Bot +from bot.constants import Keys +from bot.errors import BrandingMisconfiguration + +# Base URL for requests into the branding repository. +BRANDING_URL = "https://api.github.com/repos/python-discord/branding/contents" + +PARAMS = {"ref": "main"} # Target branch. +HEADERS = {"Accept": "application/vnd.github.v3+json"} # Ensure we use API v3. + +# A GitHub token is not necessary. However, unauthorized requests are limited to 60 per hour. +if Keys.github: + HEADERS["Authorization"] = f"token {Keys.github}" + +# Since event periods are year-agnostic, we parse them into `datetime` objects with a manually inserted year. +# Please note that this is intentionally a leap year to allow Feb 29 to be valid. +ARBITRARY_YEAR = 2020 + +# Format used to parse date strings after we inject `ARBITRARY_YEAR` at the end. +DATE_FMT = "%B %d %Y" # Ex: July 10 2020 + +log = logging.getLogger(__name__) + + +class RemoteObject: + """ + Remote file or directory on GitHub. + + The annotations match keys in the response JSON that we're interested in. + """ + + sha: str # Hash helps us detect asset change. + name: str # Filename. + path: str # Path from repo root. + type: str # Either 'file' or 'dir'. + download_url: t.Optional[str] # If type is 'dir', this is None! + + def __init__(self, dictionary: t.Dict[str, t.Any]) -> None: + """Initialize by grabbing annotated attributes from `dictionary`.""" + missing_keys = self.__annotations__.keys() - dictionary.keys() + if missing_keys: + raise KeyError(f"Fetched object lacks expected keys: {missing_keys}") + for annotation in self.__annotations__: + setattr(self, annotation, dictionary[annotation]) + + +class MetaFile(t.NamedTuple): + """Attributes defined in a 'meta.md' file.""" + + is_fallback: bool + start_date: t.Optional[date] + end_date: t.Optional[date] + description: str # Markdown event description. + + +class Event(t.NamedTuple): + """Event defined in the branding repository.""" + + path: str # Path from repo root where event lives. This is the event's identity. + meta: MetaFile + banner: RemoteObject + icons: t.List[RemoteObject] + + def __str__(self) -> str: + return f"<Event at '{self.path}'>" + + +class BrandingRepository: + """ + Branding repository abstraction. + + This class represents the branding repository's main branch and exposes available events and assets + as objects. It performs the necessary amount of validation to ensure that a misconfigured event + isn't returned. Such events are simply ignored, and will be substituted with the fallback event, + if available. Warning logs will inform core developers if a misconfigured event is encountered. + + Colliding events cause no special behaviour. In such cases, the first found active event is returned. + We work with the assumption that the branding repository checks for such conflicts and prevents them + from reaching the main branch. + + This class keeps no internal state. All `get_current_event` calls will result in GitHub API requests. + The caller is therefore responsible for being responsible and caching information to prevent API abuse. + + Requests are made using the HTTP session looked up on the bot instance. + """ + + def __init__(self, bot: Bot) -> None: + self.bot = bot + + async def fetch_directory(self, path: str, types: t.Container[str] = ("file", "dir")) -> t.Dict[str, RemoteObject]: + """ + Fetch directory found at `path` in the branding repository. + + Raise an exception if the request fails, or if the response lacks the expected keys. + + Passing custom `types` allows getting only files or directories. By default, both are included. + """ + full_url = f"{BRANDING_URL}/{path}" + log.debug(f"Fetching directory from branding repository: '{full_url}'.") + + async with self.bot.http_session.get(full_url, params=PARAMS, headers=HEADERS) as response: + if response.status != 200: + raise RuntimeError(f"Failed to fetch directory due to status: {response.status}") + + log.debug("Fetch successful, reading JSON response.") + json_directory = await response.json() + + return {file["name"]: RemoteObject(file) for file in json_directory if file["type"] in types} + + async def fetch_file(self, download_url: str) -> bytes: + """ + Fetch file as bytes from `download_url`. + + Raise an exception if the request does not succeed. + """ + log.debug(f"Fetching file from branding repository: '{download_url}'.") + + async with self.bot.http_session.get(download_url, params=PARAMS, headers=HEADERS) as response: + if response.status != 200: + raise RuntimeError(f"Failed to fetch file due to status: {response.status}") + + log.debug("Fetch successful, reading payload.") + return await response.read() + + def parse_meta_file(self, raw_file: bytes) -> MetaFile: + """ + Parse a 'meta.md' file from raw bytes. + + The caller is responsible for handling errors caused by misconfiguration. + """ + attrs, description = frontmatter.parse(raw_file, encoding="UTF-8") + + if not description: + raise BrandingMisconfiguration("No description found in 'meta.md'!") + + if attrs.get("fallback", False): + return MetaFile(is_fallback=True, start_date=None, end_date=None, description=description) + + start_date_raw = attrs.get("start_date") + end_date_raw = attrs.get("end_date") + + if None in (start_date_raw, end_date_raw): + raise BrandingMisconfiguration("Non-fallback event doesn't have start and end dates defined!") + + # We extend the configured month & day with an arbitrary leap year, allowing a datetime object to exist. + # This may raise errors if misconfigured. We let the caller handle such cases. + start_date = datetime.strptime(f"{start_date_raw} {ARBITRARY_YEAR}", DATE_FMT).date() + end_date = datetime.strptime(f"{end_date_raw} {ARBITRARY_YEAR}", DATE_FMT).date() + + return MetaFile(is_fallback=False, start_date=start_date, end_date=end_date, description=description) + + async def construct_event(self, directory: RemoteObject) -> Event: + """ + Construct an `Event` instance from an event `directory`. + + The caller is responsible for handling errors caused by misconfiguration. + """ + contents = await self.fetch_directory(directory.path) + + missing_assets = {"meta.md", "banner.png", "server_icons"} - contents.keys() + + if missing_assets: + raise BrandingMisconfiguration(f"Directory is missing following assets: {missing_assets}") + + server_icons = await self.fetch_directory(contents["server_icons"].path, types=("file",)) + + if len(server_icons) == 0: + raise BrandingMisconfiguration("Found no server icons!") + + meta_bytes = await self.fetch_file(contents["meta.md"].download_url) + + meta_file = self.parse_meta_file(meta_bytes) + + return Event(directory.path, meta_file, contents["banner.png"], list(server_icons.values())) + + async def get_events(self) -> t.List[Event]: + """ + Discover available events in the branding repository. + + Misconfigured events are skipped. May return an empty list in the catastrophic case. + """ + log.debug("Discovering events in branding repository.") + + try: + event_directories = await self.fetch_directory("events", types=("dir",)) # Skip files. + except Exception: + log.exception("Failed to fetch 'events' directory.") + return [] + + instances: t.List[Event] = [] + + for event_directory in event_directories.values(): + log.trace(f"Attempting to construct event from directory: '{event_directory.path}'.") + try: + instance = await self.construct_event(event_directory) + except Exception as exc: + log.warning(f"Could not construct event '{event_directory.path}'.", exc_info=exc) + else: + instances.append(instance) + + return instances + + async def get_current_event(self) -> t.Tuple[t.Optional[Event], t.List[Event]]: + """ + Get the currently active event, or the fallback event. + + The second return value is a list of all available events. The caller may discard it, if not needed. + Returning all events alongside the current one prevents having to query the API twice in some cases. + + The current event may be None in the case that no event is active, and no fallback event is found. + """ + utc_now = datetime.utcnow() + log.debug(f"Finding active event for: {utc_now}.") + + # Construct an object in the arbitrary year for the purpose of comparison. + lookup_now = date(year=ARBITRARY_YEAR, month=utc_now.month, day=utc_now.day) + log.trace(f"Lookup object in arbitrary year: {lookup_now}.") + + available_events = await self.get_events() + log.trace(f"Found {len(available_events)} available events.") + + for event in available_events: + meta = event.meta + if not meta.is_fallback and (meta.start_date <= lookup_now <= meta.end_date): + return event, available_events + + log.trace("No active event found. Looking for fallback event.") + + for event in available_events: + if event.meta.is_fallback: + return event, available_events + + log.warning("No event is currently active and no fallback event was found!") + return None, available_events diff --git a/bot/exts/backend/error_handler.py b/bot/exts/backend/error_handler.py index c643d346e..da0e94a7e 100644 --- a/bot/exts/backend/error_handler.py +++ b/bot/exts/backend/error_handler.py @@ -1,4 +1,5 @@ import contextlib +import difflib import logging import typing as t @@ -8,10 +9,10 @@ from sentry_sdk import push_scope from bot.api import ResponseCodeError from bot.bot import Bot -from bot.constants import Channels, Colours +from bot.constants import Colours, Icons, MODERATION_ROLES from bot.converters import TagNameConverter -from bot.errors import LockedResourceError -from bot.utils.checks import InWhitelistCheckFailure +from bot.errors import InvalidInfractedUser, LockedResourceError +from bot.utils.checks import ContextCheckFailure log = logging.getLogger(__name__) @@ -47,7 +48,6 @@ class ErrorHandler(Cog): * If CommandNotFound is raised when invoking the tag (determined by the presence of the `invoked_from_error_handler` attribute), this error is treated as being unexpected and therefore sends an error message - * Commands in the verification channel are ignored 2. UserInputError: see `handle_user_input_error` 3. CheckFailure: see `handle_check_failure` 4. CommandOnCooldown: send an error message in the invoking context @@ -63,10 +63,9 @@ class ErrorHandler(Cog): if isinstance(e, errors.CommandNotFound) and not hasattr(ctx, "invoked_from_error_handler"): if await self.try_silence(ctx): return - if ctx.channel.id != Channels.verification: - # Try to look for a tag with the command's name - await self.try_get_tag(ctx) - return # Exit early to avoid logging. + # Try to look for a tag with the command's name + await self.try_get_tag(ctx) + return # Exit early to avoid logging. elif isinstance(e, errors.UserInputError): await self.handle_user_input_error(ctx, e) elif isinstance(e, errors.CheckFailure): @@ -78,11 +77,19 @@ class ErrorHandler(Cog): await self.handle_api_error(ctx, e.original) elif isinstance(e.original, LockedResourceError): await ctx.send(f"{e.original} Please wait for it to finish and try again later.") + elif isinstance(e.original, InvalidInfractedUser): + await ctx.send(f"Cannot infract that user. {e.original.reason}") + else: + await self.handle_unexpected_error(ctx, e.original) + return # Exit early to avoid logging. + elif isinstance(e, errors.ConversionError): + if isinstance(e.original, ResponseCodeError): + await self.handle_api_error(ctx, e.original) else: await self.handle_unexpected_error(ctx, e.original) return # Exit early to avoid logging. elif not isinstance(e, errors.DisabledCommand): - # ConversionError, MaxConcurrencyReached, ExtensionError + # MaxConcurrencyReached, ExtensionError await self.handle_unexpected_error(ctx, e) return # Exit early to avoid logging. @@ -156,10 +163,46 @@ class ErrorHandler(Cog): ) else: with contextlib.suppress(ResponseCodeError): - await ctx.invoke(tags_get_command, tag_name=tag_name) + if await ctx.invoke(tags_get_command, tag_name=tag_name): + return + + if not any(role.id in MODERATION_ROLES for role in ctx.author.roles): + await self.send_command_suggestion(ctx, ctx.invoked_with) + # Return to not raise the exception return + async def send_command_suggestion(self, ctx: Context, command_name: str) -> None: + """Sends user similar commands if any can be found.""" + # No similar tag found, or tag on cooldown - + # searching for a similar command + raw_commands = [] + for cmd in self.bot.walk_commands(): + if not cmd.hidden: + raw_commands += (cmd.name, *cmd.aliases) + if similar_command_data := difflib.get_close_matches(command_name, raw_commands, 1): + similar_command_name = similar_command_data[0] + similar_command = self.bot.get_command(similar_command_name) + + if not similar_command: + return + + log_msg = "Cancelling attempt to suggest a command due to failed checks." + try: + if not await similar_command.can_run(ctx): + log.debug(log_msg) + return + except errors.CommandError as cmd_error: + log.debug(log_msg) + await self.on_command_error(ctx, cmd_error) + return + + misspelled_content = ctx.message.content + e = Embed() + e.set_author(name="Did you mean:", icon_url=Icons.questionmark) + e.description = f"{misspelled_content.replace(command_name, similar_command_name, 1)}" + await ctx.send(embed=e, delete_after=10.0) + async def handle_user_input_error(self, ctx: Context, e: errors.UserInputError) -> None: """ Send an error message in `ctx` for UserInputError, sometimes invoking the help command too. @@ -191,10 +234,12 @@ class ErrorHandler(Cog): elif isinstance(e, errors.BadUnionArgument): embed = self._get_error_embed("Bad argument", f"{e}\n{e.errors[-1]}") await ctx.send(embed=embed) + await prepared_help_command self.bot.stats.incr("errors.bad_union_argument") elif isinstance(e, errors.ArgumentParsingError): embed = self._get_error_embed("Argument parsing error", str(e)) await ctx.send(embed=embed) + prepared_help_command.close() self.bot.stats.incr("errors.argument_parsing_error") else: embed = self._get_error_embed( @@ -229,7 +274,7 @@ class ErrorHandler(Cog): await ctx.send( "Sorry, it looks like I don't have the permissions or roles I need to do that." ) - elif isinstance(e, (InWhitelistCheckFailure, errors.NoPrivateMessage)): + elif isinstance(e, (ContextCheckFailure, errors.NoPrivateMessage)): ctx.bot.stats.incr("errors.wrong_channel_or_dm_error") await ctx.send(e) diff --git a/bot/exts/backend/logging.py b/bot/exts/backend/logging.py index 94fa2b139..823f14ea4 100644 --- a/bot/exts/backend/logging.py +++ b/bot/exts/backend/logging.py @@ -29,7 +29,7 @@ class Logging(Cog): url="https://github.com/python-discord/bot", icon_url=( "https://raw.githubusercontent.com/" - "python-discord/branding/master/logos/logo_circle/logo_circle_large.png" + "python-discord/branding/main/logos/logo_circle/logo_circle_large.png" ) ) diff --git a/bot/exts/backend/sync/_syncers.py b/bot/exts/backend/sync/_syncers.py index 2eb9f9971..c9f2d2da8 100644 --- a/bot/exts/backend/sync/_syncers.py +++ b/bot/exts/backend/sync/_syncers.py @@ -5,12 +5,15 @@ from collections import namedtuple from discord import Guild from discord.ext.commands import Context +from more_itertools import chunked import bot from bot.api import ResponseCodeError log = logging.getLogger(__name__) +CHUNK_SIZE = 1000 + # These objects are declared as namedtuples because tuples are hashable, # something that we make use of when diffing site roles against guild roles. _Role = namedtuple('Role', ('id', 'name', 'colour', 'permissions', 'position')) @@ -207,10 +210,13 @@ class UserSyncer(Syncer): @staticmethod async def _sync(diff: _Diff) -> None: """Synchronise the database with the user cache of `guild`.""" + # Using asyncio.gather would still consume too many resources on the site. log.trace("Syncing created users...") if diff.created: - await bot.instance.api_client.post("bot/users", json=diff.created) + for chunk in chunked(diff.created, CHUNK_SIZE): + await bot.instance.api_client.post("bot/users", json=chunk) log.trace("Syncing updated users...") if diff.updated: - await bot.instance.api_client.patch("bot/users/bulk_patch", json=diff.updated) + for chunk in chunked(diff.updated, CHUNK_SIZE): + await bot.instance.api_client.patch("bot/users/bulk_patch", json=chunk) |