| Commit message (Collapse) | Author | Age | Lines |
|---|
| |
|
|
|
| |
Hannukah was previously in the Christmas folder, which was weird.
This now moves it to its own folder under Holidays.
|
| |
|
|
|
|
|
|
|
|
| |
Moves Advent of Code and Hacktoberfest into an events
folder. Although these are roughly associated with holidays,
they are standalone events that we have participated in
in the past.
Therefore they're being moved to an events folder
separate from the "fun" or "holidays" folders.
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
Timezones are hard :(
|
| | |
|
| |
|
|
|
|
|
|
| |
This is the only place in the codebase that uses pytz, so we can remove the
dependancy by chaging this to use arrow.
I chose to use America/Chicago arbitrarily, for no other reason that being the
first tz google returned that was in EST.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
| |
Removes roles and channels from constants.py that are not used anywhere
in the project.
Signed-off-by: Hassan Abouelela <[email protected]>
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
;
|
| | |
|
| | |
|
| |
|
|
| |
The old name was still in use in the cog_unload, making it fail
|
| |
|
|
|
|
| |
Switches all instances of override_in_channel to whitelist override.
Signed-off-by: Hassan Abouelela <[email protected]>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The Advent of Code Status Countdown task needs to wait for two things to
happen to prevent it from failing during the startup sequence:
1. The Websocket instance discord.py creates needs to be available as
an attribute of the bot, otherwise discord.py fails internally:
Traceback (most recent call last):
File "discord/client.py", line 1049, in change_presence
await self.ws.change_presence(
activity=activity, status=status, afk=afk
)
File "advent_of_code/_cog.py", line 52, in countdown_status
await bot.change_presence(activity=discord.Game(playing))
AttributeError: 'NoneType' object has no attribute 'change_presence'
2. Allegedly, according to the discord.py community, trying to change
the status too early in the sequence to establish a connection with
Discord may result ub the Discord API aborting the connection.
To solve this, I've added a `wait_until_guild_available` waiter, as it
guarantees that the websocket is available and the connections is
mature.
Kaizen: I've changed the name `new_puzzle_announcement` to
`new_puzzle_notification` to better reflect its function.
|
| |\
| |
| |
| |
| |
| | |
# Conflicts:
# bot/exts/christmas/advent_of_code/_cog.py
# bot/exts/christmas/advent_of_code/_helpers.py
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Currently, our Advent of Code background tasks fail without logging
errors or printing error messages. This makes it difficult to debug the
errors and means that they may fail silently.
While we should ideally find the root cause that hides such errors, I've
added a done_callback function in the meantime to help us debug the
current issues with the Advent of Code Notification Task.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| | |
AOC_WHITELIST was changed to AOC_WHITELIST_RESTRICTED because it is
clearer that commands
with this parameter in the `@override_in_channel()` decorator
will be restricted to the aoc commands channel and not be
allowed in the main aoc channel.
In the same vein, AOC_WHITELIST_PLUS was changed to AOC_WHITELIST.
|
| | |
| |
| |
| | |
Please -= 1
|
| | |
| |
| |
| | |
Per Mark's comment, re-raising the error isn't necessary.
|
| | |
| |
| |
| |
| | |
I'm a bit ahead of the game and changing the error handler to match
the new style that Iceman will PR shortly.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| | |
If any of the "spammier" commands (stats, leaderboard) are used within
the primary advent of code channel, rather than a non-specific embed
we instead reply with the channel they should be using.
This also adds a "AOC_WHITELIST_PLUS" constant that makes it easier to
adjust what channels the non-spammier aoc commands can be used in.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Commands like `.aoc leaderboard` and `.aoc stats` proved to be spammy
in the main advent of code channel.
An aoc_commands channel has been added for aoc commands
and this update prohibits aoc commands from being used in the primary
aoc channel and adds the comands channel to the whitelist.
This also specifically allows the less spammier commands: join,
subscribe, unsubscribe, and countdown in the primary channel to foster
discussion though.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Unfortunately, an expired session cookie wreaked havoc to our Advent of
Code commands: All commands that relied on leaderboard data failed
because we couldn't refresh our data and the cache had expired.
To mitigate an expired session, I've added a fallback session feature
that enables us to try again with a different session. While it will
issue an error message to inform us to refresh the expired session
cookie, it does mean that the functionality should continue to work in
the mean time.
The fallback session cookie is currently set to my session cookie, using
an environment variable, `AOC_FALLBACK_SESSION`. It is important that
the user connected to the session is a member of all boards and that
it's a fresh session: We don't want our fallback to expire!
At the same time, while a single fallback session works, the AoC website
also does not like too many requests from a single user. That's why
we'll still use a multi-session model under normal circumstances.
To check for expired sessions, I've added a URL check: The Advent of
Code website will silently redirect people with an expired session,
issuing an 200: OK status as usual. The only way to really check for it
is by comparing the final URL in the response object to the URL we set
out to GET. I've added a custom exception to signal such an unexpected
redirect.
Finally, instead of having the commands just break, I've added an
Exception signal that propagates back to the caller. The solution, with
try-except, is a bit hacky and could benefit from an actual error
handler, but I wanted to get things fixed first; polish can be added
later.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
We noticed that some entries on our leaderboard had an incorrect star
count attached to their name.
After a bit of digging, @HassanAbouelela discovered that this was caused
by the use of the member's name as the key for the leaderboard
dictionary: If different accounts used the same display name for the
leaderboard, they'd be combined into one glitched score dict.
The fix @HassanAbouelela wrote is to use the member id instead of the
name as the key for the leaderboard. I've changed a few names here and
there, but nothing major.
This commit closes #536
|
| | |
| |
| |
| |
| |
| |
| |
| |
| | |
I've updated some docstrings to include more information about the inner
workings of some of the functions. In addition, I've also slightly
reformulated some block comments to improve their grammar.
Kaizen change: There was a redundant list comprehension in the Advent
of Code section of the constants. I've removed it.
|
| | |
| |
| |
| |
| |
| |
| | |
The helper function calculates the time left until the next midnight in
the EST timezone, not necessarily the next midnight during and Advent of
Code event. To prevent confusion, I've clarified its function by
changing the name of the function and its docstring.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Instead of cancelling the task when it starts up outside of the
boundaries of the Advent of Code, the task will now hibernate until just
before the event starts if starts up before December. This allows it to
actually announce the first puzzle.
After the announcement for the last day of the current event is made, it
will terminate itself. It will only start hibernating again when we've
updated the environment variables for next year's event, ensuring that
it does not run unnecessarily.
To prevent issues with the guild cache not being available, I've added
our new `wait_until_guild_available` waiting function. I've also moved
the calls that get teh role/channel to before the loop, as there's no
need to get them each time the loop goes around.
I've also changed the way we calculate the time we need to sleep, as the
old way used truncated seconds, meaning that we would always wake up
relatively early. Instead, I'm now using fractional seconds, which means
we can reduce the safety padding to a fraction of second. More accurate
announcement timing!
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Trying to change the rich presence status of the bot too early in the
bot's start-up sequence will cause the task to fail. To make it wait,
I've added a `bot.wait_until_guild_available` point before the main loop
of the task starts.
This seems to solve the issue reliably, despite the `guild_available`
event not being directly related to when the bot's connection with the
API is ready to accept presence updates. However, the `ready` event is
know to fire too early, according to the discord.py community.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
I've refactored the rich presence countdown task by making it hibernate
until 2 hours before the next Advent of Code starts if the task starts
up before the event has started. This ensures that the task will run
when the event starts and allows it to countdown to the first challenge.
After the event for the configured Advent of Code year has finished, the
task will terminate. This also means that the task will terminate
immediately in the year following the currently configured Advent of
Code; it will only start hibernating again once we configure the bot for
the next event.
No unnecessary, year-long hibernation.
|
| |/
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Our Advent of Code background tasks were written in such a way that
they relied on the extension being loaded in December and only in
December.
However, since the deseasonification, the extension is loaded prior to
December, with some commands being locked to that month with a check
instead. This meant that our background tasks immediately cancelled
themselves, as they observed themselves to be running outside of the
boundaries of the event. As there was no mechanism for starting them
back up again, these tasks would only start running again after
redeployment of Sir Lancebot.
To solve this issue, I've added a helper function that allows tasks to
wait until a x hours before the event starts. This allows them to wake
up in time to prepare for the release of the first puzzle.
|
| |
|
|
|
|
|
| |
The daily stats function contained a bug that prevented it from working
correctly. The reason was that I was looking for `int` keys where the
actual keys were strings. I now make sure to create a `str` from the
`int` I get back from `range`.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
I've added support for ignoring scores from specific days. A list of
days to ignore can be provided using the environment variable
`AOC_IGNORED_DAYS` as a comma-separated list.
This example would ignore day 1 and day 23:
AOC_IGNORED_DAYS=1,23
I've also added a helper function to sort the leaderboard not only on
the achieved score, but also on the number of stars an individual has
completed.
|
| | |
|
| | |
|
| |
|
|
|
| |
Note: This won't start the countdown functions yet, they still rely on
the cog being loaded in december.
|
| |
|
|
|
| |
I've moved the helper functions to the _helpers.py module and clarified
the docstring of the `is_in_advent` helper function.
|
| |
|
|
|
| |
I accidentally removed the global leaderboard command. I've added it
back!
|
Janine vN
Xithrius
decorator-factory
wookie184
Chris
ToxicKidz
ToxicKidz
Hassan Abouelela
Joe Banks
Matteo Bertucci
Sebastiaan Zeeff