* src: WIP Internal Sharding refactor
* src: Refactor unavailable guild check
Co-Authored-By: kyranet <kyradiscord@gmail.com>
* src: More WIP Code
F in the chat to the old manager
* src: It should work but Discord says no.
Seriously why is this not working!
* fix: Inflator causing issues
* src: Finishing touches and typings
* misc: Proper debug message
* fix: Making things hidden needs writable: true as well
* fix: Sharding allowing multiple of the same shard, negative shards or strings
* fix: Again... edge cases
I love you guys .w.
* misc: Touchups
* misc: Better error?
* docs: Typo
* typings: Requested changes
* src: Requested changes
* src: Fix issues, validate provided shard options and more
* src: Forgot to remove the listener
* lint: eslint complaining
* fix: Setting shardCount to auto crashing the process
* misc: Requested changes
* typings: Correct typings for shardCount client option
* typings: Add invalidSession event to the shard and correct typings
* src: Minor docs adjustements, and code consistency between setHelloTimeout and setHeartbeatTimeout
* src: Don't block reconnect while creating shards
Might fix silent disconnects *again*
* src: Prevent reconnect from running if the Manager isn't READY
That way, if a shard dies while we're still spawning, it won't cause issues
* fix: Retry to reconnect if there's a network error going on.
The manager *should* keep reconnecting unless the token is invalid
* src: Enhance onClose handler for shards in the manager
- If the close code is between 1000 and 2000 (inclusive), you cannot resume
I tested this locally
- If there's a session ID still present, immediately try to resume
Faster resumes :papaBless:
Otherwise, the invalid session event will trigger and it'll handle accordingly
I swear if I see a SINGULAR Silent DC I'm yeeting
* src: Fix error check
* src: Make sure message exists on the error
* src: Used the wrong property for the shardQueue
* src: Make the hello timeout be made by the client
god help
* docs: Correct docs for WSEvents
* misc: Remove old events from the Events constant
* src: Throw the HTTP error if we don't get a 401
* typings: Can't forget about them
* src: Implement some more fail safes just in case
Seriously, better safe than sorry! Gotta failproof it completely
* edit shouldn't remove content
If undefined is passed to the api, content isn't removed in such a case where you are only editing the embed.
* fix a related doc string
* update typings
* requested changes
* Remove GroupDMChannels
they sparked no joy
* Start partials for message deletion
* MessageUpdate partials
* Add partials as an opt-in client option
* Add fetch() to Message
* Message.author should never be undefined
* Fix channels being the wrong type
* Allow fetching channels
* Refactor and add reaction add partials
* Reaction remove partials
* Check for emoji first
* fix message fetching
janky
* User partials in audit logs
* refactor overwrite code
* guild member partials
* partials as a whitelist
* document GuildMember#fetch
* fix: check whether a structure is a partial, not whether cache is true
* typings: Updated for latest commit (#3075)
* partials: fix messageUpdate behaviour (now "old" message can be partial)
* partials: add warnings and docs
* partials: add partials to index.yml
* partials: tighten "partial" definitions
* partials: fix embed-only messages counting as partials
* fix: Sharding bugs, silent disconnects and cleanup code
* typings
* fix: Destroy connecting with close code different from 1000
Per `If a client does not receive a heartbeat ack between its attempts at sending heartbeats, it should immediately terminate the connection with a non-1000 close code, reconnect, and attempt to resume.`
* misc: Wait x ms before reconnecting
Per https://discordapp.com/developers/docs/topics/gateway#resuming
* docs
* nit: docs
* misc: Prevent multiple calls to WebSocketManager#destroy
* fix: Implement destroying if you reset the token
* misc: Clear the WS packet queue on WebSocketShard#destroy
You can't send those packets anywhere anymore, so no point in keeping them
* fix: Handle session limits when reconnecting a full shard, cleanup
* misc: No need to create a new shard instance
* fix: closeSequence being null, thus emitting null on Client#resumed
* misc: Remove GUILD_SYNC Gateway handler and add missing dot to string
* misc: Close WS with code 4000 if we didn't get a heartbeat in time
As said in the Discord API server
* fix: Handle ready emitting in onPacket
Doesn't allow broken packets
* misc: Close the connection if Discord asks for a reconnect
Prevents double triggers
* testing: Prevent multiple reconnect attempts on a shard
Should fix some issues some people have had.
* fix: Prevent multiple reconnect calls on the shard, re-use conn to identify, remove reconnect function
Note: Closing the WS with 1000 makes the session invalid
* misc: Forgot to remove 2 unneeded setters
* docs: Wrong param docstring for WebSocketShard#destroy
* misc: Set status to reconnecting after destroying
* misc: Close connection with code 1000 on session invalidated
Allows us to cleanup the shard and do a full reconnect
Also remove identify wait delay, not used anywhere
* fix: Fix zlib crash on node
And with that, the PR is done!
* misc: Implement a reconnect queue
And that is all there was to be done in this PR.
Shards now queue up for a reconnect
* nit: Debug the queue after destroying
* docs: Make the invalidated event clearer
* lint: I'm good at my job
* docs
* docs: Make description for isReconnectingShards accurate
*can I stop finding issues, this PR is meant to be done*
* misc: Remove shard from bind params
* misc: Code re-ordering and cleanup
Resumes do not need to be queued up, as they do not count to the identify limit, and after some testing, they don't have the 5 second delay required, like in identify
* fix: Issues with token regeneration and shards not properly handling them
We close the ws connection with code 1000 if we get an invalid session payload,
that way we can queue the reconnects and handle any issues
* misc: Remove useless delays on session invalidated
They get handled by the rest of the code already
* lint
* misc: reset the sequence on Shard#destroy
This especially is a problem if you need to re-identify, as the sequence doesn't get set to the current one,
causing the sequence to be wrong
* fix: GitHub rebase and minor tweak
* Implement a 15 second timeout if shards don't connect till then
Should prevent shards that never reconnect
* revert: Make WebSocketShard#send and WebSocketManager#broadcast public
* typings: Set type to void instead of undefined
* docs: Requested Changes
- Fixed a bug where `GuildMemberRoleStore#{add,remove}` would create a rejected promise inside a promise, instead of rejecting the first one.
- Fixed a bug where `GuildMember#edit` with a specified unknown channel would throw synchronously, instead of rejecting asynchronously.
* feat({Role,Channel}Store): fetch method
* docs: Add usage examples to the new methods
* misc: Add note of why we are fetching all roles even for a single one
This PR attempts to fix the reported resumed event count in the debug output (where it is always displayed only as 1 event replayed) and in the emitted `resumed` event, where it passed the current sequence instead of passing the actual replayed event count (which was an utopic high number for smaller bots on resume).
* fix(Util): throw an explicit error if a chunk exceeds the max length
* refactor(Util): consolidate both errors in splitMessage into one
* revert(Messages): do not unnecessarily change the error code
* revert(Messages): do not remove the word 'the'
* Add explicit error to setting invalid voice channel
* restrict to guild
Co-Authored-By: Darqam <anhim2@gmail.com>
* add a more explicit error and channel type check
* bad tab
* fix: Sharding causing constant heartbeat / identify spam
* misc: Remove wait param in connect
* misc: Wait 2.5 seconds before sending identify again if session is resumable
* misc: Remove useless destroy call
* nit: Capitalization
* fix: Identify on HELLO not connectionOpen
* misc: Add different intervals for identify after invalid session
- 2500 if we couldn't resume in time
- 5000 if we didn't have a session ID (per the docs on identify, that a client can only connect every 5 seconds)
- Otherwise, just identify again
* misc: Only clear heartbeat if shard is fully dead
Reconnect clears it otherwise
* fix: Accessing .length on a Collection
* internal sharding
* ready event
* the square deal
* the new deal
* the second new deal
* add actual documentation
* the new freedom
* the great society
* federal intervention
* some of requested changes
* i ran out of things to call these
* destroy this
* fix: Client#uptime went missing
* fix(Client): destroy the client on login failure
This may happen duo invalid sharding config / invalid token / user requested destroy
* fix(Client): reject login promise when the client is destroyed before ready
* fix(WebSocketManager): remove redundancy in destroy method (#2491)
* typo(ErrorMessages): duo -> duo to
* typo(ErrorMessages): duo -> due
* fix: docs and options
* docs(WebSocketManager): WebSockethard -> WebSocketShard (#2502)
* fix(ClientUser): lazily load to account for extended user structure (#2501)
* docs(WebSocketShard): document class to make it visible in documentation (#2504)
* fix: WebSocketShard#reconnect
* fix: presenceUpdate & userUpdate
* presenceUpdate wasn't really being handled at all
* userUpdate handled incorrectly because as of v7 in the Discord API, it comes inside presenceUpdate
* re-add raw event
* member is now part of message create payload
* feat: Add functionality to support multiple servers with different shards (#2395)
* Added functionallity to spawn multiple sharding managers due to adding start and end shards
* Small fixes and limiting shard amount to max recommended
* Forgot a check in spawn()
* Fixed indentation
* Removed optiosn object documentation for totalShards
* More fixes and a check that the startShard + amount doesnt go over the recommended shard amount
* fix getting max recommended
* Removed async from constructor (my fault)
* Changed start and end shard to a shardList or "auto" + fixed some brainfarts with isNaN
* Changed the loop and totalShard count calculation
* shards are actually 0 based
* Fixed a problem with the gateway and handled some range errors and type errors
* Changed Number.isNan to isNaN and changed a few Integer checks to use Number.isInteger
* Added check if shardList contains smth greater than totalShards; made spawn use totalShards again; shardList will be ignored and rebuild if totalShards is 'auto'; fixed docs
* ShardingManager#spawn now uses a for..of loop; fixed the if statement inside the new for..of loop to still work as intended; made the totalShards be set to a new amount if smth manual is put into ShardingManager#spawn just like before; Fixed some spelling
* internal sharding
* ready event
* the square deal
* the new deal
* the second new deal
* add actual documentation
* the new freedom
* the great society
* federal intervention
* some of requested changes
* i ran out of things to call these
* destroy this
* fix: Client#uptime went missing
* fix(Client): destroy the client on login failure
This may happen duo invalid sharding config / invalid token / user requested destroy
* fix(Client): reject login promise when the client is destroyed before ready
* fix(WebSocketManager): remove redundancy in destroy method (#2491)
* typo(ErrorMessages): duo -> duo to
* typo(ErrorMessages): duo -> due
* fix: docs and options
* docs(WebSocketManager): WebSockethard -> WebSocketShard (#2502)
* fix(ClientUser): lazily load to account for extended user structure (#2501)
* docs(WebSocketShard): document class to make it visible in documentation (#2504)
* fix: WebSocketShard#reconnect
* fix: presenceUpdate & userUpdate
* presenceUpdate wasn't really being handled at all
* userUpdate handled incorrectly because as of v7 in the Discord API, it comes inside presenceUpdate
* Internal Sharding adaptation
Adapted to internal sharding
Fixed a bug where non ready invalidated sessions wouldnt respawn
* Fixed shardCount not retrieving
* Fixing style
removed unnecessary parenthesis
* Fixing and rebasing
lets hope i didnt dun hecklered it
* Fixing my own retardation
* Thanks git rebase
* fix: assigning member in message create payload
* fix: resumes
* fix: IS wont give up reconnecting now
* docs: add missing docs mostly
* fix: found lost methods
* fix: WebSocketManager#broadcast check if shard exists
* fix: ShardClientUtil#id returning undefined
* feat: handle new session rate limits (#2796)
* feat: handle new session rate limits
* i have no idea what i was doing last night
* fix if statement weirdness
* fix: re-add presence parsing from ClientOptions (#2893)
* resolve conflicts
* typings: missing typings
* re-add missing linter rule
* fix: replacing ClientUser wrongly
* address unecessary performance waste
* docs: missing disconnect event
* fix(typings): Fix 2 issues with typings (#2909)
* (Typings) Update typings to reflect current ClientOptions
* fix(Typings) fixes a bug with Websockets and DOM Types
* fix travis
* feat: allow setting presence per shard
* add WebSocketManager#shardX events
* adjust typings, docs and performance issues
* readjust shard events, now provide shardId parameter instead
* fix: ready event should check shardCount, not actualShardCount
* fix: re-add replayed parameter of Client#resume
* fix(Sharding): fixes several things in Internal Sharding (#2914)
* fix(Sharding) fixes several things in Internal Sharding
* add default value for shards property
* better implement checking for shards array
* fix travis & some casing
* split shard count into 2 words
* update to latest Internal Sharding, fix requested changes
* make sure totalShardCount is a number
* fix comment
* fix small typo
* dynamically set totalShardCount if either shards or shardCount is provided
* consistency: rename shardID to shardId
* remove Client#shardIds
* fix: typo in GuildIntegrationsUpdate handler
* fix: incorrect packet data being passed in some events (#2919)
* fix: edgecase of ShardingManager and totalShardCount (#2918)
* fix: Client#userUpdate being passed wrong parameter
and fix a potential edgecase of returning null in ClientUser#edit from this event
* fix consistency and typings issues
* consistency: shardId instances renamed to shardID
* typings: fix typings regarding WebSocket
* style(.eslintrc): remove additional whitespace
* fix(Client): remove ondisconnect handler on timeout
* docs(BaseClient): fix typo of Immediate
* nitpick: typings, private fields and methods
* typo: improve grammar a bit
* fix: error assigning client in WebSocketManager
* typo: actually spell milliseconds properly
* Add worker-based sharding mode to ShardingManager
* Fix ClientShardUtil mode
* Fix worker not being cleared on shard death
* Update docs and typings
* Clean up Client sharding logic a bit
* Add info about requirements for worker mode
* feat: Add TextChannel#rateLimitPerUser
Rename parameter in TextChannel#setRateLimitPerUser
feat: Add `rateLimitPerUser` param to ChannelData
fix: eslint
* docs: Updated typings
* fix: Requested changes
* fix: rateLimitPerUser being undefined when 0
When `rate_limit_per_user` is 0, the gateway does not send it (but REST does). When this is set to a non-zero number, this property starts to exist. Otherwise this will be `0`. Adding `|| 0` should do the trick changing `undefined` to `0`.
* fix: eslint
* wip: comprehensive permissionOverwrites refactor
* PermissionOverwrites.resolve should Promise.reject()
where a promise is the expected return value
* On second thought, async rewrite to automatically reject on throw
* Fix some docs
* Fix a bug
* fix 2 more bugs
* typings: Updated for latest commit
* typings: Add missing method in GuildChannel
* typings: Add missing `| null` in PermissionOverwriteOption type
* Suggested changes
* feat: add ClientOptions#retryLimit
* hydra needs to learn how to code right
* a default would probably help
* move incrementor & update comment
* clarify docs on Infinity
* fix: better global ratelimit handling in RequestHandler
fix: Remove useless line
fix: Better global ratelimit management
* refactor: Changed RESTManager#globallyRateLimited to be a getter
* refactor: Remove RESTManager#globallyRateLimited getter
* docs: Updated comments to reflect latest changes
* Docs: merge PermissionOverwriteOptions and OverwriteData
* fix user-member and replace eslint-disable with linebreaks
* consistency fix
* Fix: remove empty line between jsdocs and method
* Fix: allowed/denied to allow/deny to conform with recent master changes
* refactor: clean up rate limit handling
* requested changes
* remove request mode option
* fix dupe requests
* hardcode reaction ratelimits
* suggested changes
* fix small thing
* re-add restTimeOffset
* move restTimeOffset a bit
* i swear i know english its my native language ok
* requested changes
* fix: a bit *too* pre-emptive with ratelimits, now less so
* fix: dapi error shoudl reject with path
* fix: make errors in execute catchable
* fix promise return
* rebase is hard
* misc: Update ClientApplication for the current API
* cleanup: ClientApplication#fetchAssets, removed createAsset
* Major cleanup time
* Merge to kyra's branch
* docs: Updated typings
* fix: re-add ClientApplication#{cover,coverImage()}
* typings(ClientApplication): move coverImage declaration up
* Rework createMessage
- MessageAttachment is now structurally similar to FileOptions
- No longer mutates the object passed as options
- Supports more permutations of arguments
* Ignore complexity warning
* Refactor name finding
* Fix typo
* Update typings
* Default name to null for MessageAttachment
* Make Message#reply use transformOptions
* Move transformOptions
* Fix Message#reply
* Fix mutation
* Update tests
* Fix options passing
* Refactor into APIMessage
* Fix webhook send
* Expose APIMessage
* Add documentation
* Add types
* Fix type doc
* Fix another type doc
* Fix another another type doc (is this one even right!?)
* Remove trailing comma
* Properly clone split options
* Add support for sending file as stream
* Missed a doc
* Resolve files only once when splitting messages
* This looks nicer
* Assign directly
* Don't cache data and files
* Missing return type
* Use object spread instead Object.assign
* Document constructors
* Crawl is a little dot
* comp pls
* tests: sanitize local file path, disable no-await-in-loop
* abstract BitField from Permissions
* reduce useless code, improve docs
* add a ReadOnly identifier to the return type of Bitfield#freeze()
https://www.typescriptlang.org/docs/handbook/release-notes/typescript-2-1.html#partial-readonly-record-and-pick
* fix the RangeError
* update docs, convert Speaking and ActivityFlags to bitfields
* fix some docs
* Fix Speaking BitField oops
* docs for oops
* more incorrect docs
* Fix incorrectly named property
* add new classes to index
* fix missing @extends docs
* default bitfield resolve to 0, and cleanup defaulting everywhere
Also removes GuildMember#missiongPermissions() alias that had incorrect behavior
* Breaking: Rename Overwrite allowed and denied to allow and deny
To be consistent with the api's naming
* fix setSpeaking usage to bitfields instead of booleans
* fix speaking bug in playChunk
* docs: Updated typings
* fix: BitFieldResolvable should use RecursiveArray
* bugfix/requested change
* typings: Cleanup (#2)
* typings: Fix BitField#{toArray,@@iterator} output type
* typings: correct PermissionOverwrites property names and nitpicks
* src: Handle WEBHOOK_UPDATE events
* Commit rename of 77'
Or adding the letter S
* I missed this
* Properly do this now
Typos everywhere
* Typings
* refactor: remove now unnecessary guild variable
* Error for guild.fetchVanityURL feature not enabled
* added GET/guilds/{guild.id}/vanity-url endpoint
* fix: code conventions
* adopted suggestion
* Changed error message according to change request
* Renamed method
to indicate that only the code is fetched, not the entire url.
* Update Guild.js
* fix: Capture stack traces in APIRouter to preserve async stack
* fix: Handle the stack trace better
* fix: Check if error is an instance of Error (5XX rejects with Result)
* fix: Error.captureStackTrace is not supported in all browsers
* Fix guild being a guild and not an AbstractHandler in PresenceUpdate
* update the default guild to be a Guild, and not GuildMember
* getters return null instead of undefined
* fix lint
* fix: Util.basename being unreliable
new URL for WHATWG parsing was not chosen because it only works for URLs and threw in local pathes, path.basename is unreliable (according to the devs' note), and path.parse seems to work well.
* docs: Update Util.basename's description
* add Collection as possible param to GuildChannel#overwritePermissions
* change PermissionoverwriteOptions desc. default to unset
* fix eslint hiccup
* fix spelling of snowflake
* [WIP] Remove user bots
* more backend userbot removal
* Add mfaEnabled back
* revert client presences store removal
* partially revert getAuth changes
* remove more no longer used children of ClientUserGuildSettings
* fix a bug with this pr and TextBasedChannel.applyToClass
* remove a syncGuilds reference
* more user bot data handling
* various guildSync cleanup
* bots can't call logout
Had the user/bot portions of the code mixed up. Though, does this need to be a promise anymore?
* make ClientManager#destroy() sync
It nolonger needs to be a promise, and nothing depended on it being a promise that I can tell.
* requested change
* Fix massive error
* no longer used as it's userbot only
* Hide Webhook#token (consistency with Client#token)
* Make `Webhook#token` writable
* fix: devsnek's requested change
Webhook#token must be both writable and configurable.
* Add timeDifference calcs back into ratelimits
And fix x-ratelimit-reset is in seconds, not ms: https://puu.sh/AIXxY/9b3989b248.png
* mutate reset time with date header instead
* fix defaulting of reset and remaining
if the reset header is not available, then the reset time should be Date.now() instead of the difference between the date header and Date.now()
If the date header is null, fall back to Date.now() since it's the best we can do (this should never happen, but safer to handle it just in case)
if remaining is 0 we don't want to default it back to 1 since 0 is falsy
The node documentation fails to correctly document when the backlog of unsent messages exceeds a certain threshhold the function will return false. This does not mean it will refuse to send- simply that it will take time. Issue in point: https://github.com/nodejs/node/issues/7657#issuecomment-240581726
* Fixed a bug where Guild#equals would cause the given guild to lose its features
* Fix Util.arraysEqual
* Fixed docs for Util.arraysEqual
* Remove Util.arraysEqual
* Rework collection
* Fix docs
* Remove (prop, value) behavior
* Add back thisArg
* Put examples on same line
* Revert map to return array
* Make examples less specific
* Add thisArg to sweep and partition
* feat(GuildEmoji#deletable)
* Update: Discord API returns 50001
When you try to get or delete a managed emoji from the API, you get a 50001 (Missing Access) error code.
* refactor: reduced duplication in role stores
* fix docs
* fix: typo
* most of requested changes
* rest of changes
* ACTUAL rest of changes
* docs
* doooocs
* reason
* fix message embed json serialization
remove the toJSON method on message embeds so the raw data is exposed
for JSON seralization. this removes the hexColor property, but it
probably should not have been there in the first place. fixes#2419
* change api transform to tojson
Fixes#2364
Discord sends those timestamps packed as SMALL_BIG_EXT, which get converted to strings in js.
~~Maybe they are already preparing their timestamps for 2038.~~
* docs/feat(WebsocketOptions): Parse ws options presence
Allow the `presence` property in `WebsocketOptions` to be used the same way
as `ClientUser#setPresence`.
* Move presence options to top level
* Prepare to work on moving all role functions to a Store
And yes, this is *another* patch branch cause I messed up my master branch to hell
* Move all emoji role related functions to its own store
Tested everything and it works! (With a reload of the client)
Also had to change a value in DataStore#holds
holds.name for GuildEmojis would return the emoji name instead of the class name
* New Line
* Thanks JS for circular dependency!
Because we can't have nice things...
* Do space's request
* Fix equals
* Fix space's point.
Raw API data has the role property as an array of IDs
* Ignore this I need a patch branch for Git
* Move all member role related stuff to a new DataStore
GuildMemberRoleStore is a new store that holds all member role related stuffs
Because its consistent!
* Minorest doc fix ever
To whoever did this object, they forgot a `{`
* Fix the spacing in the docs
* Resue the stores resolve rather than copy paste
Cause I'm dum and it overwrite resolve to the guild role stores resolve soo
* Fix some requests
- Removed the bs private functions
- Set the roles in the constructor
But, I need feedback. There is no way, that I saw, to make a member have roles whenever GuildMmber#_patch is called,
due to the roles being null on the guild. So the only way might be the for loop and getter.
* Fix an issue that I caused in #add
I was testing some other things, and changed that to test. Forgot to change it back
* Actually make the store generate just once when the member is created by first initializing the roles in the guild
Also replaces GuildMember#_roles with GuildMemberRoleStore#_roles
Tested all functions to make sure the expected result happens.
* I missed this from moving remove from GuildMember to GuildMemberRoleStore
* Fix RoleStore#create docs
For real this time
* Do all the requested changes
- Rename all `somethingRole` to `something` (hoistRole => hoist)
- Refactor add and remove to be cleaner and re-use set
* Fix a bug where the store would loose some roles due to null roles that can throw an error in the for loop after they've been deleted
* Remove the `role.id || role` part of the add and remove functions as Appel suggested
* Replace roles.resolve with roles.resolveID for GuildMemberRoleStore#remove
* Don't use Array.isArray in checks
Use instanceof instead
* Woops, I forgot to change this
Renamed colorRole to color
* The docs have dots, so we place the dots back
This leads to GuildMemberStore#_fetchMany to always reject
because it expects more member than possible.
Also no longer call the GuildMemberRemove handler locally
to not decrement twice.
The error from something like client.on('ready', () => undefined.f);
would just be emitted as debug event instead of being thrown.
Simply moving the emitting part out of the try catch again solves this.
* make EmojiStore not Private anymore.
because why have something private if there is priority functionality on that class? also that causes that the docs wont show it directly
* make GuildChannelStore not private anymore
because why have something private if there is priority functionality on that class? also that causes that the docs wont show it directly
* make RoleStore not private anymore
because why have something private if there is priority functionality on that class? also that causes that the docs wont show it directly
* make ReactionStore not private anymore
because why have something private if there is priority functionality on that class? also that causes that the docs wont show it directly
* make all non private to stay consistent
* fix merge conflicts because of other PRs.
