-
-
Notifications
You must be signed in to change notification settings - Fork 736
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
v3: API design review #1340
base: blank
Are you sure you want to change the base?
v3: API design review #1340
Conversation
public interface IMessageChannel : IChannel | ||
{ | ||
Task<IUserMessage> SendMessageAsync(string? text = null, bool isTTS = false, IEmbed? embed = null, RequestOptions? options = default); | ||
Task<IUserMessage> SendFileAsync(Stream fileStream, string filename, string? text = null, bool isTTS = false, IEmbed? embed = null, RequestOptions? options = default); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Could maybe considering a FileProperties
class? This would mean you wouldn't need the SendFileAsync
overloads/extensions.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Or as simple as SendFileOptions
. Make it a struct
if you're terribly worried about allocations.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Can we have a way to upload multiple files per message, too? I know this is supported (mobile clients do it!). Perhaps an overload taking an IEnumerable<SendFileOptions>
.
Task SyncPermissionsAsync(RequestOptions? options = default); | ||
|
||
// review: do we want long, paramaterized methods like this, or would we rather a CreateInviteProperties | ||
Task CreateInviteAsync(int? maxAge = 86400, |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think a properties class is much cleaner.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I also lean towards CreateInviteProperties
or CreateInviteOptions
.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
So, should we do the same for stuff like ChannelPermissions, lest we forget the best url: https://docs.stillu.cc/api/Discord.ChannelPermissions.html#Discord_ChannelPermissions__ctor_System_Boolean_System_Boolean_System_Boolean_System_Boolean_System_Boolean_System_Boolean_System_Boolean_System_Boolean_System_Boolean_System_Boolean_System_Boolean_System_Boolean_System_Boolean_System_Boolean_System_Boolean_System_Boolean_System_Boolean_System_Boolean_System_Boolean_System_Boolean_System_Boolean_
// todo: docs | ||
namespace Discord | ||
{ | ||
// review: marker interfaces suck! should we just have an "IsNews" property on ITextChannel? |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
IsNews
property, yes.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Same comment as above. If there is no functional/operational difference in the API, a property is simpler.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
So what happens when discord decides to add (more) special functionality to news channels? (they can switch between news and text types) Or, are we still sticking to treating them just like text channels and not supporting them (like store channels)?
|
||
Task ModifyAsync(MemberProperties properties, RequestOptions? options = default); | ||
|
||
// review: should these two be extensions on IGuild? |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
There should at least be an id version of it on IGuild.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
That thought should apply to almost all API functionality. Except, perhaps, when there should be a "does this ID exist" check, in which case performing the Get()
call serves that purpose.
{ | ||
string Username { get; } | ||
string Discriminator { get; } | ||
// review: does anyone use DiscriminatorValue? seems like a waste of memory |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
It is a waste, yes.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I can find no usage of it that I wouldn't be happy to replace.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Would it be better to make this a ushort and people can just ToString() it if needed? I feel like it could stay consistent since discriminators are numbers.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think that it would be most appropriate to keep Discriminators strings to match the API: https://discordapp.com/developers/docs/resources/user#user-object
There's less risk of things breaking should they change what discriminators can be.
public Optional<IEnumerable<SnowflakeOrEntity<IRole>>> Roles { get; set; } | ||
public Optional<SnowflakeOrEntity<IVoiceChannel>?> VoiceChannel { get; set; } | ||
|
||
public Model ToWumpus() |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Isn't this meant to be internal?
src/Discord.Net/IDiscordClient.cs
Outdated
|
||
event Action<IMessage> MessageCreated; | ||
// todo: implement the rest of the events | ||
// review: event classes or just Action<A,B,C> types |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
the EventArgs pattern is probably better.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
EventArgs
or similar please. For one, we do basically that anyway, in a wrapper layer around the client. For two, it's WAY better from a documentation standpoint. For example, GuildMemberUpdated
today has a signature of
public event Func<SocketGuildUser, SocketGuildUser, Task> GuildMemberUpdated;
If I don't already know how this event works, and want to know, which user is the "old" version and which is the "new" one, the only way to really know for sure is to go dig through the source on github and see that when it's actually invoked, it's...
await TimedInvokeAsync(_guildMemberUpdatedEvent, nameof(GuildMemberUpdated), before, user).ConfigureAwait(false);
|
||
ValueTask<IUser> GetUserAsync(ulong id, StateBehavior stateBehavior, RequestOptions? options); | ||
// review: IReadOnlyCollection is proper collection type here? | ||
ValueTask<IReadOnlyCollection<IUser>> GetUsersAsync(StateBehavior stateBehavior, RequestOptions? options); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yes. As long as the immutable collections aren't used (they're very slow and memory heavy).
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Not sure I agree with that assessment of immutables, in general, but I definitely agree that IReadOnlyCollection
is the best bet here. Immutables would be appropriate if you KNOW your consumer is going to be doing a lot of collection manipulation, and also wants immutability.
The other option would be IReadOnlyList
but do we really expect consumers to want indexing by position?
{ | ||
public class DiscordClientConfig | ||
{ | ||
public string? Token { get; set; } |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Does this mean that TokenType is dead, and will be assumed based on the type of client that is used?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Not specifically, I just hadn't implemented it yet. Though, this client will only support bot tokens, as it makes more sense for OAuth users to just use Wumpus directly.
{ | ||
public string? Token { get; set; } | ||
|
||
public int Shard { get; set; } = 0; |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Are the sharded client and websocket client going to be merged?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Since we're just wrapping over a WumpusGatewayClient and not actually implementing any client logic ourselves here, I think it seems reasonable to allow an IDiscordClient to support multiple shards.
Task SyncPermissionsAsync(RequestOptions? options = default); | ||
|
||
// review: do we want long, paramaterized methods like this, or would we rather a CreateInviteProperties | ||
Task CreateInviteAsync(int? maxAge = 86400, |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
So, should we do the same for stuff like ChannelPermissions, lest we forget the best url: https://docs.stillu.cc/api/Discord.ChannelPermissions.html#Discord_ChannelPermissions__ctor_System_Boolean_System_Boolean_System_Boolean_System_Boolean_System_Boolean_System_Boolean_System_Boolean_System_Boolean_System_Boolean_System_Boolean_System_Boolean_System_Boolean_System_Boolean_System_Boolean_System_Boolean_System_Boolean_System_Boolean_System_Boolean_System_Boolean_System_Boolean_System_Boolean_
ulong EveryoneRoleId { get; } | ||
ulong OwnerId { get; } | ||
|
||
int AFKTimeout { get; } |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
timespan?
|
||
ValueTask<IReadOnlyCollection<IAttachedEmote>> GetEmotesAsync(StateBehavior stateBehavior = default, RequestOptions? options = default); | ||
ValueTask<IAttachedEmote> GetEmoteAsync(StateBehavior stateBehavior = default, RequestOptions? options = default); | ||
Task CreateEmoteAsync(string name, Image image, RequestOptions? options = default); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Should this consume an EmoteBuilder instead?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
EmoteBuilder is a static class for constructing IEmote/IGuildEmote from strings; see above comment regarding DiscordClientBuilder. I'll probably just replace this with IEmote.Parse/IGuildEmote.Parse
|
||
bool IsRevoked { get; } | ||
bool IsTemporary { get; } | ||
int? MaxAge { get; } |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Timespan?
bool IsManaged { get; } | ||
bool IsMentionable { get; } | ||
|
||
GuildPermissions Permissions { get; } // review: use wump permissions directly or wrap? |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I don't see any point in wrapping imo
IMemberPresence Presence { get; } | ||
|
||
DateTimeOffset? JoinedAt { get; } | ||
string Nickname { get; } |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
nit: thoughts on a DisplayName extension that does Nickname ?? Username
, since it's asked for so often
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Seconded.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
sounds good to me, I guess we can just make this a default interface implementation, since it doesn't look like extension everything is going to make the cut for c#8
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks a ton for the opportunity to comment.
// todo: docs | ||
namespace Discord | ||
{ | ||
// review: another marker interface 😕 |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think it's appropriate to split the different types of entities via the typing system, rather than simple properties, when the set of functionality for that entity diverges. I.E. when there are operations that can be performed only upon a category channel, and operations that cannot be performed upon a category channel, even if they don't appear directly upon the interface. Such operations can then be defined to operate specifically upon ICategoryChannel
, which provides a more self-documented, foolproof API.
If no such operations exist for ICategoryChannel
then it probably makes more sense as just a property.
public interface IMessageChannel : IChannel | ||
{ | ||
Task<IUserMessage> SendMessageAsync(string? text = null, bool isTTS = false, IEmbed? embed = null, RequestOptions? options = default); | ||
Task<IUserMessage> SendFileAsync(Stream fileStream, string filename, string? text = null, bool isTTS = false, IEmbed? embed = null, RequestOptions? options = default); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Or as simple as SendFileOptions
. Make it a struct
if you're terribly worried about allocations.
Task SyncPermissionsAsync(RequestOptions? options = default); | ||
|
||
// review: do we want long, paramaterized methods like this, or would we rather a CreateInviteProperties | ||
Task CreateInviteAsync(int? maxAge = 86400, |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I also lean towards CreateInviteProperties
or CreateInviteOptions
.
// todo: docs | ||
namespace Discord | ||
{ | ||
// review: marker interfaces suck! should we just have an "IsNews" property on ITextChannel? |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Same comment as above. If there is no functional/operational difference in the API, a property is simpler.
src/Discord.Net/IDiscordClient.cs
Outdated
|
||
event Action<IMessage> MessageCreated; | ||
// todo: implement the rest of the events | ||
// review: event classes or just Action<A,B,C> types |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
EventArgs
or similar please. For one, we do basically that anyway, in a wrapper layer around the client. For two, it's WAY better from a documentation standpoint. For example, GuildMemberUpdated
today has a signature of
public event Func<SocketGuildUser, SocketGuildUser, Task> GuildMemberUpdated;
If I don't already know how this event works, and want to know, which user is the "old" version and which is the "new" one, the only way to really know for sure is to go dig through the source on github and see that when it's actually invoked, it's...
await TimedInvokeAsync(_guildMemberUpdatedEvent, nameof(GuildMemberUpdated), before, user).ConfigureAwait(false);
src/Discord.Net/IDiscordClient.cs
Outdated
event Action Disconnected; | ||
event Action Ready; | ||
|
||
event Action<IChannel> ChannelCreated; |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Are events no longer asynchronous? If not, presumably we become responsible for dispatching the event to a new async
context, if desired?
I'm totally cool with that, since that's what we do currently anyway, but it might be a good idea to have a built-in mechanism to do said dispatching, for beginner consumers. Something like an injectable IDiscordEventDispatcher
, with a couple built-in implementations?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yes, users are now responsible for dispatching to an async context. (see the included example)
I'm not familiar with what you mean by providing something like an IDiscordEventDispatcher
, if you could link me to an example of how that might look I would be happy to include it.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
In my mind, the ideal way to handle discord events is the same way ASP.NET handles HTTP requests: each incoming message spawns a new "logical" thread. I.E. a new IServiceScope
in the DI system, and a spawned Task
that is simply not await
ed, so that it runs in parallel to other "logical" threads.
For the current Discord.NET version, we do this with several pieces fitting together:
https://github.com/discord-csharp/MODiX/blob/master/Modix.Common/Messaging/MessageDispatcher.cs
https://github.com/discord-csharp/MODiX/blob/master/Modix.Common/Messaging/INotificationHandler.cs
https://github.com/discord-csharp/MODiX/blob/master/Modix.Services/Core/DiscordSocketListeningBehavior.cs
You can see how subscribing to each event we need, wrapping up the args into an object, and then handing off the work to the dispatcher is a tad tedious.
An IDiscordEventDispatcher
could look like the following...
public interface IDiscordEventDispatcher
{
Task DispatchEvent(IEnumerable<Func<Task>> eventHandlers);
}
Where you could have the following built-in implementations that provide behavior similar to Discord.Commands.RunMode
.
public class SyncDiscordEventDispatcher
: IDiscordEventDispatcher
{
public Task DispatchEvent(IEnumerable<Func<Action>> eventHandlers)
{
foreach(var eventHandler in eventHandlers)
await eventHandler.Invoke();
}
}
public class ParallelDiscordEventDispatcher
: IDiscordEventDispatcher
{
public Task DispatchEvent(IEnumerable<Func<Task>> eventHandlers)
{
foreach(var eventHandler in eventHandlers)
_ = eventAction.Invoke();
return Task.CompletedTask;
}
}
You can throw in some basic error handling around the Task
s that aren't await
ed, or leave that to the consumer.
In our case, we'd implement a custom dispatcher that also handles DI scoping, since Discord.NET doesn't have any built-in DI. Which, actually, now that I think about it, wouldn't fit with the above interface definition.
So, if you're interested at all, it definitely needs some brainstorming. But, I think the concept of handling messages the way ASP.NET handles HTTP requests would be appealing to a lot of consumers.
|
||
namespace Discord | ||
{ | ||
internal class ConcurrentMemoryStateProvider : IStateProvider |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This makes me smile.
|
||
ValueTask<IUser> GetUserAsync(ulong id, StateBehavior stateBehavior, RequestOptions? options); | ||
// review: IReadOnlyCollection is proper collection type here? | ||
ValueTask<IReadOnlyCollection<IUser>> GetUsersAsync(StateBehavior stateBehavior, RequestOptions? options); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Not sure I agree with that assessment of immutables, in general, but I definitely agree that IReadOnlyCollection
is the best bet here. Immutables would be appropriate if you KNOW your consumer is going to be doing a lot of collection manipulation, and also wants immutability.
The other option would be IReadOnlyList
but do we really expect consumers to want indexing by position?
/// abide by design implications of this flag. Once Discord.Net has called out to the state provider with this | ||
/// flag, it is out of our control whether or not an async method is evaluated. | ||
/// </remarks> | ||
SyncOnly = 1, |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Possible need for clarification: What's the difference between SyncOnly
and CacheOnly
?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
SyncOnly
requests (can't enforce) that a state provider never generates an async-state machine, since all state lookups are ValueTasks.
CacheOnly
requests that a state provider doesn't contact Discord when retrieving an entity.
The motivation for implementing this might be if you're using an additional caching layer for long-term storage, such as redis, where a state provider might need to make an async call to retrieve information. If you're in a context where the ValueTask must remain synchronous (maybe in a TypeReader, or if you're .Result
'ing), you could use SyncOnly.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
That's an excellent thought. You should probably just take that last paragraph and inject it into the XML.
src/Discord.Net/IDiscordClient.cs
Outdated
{ | ||
public interface IDiscordClient | ||
{ | ||
event Action Connected; |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
What's the logic behind having these defined as event
s? Personally, I'd much prefer something like exposing the AsyncEvent<T>
fields from v2 (which are private, and used internally anyway). This way, you can work with events as first-class objects, and e.g. do something like public async Task<T> WaitFor(this AsyncEvent<T>)
like in discord.py - super useful for making interactive stuff.
incomplete, still needs receive handling gateway is yet to be designed
basic enough to test some stuff with fixed some bugs with disposables in the process
Optional struct needs work still, + writing the converter for it is going to be a headache
does not support property omission at this time, will need to be added later using a separate converter and base marker class. -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA512 "git failing to recognize gpg key, this identity is still valid" -----BEGIN PGP SIGNATURE----- iQIzBAEBCgAdFiEErbDRxgZ77MPT+ajAOrLKmA3cYakFAl4bthEACgkQOrLKmA3c YamuWw/7Bn/Ks0mTRN3tg3Z/voETJ/8JQZXJEiW7wwv8c7nSOemxRNB/Tmzo3kzC N6T5fH7Gep4o4iA7CfJ5CZtx+OY92OpyBwsJgkNvANVpjXWCeDaww0Ci5dyVwFUk fFq21l6p2sbM6PB9sEOCvryeIOrgkqBl915MkAlj+/UtnAQ9qFhomIGNLPPFeYOS eaHWjZF6ArbF5NMaOhboDDCIl2nCf+RGEetDoBP2BRaIf+eOyl0lGyQqiY1mNqkD DX8nmcaY5/Lnxhf3pwmYZbqKBPQt5R2FxmqWTg5ey0R4//izE4TJ54nlhdSnTZpH 7ZligmR9rQFdQ5jbSq6cIclo9i988ELHKBgt8mG3SiC4AT0+SBXRpPRBitkA0CPb O4W8J0HrbSFmILx9Zvuy72KC/Zzo+SOS8257S35ihosrlyupcR4zladVcIviAPWk Ovpy85W4uxPdWc6zkMOZSx9OiYFYkNlK/QdNJBXGg7LLcaLf8p33lj+T8UXa7dyC Sw/pW5RL1FYalh7iXF55ylJrKo+oySBejods+ATnmYG4JMywO+GNCE+XLCcDpoBx 9H2z0qJNb5Dgkc4cRulKwYEoT+LQKUhLFdj4wNEqE8mBw0ZoxUiBBqOD1TiZr2mf 1AFQVS/AeOc03t25OfmhNz026OAGy01bjeHr09deT20dsssEpQY= =n76m -----END PGP SIGNATURE-----
@quinchs are there any plans to act on this PR in the future? |
Hi all,
Please use this PR to leave any comments, feedback, and concerns about the new API design for v3. You are welcome to leave a comment directly on this thread, or use a GitHub suggestion on the relevant file.
Design feedback may also be left in the Discord.Net channel in Discord API.
Note that this review is focused primarily around the public API only; none of the interfaces exposed (sans the IDiscordClient) have any actual implementations yet. Implementation of the interfaces will come during a later pass.