cchat/cchat.go

927 lines
38 KiB
Go

// Code generated by ./cmd/internal. DO NOT EDIT.
// Package cchat is a set of stabilized interfaces for cchat implementations,
// joining the backend and frontend together.
//
//
// Backend
//
// Almost anything in the backend comes with an ID. For example, a Server must
// have an ID, or a Session must have a user ID. The backend is required to
// guarantee that IDs are somehow unique. This should already be the case for
// most chat services; for example, Discord provides IDs for guilds, channels,
// members, and more. The only time that the backend should not guarantee ID
// uniqueness is across Sessions, because it doesn't make sense to do so. In
// this case, the frontend should guarantee uniqueness instead, either by
// discarding duplicated items, overriding them, or anything reasonable and
// explicit.
//
// Methods implemented by the backend that have frontend containers as arguments
// can do IO. Frontends must NOT rely on individual backend states and should
// always assume that they will block.
//
// Methods that do not return an error must NOT do any IO to prevent blocking
// the main thread. As such, ID() and Name() must never do any IO. Methods that
// do return an error may do IO, but they should be documented per method.
//
// Backend implementations have certain conditions that should be adhered to:
//
// - Storing MessagesContainer and ServersContainer are advised
// against; however, they should be done if need be.
// - Other containers such as LabelContainer and IconContainer
// should also not be stored; however, the same rule as above
// applies.
// - For the server list, icon updates and such that happen after
// their calls should use SetServers().
// - For the nickname of the current server, the backend can store
// the state of the label container. It must, however, remove the
// container when the stop callback from JoinServer() is called.
// - Some methods that take in a container may take in a context as
// well. Although implementations don't have to use this context,
// it should try to.
//
// Note: IO in most cases usually refer to networking, but they should files and
// anything that is blocking, such as mutexes or semaphores.
//
// Note: As mentioned above, contexts are optional for both the frontend and
// backend. The frontend may use it for cancellation, and the backend may ignore
// it.
//
// Some interfaces can be extended. Interfaces that are extendable will have
// methods starting with "As" and returns another interface type. The
// implementation may or may not return the same struct as the interface, but
// the caller should not have to type assert it to a struct. They can also
// return nil, which should indicate the backend that the feature is not
// implemented.
//
// To avoid confusing, when said "A implements B," it is mostly assumed that A
// has a method named "AsB." It does not mean that A can be type-asserted to B.
//
// For future references, these "As" methods will be called asserter methods.
//
// Note: Backends must not do IO in the "As" methods. Most of the time, it
// should only conditionally check the local state and return value or nil.
//
// Below is an example of checking for an extended interface.
//
// if iconer := server.AsIconer(); iconer != nil {
// println("Server implements Iconer.")
// }
//
//
// Frontend
//
// Frontend contains all interfaces that a frontend can or must implement. The
// backend may call these methods any time from any goroutine. Thus, they should
// be thread-safe. They should also not block the call by doing so, as backends
// may call these methods in its own main thread.
//
// It is worth pointing out that frontend container interfaces will not have an
// error handling API, as frontends can do that themselves. Errors returned by
// backend methods will be errors from the backend itself and never the frontend
// errors.
package cchat
import (
"context"
"fmt"
text "github.com/diamondburned/cchat/text"
"io"
"time"
)
// ID is the type alias for an ID string. This type is used for clarification
// and documentation purposes only. Implementations could either use this type
// or a string type.
type ID = string
// Status represents a user's status. This might be used by the frontend to
// visually display the status.
type Status uint8
const (
StatusUnknown Status = iota
StatusOnline
StatusIdle
StatusBusy
StatusAway
StatusOffline
// Invisible is reserved.
StatusInvisible
)
func (s Status) Is(is Status) bool {
return s == is
}
// AuthenticateEntry represents a single authentication entry, usually an email
// or password prompt. Passwords or similar entries should have Secrets set to
// true, which should imply to frontends that the fields be masked.
type AuthenticateEntry struct {
Name string
Placeholder string
Description string
Secret bool
Multiline bool
}
// CompletionEntry is a single completion entry returned by CompleteMessage. The
// icon URL field is optional.
type CompletionEntry struct {
Raw string
Text text.Rich
Secondary text.Rich
IconURL string
Image bool
}
// MessageAttachment represents a single file attachment. If needed, the
// frontend will close the reader after the message is sent, that is when the
// SendMessage function returns. The backend must not use the reader after that.
type MessageAttachment struct {
io.Reader
Name string
}
// ReadIndication represents a read indication of a user/author in a messager
// server. It relates to a message ID within the server and is meant to imply
// that the user/author has read up to the given message ID.
//
// The frontend should override an existing author with the received ones. This
// could be treated as upsert operations.
type ReadIndication struct {
User User
MessageID ID
}
// ErrInvalidConfigAtField is the structure for an error at a specific
// configuration field. Frontends can use this and highlight fields if the
// backends support it.
type ErrInvalidConfigAtField struct {
Key string
Err error
}
func (e ErrInvalidConfigAtField) Error() string {
return fmt.Sprintf("Error at %s: %s", e.Key, e.Err.Error())
}
func (e ErrInvalidConfigAtField) Unwrap() error {
return e.Err
}
// Actioner adds custom message actions into each message. Similarly to
// ServerMessageEditor, some of these methods may do IO.
type Actioner interface {
// Do executes a message action on the given messageID, which would be taken
// from MessageHeader.ID(). This method is allowed to do IO; the frontend should
// take care of running it asynchronously.
Do(ctx context.Context, action string, id ID) error // Blocking
// MessageActions returns a list of possible actions to a message in pretty
// strings that the frontend will use to directly display. This method must not
// do IO.
//
// The string slice returned can be nil or empty.
Actions(id ID) []string
}
// Attacher adds attachments into the message being sent.
type Attacher interface {
Attachments() []MessageAttachment
}
// AuthenticateError is the error returned when authenticating. This error
// interface extends the normal error to allow backends to implement multi-stage
// authentication if needed in a clean way without needing any loops.
//
// This interface satisfies the error interface.
type AuthenticateError interface {
// NextStage optionally returns a slice of Authenticator interfaces if the
// authentication process requires another stage. It works similarly to
// Service's Authenticate method, both of which returns a slice of
// Authenticators.
//
// If the error returned is an actual error, and that the user should retry any
// of the authentication fields, then NextStage could return nil to signify the
// error. The frontend could reliably check nil on this field to determine
// whether or not it should recreate the authentication fields.
NextStage() []Authenticator
// Error returns the error as a string. This method makes AuthenticateError
// satisfy the built-in error interface.
Error() string
}
// The authenticator interface allows for a multistage initial authentication
// API that the backend could use. Multistage is done by calling Authenticate
// and check for AuthenticateError's NextStage method.
type Authenticator interface {
// Authenticate will be called with a list of values with indices correspond to
// the returned slice of AuthenticateEntry.
Authenticate(context.Context, []string) (Session, AuthenticateError) // Blocking
// AuthenticateForm should return a list of authentication entries for the
// frontend to render.
AuthenticateForm() []AuthenticateEntry
// Description returns the description of this authenticator method.
Description() text.Rich
// Name returns a short and concise name of this Authenticator method. The name
// should not include the name of the Service.
Name() text.Rich
}
// Backlogger adds message history capabilities into a message container. The
// backend should send old messages using the MessageCreate method of the
// MessagesContainer, and the frontend should automatically sort messages based
// on the timestamp.
//
// As there is no stop callback, if the backend needs to fetch messages
// asynchronously, it is expected to use the context to know when to cancel.
//
// The frontend should usually call this method when the user scrolls to the
// top. It is expected to guarantee not to call Backlogger more than once on the
// same ID. This can usually be done by deactivating the UI.
//
// Note that the optional usage of contexts also apply here. The frontend should
// deactivate the UI when the backend is working. However, the frontend can
// accomodate this by not deactivating until another event is triggered, then
// freeze the UI until the method is cancelled. This works even when the backend
// does not use the context.
type Backlogger interface {
// Backlog fetches messages before the given message ID into the
// MessagesContainer.
//
// This method is technically a ContainerMethod, but is listed as an IOMethod
// because of the additional message ID parameter.
Backlog(ctx context.Context, before ID, msgc MessagesContainer) error // Blocking
}
// Commander is an optional interface that a session could implement for command
// support. This is different from just intercepting the SendMessage() API, as
// this extends globally to the entire session.
//
// A very primitive use of this API would be to provide additional features that
// are not in cchat through a very basic terminal interface.
type Commander interface {
// Run executes the given command, with the slice being already split arguments,
// similar to os.Args. The function can return both a []byte and an error value.
// The frontend should render the byte slice's value first, then display the
// error.
//
// This function can do IO.
//
// The client should make guarantees that an empty string (and thus a
// zero-length string slice) should be ignored. The backend should be able to
// assume that the argument slice is always length 1 or more.
//
//
// Words
//
// This interface and everything else inside this interface must abide by shell
// rules when splitting words. This is in contrary to the default behavior
// elsewhere, such as in Sender's Completer, where words are split by whitespace
// without care for quotes.
//
// For example, provided this string:
//
// echo "This is a string"
//
// This is the correct output:
//
// []string{"echo", "This is a string"}
//
// This is the incorrect output:
//
// []string{"echo", "\"This", "is", "a", "string\""}
//
// A helper function for this kind of behavior is available in package split,
// under the ArgsIndexed function. This implementation also provides the rough
// specifications.
Run(ctx context.Context, words []string) ([]byte, error) // Blocking
// Asserters.
AsCompleter() Completer // Optional
}
// Completer adds autocompletion into the message composer. IO is not allowed,
// and the backend should do that only in goroutines and update its state for
// future calls.
//
// Frontends could utilize the split package inside utils for splitting words
// and index. This is the de-facto standard implementation for splitting words,
// thus backends can rely on their behaviors.
type Completer interface {
// Complete returns the list of possible completion entries for the given word
// list and the current word index. It takes in a list of whitespace-split slice
// of string as well as the position of the cursor relative to the given string
// slice.
Complete(words []string, current int64) []CompletionEntry
}
// Configurator is an interface which the backend can implement for a primitive
// configuration API.
type Configurator interface {
SetConfiguration(map[string]string) error
Configuration() map[string]string
}
// Editor adds message editing to the messenger. Only EditMessage can do IO.
type Editor interface {
// Edit edits the message with the given ID to the given content, which is the
// edited string from RawMessageContent. This method can do IO.
Edit(ctx context.Context, id ID, content string) error // Blocking
// RawContent gets the original message text for editing. This method must not
// do IO.
RawContent(id ID) (string, error)
// IsEditable returns whether or not a message can be edited by the client. This
// method must not do IO.
IsEditable(id ID) bool
}
// Identifier requires ID() to return a uniquely identifiable string for
// whatever this is embedded into. Typically, servers and messages have IDs. It
// is worth mentioning that IDs should be consistent throughout the lifespan of
// the program or maybe even forever.
type Identifier interface {
ID() ID
}
// LabelContainer is a generic interface for any container that can hold texts.
// It's typically used for rich text labelling for usernames and server names.
//
// Methods that takes in a LabelContainer typically holds it in the state and
// may call SetLabel any time it wants. Thus, the frontend should synchronize
// calls with the main thread if needed.
//
// Labels given to the frontend may contain images or avatars, and the frontend
// has the choice to display them or not.
type LabelContainer interface {
SetLabel(text.Rich)
}
// ListMember represents a single member in the member list. Note that this
// interface should be treated as a static container: updating a member will
// involve a completely new ListMember instance with the same ID.
//
// Note that the frontend may give everyone an avatar regardless, or it may not
// show any avatars at all.
type ListMember interface {
Identifier
// Secondary returns the subtext of this member. This could be anything, such as
// a user's custom status or away reason.
Secondary() text.Rich
// Status returns the status of the member. The backend does not have to show
// offline members with the offline status if it doesn't want to show offline
// menbers at all.
Status() Status
// Name returns the username or the nickname of the member, whichever the
// backend should prefer.
Name() text.Rich
}
// Lister is for servers that contain children servers. This is similar to
// guilds containing channels in Discord, or IRC servers containing channels.
//
// There isn't a similar stop callback API unlike other interfaces because all
// servers are expected to be listed. However, they could be hidden, such as
// collapsing a tree.
//
// The backend should call both the container and other icon and label
// containers, if any.
type Lister interface {
// Servers should call SetServers() on the given ServersContainer to render all
// servers. This function can do IO, and the frontend should run this in a
// goroutine.
Servers(context.Context, ServersContainer) error
// Columnate is optionally used by servers to tell the frontend whether or not
// its children should be put onto a new column instead of underneath it within
// the same tree. If the method returns false, then the frontend can treat its
// children as normal and show it as children within the same tree.
//
// For example, in Discord, guilds can be placed in guild folders, but guilds
// and guild folders are put in the same column while guilds are actually
// children of the folders. To replicate this behavior, guild folders should
// return false, and guilds should return true. Both channels and categories can
// return false.
Columnate() bool
}
// MemberDynamicSection represents a dynamically loaded member list section. The
// section behaves similarly to MemberSection, except the information displayed
// will be considered incomplete until LoadMore returns false.
//
// LoadLess can be called by the client to mark chunks as stale, which the
// server can then unsubscribe from.
type MemberDynamicSection interface {
// LoadLess is a method which the client must call after it is done displaying
// entries that were added from calling LoadMore.
//
// The client can call this method exactly as many times as it has called
// LoadMore. However, false should be returned if the client should stop, and
// future calls without LoadMore should still return false.
LoadLess(context.Context) bool // Blocking
// LoadMore is a method which the client can call to ask for more members. This
// method can do IO.
//
// Clients may call this method on the last section in the section slice;
// however, calling this method on any section is allowed. Clients may not call
// this method if the number of members in this section is equal to Total.
LoadMore(context.Context) bool // Blocking
}
// MemberListContainer is a generic interface for any container that can display
// a member list. This is similar to Discord's right-side member list or IRC's
// users list. Below is a visual representation of a typical member list
// container:
//
// +-MemberList-----------\
// | +-Section------------|
// | | |
// | | Header - Total |
// | | |
// | | +-Member-----------|
// | | | Name |
// | | | Secondary |
// | | \__________________|
// | | |
// | | +-Member-----------|
// | | | Name |
// | | | Secondary |
// | | \__________________|
// \_\____________________/
type MemberListContainer interface {
// RemoveMember removes a member from a section. If neither the member nor the
// section exists, then the client should ignore it.
RemoveMember(sectionID ID, memberID ID)
// SetMember adds or updates (or upsert) a member into a section. This operation
// must not change the section's member count. As such, changes should be done
// separately in SetSection. If the section does not exist, then the client
// should ignore this member, so, backends must call SetSections first before
// SetMember on a new section.
//
// Typically, the backend should try and avoid calling this method and instead
// update the labeler in the name. This method should only be used for adding
// members.
SetMember(sectionID ID, member ListMember)
// SetSections (re)sets the list of sections to be the given slice. Members from
// the old section list should be transferred over to the new section entry if
// the section name's content is the same. Old sections that don't appear in the
// new slice should be removed.
SetSections(sections []MemberSection)
}
// MemberLister adds a member list into a message server.
type MemberLister interface {
// ListMembers assigns the given container to the channel's member list. The
// given context may be used to provide HTTP request cancellations, but
// frontends must not rely solely on this, as the general context rules applies.
//
// Further behavioral documentations may be in Messenger's JoinServer method.
ListMembers(context.Context, MemberListContainer) error
}
// MemberSection represents a member list section. The section name's content
// must be unique among other sections from the same list regardless of the rich
// segments.
type MemberSection interface {
Identifier
Namer
// Total returns the total member count.
Total() int
// Asserters.
AsMemberDynamicSection() MemberDynamicSection // Optional
}
// MessageCreate is the interface for an incoming message.
type MessageCreate interface {
MessageHeader
// Noncer is optional.
Noncer
// Mentioned returns whether or not the message mentions the current user. If a
// backend does not implement mentioning, then false can be returned.
Mentioned() bool
Content() text.Rich
Author() User
}
// MessageDelete is the interface for a message delete event.
type MessageDelete interface {
MessageHeader
}
// MessageHeader implements the minimum interface for any message event.
type MessageHeader interface {
Identifier
Time() time.Time
}
// MessageUpdate is the interface for a message update (or edit) event. It is
// only responsible for updating a message's content. The author's name should
// be updated using MessageCreate's Author.
type MessageUpdate interface {
MessageHeader
Content() text.Rich
}
// MessagesContainer is a view implementation that displays a list of messages
// live. This implements the 3 most common message events: CreateMessage,
// UpdateMessage and DeleteMessage. The frontend must handle all 3.
//
// Since this container interface extends a single Server, the frontend is
// allowed to have multiple views. This is usually done with tabs or splits, but
// the backend should update them all nonetheless.
type MessagesContainer interface {
DeleteMessage(MessageDelete)
UpdateMessage(MessageUpdate)
// CreateMessage inserts a message into the container. The frontend must
// guarantee that the messages are in order based on what's returned from
// Time().
CreateMessage(MessageCreate)
}
// Messenger is for servers that contain messages. This is similar to Discord or
// IRC channels.
type Messenger interface {
// JoinServer joins a server that's capable of receiving messages. The server
// may not necessarily support sending messages.
//
// Frontends must never call JoinServer on the same server more than twice
// without calling the stop function first. This is the best of both worlds, as
// it greatly reduces complexity on both sides in most cases, therefore the
// backend can safely assume that there will only ever be one active JoinServer.
// If the frontend wishes to do this, it must keep its own shared message
// buffer.
JoinServer(context.Context, MessagesContainer) error
// Asserters.
AsSender() Sender // Optional
AsEditor() Editor // Optional
AsActioner() Actioner // Optional
AsNicknamer() Nicknamer // Optional
AsBacklogger() Backlogger // Optional
AsMemberLister() MemberLister // Optional
AsUnreadIndicator() UnreadIndicator // Optional
AsTypingIndicator() TypingIndicator // Optional
}
// Namer requires Name() to return the name of the object. Typically, this
// implies usernames for sessions or service names for services.
//
// Frontends can show the ID of the object when a name hasn't yet been set. The
// backend may immediately update the name afterwards, but assumptions should
// not be made.
type Namer interface {
// Name sets the given container to contain the name of the parent context. The
// method has no stop method; stopping is implied to be dependent on the parent
// context. As such, it's only used for updating.
Name(context.Context, LabelContainer) error
}
// Nicknamer adds the current user's nickname.
//
// The frontend will not traverse up the server tree, meaning the backend must
// handle nickname inheritance. This also means that servers that don't
// implement ServerMessage also don't need to implement ServerNickname. By
// default, the session name should be used.
type Nicknamer interface {
Namer
}
// Noncer adds nonce support. A nonce is defined in this context as a unique
// identifier from the frontend. This interface defines the common nonce getter.
//
// Nonces are useful for frontends to know if an incoming event is a reply from
// the server backend. As such, nonces should be roundtripped through the
// server. For example, IRC would use labeled responses.
//
// The Nonce method can return an empty string. This indicates that either the
// frontend or backend (or neither) supports nonces.
//
// Contrary to other interfaces that extend with an "Is" method, the Nonce
// method could return an empty string here.
type Noncer interface {
Nonce() string
}
// ReadContainer is an interface that a frontend container can implement to show
// the read bubbles on messages. This container typically implies the message
// container, but that is up to the frontend's implementation.
type ReadContainer interface {
// DeleteIndications deletes a list of unused users/authors associated with
// their read indicators. The backend can use this to free up users/authors that
// are no longer in the server, for example when they are offline or have left
// the server.
DeleteIndications(authorIDs []ID)
// AddIndications adds a map of users/authors to the respective message ID of
// the server that implements ReadIndicator.
AddIndications([]ReadIndication)
}
// ReadIndicator adds a read indicator API for frontends to show. An example of
// the read indicator is in Matrix, where each message can have a small avatar
// indicating that the user in the room has read the message.
type ReadIndicator interface {
// ReadIndicate subscribes the given container for read activities. The backend
// must keep track of which read states to send over to not overwhelm the
// frontend, and the frontend must either keep track of them, or it should not
// display it at all.
ReadIndicate(context.Context, ReadContainer) error
}
// Replier indicates that the message being sent is a reply to something.
// Frontends that support replies can assume that all messages in a Sender can
// be replied to, and the backend can choose to do nothing to the replied ID.
type Replier interface {
ReplyingTo() ID
}
// SendableMessage is the bare minimum interface of a sendable message, that is,
// a message that can be sent with SendMessage(). This allows the frontend to
// implement its own message data implementation.
//
// An example of extending this interface is MessageNonce, which is similar to
// IRCv3's labeled response extension or Discord's nonces. The frontend could
// implement this interface and check if incoming MessageCreate events implement
// the same interface.
type SendableMessage interface {
Content() string
// Asserters.
AsNoncer() Noncer // Optional
AsReplier() Replier // Optional
AsAttacher() Attacher // Optional
}
// Sender adds message sending to a messenger. Messengers that don't implement
// MessageSender will be considered read-only.
type Sender interface {
// CanAttach returns whether or not the client is allowed to upload files.
CanAttach() bool
// Send is called by the frontend to send a message to this channel.
Send(context.Context, SendableMessage) error // Blocking
// Asserters.
AsCompleter() Completer // Optional
}
// Server is a single server-like entity that could translate to a guild, a
// channel, a chat-room, and such. A server must implement at least ServerList
// or ServerMessage, else the frontend must treat it as a no-op.
//
// Note that the Server is allowed to implement both Lister and Messenger. This
// is useful when the messenger contains sub-servers, such as threads.
type Server interface {
Identifier
Namer
// Asserters.
AsLister() Lister // Optional
AsMessenger() Messenger // Optional
AsCommander() Commander // Optional
AsConfigurator() Configurator // Optional
}
// ServerUpdate represents a server update event.
type ServerUpdate interface {
// Server embeds a complete server. Unlike MessageUpdate, which only returns
// data on methods that are changed, ServerUpdate's methods must return the
// complete data even if they stay the same. As such, zero-value returns are
// treated as not updated, including the name.
Server
// PreviousID returns the ID of the item, either to be replaced or to be
// inserted in front of.
//
// If replace is true, then the returned ID is the ID of the item to be
// replaced, and the frontend should only try to use the ID as-is to find the
// old server and replace.
//
// If replace is false, then the returned ID will be the ID of the item in front
// of the embedded server. If the ID is empty or the frontend cannot find the
// server from this ID, then it should assume and prepend the server to the
// start.
PreviousID() (serverID ID, replace bool)
}
// ServersContainer is any type of view that displays the list of servers. It
// should implement a SetServers([]Server) that the backend could use to call
// anytime the server list changes (at all).
//
// Typically, most frontends should implement this interface onto a tree node,
// as servers can be infinitely nested. Frontends should also reset the entire
// node and its children when SetServers is called again.
type ServersContainer interface {
UpdateServer(ServerUpdate)
// SetServer is called by the backend service to request a reset of the server
// list. The frontend can choose to call Servers() on each of the given servers,
// or it can call that later. The backend should handle both cases.
//
// If the backend sets a nil server slice, then the frontend should take that as
// an unavailable server list rather than an empty server list. The server list
// should only be considered empty if it's an empty non-nil slice. An
// unavailable list, on the other hand, can be treated as backend issues, e.g. a
// connection issue.
SetServers([]Server)
}
// Service is a complete service that's capable of multiple sessions. It has to
// implement the Authenticate() method, which returns multiple implementations
// of Authenticator.
//
// A service can implement SessionRestorer, which would indicate the frontend
// that it can restore past sessions. Sessions are saved using the SessionSaver
// interface that Session can implement.
//
// A service can also implement Configurator if it has additional
// configurations. The current API is a flat key-value map, which can be parsed
// by the backend itself into more meaningful data structures. All
// configurations must be optional, as frontends may not implement a
// configurator UI.
type Service interface {
// Identifier returns the unique identifier for the service. There is no
// enforced representation, but services are recommended to follow the Reverse
// Domain Name Notation for consistency. An example of that would be:
//
// com.github.diamondburned.cchat-discord
// com.github.username.service
Identifier
// Namer returns the name of the service.
Namer
Authenticate() []Authenticator
// Asserters.
AsConfigurator() Configurator // Optional
AsSessionRestorer() SessionRestorer // Optional
}
// Session is returned after authentication on the service. It implements
// Name(), which should return the username most of the time. It also implements
// ID(), which might be used by frontends to check against User.ID() and other
// things.
//
// A session can implement SessionSaver, which would allow the frontend to save
// the session into its keyring at any time. Whether the keyring is completely
// secure or not is up to the frontend. For a GTK client, that would be using
// the GNOME Keyring daemon.
type Session interface {
// Identifier should typically return the user ID.
Identifier
// Namer gives the name of the session, which is typically the username.
Namer
Lister
// Disconnect asks the service to disconnect. It does not necessarily mean
// removing the service.
//
// The frontend must cancel the active ServerMessage before disconnecting. The
// backend can rely on this behavior.
//
// The frontend will reuse the stored session data from SessionSaver to
// reconnect.
//
// When this function fails, the frontend may display the error upfront.
// However, it will treat the session as actually disconnected. If needed, the
// backend must implement reconnection by itself.
Disconnect(context.Context) error // Blocking, Disposer
// Asserters.
AsCommander() Commander // Optional
AsSessionSaver() SessionSaver // Optional
}
// SessionRestorer extends Service and is called by the frontend to restore a
// saved session. The frontend may call this at any time, but it's usually on
// startup.
//
// To save a session, refer to SessionSaver.
type SessionRestorer interface {
RestoreSession(context.Context, map[string]string) (Session, error) // Blocking
}
// SessionSaver extends Session and is called by the frontend to save the
// current session. This is typically called right after authentication, but a
// frontend may call this any time, including when it's closing.
//
// The frontend can ask to restore a session using SessionRestorer, which
// extends Service.
//
// The SaveSession method must not do IO; if there are any reasons that cause
// SaveSession to fail, then a nil map should be returned.
type SessionSaver interface {
SaveSession() map[string]string
}
// TypingContainer is a generic interface for any container that can display
// users typing in the current chatbox. The typing indicator must adhere to the
// TypingTimeout returned from ServerMessageTypingIndicator. The backend should
// assume that to be the case and send events appropriately.
//
// For more documentation, refer to TypingIndicator.
type TypingContainer interface {
// RemoveTyper explicitly removes the typer with the given user ID from the list
// of typers. This function is usually not needed, as the client will take care
// of removing them after TypingTimeout has been reached or other conditions
// listed in ServerMessageTypingIndicator are met.
RemoveTyper(authorID ID)
// AddTyper appends the typer (author) into the frontend's list of typers, or it
// pushes this typer on top of others. The frontend should assume current time
// every time AddTyper is called.
AddTyper(User)
}
// TypingIndicator optionally extends ServerMessage to provide bidirectional
// typing indicating capabilities. This is similar to typing events on Discord
// and typing client tags on IRCv3.
//
// The client should remove a typer when a message is received with the same
// user ID, when RemoveTyper() is called by the backend or when the timeout
// returned from TypingTimeout() has been reached.
type TypingIndicator interface {
// TypingSubscribe subscribes the given indicator to typing events sent by the
// backend. The added event handlers have to be removed by the backend when the
// stop() callback is called.
//
// This method does not take in a context, as it's supposed to only use event
// handlers and not do any IO calls. Nonetheless, the client must treat it like
// it does and call it asynchronously.
TypingSubscribe(context.Context, TypingContainer) error
// TypingTimeout returns the interval between typing events sent by the client
// as well as the timeout before the client should remove the typer. Typically,
// a constant should be returned.
TypingTimeout() time.Duration
// Typing is called by the client to indicate that the user is typing. This
// function can do IO calls, and the client must take care of calling it in a
// goroutine (or an asynchronous queue) as well as throttling it to
// TypingTimeout.
Typing(context.Context) error // Blocking
}
// UnreadContainer is an interface that a single server container (such as a
// button or a tree node) can implement if it's capable of indicating the read
// and mentioned status for that channel.
//
// Server containers that implement this has to represent unread and mentioned
// differently. For example, a mentioned channel could have a red outline, while
// an unread channel could appear brighter.
//
// Server containers are expected to represent this information in their parent
// nodes as well. For example, if a server is unread, then its parent servers as
// well as the session node should indicate the same status. Highlighting the
// session and service nodes are, however, implementation details, meaning that
// this decision is up to the frontend to decide.
type UnreadContainer interface {
// SetUnread sets the container's unread state to the given boolean. The
// frontend may choose how to represent this.
SetUnread(unread bool, mentioned bool)
}
// UnreadIndicator adds an unread state API for frontends to use. The unread
// state describes whether a channel has been read or not by the current user.
// It is not to be confused with ReadIndicator, which indicates the unread state
// of others.
type UnreadIndicator interface {
// UnreadIndicate subscribes the given unread indicator for unread and mention
// events. Examples include when a new message is arrived and the backend needs
// to indicate that it's unread.
//
// This function must provide a way to remove callbacks, as clients must call
// this when the old server is destroyed, such as when Servers is called.
UnreadIndicate(context.Context, UnreadContainer) error
// MarkRead marks a message in the server messenger as read. Backends that
// implement the UnreadIndicator interface must give control of marking messages
// as read to the frontend if possible.
//
// This method is assumed to be a setter method that does not error out, because
// the frontend has no use in knowing the error. As such, marking messages as
// read is best-effort. The backend is in charge of synchronizing the read state
// with the server and coordinating it with reasonable rate limits, if needed.
MarkRead(messageID ID)
}
// User is the interface for an identifiable author. The interface defines that
// an author always have an ID and a name.
//
// An example of where this interface is used would be in MessageCreate's User
// method or embedded in Typer. The returned ID may or may not be used by the
// frontend, but backends must guarantee that the User's ID is in fact a user
// ID.
//
// The frontend may use the ID to squash messages with the same author together.
type User interface {
Identifier
Namer
}