| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| | |
|
| | |
|
| |
|
|
|
| |
Some lints were not resolved due to causing API changes. Most lints in the
framework were left unfixed.
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
| |
By negating hashing altogether.
The increase is around 1000-ish nanoseconds saved.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The `model` module has historically been one giant module re-exporting
all of the model types, which is somewhere around 100 types. This can be
a lot to look at for a new user and somewhat overwhelming, especially
with a large number of fine-grained imports from the module.
The module is now neatly split up into submodules, mostly like it has
been internally since the early versions of the library. The submodules
are:
- application
- channel
- error
- event
- gateway
- guild
- id
- invite
- misc
- permissions
- prelude
- user
- voice
- webhook
Each submodule contains types that are "owned" by the module. For
example, the `guild` submodule contains, but not limited to, Emoji,
AuditLogsEntry, Role, and Member. `channel` contains, but not limited
to, Attachment, Embed, Message, and Reaction.
Upgrade path:
Instead of glob importing the models via `use serenity::model::*;`,
instead glob import via the prelude:
```rust
use serenity::model::prelude::*;
```
Instead of importing from the root model module:
```rust
use serenity::model::{Guild, Message, OnlineStatus, Role, User};
```
instead import from the submodules like so:
```rust
use serenity::model::channel::Message;
use serenity::model::guild::{Guild, Role};
use serenity::model::user::{OnlineStatus, User};
```
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Builders would keep a `serde_json::Map<String, Value>`, which would
require re-creating owned strings for the same parameter multiple times
in some cases, depending on builder defaults and keying strategies.
This commit uses a `std::collections::HashMap<&'static str, Value>`
internally, and moves over values to a `serde_json::Map<String, Value>`
when it comes time to sending them to the appropriate `http` module
function.
This saves the number of heap-allocated string creations on most
builders, with specific performance increase on `builder::CreateMessage`
and `builder::CreateEmbed` & co.
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
Also not quite sure if they goofed rustfmt or something, but its changes it did were a bit bizarre.
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
| |
The 'wait' parameter allows you to specify waiting for the Message to be
sent prior to receiving a response, which will have Discord include the
JSON representation of the Message in the body.
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Modules are now separated into a fashion where the library can be used
for most use cases, without needing to compile the rest.
The core of serenity, with no features enabled, contains only the
struct (model) definitions, constants, and prelude. Models do not have
most functions compiled in, as that is separated into the `model`
feature.
The `client` module has been split into 3 modules: `client`, `gateway`,
and `http`.
`http` contains functions to interact with the REST API. `gateway`
contains the Shard to interact with the gateway, requiring `http` for
retrieving the gateway URL. `client` requires both of the other features
and acts as an abstracted interface over both the gateway and REST APIs,
handling the event loop.
The `builder` module has been separated from `utils`, and can now be
optionally compiled in. It and the `http` feature are required by the
`model` feature due to a large number of methods requiring access to
them.
`utils` now contains a number of utilities, such as the Colour struct, the
`MessageBuilder`, and mention parsing functions.
Each of the original `ext` modules are still featured, with `cache` not
requiring any feature to be enabled, `framework` requiring the `client`,
`model`, and `utils`, and `voice` requiring `gateway`.
In total the features and their requirements are:
- `builder`: none
- `cache`: none
- `client`: `gateway`, `http`
- `framework`: `client`, `model`, `utils`
- `gateway`: `http`
- `http`: none
- `model`: `builder`, `http`
- `utils`: none
- `voice`: `gateway`
The default features are `builder`, `cache`, `client`, `framework`,
`gateway`, `model`, `http`, and `utils`.
To help with forwards compatibility, modules have been re-exported from
their original locations.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The current build system is rudimentary, incomplete, and rigid, offering
little in the way of customizing decoding options.
To solve this, switch to using serde-derive with custom Deserialization
implementations. This allows very simple deserialization when special
logic does not need to be applied, yet allows us to implement our own
deserialization logic when required.
The problem with the build system was that it built enums and structs
from YAML files. This is not so good, because it requires creating a
custom build system (which was rudimentary), creating "special struct
configs" when logic needed to be ever so slightly extended (rigid), and
if special logic needed to be applied, a custom deserialization method
would have been needed to be made anyway (incomplete).
To solve this, switch to serde-derive and implementing Deserialize
ourselves where required. This reduces YAML definitions that might
look like:
```yaml
---
name: Group
description: >
A group channel, potentially including other users, separate from a [`Guild`].
[`Guild`]: struct.Guild.html
fields:
- name: channel_id
description: The Id of the group channel.
from: id
type: ChannelId
- name: icon
description: The optional icon of the group channel.
optional: true
type: string
- name: last_message_id
description: The Id of the last message sent.
optional: true
type: MessageId
- name: last_pin_timestamp
description: Timestamp of the latest pinned message.
optional: true
type: string
- name: name
description: The name of the group channel.
optional: true
type: string
- name: owner_id
description: The Id of the group channel creator.
type: UserId
- name: recipients
description: Group channel's members.
custom: decode_users
t: UserId, Arc<RwLock<User>>
type: hashmap
```
to:
```rs
/// A group channel - potentially including other [`User`]s - separate from a
/// [`Guild`].
///
/// [`Guild`]: struct.Guild.html
/// [`User`]: struct.User.html
pub struct Group {
/// The Id of the group channel.
#[serde(rename="id")]
pub channel_id: ChannelId,
/// The optional icon of the group channel.
pub icon: Option<String>,
/// The Id of the last message sent.
pub last_message_id: Option<MessageId>,
/// Timestamp of the latest pinned message.
pub last_pin_timestamp: Option<String>,
/// The name of the group channel.
pub name: Option<String>,
/// The Id of the group owner.
pub owner_id: UserId,
/// A map of the group's recipients.
#[serde(deserialize_with="deserialize_users")]
pub recipients: HashMap<UserId, Arc<RwLock<User>>>,
}
```
This is much simpler and does not have as much boilerplate.
There should not be any backwards incompatible changes other than the
old, public - yet undocumented (and hidden from documentation) - decode
methods being removed. Due to the nature of this commit, field names may
be incorrect, and will need to be corrected as deserialization errors
are found.
|
| |
|
|
|
|
|
| |
A lot of the `rest` methods took - for example a Map - by value, where
they could instead take a reference. While this only prevents one clone
in the library, user-land code should no longer need to clone maps when
using the `rest` module.
|
| | |
|
| |
|
|
|
|
| |
The context is now strictly in relation to the context of the current
channel related to the event, if any. See Context::say for a list of
events that the context can be used for.
|
| |
|
|
|
| |
Add documentation for some missing methods - such as Game methods - and
add more methods to the Message Builder.
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This adds conditional compilation for the following features, in
addition to the voice conditional compilation flag:
- extras (message builder)
- framework
- methods
- state
These 4 are enabled _by default_, while the `voice` feature flag is
disabled.
Disabling the state will allow incredibly low-memory bots.
|
| |
|
|
|
| |
Create a general `internal` module, and move `prelude_internal` to
`internal::prelude`.
|
| |
|
|
|
|
|
|
|
| |
The builders aren't a large enough portion of the library to deserve
their own root-level module, so move them to the `utils` module.
Additionally, split them into separate files, as the library will be
receiving more builders and the single-file pattern was getting rather
large.
|
| | |
|
| |
|