slack-ruby/slack-ruby-client

View on GitHub
UPGRADING.md

Summary

Maintainability
Test Coverage
Upgrading Slack-Ruby-Client
===========================

### Upgrading to >= 2.0.0

[#416](https://github.com/slack-ruby/slack-ruby-client/pull/416) Removes default values for Faraday's SSL settings `ca_file` and `ca_path`.

If you previously relied on `OpenSSL::X509::DEFAULT_CERT_FILE` or `OpenSSL::X509::DEFAULT_CERT_DIR` to set these values you must now do so explicitly. E.g.:

```ruby
Slack::Web::Client.configure do |config|
  config.ca_file = OpenSSL::X509::DEFAULT_CERT_FILE
  config.ca_path = OpenSSL::X509::DEFAULT_CERT_DIR
end
```

or

```ruby
client = Slack::Web::Client.new(ca_file: OpenSSL::X509::DEFAULT_CERT_FILE, ca_path: OpenSSL::X509::DEFAULT_CERT_DIR)
```

#### RealTime Stores

Slack has deprecated the `rtm.start` method and [replaced its response](https://api.slack.com/changelog/2021-10-rtm-start-to-stop) with that of `rtm.connect`, so it has been removed from the library along with the `start_method` configuration ([#419](https://github.com/slack-ruby/slack-ruby-client/pull/419)) and `Slack::RealTime::Stores::Starter` is now the default store.
If your app still relies on the old `rtm.start` behavior, you can achieve a similar result with `Slack::RealTime::Stores::Store`, which initializes itself using Web API methods (see [README](README.md#slackrealtimestoresstore)).

The names of the Store caches `channels` and `groups` have been changed to `public_channels` and `private_channels` to reflect their true meaning.
This is consistent with the [Slack Web API](https://api.slack.com/methods/conversations.list#arg_types), although [event types for private channels](https://api.slack.com/events?query=group) still use the term `group`.
In both cases, private channels consist of both [conversation types](https://api.slack.com/types/conversation) (`id: 'C...'', is_channel: true, is_group: false, is_private: true`) and [group types](https://api.slack.com/types/group) (`id: 'G...', is_channel: false, is_group: true, is_private: true`) depending whether the object was created before or after March 2021.

### Upgrading to >= 1.0.0

#### Deprecated Methods

Slack has deprecated all `channel` and `group` methods, which have been removed from the library.

See [this announcement from Slack](https://api.slack.com/changelog/2020-01-deprecating-antecedents-to-the-conversations-api) for details.

#### Error Handling

As of 1.0.0 `Slack::Web::Api::Errors::ServerError` and its subclasses (introduced in 0.16.0) no longer extend `Slack::Web::Api::Errors::InternalError` or its parent `Slack::Web::Api::Errors::SlackError`. If you are rescuing `SlackError` or `InternalError` with the intention of including `ServerError` and its subclasses you should adjust your code to explicitly rescue `Slack::Web::Api::Errors::ServerError`.

```ruby
# Before
begin
  client.auth_test
rescue Slack::Web::Api::Errors::SlackError
  # Includes all server errors
end

# After
begin
  client.auth_test
rescue Slack::Web::Api::Errors::SlackError, Slack::Web::Api::Errors::ServerError
  # Need to rescue the server errors separately from SlackError
end
```

Additionally the `initialize` method for `ParsingError`, `TimeoutError`, and `UnavailableError` have changed from `new(message, response)` to `new(response)`. The `message` is now built into the definition of these classes. If you are instantiating or raising these errors in your code (perhaps in tests) you will need to update your code.

```ruby
# Before
error = Slack::Web::Api::Errors::TimeoutError.new('timeout_error', response)
error.message
# => 'timeout_error'

# After
error = Slack::Web::Api::Errors::TimeoutError.new(response)
error.message
# => 'timeout_error'
```

### Upgrading to >= 0.16.0

#### Removed Celluloid and Faye-Websocket Concurrency Support

Concurrency support for `celluloid-io` and `faye-websocket` has been removed. If you are running a RealTime bot on Celluloid or Faye, you must upgrade to `async-websocket`. Please note that RealTime bots are deprecated by Slack, and we generally recommend you [migrate your classic, RealTime bot, to granular permissions](https://code.dblock.org/2020/11/30/migrating-classic-slack-ruby-bots-to-granular-permissions.html).

See [#338](https://github.com/slack-ruby/slack-ruby-client/issues/338) for more information.

#### Error Handling

As of 0.16.0 `Faraday::Error` exceptions sans `Faraday::ClientError` are wrapped into `Slack::Web::Api::Errors::ServerError`, so if you're rescuing `Faraday::Error` — you should adjust your code to use `Slack::Web::Api::Errors::ServerError` and use `exception.cause` if underlying `Faraday::Error` is needed.

See [README#other-errors](README.md#other-errors) and [#350](https://github.com/slack-ruby/slack-ruby-client/pull/350) for more information.

### Upgrading to >= 0.15.0

As of 0.15.0, `activesupport` is no longer required. Add `gem 'activesupport'` to your Gemfile if you required ActiveSupport via this library.

See [#325](https://github.com/slack-ruby/slack-ruby-client/pull/325) for more information.

### Upgrading to >= 0.14.0

If you are using async-websocket, lock down its version to 0.8.0 as newer versions are currently incompatible.

```
gem 'async-websocket', '~> 0.8.0'
```

See [#282](https://github.com/slack-ruby/slack-ruby-client/issues/282) for more information.

### Upgrading to >= 0.13.0

#### Recommended Async Library

The RealTime client now supports [async-websocket](https://github.com/socketry/async-websocket), which is actively maintained and is now the recommended library.

See [#219](https://github.com/slack-ruby/slack-ruby-client/pull/219) for implementation details.

#### Async Library Threading and Reactor Changes

The RealTime celluloid-io implementation now uses a `Thread` per client. Previous versions used an `Actor`.

The faye-websocket implementation with EventMachine will attempt to shutdown EventMachine via `EventMachine.stop` upon shutdown if a reactor wasn't already running.

See [#224](https://github.com/slack-ruby/slack-ruby-client/pull/224) for more information.

### Upgrading to >= 0.9.0

#### Changes in How the RTM Client Connects

The RealTime client now automatically chooses either [rtm.start](https://api.slack.com/methods/rtm.start) or [rtm.connect](https://api.slack.com/methods/rtm.connect) to open a connection. The `rtm.connect` method is a newer, preferred method, which serves connection purposes and returns some basic team info. The `rtm.start` method additionally returns a lot of data about the team, its channels, and members, and is required to use the full `Slack::RealTime::Stores::Store` storage class.

Prior versions always used `rtm.start`, to restore this behavior, configure `start_method`.

```ruby
Slack::RealTime::Client.config do |config|
  config.start_method = :rtm_start
end
```

See [#145](https://github.com/slack-ruby/slack-ruby-client/pull/145) for more information.

### Upgrading to >= 0.8.0

The default timeout for `rtm.start` has been increased from 60 to 180 seconds via `Slack::RealTime::Client.config.start_options[:request][:timeout]`. If you're explicitly setting `start_options` in your application, preserve the value by merging settings instead of replacing the entire `start_options` value.

Before:

```ruby
Slack::RealTime::Client.config do |config|
  config.start_options = { no_unreads: true }
end
```

After:

```ruby
Slack::RealTime::Client.config do |config|
  config.start_options[:no_unreads] = true # keeps config.start_options[:request] intact
end
```

See [#136](https://github.com/slack-ruby/slack-ruby-client/pull/136) for more information.

### Upgrading to >= 0.6.0

#### Changes to API Response Data

API responses in both Web and RealTime clients are now instances of [Slack::Messages::Message](lib/slack/messages/message.rb), which provides method access to properties.

Before:

```ruby
puts "Welcome '#{client.self['name']}' to the '#{client.team['name']}' team."
```

After:

```ruby
puts "Welcome #{client.self.name} to the #{client.team.name} team."
```

See [#56](https://github.com/slack-ruby/slack-ruby-client/issues/56) for more information.

#### Changes to RealTime Local Store

Upon a successful `rtm.start` the RealTime client keeps a local cache of objects, including `self` or `users`. It will now also track changes to these objects. The following changes have been made to the data structures.

##### client.self

The `client.self` object is now a `Slack::RealTime::Models::User`, which is a child of `Hashie::Mash`, so no code changes should be required.

##### client.team

The `client.team` object is now a `Slack::RealTime::Models::Team`, which is a child of `Hashie::Mash`, so no code changes should be required.

##### client .users, .channels, .groups, .ims

The `client.users`, `.channels`, `.groups` and `.ims` collections have been changed from `Array` to `Hash`, with object ID as key. Replace any code iterating over the array, eg. `client.users.values.each` or `client.channels.each_pair { |id, channel| ... }`.

See [#55](https://github.com/slack-ruby/slack-ruby-client/issues/55) for more information.

### Upgrading to >= 0.5.0

#### Changes to Real Time Concurrency

Since 0.5.0 `Slack::RealTime::Client` supports [Celluloid](https://github.com/celluloid/celluloid) and no longer defaults to [Faye::WebSocket](https://github.com/faye/faye-websocket-ruby) with [Eventmachine](https://github.com/eventmachine/eventmachine). It will auto-detect one or the other depending on the gems in your Gemfile, which means you may need to add one or the other to your Gemfile.

##### Faye::WebSocket with Eventmachine

```
gem 'faye-websocket'
```

##### Celluloid

```
gem 'celluloid-io'
```

When in doubt, use `faye-websocket`.

See [#5](https://github.com/slack-ruby/slack-ruby-client/issues/5) for more information.