Skip to main content
Version: nightly

Developing Ignite Apps

It's easy to create an app and use it immediately in your project. First choose a directory outside your project and run:

$ ignite app scaffold my-app

This will create a new directory my-app that contains the app's code and will output some instructions about how to use your app with the ignite command. An app path can be a local directory which has several benefits:

  • You don't need to use a Git repository during the development of your app.
  • The app is recompiled each time you run the ignite binary in your project if the source files are older than the app binary.

Thus, app development workflow is as simple as:

  1. Scaffold an app with ignite app scaffold my-app
  2. Add it to your config via ignite app install -g /path/to/my-app
  3. Update app code
  4. Run ignite my-app binary to compile and run the app
  5. Go back to 3

Once your app is ready you can publish it to a Git repository and the community can use it by calling ignite app install github.com/foo/my-app.

Now let's detail how to update your app's code.

App interface

Under the hood Ignite Apps are implemented using a plugin system based on github.com/hashicorp/go-plugin.

All apps must implement a predefined interface:

ignite/services/plugin/interface.go
type Interface interface {
// Manifest declares app's Command(s) and Hook(s).
Manifest(context.Context) (*Manifest, error)

// Execute will be invoked by ignite when an app Command is executed.
// It is global for all commands declared in Manifest, if you have declared
// multiple commands, use cmd.Path to distinguish them.
// The ClientAPI argument can be used by plugins to get chain app analysis info.
Execute(context.Context, *ExecutedCommand, ClientAPI) error

// ExecuteHookPre is invoked by ignite when a command specified by the Hook
// path is invoked.
// It is global for all hooks declared in Manifest, if you have declared
// multiple hooks, use hook.Name to distinguish them.
// The ClientAPI argument can be used by plugins to get chain app analysis info.
ExecuteHookPre(context.Context, *ExecutedHook, ClientAPI) error

// ExecuteHookPost is invoked by ignite when a command specified by the hook
// path is invoked.
// It is global for all hooks declared in Manifest, if you have declared
// multiple hooks, use hook.Name to distinguish them.
// The ClientAPI argument can be used by plugins to get chain app analysis info.
ExecuteHookPost(context.Context, *ExecutedHook, ClientAPI) error

// ExecuteHookCleanUp is invoked by ignite when a command specified by the
// hook path is invoked. Unlike ExecuteHookPost, it is invoked regardless of
// execution status of the command and hooks.
// It is global for all hooks declared in Manifest, if you have declared
// multiple hooks, use hook.Name to distinguish them.
// The ClientAPI argument can be used by plugins to get chain app analysis info.
ExecuteHookCleanUp(context.Context, *ExecutedHook, ClientAPI) error
}

The scaffolded code already implements this interface, you just need to update the method's body.

Defining app's manifest

Here is the Manifest proto message definition:

proto/ignite/services/plugin/grpc/v1/types.proto
message Manifest {
// App name.
string name = 1;

// Commands contains the commands that will be added to the list of ignite commands.
// Each commands are independent, for nested commands use the inner Commands field.
bool shared_host = 2;

// Hooks contains the hooks that will be attached to the existing ignite commands.
repeated Command commands = 3;

// Enables sharing a single app server across all running instances of an Ignite App.
// Useful if an app adds or extends long running commands.
//
// Example: if an app defines a hook on `ignite chain serve`, a server is instantiated
// when the command is run. Now if you want to interact with that instance
// from commands defined in that app, you need to enable shared host, or else the
// commands will just instantiate separate app servers.
//
// When enabled, all apps of the same path loaded from the same configuration will
// attach it's RPC client to a an existing RPC server.
//
// If an app instance has no other running app servers, it will create one and it
// will be the host.
repeated Hook hooks = 4;
}

In your app's code the Manifest method already returns a predefined Manifest struct as an example. You must adapt it according to your need.

If your app adds one or more new commands to ignite, add them to the Commands field.

If your app adds features to existing commands, add them to the Hooks field.

Of course an app can declare both, Commands and Hooks.

An app may also share a host process by setting SharedHost to true. SharedHost is desirable if an app hooks into, or declares long running commands. Commands executed from the same app context interact with the same app server. Allowing all executing commands to share the same server instance, giving shared execution context.

Adding new commands

App commands are custom commands added to Ignite CLI by an installed app. Commands can use any path not defined already by the CLI.

For instance, let's say your app adds a new oracle command to ignite scaffold, then the Manifest method will look like :

func (app) Manifest(context.Context) (*plugin.Manifest, error) {
return &plugin.Manifest{
Name: "oracle",
Commands: []*plugin.Command{
{
Use: "oracle [name]",
Short: "Scaffold an oracle module",
Long: "Long description goes here...",
// Optional flags is required
Flags: []*plugin.Flag{
{Name: "source", Type: plugin.FlagTypeString, Usage: "the oracle source"},
},
// Attach the command to `scaffold`
PlaceCommandUnder: "ignite scaffold",
},
},
}, nil
}

To update the app execution, you have to change the Execute command. For example:

func (app) Execute(_ context.Context, cmd *plugin.ExecutedCommand, _ plugin.ClientAPI) error {
if len(cmd.Args) == 0 {
return fmt.Errorf("oracle name missing")
}

flags, err := cmd.NewFlags()
if err != nil {
return err
}

var (
name = cmd.Args[0]
source, _ = flags.GetString("source")
)

// Read chain information
c, err := getChain(cmd)
if err != nil {
return err
}

//...
}

Then, run ignite scaffold oracle to execute the app.

Adding hooks

App Hooks allow existing CLI commands to be extended with new functionality. Hooks are useful when you want to streamline functionality without needing to run custom scripts after or before a command has been run. This can streamline processes that where once error prone or forgotten all together.

The following are hooks defined which will run on a registered ignite command:

NameDescription
PreRuns before a commands main functionality is invoked in the PreRun scope
PostRuns after a commands main functionality is invoked in the PostRun scope
Clean UpRuns after a commands main functionality is invoked. If the command returns an error it will run before the error is returned to guarantee execution.

Note: If a hook causes an error in the pre step the command will not run resulting in post and clean up not executing.

The following is an example of a hook definition.

func (app) Manifest(context.Context) (*plugin.Manifest, error) {
return &plugin.Manifest{
Name: "oracle",
Hooks: []*plugin.Hook{
{
Name: "my-hook",
PlaceHookOn: "ignite chain build",
},
},
}, nil
}

func (app) ExecuteHookPre(_ context.Context, h *plugin.ExecutedHook, _ plugin.ClientAPI) error {
switch h.Hook.GetName() {
case "my-hook":
fmt.Println("I'm executed before ignite chain build")
default:
return fmt.Errorf("hook not defined")
}
return nil
}

func (app) ExecuteHookPost(_ context.Context, h *plugin.ExecutedHook, _ plugin.ClientAPI) error {
switch h.Hook.GetName() {
case "my-hook":
fmt.Println("I'm executed after ignite chain build (if no error)")
default:
return fmt.Errorf("hook not defined")
}
return nil
}

func (app) ExecuteHookCleanUp(_ context.Context, h *plugin.ExecutedHook, _ plugin.ClientAPI) error {
switch h.Hook.GetName() {
case "my-hook":
fmt.Println("I'm executed after ignite chain build (regardless errors)")
default:
return fmt.Errorf("hook not defined")
}
return nil
}

Above we can see a similar definition to Command where a hook has a Name and a PlaceHookOn. You'll notice that the Execute* methods map directly to each life cycle of the hook. All hooks defined within the app will invoke these methods.