Canister SDK

SDK for writing and testing canisters for the Internet Computer in Rust. This repo includes a few crates that help to simplify the tricky aspects of IC canisters development:

• Canister SDK
  • Crates
  • canister-sdk
    • ic-exports
    • ic-canister
      • Inter-canister calls and testing
      • Canister traits and composition
      • Dependencies between canisters
    • ic-factory
    • ic-helpers
    • ic-storage
      • In-memory storage
      • Versioned state
      • Canister state and upgrades
    • ic-metrics
    • ic-auction
    • ic-log

This project builds on top of ic-cdk and ic-kit crates. It is not intended to replace them, but adds some types and macros to simplify things that are not dealt with by those crates.

Crates

The following crates are included in this repo:

• ic-auction - This crate provides an API trait for auction canisters. This crate is optional.
• ic-canister - This crate provides a set of macros and types to simplify the development of canisters.
• ic-export - This crate mainly re-exports commonly used dependencies from ic dependencies.
• ic-log - This crate provides a simple logger implementation of the log crate. Log records are saved into memory and can be inspected with a canister query.
• ic-metrics - This crate provides an API trait to simplify the collection of metrics from canisters.
• ic-storage - This crate provides a simple in-memory storage for canisters.

canister-sdk

A wrapper crate among all the canisters in this repository with all the necessary re-exports. It some features for exporting factory and auction traits as well as their APIs.

Note: Currently we require each canister to have its own export-api feature that will export the wasm definitions of the canister (and hide them if this canister is used as a dependency). This is not the case with trait canisters as rust features are additive even in transitive dependencies we're forced to introduce different export-api features for different traits in this repo (auction-api and factory-api respectively). The best approach will be to use canister-sdk dependency with dependent features that will trigger the necessary APIs transitively, e.g.

[ package ]
  edition = "2021"
  name = "awesome-canister"
  version = "0.1.0"

  [ features ]
  default = []
  export-api = ["canister-sdk/auction-api"]
  auction = ["canister-sdk/auction"]

  [ dependencies ]
  canister-sdk = { git = "https://github.com/infinity-swap/canister-sdk", package = "canister-sdk" }

There's an example crate on how to use the export-api feature in the canister-e directory. This crate shows how to use the export-api features to export the methods in the canister and how to generate the idl for the canister with the methods included.

ic-exports

The ic-exports crate re-exports commonly used ic dependencies. This means that you can use the ic-exports crate to import the ic dependencies without having to specify the version of the ic dependencies in your Cargo.toml file.

use ic_exports::*;

The common ic dependencies that are re-exported are:

• ic-cdk
• ic-cdk-macros
• ic-kit
• ic-agent
• ic-icrc1
• ic-icrc1-index
• ledger-canister
• ic-ledger-core
• ic-stable-structures
• cycles-minting-canister
• ic-base-types
• ic-ic00-types

ic-canister

This crate introduces a framework to write easily testable canisters, including testing inter-canister calls, as well as a way to compose canister APIs using rust traits. There are few examples below, but for the details you can check out the crate documentation.

Inter-canister calls and testing

For example, you can have a canister with simple API:

#[derive(Clone, Canister)]
struct MyCanister {
    #[id]
    principal: Principal,

    #[state]
    state: Rc<RefCell<MyCanisterState>>,
}

impl MyCanister {
    #[query]
    fn get_counter(&self) -> u64 {
        self.state.borrow().counter
    }

    #[update]
    fn add(&self, value: u64) {
        self.state.borrow_mut().counter += value;
    }
}

Now instead of using ic_cdk::api::call to query these APIs from another canister doing manual serialization/deserialization:

impl MyOtherCanister {
    #[update]
    async fn increment_get(&self, id: Principal) -> u64 {
        let my_canister = MyCanister::from_principal(id);
        canister_call!(my_canister.add(1), ()).await.unwrap();
        let updated_value = canister_call!(my_canister.get_counter(), u64).await.unwrap();

        updated_value
    }
}

Now, if you want to test the increment_get method of you canister, all you need to do is to write a unit test:

#[tokio::test]
async fn test_increment_get() {
    let my_canister = MyCanister::init_instance();
    let my_other_canister = MyOtherCanister::init_instance();

    assert_eq!(my_other_canister.increment_get(my_canister.principal()), 1);
    assert_eq!(my_other_canister.increment_get(my_canister.principal()), 2);
}

Even though the canisters use statics internally to store the canister state, the tests can initialize multiple instances of canisters with init_instance method, and each one of them will have a separate state.

Canister traits and composition

It is also possible to write a trait to store some part of the canister API that can be reused in different canisters. This also allows to compose a canister from different traits just by implementing the needed traits for your canister Struct:

impl Factory for MyCanister {
    fn state(&self) -> Rc<RefCell<FactoryState>> {
        <Self as Factory>::FactoryStateState::get()
    }
}

impl OtherCanisterTrait for MyCanister {
    // impl
}

For other examples you can look into the tests

Dependencies between canisters

If you use ic-cdk to create your canister's API's, you cannot simply use a canister as a rust dependency for another canister, because all the APIs of that dependency will not be included into the canister you're implementing. If you create a canister using canister-sdk, don't forget to add a export-api feature to the dependency so that it will export all the wasm from the canister trait definition (more specifically from one that is auto-generated by generate_exports! macro).

ic-helpers

Some helpers and wrappers over principals we use on day-to-day basis for ledger and management canister calls.

ic-storage

Introduces traits IcStorage and Versioned for in-memory and stable storage management respectively.

In-memory storage

In the past, the ic-cdk crate provided methods in the ic_cdk::storage module to store and get structs from the canister's memory, but they were removed in version 0.5.0 (for a good reason). The recommended way to store the data in the canister memory is to use thread_local storage with RefCell controlling access to the struct.

The ic_storage::IcStorage deriving macro does exactly that, but saving you some boilerplate. Using it is quite straightforward:

use ic_storage::IcStorage;

#[derive(IcStorage, Default)]
struct MyCanisterState {
    value: u32,
}

let local_state = MyCanisterState::get();
assert_eq!(local_state.borrow().value, 0);
local_state.borrow_mut().value = 42;
assert_eq!(local_state.borrow().value, 42);

It also allows having generic state structures. For detailed information, check out the crate level documentation.

Versioned state

The ic_storage::stable module introduces Versioned trait that allows transparent upgrades for the state during canister upgrades (even over several versions of state at once). When using this trait, the state structure can be serialized into the stable storage using ic_storage::stable::write method. Then after the upgrade, simply use ic_storage::stable::read::<NewStateType>(). This will read the serialized previous version of the state, check its version and run the upgrade methods until the current version of the type (the NewStateType struct) is reached.

Check out the module level documentation for more details.

Canister state and upgrades

When using Canister deriving macro, the fields that are marked with #[state] attribute are all preserved over canister upgrades. This is done using Versioned trait. This means that at this moment you can have only one #[state] in a canister. If the state type is changed, the new state must have the previous state type as its Versioned::Previous type. The Canister deriving macro takes care of generating the pre_upgrade and post_upgrade functions and updating the state to the new type when needed.

If a canister needs to have a state that is not preserved during the upgrade process (like caches or some other temporary data), #[state(stable_store = false)] can be used in addition to the #[state] field. Any number of non-stable state fields can be added to a canister.

ic-metrics

Metrics trait that the canister can implement to store a history of necessary metrics (stored in MetricsData) for the canister that also allows to overwrite the update_metrics call to store custom metrics for a canister. For an example you can refer to the tests.

This crate currently provides two APIs for metrics, these are :-

• get_metrics - This API returns the metrics data for the canister. It returns snapshot of the metrics data over a certain period of the interval.
• get_curr_metrics - This API returns the current metrics data for the canister.

ic-auction

The IC canisters needs cycles to run and as well as to pay for the storage. This crate provides a mechanism of cycle auctions, that doesn't require owner's attention for the canister cycle management. Cycle auctions are run in a set of intervals, and allow any user to add cycles to the canister and retrieve a reward set up via disburse_rewards call as a reward for doing so. You can find more information about the cycle auctions in the README.

This crate is optional and can be used by adding the feature auction to your Cargo.toml file.

[ dependencies ]
  canister-sdk = { git = "https://github.com/infinity-swap/canister-sdk", features = ["auction"] tag = "vx.x.xx"}

Note: When you're using the canister-sdk with these crates, make sure to include export-api feature as well, and when building the canister, make sure to include the --export-api flag.

ic-log

IC log is a log library for IC canisters. It allows you to use the log crate macros (debug!, info!, error!...) into a canister. The records are written into memory with a configurable history size and can be later be retrieved by a query to the canister.

You can see ic-log in action in the examples