Developers Home»How-to Guide»Implement a randomness function inside a pallet

Implement a randomness function inside a pallet

Randomness is used in computer programs for many applications. For example, gaming applications, NFT creation, and selecting block authors all require a degree of randomness.

True randomness is hard to come by in deterministic computers. This is particularly true in the context of a blockchain, when all the nodes in the network must agree on the state of the chain. FRAME provides runtime engineers with a source of randomness, using the Randomness trait.

This guide explains how to make use of FRAME's Randomness trait by using the random method and a nonce as a subject. The guide also illustrates how to add entropy to the randomness value by assigning the RandomCollectiveFlip pallet to the configuration trait of a pallet that exposes a "random" type.

Import Randomness

  1. In the pallet you want to use, import the Randomness trait from frame_support:

    use frame_support::traits::Randomness;
    
  2. Include it in your pallet's configuration trait:

    #[pallet::config]
    pub trait frame_system::Config {
        type MyRandomness: Randomness<Self::Hash, Self::BlockNumber>;
    }
    

    Note that the Randomness trait specifies a generic return of type Output and BlockNumber. Use BlockNumber and Hash from frame_system in your pallet to satisfy that trait requirement.

    As stated in this trait's documentation, at best, this trait can give you randomness which was hard to predict a long time ago but that has become easy to predict recently. Keep this in mind when you evaluate your use of it.

Create a nonce and use it in your randomness implementation

Use a nonce to serve as a subject for the frame_support::traits::Randomness::random(subject: &[u8]) method.

  1. There are two steps to including a nonce in your pallet:

    • Create a Nonce storage item. The storage item can be type u32 or u64.

    • Create a private nonce function. This function increments the nonce each time it's used.

    The increment_nonce() private function can be implemented in such a way that it both returns and updates the nonce. For example:

    fn get_and_increment_nonce() -> Vec<u8> {
        let nonce = Nonce::<T>::get();
        Nonce::<T>::put(nonce.wrapping_add(1));
        nonce.encode()
    }
    

    To learn more about the wrapping and encoding methods, see wrapping_add and encode() in the Rust documentation.

  2. Use Randomness in a dispatchable.

    Using the nonce, you can call the random() method that Randomness exposes. The code snippet below is a mock example that assumes relevant events and storage items have been implemented:

    #[pallet::weight(100)]
    pub fn create_unique(
        origin: OriginFor<T>)
        -> DispatchResultWithPostInfo {
        // Account calling this dispatchable.
        let sender = ensure_signed(origin)?;
            // Random value.
            let nonce = Self::get_and_increment_nonce();
            let (randomValue, _) = T::MyRandomness::random(&nonce);
        // Write the random value to storage.
        <MyStorageItem<T>>::put(randomValue);
        Self::deposit_event(Event::UniqueCreated(randomValue));
    }
    
  3. Update your pallet's runtime implementation.

    Because you have added a type to your pallet's configuration trait, Config opens up the opportunity to further enhance the randomness derived by the Randomness trait. This is accomplished by using the Randomness Collective Flip pallet.

    Using this pallet alongside the Randomness trait will significantly improve the entropy being processed by random().

    In runtime/src/lib.rs, assuming pallet_random_collective_flip is instantiated in construct_runtime as RandomCollectiveFlip, specify your exposed type in the following way:

    impl my_pallet::Config for Runtime{
        type Event;
        type MyRandomness = RandomCollectiveFlip;
    }
    

Examples

Verifiable Random Functions

Last edit: on

Run into problems?
Let us Know