Events

An ink! smart contract may define events that it can emit during contract execution. Emitting events can be used by third party tools to query information about a contract's execution and state.

Example

The following example ink! contract shows how an event Transferred is defined and emitted in the #[ink(constructor)].

#[ink::contract]
mod erc20 {
    /// Defines an event that is emitted
    /// every time value is transferred.
    #[ink(event)]
    pub struct Transferred {
        from: Option<AccountId>,
        to: Option<AccountId>,
        value: Balance,
    }

    #[ink(storage)]
    pub struct Erc20 {
        total_supply: Balance,
        // more fields ...
    }

    impl Erc20 {
        #[ink(constructor)]
        pub fn new(initial_supply: Balance) -> Self {
            let caller = Self::env().caller();
            Self::env().emit_event(Transferred {
                from: None,
                to: Some(caller),
                value: initial_supply,
            });
            Self { total_supply: initial_supply }
        }

        #[ink(message)]
        pub fn total_supply(&self) -> Balance {
            self.total_supply
        }
    }
}

See our ERC20 example contract for an elaborate example which uses events.

Event Definition

Since ink! version 5.0, events can be defined independently of the contract which emits them. Events can now be defined once and shared across multiple contracts.

This is useful for events for contracts which conform to standards such as ERC-20: contract indexers/explorers are now able to group all e.g. Transferred events.

This is how an event definition looks:

use ink::primitives::AccountId;

#[ink::event]
pub struct Transferred {
    #[ink(topic)]
    from: Option<AccountId>,
    #[ink(topic)]
    to: Option<AccountId>,
    amount: u128,
}

Note that generics are not currently supported , so the concrete types of Environment specific types such as AccountId must match up with the types used in the contract.

This definition can exist within a contract definition module (inline events), in a different module in the same crate or even in a different crate to be shared by multiple contracts.

Legacy syntax for inline Event definitions

Events defined within a #[ink::contract] module can continue to use the original syntax for an event definition, using the #[ink(event)] attribute. Under the covers this is simply expanded to the new top level #[ink::event] macro, so both events defined using the legacy style and using the new event attribute macro directly will behave exactly the same.

Topics

When an event is emitted, 0 or more topics can be associated with it. The event is then indexed together with other events with the same topic value.

An event's fields can be annotated with #[ink(topic)] (see example), which will result in a topic derived from the value of that field being emitted together with the event.

Topics are by default a 32 byte array ([u8; 32]), although this is configurable on the Polkadot SDK runtime level. If the SCALE encoded bytes of a field value are <= 32, then the encoded bytes are used directly as the topic value.

For example, in the common case of indexing a field of type AccountId, where the default AccountId type is 32 bytes in length, the topic value will be the encoded account id itself. This makes it easy to filter for all events which have a topic of a specific AccountId.

If however the size of the encoded bytes of the value of a field exceeds 32, then the encoded bytes will be hashed using the Blake2x256 hasher.

Topics are a native concept in the Polkadot SDK, and can be queried via EventTopics

How to choose which fields to make topics? A good rule of thumb is to ask yourself if somebody might want to search for this topic. For this reason the amount in the example Transferred event above was not made indexable ‒ there will most probably be a lot of different events with differing amounts each.

Signature Topic

By default all events have a signature topic. This allows indexing of all events of the same type, emitted by different contracts. The #[ink::event] macro generates a signature topic at compile time by hashing the name of the event concatenated with the names of the types of the all the field names:

blake2b("Event(field1_type,field2_type)")`

So for our Transferred example it will be:

blake2b("Transferred(Option<AccountId>,Option<AccountId>,u128)")`

caution

Important caveat: because the name of the field type is used, refactoring an event definition to use a type alias or a fully qualified type will change the signature topic, even though the underlying type is the same. Two otherwise identical definitions of an event with the same name and same field types but different field type names will have different signature topics.

When decoding events emitted from a contract, signature topics are now required to determine which type of event to decode into.

Anonymous Events

Events annotated with anonymous will not have a signature topic generated and published with the event.

For inline events, this can be done by marking the event with the anonymous attribute e.g.

#[ink(event, anonymous)]
pub struct Event { .. }

or

#[ink(event)]
#[ink(anonymous)]
pub struct Event { .. }

For events defined using the #[ink::event] macro, the anonymous flag needs to be added as an argument:

#[ink::event(anonymous)]
pub struct Event { .. }

Without a signature topic, indexers will not be able to index over the type of the event, which may be desirable for some contracts, and would be a small gas cost optimization if necessary.

However, when interacting with the contract from a client, no signature topic means that another way is required to determine the type of the event to be decoded into (i.e. how do we know it is a Transferred event, not an Approval event. One way would be to try decoding for each type of event defined in the metadata of the contract until one succeeds. If calling a specific message, it may be known up front what type of event that message will raise, so the client code could just decode into that event directly.

Emitting Events in a Constructor

In a constructor events are emitted via Self::env().emit_event(). See this example:

#[ink(constructor)]
pub fn new(initial_value: Balance) -> Self {
    let caller = Self::env().caller();
    let mut balances = HashMap::new();
    balances.insert(caller, initial_supply);

    Self::env().emit_event(Transferred {
        from: None,
        to: Some(caller),
        amount: initial_supply
    });

    Self { total_supply: initial_supply, balances }
}

Emitting Events from Messages

In a message events are emitted via self.env().emit_event():

#[ink(message)]
pub fn transfer(&mut self, to: AccountId, amount: Balance) -> Result {
    let from = self.env().caller();
    // implementation hidden
    self.env().emit_event(Transferred {
        from: Some(from),
        to: Some(to),
        amount
    });
    Ok(())
}

Cost of using Events

When using events and topics, developers should be mindful of the costs associated. Firstly: if optimizing for contract size, using events will increase the size of the final code size. So minimizing or eliminating event usage where necessary will reduce contract size. The same can be said for the execution (aka gas) costs when using events. We recommend considering the cost of events when using them, and measuring the code size and gas costs with different usage patterns when optimizing.