In the default configuration of the contracts-pallet a smart contract can only interact with the runtime via its well defined set of basic smart contract interface. This API already allows a whole variety of interaction between the contracts-pallet and the executed smart contract. For example it is possible to call and instantiate other smart contracts on the same chain, emit events, query context information or run built-in cryptographic hashing procedures.

If this basic set of features is not enough for a particular Substrate built blockchain it is possible to easily extend this API using the so-called chain extension feature.

With chain extensions you can expose parts of your runtime logic to smart contract developers.

note

The ink! repository contains the rand-extension example. This is a complete example of a chain extension implemented in both ink! and Substrate.

Structure

The interface consists of an error code that indicates lightweight errors as well as the definition of some chain extension methods.

The overall structure follows that of a simple Rust trait definition. The error code is defined as an associated type definition of the trait definition. The methods are defined as associated trait methods without implementation.

Chain extension methods must not have a self receiver such as &self or &mut self and must have inputs and output that implement SCALE codec. Their return value follows specific rules that can be altered using the handle_status attribute and alternation between Result and Non-Result types which are described in more detail below.

Usage

Usually the chain extension definition using this proc. macro is provided by the author of the chain extension in a separate crate. ink! smart contracts using this chain extension simply depend on this crate and use its associated environment definition in order to make use of the methods provided by the chain extension.

Attributes

There are two different attributes with which the chain extension methods can be flagged:

Attribute	Required	Default Value	Description
ink(extension = N: u32)	Yes	-	Determines the unique function ID of the chain extension method.
ink(handle_status = flag: bool)	Optional	true	Assumes that the returned status code of the chain extension method always indicates success and therefore always loads and decodes the output buffer of the call.

As with all ink! attributes multiple of them can either appear in a contiguous list:

type Access = i32;

#[ink::chain_extension]
pub trait MyChainExtension {
    type ErrorCode = i32;

    #[ink(extension = 5, handle_status = false)]
    fn key_access_for_account(key: &[u8], account: &[u8]) -> Access;
}

…or as multiple standalone ink! attributes applied to the same item:

type Access = i32;

#[ink::chain_extension]
pub trait MyChainExtension {
  type ErrorCode = i32;

  #[ink(extension = 5)]
  #[ink(handle_status = false)]
  fn key_access_for_account(key: &[u8], account: &[u8]) -> Access;
}

Details: handle_status

Default value: true

By default all chain extension methods should return a Result<T, E> where E: From<Self::ErrorCode>. The Self::ErrorCode represents the error code of the chain extension. This means that a smart contract calling such a chain extension method first queries the returned status code of the chain extension method and only loads and decodes the output if the returned status code indicates a successful call. This design was chosen as it is more efficient when no output besides the error code is required for a chain extension call. When designing a chain extension try to utilize the error code to return errors and only use the output buffer for information that does not fit in a single u32 value.

A chain extension method that is flagged with handle_status = false assumes that the returned error code will always indicate success. Therefore it will always load and decode the output buffer and loses the E: From<Self::ErrorCode constraint for the call.

Note that if a chain extension method does not return Result<T, E> where E: From<Self::ErrorCode>, but handle_status = true it will still return a value of type Result<T, Self::ErrorCode>.

Usage: handle_status + Result<T, E> return type

Use both handle_status = false and non-Result return type for the same chain extension method if a call to it may never fail and never returns a Result type.

Combinations

Due to the possibility to flag a chain extension method with handle_status and either (1) return Result<T, E> or (2) return just T there are 4 different cases with slightly varying semantics:

handle_status	Returns Result<T, E>	Effects
true	true	The chain extension method is required to return a value of type Result<T, E> where E: From<Self::ErrorCode>. A call will always check if the returned status code indicates success and only then will load and decode the value in the output buffer.
true	false	The chain extension method may return any non-Result type. A call will always check if the returned status code indicates success and only then will load and decode the value in the output buffer. The actual return type of the chain extension method is still Result<T, Self::ErrorCode> when the chain extension method was defined to return a value of type T.
false	true	The chain extension method is required to return a value of type Result<T, E>. A call will always assume that the returned status code indicates success and therefore always load and decode the output buffer directly.
false	false	The chain extension method may return any non-Result type. A call will always assume that the returned status code indicates success and therefore always load and decode the output buffer directly.

Error Code

Every chain extension defines exactly one ErrorCode using the following syntax:

#[ink::chain_extension]
pub trait MyChainExtension {
    type ErrorCode = MyErrorCode;

    // more definitions ...
}

The defined ErrorCode must implement FromStatusCode which should be implemented as a more or less trivial conversion from the u32 status code to a Result<(), Self::ErrorCode>. The Ok(()) value indicates that the call to the chain extension method was successful.

By convention an error code of 0 represents success. However, chain extension authors may use whatever suits their needs.

Example: Definition

In the below example a chain extension is defined that allows its users to read and write from and to the runtime storage using access privileges:

/// Custom chain extension to read to and write from the runtime.
#[ink::chain_extension]
pub trait RuntimeReadWrite {
    type ErrorCode = ReadWriteErrorCode;

    /// Reads from runtime storage.
    ///
    /// # Note
    ///
    /// Actually returns a value of type `Result<Vec<u8>, Self::ErrorCode>`.
    /// #[ink(extension = 1, returns_result = false)]
    /// fn read(key: &[u8]) -> Vec<u8>;
    ///
    /// Reads from runtime storage.
    ///
    /// Returns the number of bytes read and up to 32 bytes of the
    /// read value. Unused bytes in the output are set to 0.
    ///
    /// # Errors
    ///
    /// If the runtime storage cell stores a value that requires more than
    /// 32 bytes.
    ///
    /// # Note
    ///
    /// This requires `ReadWriteError` to implement `From<ReadWriteErrorCode>`
    /// and may potentially return any `Self::ErrorCode` through its return value.
    #[ink(extension = 2)]
    fn read_small(key: &[u8]) -> Result<(u32, [u8; 32]), ReadWriteError>;

    /// Writes into runtime storage.
    ///
    /// # Note
    ///
    /// Actually returns a value of type `Result<(), Self::ErrorCode>`.
    #[ink(extension = 3)]
    fn write(key: &[u8], value: &[u8]);

    /// Returns the access allowed for the key for the caller.
    ///
    /// # Note
    ///
    /// Assumes to never fail the call and therefore always returns `Option<Access>`.
    #[ink(extension = 4, handle_status = false)]
    fn access(key: &[u8]) -> Option<Access>;

    /// Unlocks previously acquired permission to access key.
    ///
    /// # Errors
    ///
    /// If the permission was not granted.
    ///
    /// # Note
    ///
    /// Assumes the call to never fail and therefore does _NOT_ require `UnlockAccessError`
    /// to implement `From<Self::ErrorCode>` as in the `read_small` method above.
    #[ink(extension = 5, handle_status = false)]
    fn unlock_access(key: &[u8], access: Access) -> Result<(), UnlockAccessError>;
}

#[derive(scale::Encode, scale::Decode, scale_info::TypeInfo)]
pub enum ReadWriteErrorCode {
  InvalidKey,
  CannotWriteToKey,
  CannotReadFromKey,
}

#[derive(scale::Encode, scale::Decode, scale_info::TypeInfo)]
pub enum ReadWriteError {
  ErrorCode(ReadWriteErrorCode),
  BufferTooSmall { required_bytes: u32 },
}

impl From<ReadWriteErrorCode> for ReadWriteError {
  fn from(error_code: ReadWriteErrorCode) -> Self {
    Self::ErrorCode(error_code)
  }
}

impl From<scale::Error> for ReadWriteError {
  fn from(_: scale::Error) -> Self {
    panic!("encountered unexpected invalid SCALE encoding")
  }
}

#[derive(scale::Encode, scale::Decode, scale_info::TypeInfo)]
pub struct UnlockAccessError {
  reason: String,
}

impl From<scale::Error> for UnlockAccessError {
  fn from(_: scale::Error) -> Self {
    panic!("encountered unexpected invalid SCALE encoding")
  }
}

#[derive(scale::Encode, scale::Decode, scale_info::TypeInfo)]
pub enum Access {
  ReadWrite,
  ReadOnly,
  WriteOnly,
}

impl ink::env::chain_extension::FromStatusCode for ReadWriteErrorCode {
  fn from_status_code(status_code: u32) -> Result<(), Self> {
    match status_code {
      0 => Ok(()),
      1 => Err(Self::InvalidKey),
      2 => Err(Self::CannotWriteToKey),
      3 => Err(Self::CannotReadFromKey),
      _ => panic!("encountered unknown status code"),
    }
  }
}

All the error types and other utility types used in the chain extension definition above are often required to implement various traits such as SCALE's Encode and Decode as well as scale-info's TypeInfo trait.

A full example of the above chain extension definition can be seen here.

Example: Environment

In order to allow ink! smart contracts to use the above defined chain extension it needs to be integrated into an Environment definition as shown below:

type RuntimeReadWrite = i32;

use ink::env::{Environment, DefaultEnvironment};

pub enum CustomEnvironment {}

impl Environment for CustomEnvironment {
    const MAX_EVENT_TOPICS: usize =
        <DefaultEnvironment as Environment>::MAX_EVENT_TOPICS;

    type AccountId = <DefaultEnvironment as Environment>::AccountId;
    type Balance = <DefaultEnvironment as Environment>::Balance;
    type Hash = <DefaultEnvironment as Environment>::Hash;
    type BlockNumber = <DefaultEnvironment as Environment>::BlockNumber;
    type Timestamp = <DefaultEnvironment as Environment>::Timestamp;

    type ChainExtension = RuntimeReadWrite;
}

Above we defined the CustomEnvironment which defaults to ink!'s DefaultEnvironment for all constants and types but the ChainExtension type which is assigned to our newly defined chain extension.

Example: Usage

An ink! smart contract can use the above defined chain extension through the Environment definition defined in the last example section using the env macro parameter as shown below.

Note that chain extension methods are accessible through Self::extension() or self.extension(). For example as in Self::extension().read(..) or self.extension().read(..).

#[ink::contract(env = CustomEnvironment)]
mod read_writer {

    #[ink(storage)]
    pub struct ReadWriter {}

    impl ReadWriter {
        #[ink(constructor)]
        pub fn new() -> Self { Self {} }

        #[ink(message)]
        pub fn read(&self, key: Vec<u8>) -> Result<Vec<u8>, ReadWriteErrorCode> {
            self.env()
                .extension()
                .read(&key)
        }

        #[ink(message)]
        pub fn read_small(&self, key: Vec<u8>) -> Result<(u32, [u8; 32]), ReadWriteError> {
            self.env()
                .extension()
                .read_small(&key)
        }

        #[ink(message)]
        pub fn write(
            &self,
            key: Vec<u8>,
            value: Vec<u8>,
        ) -> Result<(), ReadWriteErrorCode> {
            self.env()
                .extension()
                .write(&key, &value)
        }

        #[ink(message)]
        pub fn access(&self, key: Vec<u8>) -> Option<Access> {
            self.env()
                .extension()
                .access(&key)
        }

        #[ink(message)]
        pub fn unlock_access(&self, key: Vec<u8>, access: Access) -> Result<(), UnlockAccessError> {
            self.env()
                .extension()
                .unlock_access(&key, access)
        }
    }

    /// Custom chain extension to read to and write from the runtime.
    #[ink::chain_extension]
    pub trait RuntimeReadWrite {
          type ErrorCode = ReadWriteErrorCode;
          #[ink(extension = 1)]
          fn read(key: &[u8]) -> Vec<u8>;
          #[ink(extension = 2)]
          fn read_small(key: &[u8]) -> Result<(u32, [u8; 32]), ReadWriteError>;
          #[ink(extension = 3)]
          fn write(key: &[u8], value: &[u8]);
          #[ink(extension = 4, handle_status = false)]
          fn access(key: &[u8]) -> Option<Access>;
          #[ink(extension = 5, handle_status = false)]
          fn unlock_access(key: &[u8], access: Access) -> Result<(), UnlockAccessError>;
    }

    #[derive(scale::Encode, scale::Decode, scale_info::TypeInfo)]
    pub enum ReadWriteErrorCode {
          InvalidKey,
          CannotWriteToKey,
          CannotReadFromKey,
    }

    #[derive(scale::Encode, scale::Decode, scale_info::TypeInfo)]
    pub enum ReadWriteError {
          ErrorCode(ReadWriteErrorCode),
          BufferTooSmall { required_bytes: u32 },
    }
    impl From<ReadWriteErrorCode> for ReadWriteError {
         fn from(error_code: ReadWriteErrorCode) -> Self {
             Self::ErrorCode(error_code)
         }
    }
    impl From<scale::Error> for ReadWriteError {
         fn from(_: scale::Error) -> Self {
             panic!("encountered unexpected invalid SCALE encoding")
         }
    }

    #[derive(scale::Encode, scale::Decode, scale_info::TypeInfo)]
    pub struct UnlockAccessError {
         reason: String,
    }
    impl From<scale::Error> for UnlockAccessError {
         fn from(_: scale::Error) -> Self {
             panic!("encountered unexpected invalid SCALE encoding")
         }
    }
    #[derive(scale::Encode, scale::Decode, scale_info::TypeInfo)]
    pub enum Access {
         ReadWrite,
         ReadOnly,
         WriteOnly,
    }
    impl ink::env::chain_extension::FromStatusCode for ReadWriteErrorCode {
         fn from_status_code(status_code: u32) -> Result<(), Self> {
             match status_code {
                 0 => Ok(()),
                 1 => Err(Self::InvalidKey),
                 2 => Err(Self::CannotWriteToKey),
                 3 => Err(Self::CannotReadFromKey),
                 _ => panic!("encountered unknown status code"),
             }
         }
    }
    pub enum CustomEnvironment {}
    impl ink::env::Environment for CustomEnvironment {
         const MAX_EVENT_TOPICS: usize =
             <ink::env::DefaultEnvironment as ink::env::Environment>::MAX_EVENT_TOPICS;

         type AccountId = <ink::env::DefaultEnvironment as ink::env::Environment>::AccountId;
         type Balance = <ink::env::DefaultEnvironment as ink::env::Environment>::Balance;
         type Hash = <ink::env::DefaultEnvironment as ink::env::Environment>::Hash;
         type BlockNumber = <ink::env::DefaultEnvironment as ink::env::Environment>::BlockNumber;
         type Timestamp = <ink::env::DefaultEnvironment as ink::env::Environment>::Timestamp;

         type ChainExtension = RuntimeReadWrite;
    }
}

Technical Limitations

• Due to technical limitations it is not possible to refer to the ErrorCode associated type using Self::ErrorCode anywhere within the chain extension and its defined methods. Instead chain extension authors should directly use the error code type when required. This limitation might be lifted in future versions of ink!.
• It is not possible to declare other chain extension traits as super traits or super chain extensions of another.