The Sway Programming Language

Sway is a domain-specific language (DSL) for the Fuel Virtual Machine (FuelVM), a blockchain-optimized VM designed for the Fuel blockchain. Sway is based on Rust, and includes syntax to leverage a blockchain VM without needlessly verbose boilerplate.

This book documents how to write smart contracts in Sway, along with how to install and use the Sway toolchain.

Before starting developing smart contracts in Sway, please keep in mind the known issues and workarounds of the language and toolchain.

Introduction

To get started with Forc and Sway smart contract development, install the Fuel toolchain and Fuel full node and set up your first project.

Installation

The Sway toolchain is sufficient to compile Sway smart contracts. Otherwise, note that if you want to run Sway smart contracts (e.g. for testing), a Fuel Core full node is required, which is packaged together with the Sway toolchain together as the Fuel toolchain.

Installing via pre-compiled release binaries is the recommended way to get up and running with the Sway toolchain. Pre-compiled binaries for Linux and macOS are available. Native Windows is currently unsupported (tracking issue for Windows support). Windows Subsystem for Linux should work but is not officially supported.

Dependencies

A prerequisite for installing and using Sway is the Rust toolchain. Platform-specific instructions for installing rustup can be found here. Then, install the Rust toolchain with:

# Install the latest stable Rust toolchain.
rustup install stable

The Sway toolchain is built and tested against the stable Rust toolchain version (https://github.com/rust-lang/rust/releases/latest). There is no guarantee it will work with the nightly Rust toolchain, or with earlier stable versions, so ensure you are using stable with:

# Update installed Rust toolchain; can be used independently.
rustup update
# Set the stable Rust toolchain as default; can be used independently.
rustup default stable

Now you're ready to install fuelup, the equivalent of Rust's rustup for the Fuel toolchain. It enables easily downloading binary releases of the Fuel toolchain.

  1. Start by installing fuelup with the following command:

    curl --proto '=https' --tlsv1.2 -sSf \
    https://fuellabs.github.io/fuelup/fuelup-init.sh | sh
    

    This downloads the fuelup-init script to a temp directory on your machine, which installs fuelup. fuelup-init will ask for permission to add ~/.fuelup/bin to your PATH. Otherwise, you can also pass --no-modify-path so that fuelup-init does not modify your PATH:

    curl --proto '=https' --tlsv1.2 -sSf \
    https://fuellabs.github.io/fuelup/fuelup-init.sh | sh -s -- --no-modify-path
    
  2. Once fuelup is installed, fuelup-init automatically runs fuelup toolchain install latest to install the latest toolchain

    You can run fuelup update at anytime to get the most up-to-date toolchain.

  3. (Optional) You can optionally install distributed toolchains optimized for different networks.

    To configure the optimal toolchain for beta-2, run the following commands:

    $ fuelup self update
    Fetching binary from https://github.com/FuelLabs/fuelup/releases/download/v0.18.0/fuelup-0.18.0-aarch64-apple-darwin.tar.gz
     Downloading component fuelup without verifying checksum
     Unpacking and moving fuelup to /var/folders/tp/0l8zdx9j4s9_n609ykwxl0qw0000gn/T/.tmpP3HfvR
     Moving /var/folders/tp/0l8zdx9j4s9_n609ykwxl0qw0000gn/T/.tmpP3HfvR/fuelup to /Users/user/.fuelup/bin/fuelup
    
    
    $ fuelup toolchain install beta-3
    Downloading: forc forc-explore forc-wallet fuel-core fuel-indexer
    
    Adding component forc v0.35.0 to 'beta-3-aarch64-apple-darwin'
    Fetching binary from https://github.com/FuelLabs/sway/releases/download/v0.35.0/forc-binaries-darwin_arm64.tar.gz
    npacking and moving forc-doc to /Users/user/.fuelup/toolchains/beta-3-aarch64-apple-darwin/bin
    Unpacking and moving forc to /Users/user/.fuelup/toolchains/beta-3-aarch64-apple-darwin/bin
    Unpacking and moving forc-deploy to /Users/user/.fuelup/toolchains/beta-3-aarch64-apple-darwin/bin
    Unpacking and moving forc-run to /Users/user/.fuelup/toolchains/beta-3-aarch64-apple-darwin/bin
    Unpacking and moving forc-lsp to /Users/user/.fuelup/toolchains/beta-3-aarch64-apple-darwin/bin
    Unpacking and moving forc-fmt to /Users/user/.fuelup/toolchains/beta-3-aarch64-apple-darwin/bin
    Fetching core forc dependencies
    Installed forc v0.35.0 for toolchain 'beta-3-aarch64-apple-darwin'
    
    Adding component forc-explore v0.28.1 to 'beta-3-aarch64-apple-darwin'
    Fetching binary from https://github.com/FuelLabs/forc-explorer/releases/download/v0.28.1/forc-explore-0.28.1-aarch64-apple-darwin.tar.gz
    Unpacking and moving forc-explore to /Users/user/.fuelup/toolchains/beta-3-aarch64-apple-darwin/bin
    Installed forc-explore v0.28.1 for toolchain 'beta-3-aarch64-apple-darwin'
    
    Adding component forc-wallet v0.1.3 to 'beta-3-aarch64-apple-darwin'
    Fetching binary from https://github.com/FuelLabs/forc-wallet/releases/download/v0.1.3/forc-wallet-0.1.3-aarch64-apple-darwin.tar.gz
    Unpacking and moving forc-wallet to /Users/user/.fuelup/toolchains/beta-3-aarch64-apple-darwin/bin
    Installed forc-wallet v0.1.3 for toolchain 'beta-3-aarch64-apple-darwin'
    
    Adding component fuel-core v0.17.1 to 'beta-3-aarch64-apple-darwin'
    Fetching binary from https://github.com/FuelLabs/fuel-core/releases/download/v0.17.1/fuel-core-0.17.1-aarch64-apple-darwin.tar.gz
    Unpacking and moving fuel-core to /Users/user/.fuelup/toolchains/    beta-3-aarch64-apple-darwin/bin
    Installed fuel-core v0.17.1 for toolchain 'beta-3-aarch64-apple-darwin'
    
    Adding component fuel-indexer v0.2.3 to 'beta-3-aarch64-apple-darwin'
    Fetching binary from https://github.com/FuelLabs/fuel-indexer/releases/download/v0.2.3/fuel-indexer-0.2.3-aarch64-apple-darwin.tar.gz
    Unpacking and moving fuel-indexer to /Users/user/.fuelup/toolchains/beta-3-aarch64-apple-darwin/bin
    Installed fuel-indexer v0.2.3 for toolchain 'beta-3-aarch64-apple-darwin'
    
    Installed:
    - forc 0.35.0
    - forc-explore 0.28.1
    - forc-wallet 0.1.3
    - fuel-core 0.17.1
    - fuel-indexer 0.2.3
    
    The Fuel toolchain is installed and up to date
    

You're all set to start building!

Need Help?

You may refer to The Fuelup Book for an in-depth look into fuelup, or check out the tooling section in the Fuel forum if you're running into problems through the installation process. If you don't see your question, post the issue you're running into with as many details as possible and the team will get back to you asap!

Installing from Cargo

The Sway toolchain and Fuel Core full node can be installed from source with Cargo with:

cargo install forc fuel-core

Updating forc from Cargo

You can update the toolchain from source with Cargo with:

cargo update forc fuel-core

Installing forc Plugins from Cargo

The Fuel ecosystem has a few plugins which can be easily installed via Cargo.

Note: forc detects anything in your $PATH prefixed with forc- as a plugin. Use forc plugins to see what you currently have installed.

# Sway Formatter
cargo install forc-fmt

# Block Explorer
cargo install forc-explore

# Sway Language Server
cargo install forc-lsp

Installing from Source

Rather than installing from cargo, the Sway toolchain can be built from a local source checkout by following instructions at https://github.com/FuelLabs/sway. The Fuel Core full node implementation can be built from source by following instructions at https://github.com/FuelLabs/fuel-core.

Enable tab completion for Bash, Fish, Zsh, or PowerShell

forc supports generating completion scripts for Bash, Fish, Zsh, and PowerShell. See forc completions --help for full details, but the gist is as simple as using one of the following:

# Bash
forc completions --shell=bash > ~/.local/share/bash-completion/completions/forc

# Bash (macOS/Homebrew)
forc completions --shell=bash > $(brew --prefix)/etc/bash_completion.d/forc.bash-completion

# Fish
mkdir -p ~/.config/fish/completions
forc completions --shell=fish > ~/.config/fish/completions/forc.fish

# Zsh
forc completions --shell=zsh > ~/.zfunc/_forc

# PowerShell v5.0+
forc completions --shell=powershell >> $PROFILE.CurrentUserCurrentHost
# or
forc completions --shell=powershell | Out-String | Invoke-Expression

Once the completions have been generated and properly installed, close and reopen your terminal for the new completions to take effect.

Sway Quickstart

Check out the Developer Quickstart Guide for a step-by-step guide on building a fullstack dapp on Fuel. The guide will walk you through writing a smart contract, setting up a wallet, and building a frontend to interact with your contract.

The Fuel Toolchain

The Fuel toolchain consists of several components.

Forc (forc)

The "Fuel Orchestrator" Forc is our equivalent of Rust's Cargo. It is the primary entry point for creating, building, testing, and deploying Sway projects.

Sway Language Server (forc-lsp)

The Sway Language Server forc-lsp is provided to expose features to IDEs. Installation instructions.

Currently, only Visual Studio Code is supported through a plugin. Vim support is forthcoming, though syntax highlighting is provided.

Note: There is no need to manually run forc-lsp (the plugin will automatically start it), however both forc and forc-lsp must be in your $PATH. To check if forc is in your $PATH, type forc --help in your terminal.

Sway Formatter (forc-fmt)

A canonical formatter is provided with forc-fmt. Installation instructions. It can be run manually with

forc fmt

The Visual Studio Code plugin will automatically format Sway files with forc-fmt on save, though you might have to explicitly set the Sway plugin as the default formatter, like this:

"[sway]": {
  "editor.defaultFormatter": "FuelLabs.sway-vscode-plugin"
}

Fuel Core (fuel-core)

An implementation of the Fuel protocol, Fuel Core, is provided together with the Sway toolchain to form the Fuel toolchain. The Rust SDK will automatically start and stop an instance of the node during tests, so there is no need to manually run a node unless using Forc directly without the SDK.

A Forc Project

To initialize a new project with Forc, use forc new:

forc new my-fuel-project

Here is the project that Forc has initialized:

$ cd my-fuel-project
$ tree .
├── Forc.toml
└── src
    └── main.sw

Forc.toml is the manifest file (similar to Cargo.toml for Cargo or package.json for Node), and defines project metadata such as the project name and dependencies.

For additional information on dependency management, see: here.

[project]
authors = ["User"]
entry = "main.sw"
license = "Apache-2.0"
name = "my-fuel-project"

[dependencies]

Here are the contents of the only Sway file in the project, and the main entry point, src/main.sw:

contract;

abi MyContract {
    fn test_function() -> bool;
}

impl MyContract for Contract {
    fn test_function() -> bool {
        true
    }
}

The project is a contract, one of four different project types. For additional information on different project types, see here.

We now compile our project with forc build, passing the flag --print-finalized-asm to view the generated assembly:

$ forc build --print-finalized-asm
...
.program:
ji   i4
noop
DATA_SECTION_OFFSET[0..32]
DATA_SECTION_OFFSET[32..64]
lw   $ds $is 1
add  $$ds $$ds $is
lw   $r0 $fp i73              ; load input function selector
lw   $r1 data_0               ; load fn selector for comparison
eq   $r2 $r0 $r1              ; function selector comparison
jnzi $r2 i12                  ; jump to selected function
movi $$tmp i123               ; special code for mismatched selector
rvrt $$tmp                    ; revert if no selectors matched
ret  $one
.data:
data_0 .word 559005003

  Compiled contract "my-fuel-project".
  Bytecode size is 60 bytes.

Standard Library

Similar to Rust, Sway comes with its own standard library.

The Sway Standard Library is the foundation of portable Sway software, a set of minimal shared abstractions for the broader Sway ecosystem. It offers core types, like Result<T, E> and Option<T>, library-defined operations on language primitives, native asset management, blockchain contextual operations, access control, storage management, and support for types from other VMs, among many other things.

The entire Sway standard library is a Forc project called std, and is available directly here: https://github.com/FuelLabs/sway/tree/master/sway-lib-std (navigate to the appropriate tagged release if the latest master is not compatible). For the latest std documentation see: https://fuellabs.github.io/sway/master/std/.

Using the Standard Library

The standard library is made implicitly available to all Forc projects created using forc new. In other words, it is not required to manually specify std as an explicit dependency. Forc will automatically use the version of std that matches its version.

Importing items from the standard library can be done using the use keyword, just as importing items from any Sway project. For example:

use std::storage::StorageVec;

This imports the StorageVec type into the current namespace.

Standard Library Prelude

Sway comes with a variety of things in its standard library. However, if you had to manually import every single thing that you used, it would be very verbose. But importing a lot of things that a program never uses isn't good either. A balance needs to be struck.

The prelude is the list of things that Sway automatically imports into every Sway program. It's kept as small as possible, and is focused on things which are used in almost every single Sway program.

The current version of the prelude lives in std::prelude, and re-exports the following:

Example

Some basic example contracts to see how Sway and Forc work.

Counter

The following is a simple example of a contract which implements a counter. Both the initialize_counter() and increment_counter() ABI methods return the currently set value.

forc template --template-name counter my_counter_project
contract;

abi TestContract {
    #[storage(write)]
    fn initialize_counter(value: u64) -> u64;

    #[storage(read, write)]
    fn increment_counter(amount: u64) -> u64;
}

storage {
    counter: u64 = 0,
}

impl TestContract for Contract {
    #[storage(write)]
    fn initialize_counter(value: u64) -> u64 {
        storage.counter = value;
        value
    }

    #[storage(read, write)]
    fn increment_counter(amount: u64) -> u64 {
        let incremented = storage.counter + amount;
        storage.counter = incremented;
        incremented
    }
}

Subcurrency

The following is a simple example of a subcurrency which implements functionality to mint and send a token. It is a ledger-based token, i.e. the contract maintains a ledger of user account balances.

Being a ledger-based token, this example does not use Fuel's native asset system. It is not recommended to actually use ledger-based tokens in production; this example is here purely for illustrative purposes.

contract;

use std::{auth::msg_sender, hash::sha256};

////////////////////////////////////////
// Event declarations
////////////////////////////////////////
//
// Events allow clients to react to changes in the contract.
// Unlike Solidity, events are simply structs.
//
/// Emitted when a token is sent.
struct Sent {
    from: Address,
    to: Address,
    amount: u64,
}

////////////////////////////////////////
// ABI method declarations
////////////////////////////////////////
/// ABI for a subcurrency.
abi Token {
    // Mint new tokens and send to an address.
    // Can only be called by the contract creator.
    #[storage(read, write)]
    fn mint(receiver: Address, amount: u64);

    // Sends an amount of an existing token.
    // Can be called from any address.
    #[storage(read, write)]
    fn send(receiver: Address, amount: u64);
}

////////////////////////////////////////
// Constants
////////////////////////////////////////
/// Address of contract creator.
const MINTER = Address::from(0x9299da6c73e6dc03eeabcce242bb347de3f5f56cd1c70926d76526d7ed199b8b);

////////////////////////////////////////
// Contract storage
////////////////////////////////////////
// Contract storage persists across transactions.
storage {
    balances: StorageMap<Address, u64> = StorageMap {},
}

////////////////////////////////////////
// ABI definitions
////////////////////////////////////////
/// Contract implements the `Token` ABI.
impl Token for Contract {
    #[storage(read, write)]
    fn mint(receiver: Address, amount: u64) {
        let sender = msg_sender().unwrap();
        let sender: Address = match sender {
            Identity::Address(addr) => {
                assert(addr == MINTER);
                addr
            },
            _ => revert(0),
        };

        // Increase the balance of receiver
        storage.balances.insert(receiver, storage.balances.get(receiver).unwrap_or(0) + amount);
    }

    #[storage(read, write)]
    fn send(receiver: Address, amount: u64) {
        let sender = msg_sender().unwrap();
        let sender = match sender {
            Identity::Address(addr) => addr,
            _ => revert(0),
        };

        // Reduce the balance of sender
        let sender_amount = storage.balances.get(sender).unwrap_or(0);
        assert(sender_amount > amount);
        storage.balances.insert(sender, sender_amount - amount);

        // Increase the balance of receiver
        storage.balances.insert(receiver, storage.balances.get(receiver).unwrap_or(0) + amount);

        log(Sent {
            from: sender,
            to: receiver,
            amount: amount,
        });
    }
}

FizzBuzz

This example is not the traditional FizzBuzz; instead it is the smart contract version! A script can call the fizzbuzz ABI method of this contract with some u64 value and receive back its fizzbuzzability as an enum.

The format for custom structs and enums such as FizzBuzzResult will be automatically included in the ABI JSON so that off-chain code can handle the encoded form of the returned data.

contract;

enum FizzBuzzResult {
    Fizz: (),
    Buzz: (),
    FizzBuzz: (),
    Other: u64,
}

abi FizzBuzz {
    fn fizzbuzz(input: u64) -> FizzBuzzResult;
}

impl FizzBuzz for Contract {
    fn fizzbuzz(input: u64) -> FizzBuzzResult {
        if input % 15 == 0 {
            FizzBuzzResult::FizzBuzz
        } else if input % 3 == 0 {
            FizzBuzzResult::Fizz
        } else if input % 5 == 0 {
            FizzBuzzResult::Buzz
        } else {
            FizzBuzzResult::Other(input)
        }
    }
}

Wallet Smart Contract

ABI Declaration

library;

abi Wallet {
    #[storage(read, write), payable]
    fn receive_funds();

    #[storage(read, write)]
    fn send_funds(amount_to_send: u64, recipient_address: Address);
}

ABI Implementation

contract;

use std::{
    auth::msg_sender,
    call_frames::msg_asset_id,
    constants::BASE_ASSET_ID,
    context::msg_amount,
    token::transfer_to_address,
};

use wallet_abi::Wallet;
const OWNER_ADDRESS = Address::from(0x8900c5bec4ca97d4febf9ceb4754a60d782abbf3cd815836c1872116f203f861);

storage {
    balance: u64 = 0,
}

impl Wallet for Contract {
    #[storage(read, write), payable]
    fn receive_funds() {
        if msg_asset_id() == BASE_ASSET_ID {
            // If we received `BASE_ASSET_ID` then keep track of the balance.
            // Otherwise, we're receiving other native assets and don't care
            // about our balance of tokens.
            storage.balance += msg_amount();
        }
    }

    #[storage(read, write)]
    fn send_funds(amount_to_send: u64, recipient_address: Address) {
        let sender = msg_sender().unwrap();
        match sender {
            Identity::Address(addr) => assert(addr == OWNER_ADDRESS),
            _ => revert(0),
        };

        let current_balance = storage.balance;
        assert(current_balance >= amount_to_send);

        storage.balance = current_balance - amount_to_send;

        // Note: `transfer_to_address()` is not a call and thus not an
        // interaction. Regardless, this code conforms to
        // checks-effects-interactions to avoid re-entrancy.
        transfer_to_address(amount_to_send, BASE_ASSET_ID, recipient_address);
    }
}

Sway Program Types

A Sway program itself has a type: it is either a contract, a predicate, a script, or a library. The first three of these things are all deployable to the blockchain. A library is simply a project designed for code reuse and is never directly deployed to the chain.

Every Sway file must begin with a declaration of what type of program it is. A project can have many libraries within it, but only one contract, script, or predicate. Scripts and predicates require main functions to serve as entry points, while contracts instead publish an ABI. This chapter will go into detail about all of these various types of programs and what purposes they serve.

Contracts are used primarily for protocols or systems that operate within a fixed set of rules. A good example would be a staking contract or a decentralized exchange.

Scripts are used for complex on-chain interactions that won't persist. An example of this may be using a DEX and Lender to create a leveraged position (borrow, swap, re-collateralize, borrow) which is a complex transaction that would usually take multiple steps.

Libraries are for code that is reusable and useful for handling common situations. A good example of this would be a library to handle fixed-point math or big number math.

What is a Smart Contract?

A smart contract is no different than a script or predicate in that it is a piece of bytecode that is deployed to the blockchain via a transaction. The main features of a smart contract that differentiate it from scripts or predicates are that it is callable and stateful. Put another way, a smart contract is analogous to a deployed API with some database state. The interface of a smart contract, also just called a contract, must be defined strictly with an ABI declaration. See this contract for an example.

Syntax of a Smart Contract

As with any Sway program, the program starts with a declaration of what program type it is. A contract must also either define or import an ABI declaration and implement it. It is considered good practice to define your ABI in a separate library and import it into your contract. This allows callers of your contract to simply import the ABI directly and use it in their scripts to call your contract. Let's take a look at an ABI declaration in a library:

library;

abi Wallet {
    #[storage(read, write), payable]
    fn receive_funds();

    #[storage(read, write)]
    fn send_funds(amount_to_send: u64, recipient_address: Address);
}

Let's focus on the ABI declaration and inspect it line-by-line.

The ABI Declaration

abi Wallet {
    #[storage(read, write), payable]
    fn receive_funds();

    #[storage(read, write)]
    fn send_funds(amount_to_send: u64, recipient_address: Address);
}

In the first line, abi Wallet {, we declare the name of this Application Binary Interface, or ABI. We are naming this ABI Wallet. To import this ABI into either a script for calling or a contract for implementing, you would use

use wallet_abi::Wallet;

In the second line,

    #[storage(read, write), payable]
    fn receive_funds();

we are declaring an ABI method called receive_funds which, when called, should receive funds into this wallet. Note that we are simply defining an interface here, so there is no function body or implementation of the function. We only need to define the interface itself. In this way, ABI declarations are similar to trait declarations. This particular ABI method does not take any parameters.


In the third line,

    #[storage(read, write)]
    fn send_funds(amount_to_send: u64, recipient_address: Address);

we are declaring another ABI method, this time called send_funds. It takes two parameters: the amount to send, and the address to send the funds to.

Note: The ABI methods receive_funds and send_funds also require the annotation #[storage(read, write)] because their implementations require reading and writing a storage variable that keeps track of the wallet balance, as we will see shortly. Refer to Purity for more information on storage annotations.

Implementing an ABI for a Smart Contract

Now that we've discussed how to define the interface, let's discuss how to use it. We will start by implementing the above ABI for a specific contract.

Implementing an ABI for a contract is accomplished with impl <ABI name> for Contract syntax. The for Contract syntax can only be used to implement an ABI for a contract; implementing methods for a struct should use impl Foo syntax.

impl Wallet for Contract {
    #[storage(read, write), payable]
    fn receive_funds() {
        if msg_asset_id() == BASE_ASSET_ID {
            // If we received `BASE_ASSET_ID` then keep track of the balance.
            // Otherwise, we're receiving other native assets and don't care
            // about our balance of tokens.
            storage.balance += msg_amount();
        }
    }

    #[storage(read, write)]
    fn send_funds(amount_to_send: u64, recipient_address: Address) {
        let sender = msg_sender().unwrap();
        match sender {
            Identity::Address(addr) => assert(addr == OWNER_ADDRESS),
            _ => revert(0),
        };

        let current_balance = storage.balance;
        assert(current_balance >= amount_to_send);

        storage.balance = current_balance - amount_to_send;

        // Note: `transfer_to_address()` is not a call and thus not an
        // interaction. Regardless, this code conforms to
        // checks-effects-interactions to avoid re-entrancy.
        transfer_to_address(amount_to_send, BASE_ASSET_ID, recipient_address);
    }
}

You may notice once again the similarities between traits and ABIs. And, indeed, as a bonus, you can specify methods in addition to the interface surface of an ABI, just like a trait. By implementing the methods in the interface surface, you get the extra method implementations For Free™.

Note that the above implementation of the ABI follows the Checks, Effects, Interactions pattern.

Calling a Smart Contract from a Script

Note: In most cases, calling a contract should be done from the Rust SDK or the TypeScript SDK which provide a more ergonomic UI for interacting with a contract. However, there are situations where manually writing a script to call a contract is required.

Now that we have defined our interface and implemented it for our contract, we need to know how to actually call our contract. Let's take a look at a contract call:

script;

use std::constants::ZERO_B256;
use wallet_abi::Wallet;

fn main() {
    let contract_address = 0x9299da6c73e6dc03eeabcce242bb347de3f5f56cd1c70926d76526d7ed199b8b;
    let caller = abi(Wallet, contract_address);
    let amount_to_send = 200;
    let recipient_address = Address::from(0x9299da6c73e6dc03eeabcce242bb347de3f5f56cd1c70926d76526d7ed199b8b);
    caller.send_funds {
        gas: 10000,
        coins: 0,
        asset_id: ZERO_B256,
    }(amount_to_send, recipient_address);
}

The main new concept is the abi cast: abi(AbiName, contract_address). This returns a ContractCaller type which can be used to call contracts. The methods of the ABI become the methods available on this contract caller: send_funds and receive_funds. We then directly call the contract ABI method as if it was just a regular method. You also have the option of specifying the following special parameters inside curly braces right before the main list of parameters:

  1. gas: a u64 that represents the gas being forwarded to the contract when it is called.
  2. coins: a u64 that represents how many coins are being forwarded with this call.
  3. asset_id: a b256 that represents the ID of the asset type of the coins being forwarded.

Each special parameter is optional and assumes a default value when skipped:

  1. The default value for gas is the context gas (i.e. the content of the special register $cgas). Refer to the FuelVM specifications for more information about context gas.
  2. The default value for coins is 0.
  3. The default value for asset_id is ZERO_B256.

Libraries

Libraries in Sway are files used to define new common behavior. The most prominent example of this is the Sway Standard Library that is made implicitly available to all Forc projects created using forc new.

Writing Libraries

Libraries are defined using the library keyword at the beginning of a file, followed by a name so that they can be imported.

library;

// library code

A good reference library to use when learning library design is the Sway Standard Library. For example, the standard library offers an implementation of enum Option<T> which is a generic type that represents either the existence of a value using the variant Some(..) or a value's absence using the variant None. The Sway file implementing Option<T> has the following structure:

  • The library keyword:
library;
  • A use statement that imports revert from another library inside the standard library:
use ::revert::revert;
  • The enum definition which starts with the keyword pub to indicate that this Option<T> is publically available outside the option library:
pub enum Option<T> {
    // variants
}
  • An impl block that implements some methods for Option<T>:
impl<T> Option<T> {

    fn is_some(self) -> bool {
        // body of is_some
    }

    // other methods
}

Now that the library option is fully written, and because Option<T> is defined with the pub keyword, we are now able to import Option<T> using use std::option::Option; from any Sway project and have access to all of its variants and methods. That being said, Option is automatically available in the standard library prelude so you never actually have to import it manually.

Libraries are composed of just a Forc.toml file and a src directory, unlike contracts which usually contain a tests directory and a Cargo.toml file as well. An example of a library's Forc.toml:

[project]
authors = ["Fuel Labs <contact@fuel.sh>"]
entry = "lib.sw"
license = "Apache-2.0"
name = "my_library"

[dependencies]

which denotes the authors, an entry file, the name by which it can be imported, and any dependencies.

For large libraries, it is recommended to have a lib.sw entry point re-export all other sub-libraries.

The mod keyword registers a submodule, making its items (such as functions and structs) accessible from the parent library. If used at the top level it will refer to a file in the src folder and in other cases in a folder named after the library in which it is defined.

For example, the lib.sw of the standard library looks like:

library;

mod block;
mod storage;
mod constants;
mod vm;
// .. Other deps

with other libraries contained in the src folder, like the vm library (inside of src/vm.sw):

library;

mod evm;
// ...

and it's own sub-library evm located in src/vm/evm.sw:

library;

// ...

Using Libraries

There are two types of Sway libraries, based on their location and how they can be imported.

Internal Libraries

Internal libraries are located within the project's src directory alongside main.sw or in the appropriate folders as shown below:

$ tree
.
├── Cargo.toml
├── Forc.toml
└── src
    ├── internal_lib.sw
    ├── main.sw
    └── internal_lib
        └── nested_lib.sw

As internal_lib is an internal library, it can be imported into main.sw as follows:

  • Use the mod keyword followed by the library name to make the internal library a dependancy
  • Use the use keyword with a :: separating the name of the library and the imported item(s)
mod internal_lib; // Assuming the library name in `internal_lib.sw` is `internal_lib`

use internal_lib::mint;

// `mint` from `internal_library` is now available in this file

External Libraries

External libraries are located outside the main src directory as shown below:

$ tree
.
├── my_project
│   ├── Cargo.toml
│   ├── Forc.toml
│   └─── src
│       └── main.sw
│
└── external_lib
    ├── Cargo.toml
    ├── Forc.toml
    └─── src
        └── lib.sw

As external_lib is outside the src directory of my_project, it needs to be added as a dependency in the Forc.toml file of my_project, by adding the library path in the dependencies section as shown below, before it can be imported:

[dependencies]
external_library = { path = "../external_library" }

Once the library dependency is added to the toml file, you can import items from it as follows:

  • Make sure the item you want imported are declared with the pub keyword (if applicable, for instance: pub fn mint() {})
  • Use the use keyword to selectively import items from the library
use external_library::mint;

// `mint` from `external_library` is now available in this file

Wildcard imports using * are supported, but it is generally recommended to use explicit imports where possible.

Note: the standard library is implicitly available to all Forc projects, that is, you are not required to manually specify std as an explicit dependency in Forc.toml.

Reference Sway Libraries

The repository sway-libs is a collection of external libraries that you can import and make use of in your Fuel applications. These libraries are meant to be learning references of common use-cases valuable for dapp development.

Some Sway Libraries to try out:

Example

You can import and use a Sway Library such as the NFT library just like any other external library.

use sway_libs::nft::{
    mint,
    transfer,
    owner_of,
    approve,
};

Once imported, you can use the following basic functionality of the library in your smart contract:

  • Minting tokens
  • Transfering tokens
  • Retrieving owner of a token
  • Approving users to transfer a token

Scripts

A script is runnable bytecode on the chain which executes once to perform some task. It does not represent ownership of any resources and it cannot be called by a contract. A script can return a single value of any type.

Scripts are state-aware in that while they have no persistent storage (because they only exist during the transaction) they can call contracts and act based upon the returned values and results.

This example script calls a contract:

script;

use std::constants::ZERO_B256;
use wallet_abi::Wallet;

fn main() {
    let contract_address = 0x9299da6c73e6dc03eeabcce242bb347de3f5f56cd1c70926d76526d7ed199b8b;
    let caller = abi(Wallet, contract_address);
    let amount_to_send = 200;
    let recipient_address = Address::from(0x9299da6c73e6dc03eeabcce242bb347de3f5f56cd1c70926d76526d7ed199b8b);
    caller.send_funds {
        gas: 10000,
        coins: 0,
        asset_id: ZERO_B256,
    }(amount_to_send, recipient_address);
}

Scripts, similar to predicates, rely on a main() function as an entry point. You can call other functions defined in a script from the main() function or call another contract via an abi cast.

An example use case for a script would be a router that trades funds through multiple DEXes to get the price for the input asset, or a script to re-adjust a Collateralized Debt Position via a flashloan.

Scripts and the SDKs

Unlike EVM transactions which can call a contract directly (but can only call a single contract), Fuel transactions execute a script, which may call zero or more contracts. The Rust and TypeScript SDKs provide functions to call contract methods as if they were calling contracts directly. Under the hood, the SDKs wrap all contract calls with scripts that contain minimal code to simply make the call and forward script data as call parameters.

Predicates

From the perspective of Sway, predicates are programs that return a Boolean value and which represent ownership of some resource upon execution to true. They have no access to contract storage. Here is a trivial predicate, which always evaluates to true:

predicate;

// All predicates require a main function which returns a Boolean value.
fn main() -> bool {
    true
}

Debugging Predicates

Because they don't have any side effects (they are pure), predicates cannot create receipts. Therefore, they cannot have logging or create a stack backtrace. This means that there is no native way to debug them aside from using a single-stepping debugger (which is a work-in-progress).

As a workaround, the predicate can be written, tested, and debugged first as a script, and then changed back into a predicate.

Sway Language basics

Sway is a programming language designed for the FuelVM. It is a statically typed, compiled language with type inference and traits. Sway aims to make smart contract development safer and more performant through the use of strong static analysis and compiler feedback.

Sway basics.

Variables

Variables in Sway are immutable by default. This means that, by default, once a variable is declared, its value cannot change. This is one of the ways how Sway encourages safe programming, and many modern languages have this same default. Let's take a look at variables in detail.

Declaring a Variable

Let's look at a variable declaration:

let foo = 5;

Great! We have just declared a variable, foo. What do we know about foo?

  1. It is immutable.
  2. Its value is 5.
  3. Its type is u64, a 64-bit unsigned integer.

u64 is the default numeric type, and represents a 64-bit unsigned integer. See the section Built-in Types for more details.

We can also make a mutable variable. Let's take a look:

let mut foo = 5;
foo = 6;

Now, foo is mutable, and the reassignment to the number 6 is valid. That is, we are allowed to mutate the variable foo to change its value.

Type Annotations

A variable declaration can contain a type annotation. A type annotation serves the purpose of declaring the type, in addition to the value, of a variable. Let's take a look:

let foo: u32 = 5;

We have just declared the type of the variable foo as a u32, which is an unsigned 32-bit integer. Let's take a look at a few other type annotations:

let bar: str[4] = "sway";
let baz: bool = true;

If the value declared cannot be assigned to the declared type, there will be an error generated by the compiler.

Configurable Constants

Configurable constants are special constants that behave like regular constants in the sense that they cannot change during program execution, but they can be configured after the Sway program has been built. The Rust and TS SDKs allow updating the values of these constants by injecting new values for them directly in the bytecode without having to build the program again. These are useful for contract factories and behave somewhat similarly to immutable variables from languages like Solidity.

Configurable constants are declared inside a configurable block and require a type ascription and an initializer as follows:

configurable {
    U8: u8 = 8u8,
    BOOL: bool = true,
    ARRAY: [u32; 3] = [253u32, 254u32, 255u32],
    STR_4: str[4] = "fuel",
    STRUCT: StructWithGeneric<u8> = StructWithGeneric {
        field_1: 8u8,
        field_2: 16,
    },
    ENUM: EnumWithGeneric<bool> = EnumWithGeneric::VariantOne(true),
}

At most one configurable block is allowed in a Sway project. Moreover, configurable blocks are not allowed in libraries.

Configurable constants can be read directly just like regular constants:

    fn return_configurables() -> (u8, bool, [u32; 3], str[4], StructWithGeneric<u8>) {
        (U8, BOOL, ARRAY, STR_4, STRUCT)
    }

Built-in Types

Every value in Sway is of a certain type. Although deep down, all values are just ones and zeroes in the underlying virtual machine, Sway needs to know what those ones and zeroes actually mean. This is accomplished with types.

Sway is a statically typed language. At compile time, the types of every value must be known. This does not mean you need to specify every single type: usually, the type can be reasonably inferred by the compiler.

Primitive Types

Sway has the following primitive types:

  1. u8 (8-bit unsigned integer)
  2. u16 (16-bit unsigned integer)
  3. u32 (32-bit unsigned integer)
  4. u64 (64-bit unsigned integer)
  5. str[] (fixed-length string)
  6. bool (Boolean true or false)
  7. b256 (256 bits (32 bytes), i.e. a hash)

All other types in Sway are built up of these primitive types, or references to these primitive types. You may notice that there are no signed integers—this is by design. In the blockchain domain that Sway occupies, floating-point values and negative numbers have smaller utility, so their implementation has been left up to libraries for specific use cases.

Numeric Types

All of the unsigned integer types are numeric types.

Numbers can be declared with binary syntax, hexadecimal syntax, base-10 syntax, and underscores for delineation. Let's take a look at the following valid numeric primitives:

0xffffff    // hexadecimal
0b10101010  // binary
10          // base-10
100_000     // underscore delineated base-10
0x1111_0000 // underscore delineated binary
0xfff_aaa   // underscore delineated hexadecimal

The default numeric type is u64. The FuelVM's word size is 64 bits, and the cases where using a smaller numeric type saves space are minimal.

Boolean Type

The boolean type (bool) has two potential values: true or false. Boolean values are typically used for conditional logic or validation, for example in if expressions. Booleans can be negated, or flipped, with the unary negation operator !. For example:

fn returns_false() -> bool {
    let boolean_value: bool = true;
    !boolean_value
}

String Type

In Sway, static-length strings are a primitive type. This means that when you declare a string, its size is a part of its type. This is necessary for the compiler to know how much memory to give for the storage of that data. The size of the string is denoted with square brackets. Let's take a look:

let my_string: str[4] = "fuel";

Because the string literal "fuel" is four letters, the type is str[4], denoting a static length of 4 characters. Strings default to UTF-8 in Sway.

Compound Types

Compound types are types that group multiple values into one type. In Sway, we have arrays and tuples.

Tuple Types

A tuple is a general-purpose static-length aggregation of types. In more plain terms, a tuple is a single type that consists of an aggregate of zero or more types. The internal types that make up a tuple, and the tuple's arity, define the tuple's type. Let's take a look at some examples.

let x: (u64, u64) = (0, 0);

This is a tuple, denoted by parenthesized, comma-separated values. Note that the type annotation, (u64, u64), is similar in syntax to the expression which instantiates that type, (0, 0).

let x: (u64, bool) = (42, true);
assert(x.1);

In this example, we have created a new tuple type, (u64, bool), which is a composite of a u64 and a bool. To access a value within a tuple, we use tuple indexing: x.1 stands for the first (zero-indexed, so the bool) value of the tuple. Likewise, x.0 would be the zeroth, u64 value of the tuple. Tuple values can also be accessed via destructuring:

struct Foo {}
let x: (u64, Foo, bool) = (42, Foo {}, true);
let (number, foo, boolean) = x;

To create one-arity tuples, we will need to add a trailing comma:

let x: u64 = (42);     // x is of type u64
let y: (u64) = (42);   // y is of type u64
let z: (u64,) = (42,); // z is of type (u64), i.e. a one-arity tuple
let w: (u64) = (42,);  // type error

Arrays

An array is similar to a tuple, but an array's values must all be of the same type. Arrays can hold arbitrary types including non-primitive types.

An array is written as a comma-separated list inside square brackets:

let x = [1, 2, 3, 4, 5];

Arrays are allocated on the stack since their size is known. An array's size is always static, i.e. it cannot change. An array of five elements cannot become an array of six elements.

Arrays can be iterated over, unlike tuples. An array's type is written as the type the array contains followed by the number of elements, semicolon-separated and within square brackets, e.g. [u64; 5]. To access an element in an array, use the array indexing syntax, i.e. square brackets.

Array elements can also be mutated if the underlying array is declared as mutable:

let mut x = [1, 2, 3, 4, 5];
x[0] = 0;
script;

struct Foo {
    f1: u32,
    f2: b256,
}

fn main() {
    // Array of integers with type ascription
    let array_of_integers: [u8; 5] = [1, 2, 3, 4, 5];

    // Array of strings
    let array_of_strings = ["Bob", "Jan", "Ron"];

    // Array of structs
    let array_of_structs: [Foo; 2] = [
        Foo {
            f1: 11,
            f2: 0x1111111111111111111111111111111111111111111111111111111111111111,
        },
        Foo {
            f1: 22,
            f2: 0x2222222222222222222222222222222222222222222222222222222222222222,
        },
    ];

    // Accessing an element of an array
    let mut array_of_bools: [bool; 2] = [true, false];
    assert(array_of_bools[0]);

    // Mutating the element of an array
    array_of_bools[1] = true;
    assert(array_of_bools[1]);
}

Commonly Used Library Types

The Sway Standard Library is the foundation of portable Sway software, a set of minimal shared abstractions for the broader Sway ecosystem. It offers core types, library-defined operations on language primitives, native asset management, blockchain contextual operations, access control, storage management, and support for types from other VMs, among many other things. Reference the standard library docs here.

Result<T, E>

Type Result is the type used for returning and propagating errors. It is an enum with two variants: Ok(T), representing success and containing a value, and Err(E), representing error and containing an error value. The T and E in this definition are type parameters, allowing Result to be generic and to be used with any types.

/// `Result` is a type that represents either success (`Ok`) or failure (`Err`).
pub enum Result<T, E> {
    /// Contains the success value.
    Ok: T,
    /// Contains the error value.
    Err: E,
}

Functions return Result whenever errors are expected and recoverable. Take the following example:

script;

enum MyContractError {
    DivisionByZero: (),
}

fn divide(numerator: u64, denominator: u64) -> Result<u64, MyContractError> {
    if (denominator == 0) {
        return Result::Err(MyContractError::DivisionByZero);
    } else {
        Result::Ok(numerator / denominator)
    }
}

fn main() -> Result<u64, str[4]> {
    let result = divide(20, 2);
    match result {
        Result::Ok(value) => Result::Ok(value),
        Result::Err(MyContractError::DivisionByZero) => Result::Err("Fail"),
    }
}

Option<T>

Type Option represents an optional value: every Option is either Some and contains a value, or None, and does not. Option types are very common in Sway code, as they have a number of uses:

  • Initial values where None can be used as an initializer.
  • Return value for otherwise reporting simple errors, where None is returned on error.

The implementation of Option matches on the variant: if it's Ok it returns the inner value, if it's None, it reverts.

/// `Option` is a type that represents an optional value.
pub enum Option<T> {
    /// No value.
    None: (),
    /// Some value of type `T`.
    Some: T,
}

Option is commonly paired with pattern matching to query the presence of a value and take action, allowing developers to choose how to handle the None case.

Below is an example that uses pattern matching to handle invalid divisions by 0 by returning an Option:

script;

fn divide(numerator: u64, denominator: u64) -> Option<u64> {
    if denominator == 0 {
        Option::None
    } else {
        Option::Some(numerator / denominator)
    }
}

fn main() {
    let result = divide(6, 2);
    // Pattern match to retrieve the value
    match result {
        // The division was valid
        Option::Some(x) => std::logging::log(x),
        // The division was invalid
        Option::None => std::logging::log("Cannot divide by 0"),
    }
}

Blockchain Types

Sway is fundamentally a blockchain language, and it offers a selection of types tailored for the blockchain use case.

These are provided via the standard library (lib-std) which both add a degree of type-safety, as well as make the intention of the developer more clear.

Address Type

The Address type is a type-safe wrapper around the primitive b256 type. Unlike the EVM, an address never refers to a deployed smart contract (see the ContractId type below). An Address can be either the hash of a public key (effectively an externally owned account if you're coming from the EVM) or the hash of a predicate. Addresses own UTXOs.

An Address is implemented as follows.

pub struct Address {
    value: b256,
}

Casting between the b256 and Address types must be done explicitly:

let my_number: b256 = 0x000000000000000000000000000000000000000000000000000000000000002A;
let my_address: Address = Address::from(my_number);
let forty_two: b256 = my_address.into();

ContractId Type

The ContractId type is a type-safe wrapper around the primitive b256 type. A contract's ID is a unique, deterministic identifier analogous to a contract's address in the EVM. Contracts cannot own UTXOs but can own assets.

A ContractId is implemented as follows.

pub struct ContractId {
    value: b256,
}

Casting between the b256 and ContractId types must be done explicitly:

let my_number: b256 = 0x000000000000000000000000000000000000000000000000000000000000002A;
let my_contract_id: ContractId = ContractId::from(my_number);
let forty_two: b256 = my_contract_id.into();

Identity Type

The Identity type is an enum that allows for the handling of both Address and ContractId types. This is useful in cases where either type is accepted, e.g. receiving funds from an identified sender, but not caring if the sender is an address or a contract.

An Identity is implemented as follows.

pub enum Identity {
    Address: Address,
    ContractId: ContractId,
}

Casting to an Identity must be done explicitly:

        let raw_address: b256 = 0xddec0e7e6a9a4a4e3e57d08d080d71a299c628a46bc609aab4627695679421ca;
        let my_identity: Identity = Identity::Address(Address::from(raw_address));

A match statement can be used to return to an Address or ContractId as well as handle cases in which their execution differs.

        let my_contract_id: ContractId = match my_identity {
            Identity::ContractId(identity) => identity,
            _ => revert(0),
        };
        match my_identity {
            Identity::Address(address) => transfer_to_address(amount, token_id, address),
            Identity::ContractId(contract_id) => force_transfer_to_contract(amount, token_id, contract_id),
        };

A common use case for Identity is for access control. The use of Identity uniquely allows both ContractId and Address to have access control inclusively.

        let sender = msg_sender().unwrap();
        require(sender == storage.owner, MyError::UnauthorizedUser(sender));

Functions

Functions in Sway are declared with the fn keyword. Let's take a look:

fn equals(first_param: u64, second_param: u64) -> bool {
    first_param == second_param
}

We have just declared a function named equals which takes two parameters: first_param and second_param. The parameters must both be 64-bit unsigned integers.

This function also returns a bool value, i.e. either true or false. This function returns true if the two given parameters are equal, and false if they are not. If we want to use this function, we can do so like this:

fn main() {
    equals(5, 5); // evaluates to `true`
    equals(5, 6); // evaluates to `false`
}

Mutable Parameters

We can make a function parameter mutable by adding ref mut before the parameter name. This allows mutating the argument passed into the function when the function is called. For example:

fn increment(ref mut num: u32) {
    let prev = num;
    num = prev + 1u32;
}

This function is allowed to mutate its parameter num because of the mut keyword. In addition, the ref keyword instructs the function to modify the argument passed to it when the function is called, instead of modifying a local copy of it.

    let mut num: u32 = 0;
    increment(num);
    assert(num == 1u32); // The function `increment()` modifies `num`

Note that the variable num itself has to be declared as mutable for the above to compile.

Note It is not currently allowed to use mut without ref or vice versa for a function parameter.

Similarly, ref mut can be used with more complex data types such as:

fn swap_tuple(ref mut pair: (u64, u64)) {
    let temp = pair.0;
    pair.0 = pair.1;
    pair.1 = temp;
}

fn update_color(ref mut color: Color, new_color: Color) {
    color = new_color;
}

We can then call these functions as shown below:

    let mut tuple = (42, 24);
    swap_tuple(tuple);
    assert(tuple.0 == 24); // The function `swap_tuple()` modifies `tuple.0`
    assert(tuple.1 == 42); // The function `swap_tuple()` modifies `tuple.1`
    let mut color = Color::Red;
    update_color(color, Color::Blue);
    assert(match color {
        Color::Blue => true,
        _ => false,
    }); // The function `update_color()` modifies the color to Blue

Note The only place, in a Sway program, where the ref keyword is valid is before a mutable function parameter.

Structs, Tuples, and Enums

Structs

Structs in Sway are a named grouping of types. You may also be familiar with structs via another name: product types. Sway does not make any significantly unique usages of structs; they are similar to most other languages which have structs. If you're coming from an object-oriented background, a struct is like the data attributes of an object.

Firstly, we declare a struct named Foo with two fields. The first field is named bar and it accepts values of type u64, the second field is named baz and it accepts bool values.

library;

// Declare a struct type
pub struct Foo {
    bar: u64,
    baz: bool,
}

// Struct types for destructuring
pub struct Point {
    x: u64,
    y: u64,
}

pub struct Line {
    p1: Point,
    p2: Point,
}

pub struct TupleInStruct {
    nested_tuple: (u64, (u32, (bool, str[2]))),
}

In order to instantiate the struct we use struct instantiation syntax, which is very similar to the declaration syntax except with expressions in place of types.

There are three ways to instantiate the struct.

  • Hardcoding values for the fields
  • Passing in variables with names different than the struct fields
  • Using a shorthand notation via variables that are the same as the field names
library;

mod data_structures;
use data_structures::{Foo, Line, Point, TupleInStruct};

fn hardcoded_instantiation() -> Foo {
    // Instantiate `foo` as `Foo`
    let mut foo = Foo {
        bar: 42,
        baz: false,
    };

    // Access and write to "baz"
    foo.baz = true;

    // Return the struct
    foo
}

fn variable_instantiation() -> Foo {
    // Declare variables with the same names as the fields in `Foo`
    let number = 42;
    let truthness = false;

    // Instantiate `foo` as `Foo`
    let mut foo = Foo {
        bar: number,
        baz: truthness,
    };

    // Access and write to "baz"
    foo.baz = true;

    // Return the struct
    foo
}

fn shorthand_instantiation() -> Foo {
    // Declare variables with the same names as the fields in `Foo`
    let bar = 42;
    let baz = false;

    // Instantiate `foo` as `Foo`
    let mut foo = Foo { bar, baz };

    // Access and write to "baz"
    foo.baz = true;

    // Return the struct
    foo
}

fn struct_destructuring() {
    let point1 = Point { x: 0, y: 0 };
    // Destructure the values from the struct into variables
    let Point { x, y } = point1;

    let point2 = Point { x: 1, y: 1 };
    // If you do not care about specific struct fields then use ".." at the end of your variable list
    let Point { x, .. } = point2;

    let line = Line {
        p1: point1,
        p2: point2,
    };
    // Destructure the values from the nested structs into variables
    let Line {
        p1: Point { x: x0, y: y0 },
        p2: Point { x: x1, y: y1 },
    } = line;
    // You may also destructure tuples nested in structs and structs nested in tuples
    let tuple_in_struct = TupleInStruct {
        nested_tuple: (42u64, (42u32, (true, "ok"))),
    };
    let TupleInStruct {
        nested_tuple: (a, (b, (c, d))),
    } = tuple_in_struct;

    let struct_in_tuple = (Point { x: 2, y: 4 }, Point { x: 3, y: 6 });
    let (Point { x: x0, y: y0 }, Point { x: x1, y: y1 }) = struct_in_tuple;
}

Note You can mix and match all 3 ways to instantiate the struct at the same time. Moreover, the order of the fields does not matter when instantiating however we encourage declaring the fields in alphabetical order and instantiating them in the same alphabetical order

Furthermore, multiple variables can be extracted from a struct using the destructuring syntax.

Struct Memory Layout

Note This information is not vital if you are new to the language, or programming in general

Structs have zero memory overhead. What that means is that in memory, each struct field is laid out sequentially. No metadata regarding the struct's name or other properties is preserved at runtime. In other words, structs are compile-time constructs. This is the same in Rust, but different in other languages with runtimes like Java.

Tuples

Tuples are a basic static-length type which contain multiple different types within themselves. The type of a tuple is defined by the types of the values within it, and a tuple can contain basic types as well as structs and enums.

You can access values directly by using the . syntax. Moreover, multiple variables can be extracted from a tuple using the destructuring syntax.

library;

fn tuple() {
    // You can declare the types youself
    let tuple1: (u8, bool, u64) = (100, false, 10000);

    // Or have the types be inferred
    let mut tuple2 = (5, true, ("Sway", 8));

    // Retrieve values from tuples
    let number = tuple1.0;
    let sway = tuple2.2.1;

    // Destructure the values from the tuple into variables
    let (n1, truthness, n2) = tuple1;

    // If you do not care about specific values then use "_"
    let (_, truthness, _) = tuple2;

    // Internally mutate the tuple
    tuple2.1 = false;

    // Or change the values all at once (must keep the same data types)
    tuple2 = (9, false, ("Fuel", 99));
}

Enums

Enumerations, or enums, are also known as sum types. An enum is a type that could be one of several variants. To declare an enum, you enumerate all potential variants.

Here, we have defined five potential colors. Each enum variant is just the color name. As there is no extra data associated with each variant, we say that each variant is of type (), or unit.

library;

// Declare the enum
enum Color {
    Blue: (),
    Green: (),
    Red: (),
    Silver: (),
    Grey: (),
}

fn main() {
    // To instantiate a variable with the value of an enum the syntax is
    let blue = Color::Blue;
    let silver = Color::Silver;
}

Enums of Structs

It is also possible to have an enum variant contain extra data. Take a look at this more substantial example, which combines struct declarations with enum variants:

library;

struct Item {
    price: u64,
    amount: u64,
    id: u64,
}

enum MyEnum {
    Item: Item,
}

fn main() {
    let my_enum = MyEnum::Item(Item {
        price: 5,
        amount: 2,
        id: 42,
    });
}

Enums of Enums

It is possible to define enums of enums:

library;

pub enum Error {
    StateError: StateError,
    UserError: UserError,
}

pub enum StateError {
    Void: (),
    Pending: (),
    Completed: (),
}

pub enum UserError {
    InsufficientPermissions: (),
    Unauthorized: (),
}

Preferred usage

The preferred way to use enums is to use the individual (not nested) enums directly because they are easy to follow and the lines are short:

library;

use ::enum_of_enums::{StateError, UserError};

fn preferred() {
    let error1 = StateError::Void;
    let error2 = UserError::Unauthorized;
}

Inadvisable

If you wish to use the nested form of enums via the Error enum from the example above, then you can instantiate them into variables using the following syntax:

library;

use ::enum_of_enums::{Error, StateError, UserError};

fn avoid() {
    let error1 = Error::StateError(StateError::Void);
    let error2 = Error::UserError(UserError::Unauthorized);
}

Key points to note:

  • You must import all of the enums you need instead of just the Error enum
  • The lines may get unnecessarily long (depending on the names)
  • The syntax is not the most ergonomic

Enum Memory Layout

Note This information is not vital if you are new to the language, or programming in general.

Enums do have some memory overhead. To know which variant is being represented, Sway stores a one-word (8-byte) tag for the enum variant. The space reserved after the tag is equivalent to the size of the largest enum variant. So, to calculate the size of an enum in memory, add 8 bytes to the size of the largest variant. For example, in the case of Color above, where the variants are all (), the size would be 8 bytes since the size of the largest variant is 0 bytes.

Methods and Associated Functions

Methods are similar to functions in that we declare them with the fn keyword and they have parameters and return a value. However, unlike functions, Methods are defined within the context of a struct (or enum), and either refers to that type or mutates it. The first parameter of a method is always self, which represents the instance of the struct the method is being called on.

Associated functions are very similar to methods, in that they are also defined in the context of a struct or enum, but they do not actually use any of the data in the struct and as a result do not take self as a parameter. Associated functions could be standalone functions, but they are included in a specific type for organizational or semantic reasons.

To declare methods and associated functions for a struct or enum, use an impl block. Here, impl stands for implementation.

script;

struct Foo {
    bar: u64,
    baz: bool,
}

impl Foo {
    // this is a _method_, as it takes `self` as a parameter.
    fn is_baz_true(self) -> bool {
        self.baz
    }

    // this is an _associated function_, since it does not take `self` as a parameter.
    fn new_foo(number: u64, boolean: bool) -> Foo {
        Foo {
            bar: number,
            baz: boolean,
        }
    }
}

fn main() {
    let foo = Foo::new_foo(42, true);
    assert(foo.is_baz_true());
}

To call a method, simply use dot syntax: foo.iz_baz_true().

Similarly to free functions, methods and associated functions may accept ref mut parameters. For example:

struct Coordinates {
    x: u64,
    y: u64,
}

impl Coordinates {
    fn move_right(ref mut self, distance: u64) {
        self.x += distance;
    }
}

and when called:

    let mut point = Coordinates { x: 1, y: 1 };
    point.move_right(5);
    assert(point.x == 6);
    assert(point.y == 1);

Comments and Logging

Comments

Comments in Sway start with two slashes and continue until the end of the line. For comments that extend beyond a single line, you'll need to include // on each line.

// hello world
// let's make a couple of lines
// commented.

You can also place comments at the ends of lines containing code.

fn main() {
    let baz = 8; // Eight is a lucky number
}

You can also do block comments

fn main() {
    /*
    You can write on multiple lines
    like this if you want
    */
    let baz = 8;
}

Logging

The logging library provides a generic log function that can be imported using use std::logging::log and used to log variables of any type. Each call to log appends a receipt to the list of receipts. There are two types of receipts that a log can generate: Log and LogData.

fn log_values(){
  // Generates a Log receipt
  log(42);

  // Generates a LogData receipt
  let string = "sway";
  log(string);
}

Log Receipt

The Log receipt is generated for non-reference types, namely bool, u8, u16, u32, and u64. For example, logging an integer variable x that holds the value 42 using log(x) may generate the following receipt:

"Log": {
  "id": "0000000000000000000000000000000000000000000000000000000000000000",
  "is": 10352,
  "pc": 10404,
  "ra": 42,
  "rb": 1018205,
  "rc": 0,
  "rd": 0
}

Note that ra will include the value being logged. The additional registers rc and rd will be zero when using log while rb may include a non-zero value representing a unique ID for the log instance. The unique ID is not meaningful on its own but allows the Rust and the TS SDKs to know the type of the data being logged, by looking up the log ID in the JSON ABI file.

LogData Receipt

LogData is generated for reference types which include all types except for the non_reference types mentioned above. For example, logging a b256 variable b that holds the value 0x1111111111111111111111111111111111111111111111111111111111111111 using log(b) may generate the following receipt:

"LogData": {
  "data": "1111111111111111111111111111111111111111111111111111111111111111",
  "digest": "02d449a31fbb267c8f352e9968a79e3e5fc95c1bbeaa502fd6454ebde5a4bedc",
  "id": "0000000000000000000000000000000000000000000000000000000000000000",
  "is": 10352,
  "len": 32,
  "pc": 10444,
  "ptr": 10468,
  "ra": 0,
  "rb": 1018194
}

Note that data in the receipt above will include the value being logged as a hexadecimal. Similarly to the Log receipt, additional registers are written: ra will always be zero when using log, while rb will contain a unique ID for the log instance.

Note The Rust SDK exposes APIs that allow you to retrieve the logged values and display them nicely based on their types as indicated in the JSON ABI file.

Control Flow

if expressions

Sway supports if, else, and else if expressions that allow you to branch your code depending on conditions.

For example:

fn main() {
    let number = 6;

    if number % 4 == 0 {
        // do something
    } else if number % 3 == 0 {
        // do something else
    } else {
        // do something else
    }
}

Using if in a let statement

Like Rust, ifs are expressions in Sway. What this means is you can use if expressions on the right side of a let statement to assign the outcome to a variable.

let my_data = if some_bool < 10 { foo() } else { bar() };

Note that all branches of the if expression must return a value of the same type.

match expressions

Sway supports advanced pattern matching through exhaustive match expressions. Unlike an if statement, a match expression asserts at compile time that all possible patterns have been matched. If you don't handle all the patterns, you will get compiler error indicating that your match expression is non-exhaustive.

The basic syntax of a match statement is as follows:

let result = match expression {
    pattern1 => code_to_execute_if_expression_matches_pattern1,
    pattern2 => code_to_execute_if_expression_matches_pattern2,
    ...
    _ => code_to_execute_if_expression_matches_no_pattern,
}

Some examples of how you can use a match statement:

script;

// helper functions for our example
fn on_even(num: u64) {
    // do something with even numbers
}
fn on_odd(num: u64) {
    // do something with odd numbers
}

fn main(num: u64) -> u64 {
    // Match as an expression
    let isEven = match num % 2 {
        0 => true,
        _ => false,
    };

    // Match as control flow
    let x = 12;
    match x {
        5 => on_odd(x),
        _ => on_even(x),
    };

    // Match an enum
    enum Weather {
        Sunny: (),
        Rainy: (),
        Cloudy: (),
        Snowy: (),
    }
    let currentWeather = Weather::Sunny;
    let avgTemp = match currentWeather {
        Weather::Sunny => 80,
        Weather::Rainy => 50,
        Weather::Cloudy => 60,
        Weather::Snowy => 20,
    };

    // match expression used for a return
    let outsideTemp = Weather::Sunny;
    match outsideTemp {
        Weather::Sunny => 80,
        Weather::Rainy => 50,
        Weather::Cloudy => 60,
        Weather::Snowy => 20,
    }
}

Loops

while

Loops in Sway are currently limited to while loops. This is what they look like:

while counter < 10 {
    counter = counter + 1;
}

You need the while keyword, some condition (value < 10 in this case) which will be evaluated each iteration, and a block of code inside the curly braces ({...}) to execute each iteration.

break and continue

break and continue keywords are available to use inside the body of a while loop. The purpose of the break statement is to break out of a loop early:

fn break_example() -> u64 {
    let mut counter = 1;
    let mut sum = 0;
    let num = 10;
    while true {
        if counter > num {
            break;
        }
        sum += counter;
        counter += 1;
    }
    sum // 1 + 2 + .. + 10 = 55
}

The purpose of the continue statement is to skip a portion of a loop in an iteration and jump directly into the next iteration:

fn continue_example() -> u64 {
    let mut counter = 0;
    let mut sum = 0;
    let num = 10;
    while counter < num {
        counter += 1;
        if counter % 2 == 0 {
            continue;
        }
        sum += counter;
    }
    sum // 1 + 3 + .. + 9 = 25
}

Nested loops

You can also use nested while loops if needed:

while condition_1 == true {
    // do stuff...
    while condition_2 == true {
        // do more stuff...
    }
}

Blockchain Development with Sway

Sway is fundamentally a blockchain language. Because of this, it has some features and requirements that you may not have seen in general-purpose programming languages.

These are also some concepts related to the FuelVM and Fuel ecosystem that you may utilize when writing Sway.

Hashing and Cryptography

The Sway standard library provides easy access to a selection of cryptographic hash functions (sha256 and EVM-compatible keccak256), and EVM-compatible secp256k1-based signature recovery operations.

Hashing

script;

use std::hash::{keccak256, sha256};

const VALUE_A = 0x9280359a3b96819889d30614068715d634ad0cf9bba70c0f430a8c201138f79f;

enum Location {
    Earth: (),
    Mars: (),
}

struct Person {
    name: str[4],
    age: u64,
    alive: bool,
    location: Location,
    stats: Stats,
    some_tuple: (bool, u64),
    some_array: [u64; 2],
    some_b256: b256,
}

struct Stats {
    strength: u64,
    agility: u64,
}

fn main() {
    let zero = b256::min();
    // Use the generic sha256 to hash some integers
    let sha_hashed_u8 = sha256(u8::max());
    let sha_hashed_u16 = sha256(u16::max());
    let sha_hashed_u32 = sha256(u32::max());
    let sha_hashed_u64 = sha256(u64::max());

    // Or hash a b256
    let sha_hashed_b256 = sha256(VALUE_A);

    // You can hash booleans too
    let sha_hashed_bool = sha256(true);

    // Strings are not a problem either
    let sha_hashed_str = sha256("Fastest Modular Execution Layer!");

    // Tuples of any size work too
    let sha_hashed_tuple = sha256((true, 7));

    // As do arrays
    let sha_hashed_array = sha256([4, 5, 6]);

    // Enums work too
    let sha_hashed_enum = sha256(Location::Earth);

    // Complex structs are not a problem
    let sha_hashed_struct = sha256(Person {
        name: "John",
        age: 9000,
        alive: true,
        location: Location::Mars,
        stats: Stats {
            strength: 10,
            agility: 9,
        },
        some_tuple: (true, 8),
        some_array: [17, 76],
        some_b256: zero,
    });

    log(sha_hashed_u8);
    log(sha_hashed_u16);
    log(sha_hashed_u32);
    log(sha_hashed_u64);
    log(sha_hashed_b256);
    log(sha_hashed_bool);
    log(sha_hashed_str);
    log(sha_hashed_tuple);
    log(sha_hashed_array);
    log(sha_hashed_enum);
    log(sha_hashed_struct);

    // Use the generic keccak256 to hash some integers
    let keccak_hashed_u8 = keccak256(u8::max());
    let keccak_hashed_u16 = keccak256(u16::max());
    let keccak_hashed_u32 = keccak256(u32::max());
    let keccak_hashed_u64 = keccak256(u64::max());

    // Or hash a b256
    let keccak_hashed_b256 = keccak256(VALUE_A);

    // You can hash booleans too
    let keccak_hashed_bool = keccak256(true);

    // Strings are not a problem either
    let keccak_hashed_str = keccak256("Fastest Modular Execution Layer!");

    // Tuples of any size work too
    let keccak_hashed_tuple = keccak256((true, 7));

    // As do arrays
    let keccak_hashed_array = keccak256([4, 5, 6]);

    // Enums work too
    let keccak_hashed_enum = keccak256(Location::Earth);

    // Complex structs are not a problem
    let keccak_hashed_struct = keccak256(Person {
        name: "John",
        age: 9000,
        alive: true,
        location: Location::Mars,
        stats: Stats {
            strength: 10,
            agility: 9,
        },
        some_tuple: (true, 8),
        some_array: [17, 76],
        some_b256: zero,
    });

    log(keccak_hashed_u8);
    log(keccak_hashed_u16);
    log(keccak_hashed_u32);
    log(keccak_hashed_u64);
    log(keccak_hashed_b256);
    log(keccak_hashed_bool);
    log(keccak_hashed_str);
    log(keccak_hashed_tuple);
    log(keccak_hashed_array);
    log(keccak_hashed_enum);
    log(keccak_hashed_struct);
}

Signature Recovery

script;

use std::{b512::B512, ecr::{ec_recover, ec_recover_address, EcRecoverError}};

const MSG_HASH = 0xee45573606c96c98ba970ff7cf9511f1b8b25e6bcd52ced30b89df1e4a9c4323;

fn main() {
    let hi = 0xbd0c9b8792876713afa8bff383eebf31c43437823ed761cc3600d0016de5110c;
    let lo = 0x44ac566bd156b4fc71a4a4cb2655d3dd360c695edb17dc3b64d611e122fea23d;
    let signature: B512 = B512::from((hi, lo));

    // A recovered public key pair.
    let public_key = ec_recover(signature, MSG_HASH);

    // A recovered Fuel address.
    let result_address: Result<Address, EcRecoverError> = ec_recover_address(signature, MSG_HASH);
    if let Result::Ok(address) = result_address {
        log(address.value);
    } else {
        revert(0);
    }
}

Note: Recovery of EVM addresses is also supported via std::vm::evm.

Storage

When developing a smart contract, you will typically need some sort of persistent storage. In this case, persistent storage, often just called storage in this context, is a place where you can store values that are persisted inside the contract itself. This is in contrast to a regular value in memory, which disappears after the contract exits.

Put in conventional programming terms, contract storage is like saving data to a hard drive. That data is saved even after the program which saved it exits. That data is persistent. Using memory is like declaring a variable in a program: it exists for the duration of the program and is non-persistent.

Some basic use cases of storage include declaring an owner address for a contract and saving balances in a wallet.

Storage Accesses Via the storage Keyword

Declaring variables in storage requires a storage declaration that contains a list of all your variables, their types, and their initial values as follows:

struct Type1 {
    x: u64,
    y: u64,
}

struct Type2 {
    w: b256,
    z: bool,
}

storage {
    var1: Type1 = Type1 { x: 0, y: 0 },
    var2: Type2 = Type2 {
        w: 0x0000000000000000000000000000000000000000000000000000000000000000,
        z: false,
    },
}

To write into a storage variable, you need to use the storage keyword as follows:

    #[storage(write)]
    fn store_something() {
        storage.var1.x = 42;
        storage.var1.y = 77;
        storage.var2.w = 0x1111111111111111111111111111111111111111111111111111111111111111;
        storage.var2.z = true;
    }

To read a storage variable, you also need to use the storage keyword as follows:

    #[storage(read)]
    fn get_something() -> (u64, u64, b256, bool) {
        (
            storage.var1.x,
            storage.var1.y,
            storage.var2.w,
            storage.var2.z,
        )
    }

Storage Maps

Generic storage maps are available in the standard library as StorageMap<K, V> which have to be defined inside a storage block and allow you to call insert() and get() to insert values at specific keys and get those values respectively. Refer to Storage Maps for more information about StorageMap<K, V>.

Manual Storage Management

It is possible to leverage FuelVM storage operations directly using the std::storage::store and std::storage::get functions provided in the standard library. With this approach you will have to manually assign the internal key used for storage. An example is as follows:

contract;

use std::storage::{get, store};

abi StorageExample {
    #[storage(write)]
    fn store_something(amount: u64);

    #[storage(read)]
    fn get_something() -> u64;
}

const STORAGE_KEY: b256 = 0x0000000000000000000000000000000000000000000000000000000000000000;

impl StorageExample for Contract {
    #[storage(write)]
    fn store_something(amount: u64) {
        store(STORAGE_KEY, amount);
    }

    #[storage(read)]
    fn get_something() -> u64 {
        let value: Option<u64> = get::<u64>(STORAGE_KEY);
        value.unwrap_or(0)
    }
}

Note: Though these functions can be used for any data type, they should mostly be used for arrays because arrays are not yet supported in storage blocks. Note, however, that all data types can be used as types for keys and/or values in StorageMap<K, V> without any restrictions.

Purity

A function is pure if it does not access any persistent storage. Conversely, the function is impure if it does access any storage. Naturally, as storage is only available in smart contracts, impure functions cannot be used in predicates, scripts, or libraries. A pure function cannot call an impure function.

In Sway, functions are pure by default but can be opted into impurity via the storage function attribute. The storage attribute may take read and/or write arguments indicating which type of access the function requires.

#[storage(read)]
fn get_amount() -> u64 {
    ...
}

#[storage(read, write)]
fn increment_amount(increment: u64) -> u64 {
    ...
}

Impure functions which call other impure functions must have at least the same storage privileges or a superset of those for the function called. For example, to call a function with write access a caller must also have write access, or both read and write access. To call a function with read and write access the caller must also have both privileges.

The storage attribute may also be applied to methods and associated functions, trait and ABI declarations.

A pure function gives you some guarantees: you will not incur excessive storage gas costs, the compiler can apply additional optimizations, and they are generally easy to reason about and audit. A similar concept exists in Solidity. Note that Solidity refers to contract storage as contract state, and in the Sway/Fuel ecosystem, these two terms are largely interchangeable.

Identifiers

Addresses in Sway are similar to EVM addresses. The two major differences are:

  1. Sway addresses are 32 bytes long (instead of 20)
  2. Sway addresses are computed with the SHA-256 hash of the public key instead of the keccak-256 hash.

Contracts, on the other hand, are uniquely identified with a contract ID rather than an address. A contract's ID is also 32 bytes long and is calculated here.

Native Support for Multiple Asset Types

The FuelVM has built-in support for working with multiple assets.

What does this mean in practice?

As in the EVM, sending ETH to an address or contract is an operation built into the FuelVM, meaning it doesn't rely on the existence of some token smart contract to update balances to track ownership.

However, unlike the EVM, the process for sending any native asset is the same. This means that while you would still need a smart contract to handle the minting and burning of fungible tokens, the sending and receiving of these tokens can be done independently of the token contract.

Liquidity Pool Example

All contracts in Fuel can mint and burn their own native token. Contracts can also receive and transfer any native asset including their own. Internal balances of all native assets pushed through calls or minted by the contract are tracked by the FuelVM and can be queried at any point using the balance_of function from the std library. Therefore, there is no need for any manual accounting of the contract's balances using persistent storage.

The std library provides handy methods for accessing Fuel's native assset operations.

In this example, we show a basic liquidity pool contract minting its own native asset LP token.

contract;

use std::{
    call_frames::{
        contract_id,
        msg_asset_id,
    },
    context::msg_amount,
    token::{
        mint_to_address,
        transfer_to_address,
    },
};

abi LiquidityPool {
    fn deposit(recipient: Address);
    fn withdraw(recipient: Address);
}

const BASE_TOKEN = ContractId::from(0x9ae5b658754e096e4d681c548daf46354495a437cc61492599e33fc64dcdc30c);

impl LiquidityPool for Contract {
    fn deposit(recipient: Address) {
        assert(msg_asset_id() == BASE_TOKEN);
        assert(msg_amount() > 0);

        // Mint two times the amount.
        let amount_to_mint = msg_amount() * 2;

        // Mint some LP token based upon the amount of the base token.
        mint_to_address(amount_to_mint, recipient);
    }

    fn withdraw(recipient: Address) {
        assert(msg_asset_id() == contract_id());
        assert(msg_amount() > 0);

        // Amount to withdraw.
        let amount_to_transfer = msg_amount() / 2;

        // Transfer base token to recipient.
        transfer_to_address(amount_to_transfer, BASE_TOKEN, recipient);
    }
}

Native Token Example

In this example, we show a native token contract with more minting, burning and transferring capabilities.

contract;

use std::{context::*, token::*};

abi NativeAssetToken {
    fn mint_coins(mint_amount: u64);
    fn burn_coins(burn_amount: u64);
    fn force_transfer_coins(coins: u64, asset_id: ContractId, target: ContractId);
    fn transfer_coins_to_output(coins: u64, asset_id: ContractId, recipient: Address);
    fn deposit();
    fn get_balance(target: ContractId, asset_id: ContractId) -> u64;
    fn mint_and_send_to_contract(amount: u64, destination: ContractId);
    fn mint_and_send_to_address(amount: u64, recipient: Address);
}

impl NativeAssetToken for Contract {
    /// Mint an amount of this contracts native asset to the contracts balance.
    fn mint_coins(mint_amount: u64) {
        mint(mint_amount);
    }

    /// Burn an amount of this contracts native asset.
    fn burn_coins(burn_amount: u64) {
        burn(burn_amount);
    }

    /// Transfer coins to a target contract.
    fn force_transfer_coins(coins: u64, asset_id: ContractId, target: ContractId) {
        force_transfer_to_contract(coins, asset_id, target);
    }

    /// Transfer coins to a transaction output to be spent later.
    fn transfer_coins_to_output(coins: u64, asset_id: ContractId, recipient: Address) {
        transfer_to_address(coins, asset_id, recipient);
    }

    /// Get the internal balance of a specific coin at a specific contract.
    fn get_balance(target: ContractId, asset_id: ContractId) -> u64 {
        balance_of(target, asset_id)
    }

    /// Deposit tokens back into the contract.
    fn deposit() {
        assert(msg_amount() > 0);
    }

    /// Mint and send this contracts native token to a destination contract.
    fn mint_and_send_to_contract(amount: u64, destination: ContractId) {
        mint_to_contract(amount, destination);
    }

    /// Mind and send this contracts native token to a destination address.
    fn mint_and_send_to_address(amount: u64, recipient: Address) {
        mint_to_address(amount, recipient);
    }
}

Access Control

Smart contracts require the ability to restrict access to and identify certain users or contracts. Unlike account-based blockchains, transactions in UTXO-based blockchains (i.e. Fuel) do not necessarily have a unique transaction sender. Additional logic is needed to handle this difference, and is provided by the standard library.

msg_sender

To deliver an experience akin to the EVM's access control, the std library provides a msg_sender function, which identifies a unique caller based upon the call and/or transaction input data.

contract;

use std::auth::msg_sender;

abi MyOwnedContract {
    fn receive(field_1: u64) -> bool;
}

const OWNER = Address::from(0x9ae5b658754e096e4d681c548daf46354495a437cc61492599e33fc64dcdc30c);

impl MyOwnedContract for Contract {
    fn receive(field_1: u64) -> bool {
        let sender = msg_sender().unwrap();
        if let Identity::Address(addr) = sender {
            assert(addr == OWNER);
        } else {
            revert(0);
        }

        true
    }
}

The msg_sender function works as follows:

  • If the caller is a contract, then Result::Ok(Sender) is returned with the ContractId sender variant.
  • If the caller is external (i.e. from a script), then all coin input owners in the transaction are checked. If all owners are the same, then Result::Ok(Sender) is returned with the Address sender variant.
  • If the caller is external and coin input owners are different, then the caller cannot be determined and a Result::Err(AuthError) is returned.

Contract Ownership

Many contracts require some form of ownership for access control. To accomplish this, it is recommended that a storage variable of type Option<Identity> is used to keep track of the owner. This allows setting and revoking ownership using the variants Some(..) and None respectively. This is better, safer, and more readable than using the Identity type directly where revoking ownership has to be done using some magic value such as std::constants::ZERO_B256 or otherwise.

The following is an example of how to properly set ownership of a contract:

    #[storage(write)]
    fn set_owner(identity: Identity) {
        storage.owner = Option::Some(identity);
    }

The following is an example of how to properly revoke ownership of a contract:

    #[storage(write)]
    fn revoke_ownership() {
        storage.owner = Option::None;
    }

Calling Contracts

Smart contracts can be called by other contracts or scripts. In the FuelVM, this is done primarily with the call instruction.

Sway provides a nice way to manage callable interfaces with its abi system. The Fuel ABI specification can be found here.

Example

Here is an example of a contract calling another contract in Sway. A script can call a contract in the same way.

// ./contract_a.sw
contract;

abi ContractA {
    fn receive(field_1: bool, field_2: u64) -> u64;
}

impl ContractA for Contract {
    fn receive(field_1: bool, field_2: u64) -> u64 {
        assert(field_1 == true);
        assert(field_2 > 0);
        return_45()
    }
}

fn return_45() -> u64 {
  45
}
// ./contract_b.sw
contract;

use contract_a::ContractA;

abi ContractB {
    fn make_call();
}

const contract_id = 0x79fa8779bed2f36c3581d01c79df8da45eee09fac1fd76a5a656e16326317ef0;

impl ContractB for Contract {
    fn make_call() {
      let x = abi(ContractA, contract_id);
      let return_value = x.receive(true, 3); // will be 45
    }
}

Note: The ABI is for external calls only therefore you cannot define a method in the ABI and call it in the same contract. If you want to define a function for a contract, but keep it private so that only your contract can call it, you can define it outside of the impl and call it inside the contract, similar to the return_45() function above.

Advanced Calls

All calls forward a gas stipend, and may additionally forward one native asset with the call.

Here is an example of how to specify the amount of gas (gas), the asset ID of the native asset (asset_id), and the amount of the native asset (coins) to forward:

script;

abi MyContract {
    fn foo(field_1: bool, field_2: u64);
}

fn main() {
    let x = abi(MyContract, 0x79fa8779bed2f36c3581d01c79df8da45eee09fac1fd76a5a656e16326317ef0);
    let asset_id = 0x7777_7777_7777_7777_7777_7777_7777_7777_7777_7777_7777_7777_7777_7777_7777_7777;
    x.foo {
        gas: 5000, asset_id: asset_id, coins: 5000
    }
    (true, 3);
}

Handling Re-entrancy

A common attack vector for smart contracts is re-entrancy. Similar to the EVM, the FuelVM allows for re-entrancy.

A stateless re-entrancy guard is included in the sway-libs library. The guard will panic (revert) at run time if re-entrancy is detected.

contract;

use reentrancy::reentrancy_guard;

abi MyContract {
    fn some_method();
}

impl ContractB for Contract {
    fn some_method() {
        reentrancy_guard();
        // do something
    }
}

CEI pattern violation static analysis

Another way of avoiding re-entrancy-related attacks is to follow the so-called CEI pattern. CEI stands for "Checks, Effects, Interactions", meaning that the contract code should first perform safety checks, also known as "pre-conditions", then perform effects, i.e. modify or read the contract storage and execute external contract calls (interaction) only at the very end of the function/method.

Please see this blog post for more detail on some vulnerabilities in case of storage modification after interaction and this blog post for more information on storage reads after interaction.

The Sway compiler implements a check that the CEI pattern is not violated in the user contract and issues warnings if that's the case.

For example, in the following contract the CEI pattern is violated, because an external contract call is executed before a storage write.

contract;

mod other_contract;

use other_contract::*;

use std::auth::msg_sender;

abi MyContract {
    #[storage(read, write)]
    fn withdraw(external_contract_id: ContractId);
}

storage {
    balances: StorageMap<Identity, u64> = StorageMap {},
}

impl MyContract for Contract {
    #[storage(read, write)]
    fn withdraw(external_contract_id: ContractId) {
        let sender = msg_sender().unwrap();
        let bal = storage.balances.get(sender).unwrap_or(0);

        assert(bal > 0);

        // External call
        let caller = abi(OtherContract, external_contract_id.into());
        caller.external_call { coins: bal }();

        // Storage update _after_ external call
        storage.balances.insert(sender, 0);
    }
}

Here, other_contract is defined as follows:

library;

abi OtherContract {
    #[payable]
    fn external_call();
}

The CEI pattern analyzer issues a warning as follows, pointing to the interaction before a storage modification:

warning
  --> /path/to/contract/main.sw:28:9
   |
26 |
27 |           let caller = abi(OtherContract, external_contract_id.into());
28 |           caller.external_call { coins: bal }();
   |  _________-
29 | |
30 | |         // Storage update _after_ external call
31 | |         storage.balances.insert(sender, 0);
   | |__________________________________________- Storage write after external contract interaction in function or method "withdraw". Consider making all storage writes before calling another contract
32 |       }
33 |   }
   |
____

In case there is a storage read after an interaction, the CEI analyzer will issue a similar warning.

In addition to storage reads and writes after an interaction, the CEI analyzer reports analogous warnings about:

  • balance tree updates, i.e. balance tree reads with subsequent writes, which may be produced by the tr and tro asm instructions or library functions using them under the hood;
  • balance trees reads with bal instruction;
  • changes to the output messages that can be produced by the __smo intrinsic function or the smo asm instruction.

Differences from the EVM

While the Fuel contract calling paradigm is similar to the EVM's (using an ABI, forwarding gas and data), it differs in two key ways:

  1. Native assets: FuelVM calls can forward any native asset not just base asset.

  2. No data serialization: Contract calls in the FuelVM do not need to serialize data to pass it between contracts; instead they simply pass a pointer to the data. This is because the FuelVM has a shared global memory which all call frames can read from.

Advanced Concepts

Advanced concepts.

Advanced Types

Creating Type Synonyms with Type Aliases

Sway provides the ability to declare a type alias to give an existing type another name. For this we use the type keyword. For example, we can create the alias Kilometers to u64 like so:

type Kilometers = u64;

Now, the alias Kilometers is a synonym for u64. Note that Kilometers is not a separate new type. Values that have the type Kilometers will be treated the same as values of type u64:

    let x: u64 = 5;
    let y: Kilometers = 5;
    assert(x + y == 10);

Because Kilometers and u64 are the same type, we can add values of both types and we can pass Kilometers values to functions that take u64 parameters. However, using this method, we don’t get the type checking benefits that we get from introducing a separate new type called Kilometers. In other words, if we mix up Kilometers and i32 values somewhere, the compiler will not give us an error.

The main use case for type synonyms is to reduce repetition. For example, we might have a lengthy array type like this:

[MyStruct<u64, b256>; 5]

Writing this lengthy type in function signatures and as type annotations all over the code can be tiresome and error prone. Imagine having a project full of code like this:

fn foo_long(array: [MyStruct<u64, b256>; 5]) -> [MyStruct<u64, b256>; 5] {
    array
}

A type alias makes this code more manageable by reducing the repetition. Below, we’ve introduced an alias named MyArray for the verbose type and can replace all uses of the type with the shorter alias MyArray:

type MyArray = [MyStruct<u64, b256>; 5];

fn foo_shorter(array: MyArray) -> MyArray {
    array
}

This code is much easier to read and write! Choosing a meaningful name for a type alias can help communicate your intent as well.

Generic Types

Basics

In Sway, generic types follow a very similar pattern to those in Rust. Let's look at some example syntax, starting with a generic function:

fn noop<T>(argument: T) -> T {
    argument
}

Here, the noop() function trivially returns exactly what was given to it. T is a type parameter, and it says that this function exists for all types T. More formally, this function could be typed as:

noop :: ∀T. T -> T

Generic types are a way to refer to types in general, meaning without specifying a single type. Our noop function would work with any type in the language, so we don't need to specify noop(argument: u8) -> u8, noop(argument: u16) -> u16, etc.

Code Generation

One question that arises when dealing with generic types is: how does the assembly handle this? There are a few approaches to handling generic types at the lowest level. Sway uses a technique called monomorphization. This means that the generic function is compiled to a non-generic version for every type it is called on. In this way, generic functions are purely shorthand for the sake of ergonomics.

Trait Constraints

Note Trait constraints have not yet been implemented

Important background to know before diving into trait constraints is that the where clause can be used to specify the required traits for the generic argument. So, when writing something like a HashMap you may want to specify that the generic argument implements a Hash trait.

fn get_hashmap_key<T>(Key : T) -> b256
    where T: Hash
{
    // Code within here can then call methods associated with the Hash trait on Key
}

Of course, our noop() function is not useful. Often, a programmer will want to declare functions over types which satisfy certain traits. For example, let's try to implement the successor function, successor(), for all numeric types.

fn successor<T>(argument: T)
    where T: Add
{
    argument + 1
}

Run forc build, and you will get:

.. |
 9 |   where T: Add
10 |   {
11 |       argument + 1                                        
   |                  ^ Mismatched types: expected type "T" but saw type "u64"
12 |   }
13 |

This is because we don't know for a fact that 1, which in this case defaulted to 1u64, actually can be added to T. What if T is f64? Or b256? What does it mean to add 1u64 in these cases?

We can solve this problem with another trait constraint. We can only find the successor of some value of type T if that type T defines some incrementor. Let's make a trait:

trait Incrementable {
    /// Returns the value to add when calculating the successor of a value.
    fn incrementor() -> Self;
}

Now, we can modify our successor() function:

fn successor<T>(argument: T)
    where T: Add,
          T: Incrementable
{
    argument + T::incrementor()
}

Generic Structs and Enums

Just like functions, structs and enums can be generic. Let's take a look at the standard library version of Option<T>:

enum Option<T> {
    Some: T,
    None: (),
}

Just like an unconstrained generic function, this type exists for all (∀) types T. Result<T, E> is another example:

enum Result<T, E> {
    Ok: T,
    Err: E,
}

Both generic enums and generic structs can be trait constrained, as well. Consider this struct:

struct Foo<T>
    where T: Add
{
    field_one: T,
}

Type Arguments

Similar to Rust, Sway has what is colloquially known as the turbofish. The turbofish looks like this: ::<> (see the little fish with bubbles behind it?). The turbofish is used to annotate types in a generic context. Say you have the following function:

fn foo<T, E>(t: T) -> Result<T, E> {
    Result::Ok(t)
}

In this code example, which is admittedly asinine, you can't possibly know what type E is. You'd need to provide the type manually, with a turbofish:

fn foo<T, E>(t: T) -> Result<T, E> {
    Result::Ok::<T, MyErrorType>(t)
}

It is also common to see the turbofish used on the function itself:

fn main() {
    foo::<Bar, Baz>()
}

Traits

Declaring a Trait

A trait opts a type into a certain type of behavior or functionality that can be shared among types. This allows for easy reuse of code and generic programming. If you have ever used a typeclass in Haskell, a trait in Rust, or even an interface in Java, these are similar concepts.

Let's take a look at some code:

trait Compare {
    fn equals(self, b: Self) -> bool;
} {
    fn not_equals(self, b: Self) -> bool {
        !self.equals(b)
    }
}

We have just declared a trait called Compare. After the name of the trait, there are two blocks of code (a block is code enclosed in { curly brackets }). The first block is the interface surface. The second block is the methods provided by the trait. If a type can provide the methods in the interface surface, then it gets access to the methods in the trait for free! What the above trait is saying is: if you can determine if two values are equal, then for free, you can determine that they are not equal. Note that trait methods have access to the methods defined in the interface surface.

Implementing a Trait

Ok, so I know that numbers can be equal. I want to implement my Compare trait for u64. Let's take a look at how that is done:

impl Compare for u64 {
    fn equals(self, b: Self) -> bool {
        self == b
    }
}

The above snippet declares all of the methods in the trait Compare for the type u64. Now, we have access to both the equals and not_equals methods for u64, as long as the trait Compare is in scope.

Supertraits

When using multiple traits, scenarios often come up where one trait may require functionality from another trait. This is where supertraits come in as they allow you to require a trait when implementing another trait (ie. a trait with a trait). A good example of this is the Ord trait of the core library of Sway. The Ord trait requires the Eq trait, so Eq is kept as a separate trait as one may decide to implement Eq without implementing other parts of the Ord trait.


trait Eq {
    fn equals(self, b: Self) -> bool;
}

trait Ord: Eq {
    fn gte(self, b: Self) -> bool;
}

impl Ord for u64 {
    fn gte(self, b: Self) -> bool {
        // As `Eq` is a supertrait of `Ord`, `Ord` can access the equals method
        self.equals(b) || self.gt(b)
    }
}

To require a supertrait, add a : after the trait name and then list the traits you would like to require and separate them with a +.

ABI supertraits

ABIs can also have supertrait annotations:

contract;

trait ABIsupertrait {
    fn foo();
}

abi MyAbi : ABIsupertrait {
    fn bar();
} {
    fn baz() {
        Self::foo() // supertrait method usage
    }
}

impl ABIsupertrait for Contract {
    fn foo() {}
}

// The implementation of MyAbi for Contract must also implement ABIsupertrait
impl MyAbi for Contract {
    fn bar() {
        Self::foo() // supertrait method usage
    }
}

The implementation of MyAbi for Contract must also implement the ABIsupertrait trait. Methods in ABIsupertrait are not available externally, i.e. they're not actually contract methods, but they can be used in the actual contract methods, as shown in the example above.

ABI supertraits are intended to make contract implementations compositional, allowing combining orthogonal contract features using, for instance, libraries.

Use Cases

Custom Types (structs, enums)

Often, libraries and APIs have interfaces that are abstracted over a type that implements a certain trait. It is up to the consumer of the interface to implement that trait for the type they wish to use with the interface. For example, let's take a look at a trait and an interface built off of it.

library;

pub enum Suit {
    Hearts: (),
    Diamonds: (),
    Clubs: (),
    Spades: (),
}

pub trait Card {
    fn suit(self) -> Suit;
    fn value(self) -> u8;
}

fn play_game_with_deck<T>(a: Vec<T>) where T: Card {
    // insert some creative card game here
}

Note Trait constraints (i.e. using the where keyword) have not yet been implemented

Now, if you want to use the function play_game_with_deck with your struct, you must implement Card for your struct. Note that the following code example assumes a dependency games has been included in the Forc.toml file.

script;

use games::*;

struct MyCard {
    suit: Suit,
    value: u8
}

impl Card for MyCard {
    fn suit(self) -> Suit {
        self.suit
    }
    fn value(self) -> u8 {
        self.value
    }
}

fn main() {
    let mut i = 52;
    let mut deck: Vec<MyCard> = Vec::with_capacity(50);
    while i > 0 {
        i = i - 1;
        deck.push(MyCard { suit: generate_random_suit(), value: i % 4}
    }
    play_game_with_deck(deck);
}

fn generate_random_suit() -> Suit {
  [ ... ]
}

Inline Assembly in Sway

While many users will never have to touch assembly language while writing sway code, it is a powerful tool that enables many advanced use-cases (ie: optimizations, building libraries, etc).

ASM Block

In Sway, the way we use assembly inline is to declare an asm block like this:

asm() {...}

Declaring an asm block is similar to declaring a function. We can specify register names to operate on as arguments, we can perform operations within the block, and we can return a value. Here's an example showing what this might look like:

pub fn add_1(num: u32) -> u32 {
    asm(r1: num, r2) {
        add r2 r1 one;
        r2: u32
    }
}

An asm block can only return a single register. If you really need to return more than one value, you can modify a tuple. Here's an example showing how can implement this (u64, u64):

script;

fn adder(a: u64, b: u64, c: u64) -> (u64, u64) {
    let empty_tuple = (0u64, 0u64);
    asm(output: empty_tuple, r1: a, r2: b, r3: c, r4, r5) {
        add  r4 r1 r2; // add a & b and put the result in r4
        add  r5 r2 r3; // add b & c and put the result in r5
        sw   output r4 i0; // store the word in r4 in output + 0 words
        sw   output r5 i1; // store the word in r5 in output + 1 word
        output: (u64, u64) // return both values
    }
}

fn main() -> bool {
    let (first, second) = adder(1, 2, 3);
    assert(first == 3);
    assert(second == 5);
    true
}

Note that this is contrived example meant to demonstrate the syntax; there's absolutely no need to use assembly to add integers!

Note that in the above example:

  • we initialized the register r1 with the value of num.
  • we declared a second register r2 (you may choose any register names you want).
  • we use the add opcode to add one to the value of r1 and store it in r2.
  • one is an example of a "reserved register", of which there are 16 in total. Further reading on this is linked below under "Semantics".
  • we return r2 & specify the return type as being u32 (the return type is u64 by default).

An important note is that the ji and jnei opcodes are not available within an asm block. For those looking to introduce control flow to asm blocks, it is recommended to surround smaller chunks of asm with control flow (if, else, and while).

For examples of assembly in action, check out the Sway standard library.

For a complete list of all instructions supported in the FuelVM: Instructions.

And to learn more about the FuelVM semantics: Semantics.

Common Collections

Sway’s standard library includes a number of very useful data structures called collections. Most other data types represent one specific value, but collections can contain multiple values. Unlike the built-in array and tuple types which are allocated on the "stack" and cannot grow in size, the data these collections point to is stored either on the "heap" or in contract "storage", which means the amount of data does not need to be known at compile time and can grow as the program runs. Each kind of collection has different capabilities and costs, and choosing an appropriate one for your current situation is a skill you’ll develop over time. In this chapter, we’ll discuss three collections that are used very often in Sway programs:

A vector on the heap allows you to store a variable number of values next to each other.

A storage vector is similar to a vector on the heap but uses persistent storage.

A storage map allows you to associate a value with a particular key.

We’ll discuss how to create and update vectors, storage vectors, and storage maps, as well as what makes each special.

Vectors on the Heap

The first collection type we’ll look at is Vec<T>, also known as a vector. Vectors allow you to store more than one value in a single data structure that puts all the values next to each other in memory. Vectors can only store values of the same type. They are useful when you have a list of items, such as the lines of text in a file or the prices of items in a shopping cart.

Vec<T> is included in the standard library prelude which means that there is no need to import it manually.

Creating a New Vector

To create a new empty vector, we call the Vec::new function, as shown below:

    let v: Vec<u64> = Vec::new();

Note that we added a type annotation here. Because we aren’t inserting any values into this vector, the Sway compiler doesn’t know what kind of elements we intend to store. Vectors are implemented using generics which means that the Vec<T> type provided by the standard library can hold any type. When we create a vector to hold a specific type, we can specify the type within angle brackets. In the example above, we’ve told the Sway compiler that the Vec<T> in v will hold elements of the u64 type.

Updating a Vector

To create a vector and then add elements to it, we can use the push method, as shown below:

    let mut v = Vec::new();

    v.push(5);
    v.push(6);
    v.push(7);
    v.push(8);

As with any variable, if we want to be able to change its value, we need to make it mutable using the mut keyword, as discussed in the section Declaring a Variable. The numbers we place inside are all of type u64, and the Sway compiler infers this from the data, so we don’t need the Vec<u64> annotation.

Reading Elements of Vectors

To read a value stored in a vector at a particular index, you can use the get method as shown below:

    let third = v.get(2);
    match third {
        Option::Some(third) => log(third),
        Option::None => revert(42),
    }

Note two details here. First, we use the index value of 2 to get the third element because vectors are indexed by number, starting at zero. Second, we get the third element by using the get method with the index passed as an argument, which gives us an Option<T>.

When the get method is passed an index that is outside the vector, it returns None without panicking. This is particularly useful if accessing an element beyond the range of the vector may happen occasionally under normal circumstances. Your code will then have logic to handle having either Some(element) or None. For example, the index could be coming as a contract method argument. If the argument passed is too large, the method get will return a None value, and the contract method may then decide to revert when that happens or return a meaningful error that tells the user how many items are in the current vector and give them another chance to pass a valid value.

Iterating over the Values in a Vector

To access each element in a vector in turn, we would iterate through all of the valid indices using a while loop and the len method as shown below:

    let mut i = 0;
    while i < v.len() {
        log(v.get(i).unwrap());
        i += 1;
    }

Note two details here. First, we use the method len which returns the length of the vector. Second, we call the method unwrap to extract the Option returned by get. We know that unwrap will not fail (i.e. will not cause a revert) because each index i passed to get is known to be smaller than the length of the vector.

Using an Enum to store Multiple Types

Vectors can only store values that are the same type. This can be inconvenient; there are definitely use cases for needing to store a list of items of different types. Fortunately, the variants of an enum are defined under the same enum type, so when we need one type to represent elements of different types, we can define and use an enum!

For example, say we want to get values from a row in a table in which some of the columns in the row contain integers, some b256 values, and some Booleans. We can define an enum whose variants will hold the different value types, and all the enum variants will be considered the same type: that of the enum. Then we can create a vector to hold that enum and so, ultimately, holds different types. We’ve demonstrated this below:

    enum TableCell {
        Int: u64,
        B256: b256,
        Boolean: bool,
    }

    let mut row = Vec::new();
    row.push(TableCell::Int(3));
    row.push(TableCell::B256(0x0101010101010101010101010101010101010101010101010101010101010101));
    row.push(TableCell::Boolean(true));

Now that we’ve discussed some of the most common ways to use vectors, be sure to review the API documentation for all the many useful methods defined on Vec<T> by the standard library. For now, these can be found in the source code for Vec<T>. For example, in addition to push, a pop method removes and returns the last element, a remove method removes and returns the element at some chosen index within the vector, an insert method inserts an element at some chosen index within the vector, etc.

Storage Vectors

The second collection type we’ll look at is StorageVec<T>. Just like vectors on the heap (i.e. Vec<T>), storage vectors allow you to store more than one value in a single data structure where each value is assigned an index and can only store values of the same type. However, unlike Vec<T>, the elements of a StorageVec are stored in persistent storage, and consecutive elements are not necessarily stored in storage slots that have consecutive keys.

In order to use StorageVec<T>, you must first import StorageVec as follows:

use std::storage::StorageVec;

Another major difference between Vec<T> and StorageVec<T> is that StorageVec<T> can only be used in a contract because only contracts are allowed to access persistent storage.

Creating a New Storage Vector

To create a new empty storage vector, we have to declare the vector in a storage block as follows:

    v: StorageVec<u64> = StorageVec {},

Just like any other storage variable, two things are required when declaring a StorageVec: a type annotation and an initializer. The initializer is just an empty struct of type StorageVec because StorageVec<T> itself is an empty struct! Everything that is interesting about StorageVec<T> is implemented in its methods.

Storage vectors, just like Vec<T>, are implemented using generics which means that the StorageVec<T> type provided by the standard library can hold any type. When we create a storage vector to hold a specific type, we can specify the type within angle brackets. In the example above, we’ve told the Sway compiler that the StorageVec<T> in v will hold elements of the u64 type.

Updating a Storage Vector

To add elements to a storage vector, we can use the push method, as shown below:

    #[storage(read, write)]
    fn push_to_storage_vec() {
        storage.v.push(5);
        storage.v.push(6);
        storage.v.push(7);
        storage.v.push(8);
    }

Note two details here. First, in order to use push, we need to first access the vector using the storage keyword. Second, because push requires accessing storage, a storage annotation is required on the ABI function that calls push. While it may seem that #[storage(write)] should be enough here, the read annotation is also required because each call to push requires reading (and then updating) the length of the storage vector which is also stored in persistent storage.

Note The storage annotation is also required for any private function defined in the contract that tries to push into the vector.

Note There is no need to add the mut keyword when declaring a StorageVec<T>. All storage variables are mutable by default.

Reading Elements of Storage Vectors

To read a value stored in a vector at a particular index, you can use the get method as shown below:

    #[storage(read)]
    fn read_from_storage_vec() {
        let third = storage.v.get(2);
        match third {
            Option::Some(third) => log(third),
            Option::None => revert(42),
        }
    }

Note three details here. First, we use the index value of 2 to get the third element because vectors are indexed by number, starting at zero. Second, we get the third element by using the get method with the index passed as an argument, which gives us an Option<T>. Third, the ABI function calling get only requires the annotation #[storage(read)] as one might expect because get does not write to storage.

When the get method is passed an index that is outside the vector, it returns None without panicking. This is particularly useful if accessing an element beyond the range of the vector may happen occasionally under normal circumstances. Your code will then have logic to handle having either Some(element) or None. For example, the index could be coming as a contract method argument. If the argument passed is too large, the method get will return a None value, and the contract method may then decide to revert when that happens or return a meaningful error that tells the user how many items are in the current vector and give them another chance to pass a valid value.

Iterating over the Values in a Vector

To access each element in a vector in turn, we would iterate through all of the valid indices using a while loop and the len method as shown below:

    #[storage(read)]
    fn iterate_over_a_storage_vec() {
        let mut i = 0;
        while i < storage.v.len() {
            log(storage.v.get(i).unwrap());
            i += 1;
        }
    }

Again, this is quite similar to iterating over the elements of a Vec<T> where we use the method len to return the length of the vector. We also call the method unwrap to extract the Option returned by get. We know that unwrap will not fail (i.e. will not cause a revert) because each index i passed to get is known to be smaller than the length of the vector.

Using an Enum to store Multiple Types

Storage vectors, just like Vec<T>, can only store values that are the same type. Similarly to what we did for Vec<T> in the section Using an Enum to store Multiple Types, we can define an enum whose variants will hold the different value types, and all the enum variants will be considered the same type: that of the enum. This is shown below:

enum TableCell {
    Int: u64,
    B256: b256,
    Boolean: bool,
}

Then we can declare a storage vector in a storage block to hold that enum and so, ultimately, holds different types:

    row: StorageVec<TableCell> = StorageVec {},

We can now push different enum variants to the storage vector as follows:

    #[storage(read, write)]
    fn push_to_multiple_types_storage_vec() {
        storage.row.push(TableCell::Int(3));
        storage.row.push(TableCell::B256(0x0101010101010101010101010101010101010101010101010101010101010101));
        storage.row.push(TableCell::Boolean(true));
    }

Now that we’ve discussed some of the most common ways to use storage vectors, be sure to review the API documentation for all the many useful methods defined on StorageVec<T> by the standard library. For now, these can be found in the source code for StorageVec<T>. For example, in addition to push, a pop method removes and returns the last element, a remove method removes and returns the element at some chosen index within the vector, an insert method inserts an element at some chosen index within the vector, etc.

Storage Maps

Another important common collection is the storage map. The type StorageMap<K, V> from the standard library stores a mapping of keys of type K to values of type V using a hashing function, which determines how it places these keys and values into storage slots. This is similar to Rust's HashMap<K, V> but with a few differences.

Storage maps are useful when you want to look up data not by using an index, as you can with vectors, but by using a key that can be of any type. For example, when building a ledger-based sub-currency smart contract, you could keep track of the balance of each wallet in a storage map in which each key is a wallet’s Address and the values are each wallet’s balance. Given an Address, you can retrieve its balance.

Similarly to StorageVec<T>, StorageMap<K, V> can only be used in a contract because only contracts are allowed to access persistent storage.

StorageMap<T> is included in the standard library prelude which means that there is no need to import it manually.

Creating a New Storage Map

To create a new empty storage map, we have to declare the map in a storage block as follows:

    map: StorageMap<Address, u64> = StorageMap {},

Just like any other storage variable, two things are required when declaring a StorageMap: a type annotation and an initializer. The initializer is just an empty struct of type StorageMap because StorageMap<K, V> itself is an empty struct! Everything that is interesting about StorageMap<K, V> is implemented in its methods.

Storage maps, just like Vec<T> and StorageVec<T>, are implemented using generics which means that the StorageMap<K, V> type provided by the standard library can map keys of any type K to values of any type V. In the example above, we’ve told the Sway compiler that the StorageMap<K, V> in map will map keys of type Address to values of type u64.

Updating a Storage Map

To insert key-value pairs into a storage map, we can use the insert method, as shown below:

    #[storage(write)]
    fn insert_into_storage_map() {
        let addr1 = Address::from(0x0101010101010101010101010101010101010101010101010101010101010101);
        let addr2 = Address::from(0x0202020202020202020202020202020202020202020202020202020202020202);

        storage.map.insert(addr1, 42);
        storage.map.insert(addr2, 77);
    }

Note two details here. First, in order to use insert, we need to first access the storage map using the storage keyword. Second, because insert requires writing into storage, a #[storage(write)] annotation is required on the ABI function that calls insert.

Note The storage annotation is also required for any private function defined in the contract that tries to insert into the map.

Note There is no need to add the mut keyword when declaring a StorageMap<K, V>. All storage variables are mutable by default.

Accessing Values in a Storage Map

We can get a value out of the storage map by providing its key to the get method, as shown below:

    #[storage(read, write)]
    fn get_from_storage_map() {
        let addr1 = Address::from(0x0101010101010101010101010101010101010101010101010101010101010101);
        let addr2 = Address::from(0x0202020202020202020202020202020202020202020202020202020202020202);

        storage.map.insert(addr1, 42);
        storage.map.insert(addr2, 77);

        let value1 = storage.map.get(addr1).unwrap_or(0);
    }

Here, value1 will have the value that's associated with the first address, and the result will be 42. The get method returns an Option<V>; if there’s no value for that key in the storage map, get will return Option::None. This program handles the Option by calling unwrap_or to set value1 to zero if map doesn't have an entry for the key.

Storage maps with multiple keys

You might find yourself needing a StorageMap<K1, V1> where the type V1 is itself another StorageMap<K2, V2>. This is not allowed in Sway. The right approach is to use a single StorageMap<K, V> where K is a tuple (K1, K2). For example:

    map_two_keys: StorageMap<(b256, bool), b256> = StorageMap {},

Limitations

It is not currently allowed to have a StorageMap<K, V> as a component of a complex type such as a struct or an enum. For example, the code below is not legal:

Struct Wrapper {
    map1: StorageMap<u64, u64>,
    map2: StorageMap<u64, u64>,
}

storage {
    w: Wrapper
}
...

storage.w.map1.insert(..);

Testing

Sway aims to provide facilities for both unit testing and integration testing.

Unit testing refers to "in-language" testing which can be triggered via the forc test command. Sway unit testing is currently a high-priority work-in-progress, you can follow along at this issue.

Integration testing refers to the testing of your Sway project's integration within some wider application. You can add integration testing to your Sway+Rust projects today using the cargo generate template and Rust SDK.

Unit Testing

Forc provides built-in support for building and executing tests for a package.

Tests are written as free functions with the #[test] attribute. For example:

#[test]
fn test_meaning_of_life() {
    assert(6 * 7 == 42);
}

Each test function is ran as if it were the entry point for a script. Tests "pass" if they return successfully, and "fail" if they revert or vice versa while testing failure.

Building and Running Tests

We can build and execute all tests within a package with the following:

forc test

The output should look similar to this:

  Compiled library "core".
  Compiled library "std".
  Compiled library "lib_single_test".
  Bytecode size is 92 bytes.
   Running 1 tests
      test test_meaning_of_life ... ok (170.652µs)
   Result: OK. 1 passed. 0 failed. Finished in 1.564996ms.

Visit the forc test command reference to find the options available for forc test.

Testing Failure

Forc supports testing failing cases for test functions declared with #[test(should_revert)]. For example:

#[test(should_revert)]
fn test_meaning_of_life() {
    assert(6 * 6 == 42);
}

Tests with #[test(should_revert)] considered to be passing if they are reverting.

Calling Contracts

Unit tests can call contract functions an example for such calls can be seen below.

contract;

abi MyContract {
    fn test_function() -> bool;
}

impl MyContract for Contract {
    fn test_function() -> bool {
        true
    }
}

To test the test_function(), a unit test like the following can be written.

#[test]
fn test_success() {
    let caller = abi(MyContract, CONTRACT_ID);
    let result = caller.test_function {}();
    assert(result == true)
}

It is also possible to test failure with contract calls as well.

#[test(should_revert)]
fn test_fail() {
    let caller = abi(MyContract, CONTRACT_ID);
    let result = caller.test_function {}();
    assert(result == false)
}

Note: When running forc test, your contract will be built twice: first without unit tests in order to determine the contract's ID, then a second time with unit tests with the CONTRACT_ID provided to their namespace. This CONTRACT_ID can be used with the abi cast to enable contract calls within unit tests.

Unit tests can call methods of external contracts if those contracts are added as contract dependencies, i.e. in the the contract-dependencies section of the manifest file. An example of such calls is shown below:

contract;

abi CallerContract {
    fn test_false() -> bool;
}

impl CallerContract for Contract {
    fn test_false() -> bool {
        false
    }
}

abi CalleeContract {
    fn test_true() -> bool;
}

#[test]
fn test_multi_contract_calls() {
    let caller = abi(CallerContract, CONTRACT_ID);
    let callee = abi(CalleeContract, callee::CONTRACT_ID);

    let should_be_false = caller.test_false();
    let should_be_true = callee.test_true();
    assert(!should_be_false);
    assert(should_be_true);
}

Example Forc.toml for contract above:

[project]
authors = ["Fuel Labs <contact@fuel.sh>"]
entry = "main.sw"
license = "Apache-2.0"
name = "caller"

[dependencies]
std = { path = "../../../sway-lib-std/" }

[contract-dependencies]
callee = { path = "../callee" }

Testing with Rust

A common use of Sway is for writing contracts or scripts that exist as part of a wider Rust application. In order to test the interaction between our Sway code and our Rust code we can add integration testing.

Adding Rust Integration Testing

To add Rust integration testing to a Forc project we can use the sway-test-rs cargo generate template. This template makes it easy for Sway devs to add the boilerplate required when setting up their Rust integration testing.

Let's add a Rust integration test to the fresh project we created in the introduction.

1. Enter the project

To recap, here's what our empty project looks like:

$ cd my-fuel-project
$ tree .
├── Forc.toml
└── src
    └── main.sw

2. Install cargo generate

We're going to add a Rust integration test harness using a cargo generate template. Let's make sure we have the cargo generate command installed!

cargo install cargo-generate

Note: You can learn more about cargo generate by visiting its repository.

3. Generate the test harness

Let's generate the default test harness with the following:

cargo generate --init fuellabs/sway templates/sway-test-rs --name my-fuel-project --force

--force forces your --name input to retain your desired casing for the {{project-name}} placeholder in the template. Otherwise, cargo-generate automatically converts it to kebab-case. With --force, this means that both my_fuel_project and my-fuel-project are valid project names, depending on your needs.

If all goes well, the output should look as follows:

⚠️   Favorite `fuellabs/sway` not found in config, using it as a git repository: https://github.com/fuellabs/sway
🤷   Project Name : my-fuel-project
🔧   Destination: /home/user/path/to/my-fuel-project ...
🔧   Generating template ...
[1/3]   Done: Cargo.toml
[2/3]   Done: tests/harness.rs
[3/3]   Done: tests
🔧   Moving generated files into: `/home/user/path/to/my-fuel-project`...
✨   Done! New project created /home/user/path/to/my-fuel-project

Let's have a look at the result:

$ tree .
├── Cargo.toml
├── Forc.toml
├── src
│   └── main.sw
└── tests
    └── harness.rs

We have two new files!

  • The Cargo.toml is the manifest for our new test harness and specifies the required dependencies including fuels the Fuel Rust SDK.
  • The tests/harness.rs contains some boilerplate test code to get us started, though doesn't call any contract methods just yet.

4. Build the forc project

Before running the tests, we need to build our contract so that the necessary ABI, storage and bytecode artifacts are available. We can do so with forc build:

$ forc build
  Creating a new `Forc.lock` file. (Cause: lock file did not exist)
    Adding core
    Adding std git+https://github.com/fuellabs/sway?tag=v0.24.5#e695606d8884a18664f6231681333a784e623bc9
   Created new lock file at /home/user/path/to/my-fuel-project/Forc.lock
  Compiled library "core".
  Compiled library "std".
  Compiled contract "my-fuel-project".
  Bytecode size is 60 bytes.

At this point, our project should look like the following:

$ tree
├── Cargo.toml
├── Forc.lock
├── Forc.toml
├── out
│   └── debug
│       ├── my-fuel-project-abi.json
│       ├── my-fuel-project.bin
│       └── my-fuel-project-storage_slots.json
├── src
│   └── main.sw
└── tests
    └── harness.rs

We now have an out directory with our required JSON files!

Note: This step may no longer be required in the future as we plan to enable the integration testing to automatically build the artifacts as necessary so that files like the ABI JSON are always up to date.

5. Build and run the tests

Now we're ready to build and run the default integration test.

$ cargo test
    Updating crates.io index
   Compiling version_check v0.9.4
   Compiling proc-macro2 v1.0.46
   Compiling quote v1.0.21
   ...
   Compiling fuels v0.24.0
   Compiling my-fuel-project v0.1.0 (/home/user/path/to/my-fuel-project)
    Finished test [unoptimized + debuginfo] target(s) in 1m 03s
     Running tests/harness.rs (target/debug/deps/integration_tests-373971ac377845f7)

running 1 test
test can_get_contract_id ... ok

test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.36s

Note: The first time we run cargo test, cargo will spend some time fetching and building the dependencies for Fuel's Rust SDK. This might take a while, but only the first time!

If all went well, we should see some output that looks like the above!

Writing Tests

Now that we've learned how to setup Rust integration testing in our project, let's try to write some of our own tests!

First, let's update our contract code with a simple counter example:

contract;

abi TestContract {
    #[storage(write)]
    fn initialize_counter(value: u64) -> u64;

    #[storage(read, write)]
    fn increment_counter(amount: u64) -> u64;
}

storage {
    counter: u64 = 0,
}

impl TestContract for Contract {
    #[storage(write)]
    fn initialize_counter(value: u64) -> u64 {
        storage.counter = value;
        value
    }

    #[storage(read, write)]
    fn increment_counter(amount: u64) -> u64 {
        let incremented = storage.counter + amount;
        storage.counter = incremented;
        incremented
    }
}

To test our initialize_counter and increment_counter contract methods from the Rust test harness, we could update our tests/harness.rs file with the following:

use fuels::{prelude::*, tx::ContractId};

// Load abi from json
abigen!(TestContract, "out/debug/my-fuel-project-abi.json");

async fn get_contract_instance() -> (TestContract, ContractId) {
    // Launch a local network and deploy the contract
    let mut wallets = launch_custom_provider_and_get_wallets(
        WalletsConfig::new(
            Some(1),             /* Single wallet */
            Some(1),             /* Single coin (UTXO) */
            Some(1_000_000_000), /* Amount per coin */
        ),
        None,
    )
    .await;
    let wallet = wallets.pop().unwrap();

    let id = Contract::deploy(
        "./out/debug/my-fuel-project.bin",
        &wallet,
        TxParameters::default(),
        StorageConfiguration::with_storage_path(Some(
            "./out/debug/my-fuel-project-storage_slots.json".to_string(),
        )),
    )
    .await
    .unwrap();

    let instance = TestContract::new(id.to_string(), wallet);

    (instance, id.into())
}

#[tokio::test]
async fn initialize_and_increment() {
    let (contract_instance, _id) = get_contract_instance().await;
    // Now you have an instance of your contract you can use to test each function

    let result = contract_instance
        .methods()
        .initialize_counter(42)
        .call()
        .await
        .unwrap();

    assert_eq!(42, result.value);

    // Call `increment_counter()` method in our deployed contract.
    let result = contract_instance
        .methods()
        .increment_counter(10)
        .call()
        .await
        .unwrap();

    assert_eq!(52, result.value);
}

Let's build our project once more and run the test:

forc build
$ cargo test
   Compiling my-fuel-project v0.1.0 (/home/mindtree/programming/sway/my-fuel-project)
    Finished test [unoptimized + debuginfo] target(s) in 11.61s
     Running tests/harness.rs (target/debug/deps/integration_tests-373971ac377845f7)

running 1 test
test initialize_and_increment ... ok

test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 1.25s

When cargo runs our test, our test uses the SDK to spin up a local in-memory Fuel network, deploy our contract to it, and call the contract methods via the ABI.

You can add as many functions decorated with #[tokio::test] as you like, and cargo test will automatically test each of them!

Application Frontend Development

TypeScript SDK

The TypeScript SDK supports common tasks like:

  • Deploying and calling contracts
  • Generating contract types with TypeChain
  • Building and sending transactions
  • Encoding and decoding contract ABI

Refer here for full documentation.

Sway Reference

Compiler Intrinsics

The Sway compiler supports a list of intrinsics that perform various low level operations that are useful for building libraries. Compiler intrinsics should rarely be used but are preferred over asm blocks because they are type-checked and are safer overall. Below is a list of all available compiler intrinsics:


__size_of_val<T>(val: T) -> u64

Description: Return the size of type T in bytes.

Constraints: None.


__size_of<T>() -> u64

Description: Return the size of type T in bytes.

Constraints: None.


__is_reference_type<T>() -> bool

Description: Returns true if T is a reference type and false otherwise.

Constraints: None.


__eq<T>(lhs: T, rhs: T) -> bool

Description: Returns whether lhs and rhs are equal.

Constraints: T is bool, u8, u16, u32, u64, or raw_ptr.


__gt<T>(lhs: T, rhs: T) -> bool

Description: Returns whether lhs is greater than rhs.

Constraints: T is u8, u16, u32, u64.


__lt<T>(lhs: T, rhs: T) -> bool

Description: Returns whether lhs is less than rhs.

Constraints: T is u8, u16, u32, u64.


__gtf<T>(index: u64, tx_field_id: u64) -> T

Description: Returns transaction field with ID tx_field_id at index index, if applicable. This is a wrapper around FuelVM's gtf instruction. The resuting field is cast to T.

Constraints: None.


__addr_of<T>(val: T) -> raw_ptr

Description: Returns the address in memory where val is stored.

Constraints: T is a reference type.


__state_load_word(key: b256) -> u64

Description: Reads and returns a single word from storage at key key.

Constraints: None.


__state_load_quad(key: b256, ptr: raw_ptr, slots: u64) -> bool

Description: Reads slots number of slots (b256 each) from storage starting at key key and stores them in memory starting at address ptr. Returns a Boolean describing whether all the storage slots were previously set.

Constraints: None.


__state_store_word(key: b256, val: u64) -> bool

Description: Stores a single word val into storage at key key. Returns a Boolean describing whether the store slot was previously set.

Constraints: None.


__state_store_quad(key: b256, ptr: raw_ptr, slots: u64) -> bool

Description: Stores slots number of slots (b256 each) starting at address ptr in memory into storage starting at key key. Returns a Boolean describing whether the first storage slot was previously set.

Constraints: None.


__log<T>(val: T)

Description: Logs value val.

Constraints: None.


__add<T>(lhs: T, rhs: T) -> T

Description: Adds lhs and rhs and returns the result.

Constraints: T is an integer type, i.e. u8, u16, u32, u64.


__sub<T>(lhs: T, rhs: T) -> T

Description: Subtracts rhs from lhs.

Constraints: T is an integer type, i.e. u8, u16, u32, u64.


__mul<T>(lhs: T, rhs: T) -> T

Description: Multiplies lhs by rhs.

Constraints: T is an integer type, i.e. u8, u16, u32, u64.


__div<T>(lhs: T, rhs: T) -> T

Description: Divides lhs by rhs.

Constraints: T is an integer type, i.e. u8, u16, u32, u64.


__and<T>(lhs: T, rhs: T) -> T

Description: Bitwise AND lhs and rhs.

Constraints: T is an integer type, i.e. u8, u16, u32, u64.


or<T>(lhs: T, rhs: T) -> T

Description: Bitwise OR lhs and rhs.

Constraints: T is an integer type, i.e. u8, u16, u32, u64.


__xor<T>(lhs: T, rhs: T) -> T

Description: Bitwise XOR lhs and rhs.

Constraints: T is an integer type, i.e. u8, u16, u32, u64.


__revert(code: u64)

Description: Reverts with error code code.

Constraints: None.


__ptr_add(ptr: raw_ptr, offset: u64)

Description: Adds offset to the raw value of pointer ptr.

Constraints: None.


__ptr_sub(ptr: raw_ptr, offset: u64)

Description: Subtracts offset to the raw value of pointer ptr.

Constraints: None.


__smo<T>(recipient: b256, data: T, output_index: u64, coins: u64)

Description: Sends a message data of arbitrary type T and coins amount of the base asset to address recipient. This intrinsic assumes that an OutputMessage is available at index output_index.

Constraints: None.


Attributes

The Sway compiler supports a list of attributes that perform various operations that are useful for building, testing and documenting Sway programs. Below is a list of all available attributes:

Allow

The #[allow(dead_code)] attribute overrides the check for dead code so that violations will go unreported.

Doc

The #[doc(..)] attribute specifies documentation.

Line doc comments beginning with exactly three slashes ///, are interpreted as a special syntax for doc attributes. That is, they are equivalent to writing #[doc("...")] around the body of the comment, i.e., /// Foo turns into #[doc("Foo")]

Line comments beginning with //! are doc comments that apply to the module of the source file they are in. That is, they are equivalent to writing #![doc("...")] around the body of the comment. //! module level doc comments should be at the top of Sway files.

Documentation can be generated from doc attributes using forc doc.

Inline

The inline attribute suggests that a copy of the attributed function should be placed in the caller, rather than generating code to call the function where it is defined.

Note: The Sway compiler automatically inlines functions based on internal heuristics. Incorrectly inlining functions can make the program slower, so this attribute should be used with care.

The #[inline(never)] attribute suggests that an inline expansion should never be performed.

The #[inline(always)] attribute suggests that an inline expansion should always be performed.

Note: #[inline(..)] in every form is a hint, with no requirements on the language to place a copy of the attributed function in the caller.

Payable

The lack of #[payable] implies the method is non-payable. When calling an ABI method that is non-payable, the compiler emits an error if the amount of coins forwarded with the call is not guaranteed to be zero. Note that this is strictly a compile-time check and does not incur any runtime cost.

Storage

In Sway, functions are pure by default but can be opted into impurity via the storage function attribute. The storage attribute may take read and/or write arguments indicating which type of access the function requires.

The #[storage(read)] attribute indicates that a function requires read access to the storage.

The #[storage(write)] attribute indicates that a function requires write access to the storage.

More details in Purity.

Test

The #[test] attribute marks a function to be executed as a test.

The #[test(should_revert)] attribute marks a function to be executed as a test that should revert.

More details in Unit Testing.

Style Guide

Capitalization

In Sway, structs, traits, and enums are CapitalCase. Modules, variables, and functions are snake_case, constants are SCREAMING_SNAKE_CASE. The compiler will warn you if your capitalization is ever unidiomatic.

Known Issues and Workarounds

Known Issues

  • #870: All impl blocks need to be defined before any of the functions they define can be called. This includes sibling functions in the same impl declaration, i.e., functions in an impl can't call each other yet.

Missing Features

  • #1182 Arrays in a storage block are not yet supported. See the Manual Storage Management section for details on how to use store and get from the standard library to manage storage slots directly. Note, however, that StorageMap<K, V> does support arbitrary types for K and V without any limitations.

  • #1796: It is not yet allowed to use StorageMap<K, V> as a component of a complex type such as a struct or an enum.

General

  • No compiler optimization passes have been implemented yet, therefore bytecode will be more expensive and larger than it would be in production. Note that eventually the optimizer will support zero-cost abstractions, avoiding the need for developers to go down to inline assembly to produce optimal code.

Differences From Solidity

This page outlines some of the critical differences between Sway and Solidity, and between the FuelVM and the EVM.

Underlying Virtual Machine

The underlying virtual machine targeted by Sway is the FuelVM, specified here. Solidity targets the Ethereum Virtual Machine (EVM), specified here.

Word Size

Words in the FuelVM are 64 bits (8 bytes), rather than the 256 bits (32 bytes) of the EVM. Therefore, primitive integers only go up to u64, and hashes (the b256 type) are not in registers but rather in memory. A b256 is therefore a pointer to a 32-byte memory region containing the hash value.

Unsigned Integers Only

Only unsigned integers are provided as primitives: u8, u16, u32, and u64. Signed integer arithmetic is not available in the FuelVM. Signed integers and signed integer arithmetic can be implemented in high-level libraries if needed.

Global Revert

Panics in the FuelVM (called "reverts" in Solidity and the EVM) are global, i.e. they cannot be caught. A panic will completely and unconditionally revert the stateful effects of a transaction, minus gas used.

Default Safe Math

Math in the FuelVM is by default safe (i.e. any overflow or exception is a panic). Safety checks are performed natively in the VM implementation, rather than at the bytecode level like Solidity's default safe math.

No* Code Size Limit

There is no practical code size limit to Sway contracts. The physical limit is governed by the VM_MAX_RAM VM parameter, which at the time of writing is 64 MiB.

Differences From Rust

Sway shares a lot with Rust, especially its syntax. Because they are so similar, you may be surprised or caught off guard when they differ. This page serves to outline, from a high level, some of the syntactic gotchas that you may encounter.

Enum Variant Syntax

In Rust, enums generally take one of three forms: unit variants, which have no inner data, struct variants, which contain named fields, and tuple variants, which contain within them a tuple of data. If you are unfamiliar with these terms, this is what they look like:

// note to those skimming the docs: this is Rust syntax! Not Sway! Don't copy/paste this into a Sway program.

enum Foo {
    UnitVariant,
    TupleVariant(u32, u64, bool),
    StructVariant {
        field_one: bool,
        field_two: bool
    }
}

In Sway, enums are simplified. Enums variants must all specify exactly one type. This type represents their interior data. This is actually isomorphic to what Rust offers, just with a different syntax. I'll now rewrite the above enum but with Sway syntax:

// This is equivalent Sway syntax for the above Rust enum.
enum Foo {
    UnitVariant: (),
    TupleVariant: (u32, u64, bool),
    StructVariant: MyStruct,
}

struct MyStruct {
    field_one: bool,
    field_two: bool,
}

Contributing To Sway

Thanks for your interest in contributing to Sway! This document outlines the process for installing and setting up the Sway toolchain for development, as well as some conventions on contributing to Sway.

If you run into any difficulties getting started, you can always ask questions on our Discourse.

Building and setting up a development workspace

See the introduction section for instructions on installing and setting up the Sway toolchain.

Getting the repository

  1. Visit the Sway repo and fork the project.
  2. Then clone your forked copy to your local machine and get to work.
git clone https://github.com/FuelLabs/sway
cd sway

Building and testing

The following steps will run the sway test suite and ensure that everything is set up correctly.

First, open a new terminal and start fuel-core with:

fuel-core

Then open a second terminal, cd into the sway repo and run:

cargo run --bin test

After the test suite runs, you should see:

Tests passed.
_n_ tests run (0 skipped)

Congratulations! You've now got everything setup and are ready to start making contributions.

Finding something to work on

There are many ways in which you may contribute to the Sway project, some of which involve coding knowledge and some which do not. A few examples include:

  • Reporting bugs
  • Adding documentation to the Sway book
  • Adding new features or bugfixes for which there is already an open issue
  • Making feature requests

Check out our Help Wanted, Sway Book or Good First Issue issues to find a suitable task.

If you are planning something big, for example, related to multiple components or changes current behaviors, make sure to open an issue to discuss with us before starting on the implementation.

Contribution flow

This is a rough outline of what a contributor's workflow looks like:

  • Make sure what you want to contribute is already tracked as an issue.
    • We may discuss the problem and solution in the issue.
  • Create a Git branch from where you want to base your work. This is usually master.
  • Write code, add test cases, and commit your work.
  • Run tests and make sure all tests pass.
  • If the PR contains any breaking changes, add the breaking label to your PR.
  • Push your changes to a branch in your fork of the repository and submit a pull request.
    • Make sure mention the issue, which is created at step 1, in the commit message.
  • Your PR will be reviewed and some changes may be requested.
    • Once you've made changes, your PR must be re-reviewed and approved.
    • If the PR becomes out of date, you can use GitHub's 'update branch' button.
    • If there are conflicts, you can merge and resolve them locally. Then push to your PR branch. Any changes to the branch will require a re-review.
  • Our CI system (Github Actions) automatically tests all authorized pull requests.
  • Use Github to merge the PR once approved.

Thanks for your contributions!

Linking issues

Pull requests should be linked to at least one issue in the same repo.

If the pull request resolves the relevant issues, and you want GitHub to close these issues automatically after it merged into the default branch, you can use the syntax (KEYWORD #ISSUE-NUMBER) like this:

close #123

If the pull request links an issue but does not close it, you can use the keyword ref like this:

ref #456

Multiple issues should use full syntax for each issue and separate by a comma, like:

close #123, ref #456

Forc Reference

Forc stands for Fuel Orchestrator. Forc provides a variety of tools and commands for developers working with the Fuel ecosystem, such as scaffolding a new project, formatting, running scripts, deploying contracts, testing contracts, and more. If you're coming from a Rust background, forc is similar to cargo.

If you are new to Forc, see the Forc Project introduction section.

For a comprehensive overview of the Forc CLI commands, see the Commands section.

Manifest Reference

The Forc.toml (the manifest file) is a compulsory file for each package and it is written in [TOML] format. Forc.toml consists of the following fields:

  • [project] — Defines a sway project.

    • name — The name of the project.
    • authors — The authors of the project.
    • organization — The organization of the project.
    • license— The project license.
    • entry — The entry point for the compiler to start parsing from.
      • For the recomended way of selecting an entry point of large libraries please take a look at: Libraries
    • implicit-std - Controls whether provided std version (with the current forc version) will get added as a dependency implicitly. Unless you know what you are doing, leave this as default.
    • forc-version - The minimum forc version required for this project to work properly.
  • [dependencies] — Defines the dependencies.

  • [network] — Defines a network for forc to interact with.

    • url — URL of the network.
  • [build-profiles] - Defines the build profiles.

  • [patch] - Defines the patches.

  • [contract-dependencies] - Defines the contract dependencies.

The [project] section

An example Forc.toml is shown below. Under [project] the following fields are optional:

  • authors
  • organization

Also for the following fields, a default value is provided so omitting them is allowed:

  • entry - (default : main.sw)
  • implicit-std - (default : true)
[project]
authors = ["user"]
entry = "main.sw"
organization = "Fuel_Labs"
license = "Apache-2.0"
name = "wallet_contract"

The [dependencies] section

The following fields can be provided with a dependency:

  • version - Desired version of the dependency
  • path - The path of the dependency (if it is local)
  • git - The URL of the git repo hosting the dependency
  • branch - The desired branch to fetch from the git repo
  • tag - The desired tag to fetch from the git repo
  • rev - The desired rev (i.e. commit hash) reference

Please see dependencies for details

The [network] section

For the following fields, a default value is provided so omitting them is allowed:

The [build-profiles-*] section

The [build-profiles] tables provide a way to customize compiler settings such as debug options.

The following fields needs to be provided for a build-profile:

  • print-ast - Whether to print out the generated AST (true) or not (false).
  • print-dca-graph - Whether to print out the computed DCA graph (in GraphViz DOT format).
  • print-finalized-asm - Whether to compile to bytecode (false) or to print out the generated ASM (true).
  • print-intermediate-asm - Whether to compile to bytecode (false) or to print out the generated ASM (true).
  • print-ir - Whether to compile to bytecode (false) or to print out the generated IR (true).
  • terse-mode - Terse mode. Limited warning and error output.

There are two default [build-profile] available with every manifest file. These are debug and release profiles. If you want to override these profiles, you can provide them explicitly in the manifest file like the following example:

[project]
authors = ["user"]
entry = "main.sw"
organization = "Fuel_Labs"
license = "Apache-2.0"
name = "wallet_contract"

[build-profiles.debug]
print-finalized-asm = false
print-intermediate-asm = false
print-ir = false
terse = false

[build-profiles.release]
print-finalized-asm = false 
print-intermediate-asm = false
print-ir = false
terse = true

Since release and debug implicitly included in every manifest file, you can use them by just passing --release or by not passing anything (debug is default). For using a user defined build profile there is --build-profile <profile name> option available to the relevant commands. (For an example see forc-build)

Note that providing the corresponding cli options (like --print-finalized-asm) will override the selected build profile. For example if you pass both --release and --print-finalized-asm, release build profile is omitted and resulting build profile would have a structure like the following:

  • print-finalized-asm - true
  • print-intermediate-asm - false
  • print-ir - false
  • terse - false

The [patch] section

The [patch] section of Forc.toml can be used to override dependencies with other copies. The example provided below patches https://github.com/fuellabs/sway source with master branch of the same repo.

[project]
authors = ["user"]
entry = "main.sw"
organization = "Fuel_Labs"
license = "Apache-2.0"
name = "wallet_contract"

[dependencies]

[patch.'https://github.com/fuellabs/sway']
std = { git = "https://github.com/fuellabs/sway", branch = "test" }

In the example above, std is patched with the test branch from std repo. You can also patch git dependencies with dependencies defined with a path.

[patch.'https://github.com/fuellabs/sway']
std = { path = "/path/to/local_std_version" }

Just like std or core you can also patch dependencies you declared with a git repo.

[project]
authors = ["user"]
entry = "main.sw"
organization = "Fuel_Labs"
license = "Apache-2.0"
name = "wallet_contract"

[dependencies]
foo = { git = "https://github.com/foo/foo", branch = "master" }

[patch.'https://github.com/foo']
foo = { git = "https://github.com/foo/foo", branch = "test" }

Note that each key after the [patch] is a URL of the source that is being patched.

The [contract-dependencies] section

The [contract-dependenices] table can be used to declare contract dependencies for a Sway contract or script. Contract dependencies are the set of contracts that our contract or script may interact with. Declaring [contract-dependencies] makes it easier to refer to contracts in your Sway source code without having to manually update IDs each time a new version is deployed. Instead, we can use forc to pin and update contract dependencies just like we do for regular library dependencies.

Contracts declared under [contract-dependencies] are built and pinned just like regular [dependencies] however rather than importing each contract dependency's entire public namespace we instead import their respective contract IDs as CONTRACT_ID constants available via each contract dependency's namespace root. This means you can use a contract dependency's ID as if it were declared as a pub const in the root of the contract dependency package as demonstrated in the example below.

Entries under [contract-dependencies] can be declared in the same way that [dependencies] can be declared. That is, they can refer to the path or git source of another contract. Note that entries under [contract-dependencies] must refer to contracts and will otherwise produce an error.

Example Forc.toml:

[project]
authors = ["user"]
entry = "main.sw"
organization = "Fuel_Labs"
license = "Apache-2.0"
name = "wallet_contract"

[contract-dependencies]
foo = { path = "../foo" }

Example usage:

script;

fn main() {
  let foo_id = foo::CONTRACT_ID;
}

Because the ID of a contract is computed deterministically, rebuilding the same contract would always result in the same contract ID. Since two contracts with same contract ID cannot be deployed on the blockchain, a "salt" factor is needed to modify the contract ID. For each contract dependency declared under [contract-dependencies], salt can be specified. An example is shown below:

[contract-dependencies]
foo = { path = "../foo", salt = "0x1000000000000000000000000000000000000000000000000000000000000000" }

For contract dependencies that do not specify any value for salt, a default of all zeros for salt is implicitly applied.

Workspaces

A workspace is a collection of one or more packages, namely workspace members, that are managed together.

The key points for workspaces are:

  • Common forc commands available for a single package can also be used for a workspace, like forc build or forc deploy.
  • All packages share a common Forc.lock file which resides in the root directory of the workspace.

Workspace manifests are declared within Forc.toml files and support the following fields:

  • members - Packages to include in the workspace.

An empty workspace can be created with forc new --workspace or forc init --workspace.

The members field

The members field defines which packages are members of the workspace:

[workspace]
members = ["member1", "path/to/member2"]

The members field accepts entries to be given in relative path with respect to the workspace root. Packages that are located within a workspace directory but are not contained within the members set are ignored.

Some forc commands that support workspaces

  • forc build - Builds an entire workspace.
  • forc deploy - Builds and deploys all deployable members (i.e, contracts) of the workspace in the correct order.
  • forc run - Builds and runs all scripts of the workspace.
  • forc check - Checks all members of the workspace.
  • forc update - Checks and updates workspace level Forc.lock file that is shared between workspace members.
  • forc clean - Cleans all output artifacts for each member of the workspace.
  • forc fmt - Formats all members of a workspace.

Dependencies

Forc has a dependency management system which can pull packages using git. This allows users to build and share Forc libraries.

Adding a dependency

If your Forc.toml doesn't already have a [dependencies] table, add one. Below, list the package name alongside its source. Currently, forc supports both git and path sources.

If a git source is specified, forc will fetch the git repository at the given URL and then search for a Forc.toml for a package with the given name anywhere inside the git repository.

The following example adds a library dependency named custom_lib. For git dependencies you may optionally specify a branch, tag, or rev (i.e. commit hash) reference.

[dependencies]
custom_lib = { git = "https://github.com/FuelLabs/custom_lib", branch = "master" }
# custom_lib = { git = "https://github.com/FuelLabs/custom_lib", tag = "v0.0.1" }
# custom_lib = { git = "https://github.com/FuelLabs/custom_lib", rev = "87f80bdf323e2d64e213895d0a639ad468f4deff" }

Depending on a local library using path:

[dependencies]
custom_lib = { path = "../custom_lib" }

Once the package is added, running forc build will automatically download added dependencies.

Updating dependencies

To update dependencies in your Forc directory you can run forc update. For path dependencies this will have no effect. For git dependencies with a branch reference, this will update the project to use the latest commit for the given branch.

Here are a list of commands available to forc:

forc-addr2line

Show location and context of an opcode address in its source file

USAGE:

forc addr2line [OPTIONS] --sourcemap-path <SOURCEMAP_PATH> --opcode-index <OPCODE_INDEX>

OPTIONS:

-c, --context <CONTEXT>

How many lines of context to show [default: 2]

-g, --sourcemap-path <SOURCEMAP_PATH>

Source file mapping in JSON format

-h, --help

Print help information

-i, --opcode-index <OPCODE_INDEX>

Opcode index

-L, --log-level <LOG_LEVEL>

Set the log level

-s, --silent

Silence all output

-S, --search-dir <SEARCH_DIR>

Where to search for the project root [default: .]

-v, --verbose

Use verbose output

forc-build

Compile the current or target project.

The output produced will depend on the project's program type.

  • script, predicate and contract projects will produce their bytecode in binary format <project-name>.bin.

  • script projects will also produce a file containing the hash of the bytecode binary <project-name>-bin-hash (using fuel_cypto::Hasher).

  • predicate projects will also produce a file containing the root hash of the bytecode binary <project-name>-bin-root (using fuel_tx::Contract::root_from_code).

  • contract and library projects will also produce the public ABI in JSON format <project-name>-abi.json.

USAGE:

forc build [OPTIONS]

OPTIONS:

--ast

Print the generated Sway AST (Abstract Syntax Tree)

--build-profile <BUILD_PROFILE>

Name of the build profile to use.

If unspecified, forc will use debug build profile.

--build-target <BUILD_TARGET>

Build target to use for code generation

[default: fuel] [possible values: fuel, evm, midenvm]

--dca-graph

Print the computed Sway DCA graph

--error-on-warnings

Treat warnings as errors

--finalized-asm

Print the finalized ASM.

This is the state of the ASM with registers allocated and optimisations applied.

-g, --output-debug <DEBUG_FILE>

If set, outputs source file mapping in JSON format

-h, --help

Print help information

--intermediate-asm

Print the generated ASM.

This is the state of the ASM prior to performing register allocation and other ASM optimisations.

--ir

Print the generated Sway IR (Intermediate Representation)

--json-abi

By default the JSON for ABIs is formatted for human readability. By using this option JSON output will be "minified", i.e. all on one line without whitespace

--json-abi-with-callpaths

Outputs json abi with callpaths instead of names for struct and enums

--json-storage-slots

By default the JSON for initial storage slots is formatted for human readability. By using this option JSON output will be "minified", i.e. all on one line without whitespace

-L, --log-level <LOG_LEVEL>

Set the log level

--locked

Requires that the Forc.lock file is up-to-date. If the lock file is missing, or it needs to be updated, Forc will exit with an error

-o, --output-bin <BIN_FILE>

If set, outputs a binary file representing the script bytes

--offline

Offline mode, prevents Forc from using the network when managing dependencies. Meaning it will only try to use previously downloaded dependencies

--output-directory <OUTPUT_DIRECTORY>

The directory in which the sway compiler output artifacts are placed.

By default, this is <project-root>/out.

-p, --path <PATH>

Path to the project, if not specified, current working directory will be used

--release

Use release build plan. If a custom release plan is not specified, it is implicitly added to the manifest file.

If --build-profile is also provided, forc omits this flag and uses provided build-profile.

-s, --silent

Silence all output

-t, --terse

Terse mode. Limited warning and error output

--tests

Also build all tests within the project

--time-phases

Output the time elapsed over each part of the compilation process

-v, --verbose

Use verbose output

EXAMPLE

Compile the sway files of the current project.

$ forc build
Compiled script "my-fuel-project".
Bytecode size is 28 bytes.

The output produced will depend on the project's program type. Building script, predicate and contract projects will produce their bytecode in binary format <project-name>.bin. Building contracts and libraries will also produce the public ABI in JSON format <project-name>-abi.json.

By default, these artifacts are placed in the out/ directory.

If a Forc.lock file did not yet exist, it will be created in order to pin each of the dependencies listed in Forc.toml to a specific commit or version.

forc-check

Check the current or target project and all of its dependencies for errors.

This will essentially compile the packages without performing the final step of code generation, which is faster than running forc build.

USAGE:

forc check [OPTIONS] [BUILD_TARGET]

ARGS:

<BUILD_TARGET> Build target to use for code generation

[default: fuel] [possible values: fuel, evm, midenvm]

OPTIONS:

--disable-tests

Disable checking unit tests

-h, --help

Print help information

-L, --log-level <LOG_LEVEL>

Set the log level

--locked

Requires that the Forc.lock file is up-to-date. If the lock file is missing, or it needs to be updated, Forc will exit with an error

--offline

Offline mode, prevents Forc from using the network when managing dependencies. Meaning it will only try to use previously downloaded dependencies

-p, --path <PATH>

Path to the project, if not specified, current working directory will be used

-s, --silent

Silence all output

-t, --terse

Terse mode. Limited warning and error output

-v, --verbose

Use verbose output

forc-clean

Removes the default forc compiler output artifact directory, i.e. <project-name>/out

USAGE:

forc clean [OPTIONS]

OPTIONS:

-h, --help

Print help information

-L, --log-level <LOG_LEVEL>

Set the log level

-p, --path <PATH>

Path to the project, if not specified, current working directory will be used

-s, --silent

Silence all output

-v, --verbose

Use verbose output

forc-completions

Generate tab-completion scripts for your shell

USAGE:

forc completions [OPTIONS] --shell

OPTIONS:

-h, --help

Print help information

-L, --log-level <LOG_LEVEL>

Set the log level

-s, --silent

Silence all output

-S, --shell <SHELL>

Specify shell to enable tab-completion for

[possible values: zsh, bash, fish, powershell, elvish]

For more info: https://fuellabs.github.io/sway/latest/forc/commands/forc_completions.html

-v, --verbose

Use verbose output

DISCUSSION

Enable tab completion for Bash, Fish, Zsh, or PowerShell The script is output on stdout, allowing one to re-direct the output to the file of their choosing. Where you place the file will depend on which shell, and which operating system you are using. Your particular configuration may also determine where these scripts need to be placed.

Here are some common set ups for the three supported shells under Unix and similar operating systems (such as GNU/Linux).

BASH

Completion files are commonly stored in /etc/bash_completion.d/ for system-wide commands, but can be stored in ~/.local/share/bash-completion/completions for user-specific commands. Run the command:

mkdir -p ~/.local/share/bash-completion/completions
forc completions --shell=bash >> ~/.local/share/bash-completion/completions/forc

This installs the completion script. You may have to log out and log back in to your shell session for the changes to take effect.

BASH (macOS/Homebrew)

Homebrew stores bash completion files within the Homebrew directory. With the bash-completion brew formula installed, run the command:

mkdir -p $(brew --prefix)/etc/bash_completion.d
forc completions --shell=bash > $(brew --prefix)/etc/bash_completion.d/forc.bash-completion

FISH

Fish completion files are commonly stored in $HOME/.config/fish/completions. Run the command:

mkdir -p ~/.config/fish/completions
forc completions --shell=fish > ~/.config/fish/completions/forc.fish

This installs the completion script. You may have to log out and log back in to your shell session for the changes to take effect.

ZSH

ZSH completions are commonly stored in any directory listed in your $fpath variable. To use these completions, you must either add the generated script to one of those directories, or add your own to this list.

Adding a custom directory is often the safest bet if you are unsure of which directory to use. First create the directory; for this example we'll create a hidden directory inside our $HOME directory:

mkdir ~/.zfunc

Then add the following lines to your .zshrc just before compinit:

fpath+=~/.zfunc

Now you can install the completions script using the following command:

forc completions --shell=zsh > ~/.zfunc/_forc

You must then either log out and log back in, or simply run

exec zsh

for the new completions to take effect.

CUSTOM LOCATIONS

Alternatively, you could save these files to the place of your choosing, such as a custom directory inside your $HOME. Doing so will require you to add the proper directives, such as sourceing inside your login script. Consult your shells documentation for how to add such directives.

POWERSHELL

The powershell completion scripts require PowerShell v5.0+ (which comes with Windows 10, but can be downloaded separately for windows 7 or 8.1).

First, check if a profile has already been set

Test-Path $profile

If the above command returns False run the following

New-Item -path $profile -type file -force

Now open the file provided by $profile (if you used the New-Item command it will be ${env:USERPROFILE}\Documents\WindowsPowerShell\Microsoft.PowerShell_profile.ps1

Next, we either save the completions file into our profile, or into a separate file and source it inside our profile. To save the completions into our profile simply use

forc completions --shell=powershell >> ${env:USERPROFILE}\Documents\WindowsPowerShell\Microsoft.PowerShell_profile.ps1

forc-contract-id

Determine contract-id for a contract. For workspaces outputs all contract ids in the workspace

USAGE:

forc contract-id [OPTIONS]

OPTIONS:

--ast

Print the generated Sway AST (Abstract Syntax Tree)

--build-profile <BUILD_PROFILE>

Name of the build profile to use.

If unspecified, forc will use debug build profile.

--dca-graph

Print the computed Sway DCA graph

--error-on-warnings

Treat warnings as errors

--finalized-asm

Print the finalized ASM.

This is the state of the ASM with registers allocated and optimisations applied.

-g, --output-debug <DEBUG_FILE>

If set, outputs source file mapping in JSON format

-h, --help

Print help information

--intermediate-asm

Print the generated ASM.

This is the state of the ASM prior to performing register allocation and other ASM optimisations.

--ir

Print the generated Sway IR (Intermediate Representation)

--json-abi

By default the JSON for ABIs is formatted for human readability. By using this option JSON output will be "minified", i.e. all on one line without whitespace

--json-abi-with-callpaths

Outputs json abi with callpaths instead of names for struct and enums

--json-storage-slots

By default the JSON for initial storage slots is formatted for human readability. By using this option JSON output will be "minified", i.e. all on one line without whitespace

-L, --log-level <LOG_LEVEL>

Set the log level

--locked

Requires that the Forc.lock file is up-to-date. If the lock file is missing, or it needs to be updated, Forc will exit with an error

-o, --output-bin <BIN_FILE>

If set, outputs a binary file representing the script bytes

--offline

Offline mode, prevents Forc from using the network when managing dependencies. Meaning it will only try to use previously downloaded dependencies

--output-directory <OUTPUT_DIRECTORY>

The directory in which the sway compiler output artifacts are placed.

By default, this is <project-root>/out.

-p, --path <PATH>

Path to the project, if not specified, current working directory will be used

--release

Use release build plan. If a custom release plan is not specified, it is implicitly added to the manifest file.

If --build-profile is also provided, forc omits this flag and uses provided build-profile.

-s, --silent

Silence all output

--salt <SALT>

Added salt used to derive the contract ID.

By default, this is 0x0000000000000000000000000000000000000000000000000000000000000000.

-t, --terse

Terse mode. Limited warning and error output

--time-phases

Output the time elapsed over each part of the compilation process

-v, --verbose

Use verbose output

forc-init

Create a new Forc project in an existing directory

USAGE:

forc init [OPTIONS]

OPTIONS:

--contract

The default program type, excluding all flags or adding this flag creates a basic contract program

-h, --help

Print help information

-L, --log-level <LOG_LEVEL>

Set the log level

--library

Create a package with a library target (src/lib.sw)

--name <NAME>

Set the package name. Defaults to the directory name

--path <PATH>

The directory in which the forc project will be initialized

--predicate

Create a package with a predicate target (src/predicate.rs)

-s, --silent

Silence all output

--script

Create a package with a script target (src/main.sw)

-v, --verbose

Use verbose output

--workspace

Adding this flag creates an empty workspace

EXAMPLE

$ mkdir my-fuel-project
$ cd my-fuel-project
$ forc init
$ tree
.
├── Forc.toml
└── src
    └── main.sw

Forc.toml is the Forc manifest file, containing information about the project and dependencies.

A src/ directory is created, with a single main.sw Sway file in it.

forc-new

Create a new Forc project at <path>

USAGE:

forc new [OPTIONS]

ARGS:

<PATH>

The path at which the project directory will be created

OPTIONS:

--contract

The default program type. Excluding all flags or adding this flag creates a basic contract program

-h, --help

Print help information

-L, --log-level <LOG_LEVEL>

Set the log level

--library

Adding this flag creates an empty library program

--name <NAME>

Set the package name. Defaults to the directory name

--predicate

Adding this flag creates an empty predicate program

-s, --silent

Silence all output

--script

Adding this flag creates an empty script program

-v, --verbose

Use verbose output

--workspace

Adding this flag creates an empty workspace

EXAMPLE

$ forc new my-fuel-project
$ cd my-fuel-project
$ tree
.
├── Forc.toml
└── src
    └── main.sw

Forc.toml is the Forc manifest file, containing information about the project and dependencies.

A src/ directory is created, with a single main.sw Sway file in it.

forc-parse-bytecode

Parse bytecode file into a debug format

USAGE:

forc parse-bytecode [OPTIONS] <FILE_PATH>

ARGS:

<FILE_PATH>

OPTIONS:

-h, --help

Print help information

-L, --log-level <LOG_LEVEL>

Set the log level

-s, --silent

Silence all output

-v, --verbose

Use verbose output

EXAMPLE

We can try this command with the initial project created using forc init, with the counter template:

forc new --template counter counter
cd counter
forc build -o obj
counter$ forc parse-bytecode obj

  half-word   byte   op                   raw           notes
          0   0      JI(4)                90 00 00 04   conditionally jumps to byte 16
          1   4      NOOP                 47 00 00 00
          2   8      Undefined            00 00 00 00   data section offset lo (0)
          3   12     Undefined            00 00 00 c8   data section offset hi (200)
          4   16     LW(63, 12, 1)        5d fc c0 01
          5   20     ADD(63, 63, 12)      10 ff f3 00
         ...
         ...
         ...
         60   240    Undefined            00 00 00 00
         61   244    Undefined            fa f9 0d d3
         62   248    Undefined            00 00 00 00
         63   252    Undefined            00 00 00 c8

forc-plugins

Find all forc plugins available via PATH.

Prints information about each discovered plugin.

USAGE:

forc plugins [OPTIONS]

OPTIONS:

-d, --describe

Prints the long description associated with each listed plugin

-h, --help

Print help information

-L, --log-level <LOG_LEVEL>

Set the log level

-p, --paths

Prints the absolute path to each discovered plugin

-s, --silent

Silence all output

-v, --verbose

Use verbose output

forc-predicate-root

Determine predicate-root for a predicate. For workspaces outputs all predicate roots in the workspace

USAGE:

forc predicate-root [OPTIONS]

OPTIONS:

--ast

Print the generated Sway AST (Abstract Syntax Tree)

--build-profile <BUILD_PROFILE>

Name of the build profile to use.

If unspecified, forc will use debug build profile.

--dca-graph

Print the computed Sway DCA graph

--error-on-warnings

Treat warnings as errors

--finalized-asm

Print the finalized ASM.

This is the state of the ASM with registers allocated and optimisations applied.

-g, --output-debug <DEBUG_FILE>

If set, outputs source file mapping in JSON format

-h, --help

Print help information

--intermediate-asm

Print the generated ASM.

This is the state of the ASM prior to performing register allocation and other ASM optimisations.

--ir

Print the generated Sway IR (Intermediate Representation)

--json-abi

By default the JSON for ABIs is formatted for human readability. By using this option JSON output will be "minified", i.e. all on one line without whitespace

--json-abi-with-callpaths

Outputs json abi with callpaths instead of names for struct and enums

--json-storage-slots

By default the JSON for initial storage slots is formatted for human readability. By using this option JSON output will be "minified", i.e. all on one line without whitespace

-L, --log-level <LOG_LEVEL>

Set the log level

--locked

Requires that the Forc.lock file is up-to-date. If the lock file is missing, or it needs to be updated, Forc will exit with an error

-o, --output-bin <BIN_FILE>

If set, outputs a binary file representing the script bytes

--offline

Offline mode, prevents Forc from using the network when managing dependencies. Meaning it will only try to use previously downloaded dependencies

--output-directory <OUTPUT_DIRECTORY>

The directory in which the sway compiler output artifacts are placed.

By default, this is <project-root>/out.

-p, --path <PATH>

Path to the project, if not specified, current working directory will be used

--release

Use release build plan. If a custom release plan is not specified, it is implicitly added to the manifest file.

If --build-profile is also provided, forc omits this flag and uses provided build-profile.

-s, --silent

Silence all output

-t, --terse

Terse mode. Limited warning and error output

--time-phases

Output the time elapsed over each part of the compilation process

-v, --verbose

Use verbose output

forc-test

Run the Sway unit tests for the current project.

NOTE: Previously this command was used to support Rust integration testing, however the provided behaviour served no benefit over running cargo test directly. The proposal to change the behaviour to support unit testing can be found at the following link: https://github.com/FuelLabs/sway/issues/1833

Sway unit tests are functions decorated with the #[test] attribute. Each test is compiled as a unique entry point for a single program and has access to the namespace of the module in which it is declared.

Unit tests decorated with the #[test(script)] attribute that are declared within contract projects may also call directly into their associated contract's ABI.

Upon successful compilation, test scripts are executed to their completion. A test is considered a failure in the case that a revert (rvrt) instruction is encountered during execution. Otherwise, it is considered a success.

USAGE:

forc test [OPTIONS] [FILTER]

ARGS:

<FILTER> When specified, only tests containing the given string will be executed

OPTIONS:

--ast

Print the generated Sway AST (Abstract Syntax Tree)

--build-profile <BUILD_PROFILE>

Name of the build profile to use.

If unspecified, forc will use debug build profile.

--build-target <BUILD_TARGET>

Build target to use for code generation

[default: fuel] [possible values: fuel, evm, midenvm]

--dca-graph

Print the computed Sway DCA graph

--error-on-warnings

Treat warnings as errors

--finalized-asm

Print the finalized ASM.

This is the state of the ASM with registers allocated and optimisations applied.

-g, --output-debug <DEBUG_FILE>

If set, outputs source file mapping in JSON format

-h, --help

Print help information

--intermediate-asm

Print the generated ASM.

This is the state of the ASM prior to performing register allocation and other ASM optimisations.

--ir

Print the generated Sway IR (Intermediate Representation)

--json-abi

By default the JSON for ABIs is formatted for human readability. By using this option JSON output will be "minified", i.e. all on one line without whitespace

--json-abi-with-callpaths

Outputs json abi with callpaths instead of names for struct and enums

--json-storage-slots

By default the JSON for initial storage slots is formatted for human readability. By using this option JSON output will be "minified", i.e. all on one line without whitespace

-l, --logs

Print Log and LogData receipts for tests

-L, --log-level <LOG_LEVEL>

Set the log level

--locked

Requires that the Forc.lock file is up-to-date. If the lock file is missing, or it needs to be updated, Forc will exit with an error

-o, --output-bin <BIN_FILE>

If set, outputs a binary file representing the script bytes

--offline

Offline mode, prevents Forc from using the network when managing dependencies. Meaning it will only try to use previously downloaded dependencies

--output-directory <OUTPUT_DIRECTORY>

The directory in which the sway compiler output artifacts are placed.

By default, this is <project-root>/out.

-p, --path <PATH>

Path to the project, if not specified, current working directory will be used

-r, --pretty-print

Pretty-print the logs emiited from tests

--release

Use release build plan. If a custom release plan is not specified, it is implicitly added to the manifest file.

If --build-profile is also provided, forc omits this flag and uses provided build-profile.

-s, --silent

Silence all output

-t, --terse

Terse mode. Limited warning and error output

--time-phases

Output the time elapsed over each part of the compilation process

-v, --verbose

Use verbose output

forc-update

Update dependencies in the Forc dependencies directory

USAGE:

forc update [OPTIONS]

OPTIONS:

-c, --check

Checks if the dependencies have newer versions. Won't actually perform the update, will output which ones are up-to-date and outdated

-d <TARGET_DEPENDENCY>

Dependency to be updated. If not set, all dependencies will be updated

-h, --help

Print help information

-L, --log-level <LOG_LEVEL>

Set the log level

-p, --path <PATH>

Path to the project, if not specified, current working directory will be used

-s, --silent

Silence all output

-v, --verbose

Use verbose output

forc-template

Create a new Forc project from a git template

USAGE:

forc template [OPTIONS] <PROJECT_NAME>

ARGS:

<PROJECT_NAME>

The name of the project that will be created

OPTIONS:

-h, --help

Print help information

-L, --log-level <LOG_LEVEL>

Set the log level

-s, --silent

Silence all output

-t, --template-name <TEMPLATE_NAME>

The name of the template that needs to be fetched and used from git repo provided

-u, --url <URL>

The template url, should be a git repo [default: https://github.com/fuellabs/sway]

-v, --verbose

Use verbose output

EXAMPLE

forc template --url https://github.com/owner/template/ --project_name my_example_project

The command above fetches the HEAD of the template repo and searches for Forc.toml at the root of the fetched repo. It will fetch the repo and prepare a new Forc.toml with the new project name. Outputs everything to current_dir/project_name.

forc template --url https://github.com/FuelLabs/sway --template_name counter --project_name my_example_project

The command above fetches the HEAD of the sway repo and searches for counter example inside it (there is an example called counter under sway/examples). It will fetch the counter example and prepare a new Forc.toml with the new project name. Outputs everything to current_dir/project_name.

Plugins

Plugins can be used to extend forc with new commands that go beyond the native commands mentioned in the previous chapter. While the Fuel ecosystem provides a few commonly useful plugins (forc-fmt, forc-client, forc-lsp, forc-explore), anyone can write their own!

Let's install a plugin, forc-explore, and see what's underneath the plugin:

cargo install forc-explore

Check that we have installed forc-explore:

$ forc plugins
Installed Plugins:
forc-explore

forc-explore runs the Fuel Network Explorer, which you can run and check out for yourself:

$ forc explore
Fuel Network Explorer 0.1.1
Running server on http://127.0.0.1:3030
Server::run{addr=127.0.0.1:3030}: listening on http://127.0.0.1:3030

You can visit http://127.0.0.1:3030 to check out the network explorer!

Note that some plugin crates can also provide more than one command. For example, installing the forc-client plugin provides the forc deploy and forc run commands. This is achieved by specifying multiple [[bin]] targets within the forc-client manifest.

Writing your own plugin

We encourage anyone to write and publish their own forc plugin to enhance their development experience.

Your plugin must be named in the format forc-<MY_PLUGIN> and you may use the above template as a starting point. You can use clap and add more subcommands, options and configurations to suit your plugin's needs.

forc-client

Forc plugin for interacting with a Fuel node.

Initializing the wallet and adding accounts

If you don't have an initialized wallet or any account for your wallet you won't be able to sign transactions.

To create a wallet you can use forc wallet new. It will ask you to choose a password to encrypt your wallet. After the initialization is done you will have your mnemonic phrase.

After you have created a wallet, you can derive a new account by running forc wallet account new. It will ask your password to decrypt the wallet before deriving an account.

Signing transactions using forc-wallet CLI

To submit the transactions created by forc deploy or forc run, you need to sign them first (unless you are using a client without UTXO validation). To sign a transaction you can use forc-wallet CLI. This section is going to walk you through the whole signing process.

By default fuel-core runs without UTXO validation, which means you can run unsigned transactions. This allows you to send invalid inputs to emulate different conditions.

If you want to run fuel-core with UTXO validation, you can pass --utxo-validation to fuel-core run. If UTXO validation is enabled, unsigned transactions will return an error.

To install forc-wallet please refer to forc-wallet's github repo.

  1. Construct the transaction by using either forc deploy or forc run. To do so simply run forc deploy or forc run with your desired parameters. For a list of parameters please refer to the forc-deploy or forc-run section of the book. Once you run either command you will be asked the address of the wallet you are going to be signing with. After the address is given the transaction will be generated and you will be given a transaction ID. At this point CLI will actively wait for you to insert the signature.
  2. Take the transaction ID generated in the first step and sign it with forc wallet sign --account <account_index> tx-id <transaction_id>. This will generate a signature.
  3. Take the signature generated in the second step and provide it to forc-deploy (or forc-run). Once the signature is provided, the signed transaction will be submitted.

Other useful commands of forc-wallet

  • You can see a list of existing accounts with accounts command.
forc wallet accounts
  • If you want to retrieve the address for an account by its index you can use account command.
forc wallet account <account_index>

If you don't want to sign the transaction generated by forc-deploy or forc-run you can pass --unsigned to them.

forc-deploy --unsigned
forc-run --unsigned

Interacting with the testnet

While using forc-deploy or forc-run to interact with the testnet you need to pass the testnet end point with --url

forc-deploy --url https://node-beta-1.fuel.network/graphql:443

Since deploying and running projects on the testnet cost gas, you will need coins to pay for them. You can get some using the testnet faucet.

Also the default value of the "gas price" parameter is 0 for both forc-deploy and forc-run. Without changing it you will get an error complaining about gas price being too low. While using testnet you can pass --gas-price 1 to overcome this issue. So a complete command for deploying to the testnet would look like:

forc-deploy --url https://node-beta-1.fuel.network/graphql:443 --gas-price 1

forc-deploy

Build profile options

USAGE:

forc deploy [OPTIONS] [SIGNING_KEY]

ARGS:

<SIGNING_KEY> Set the key to be used for signing

OPTIONS:

--ast

Print the generated Sway AST (Abstract Syntax Tree)

--build-profile <BUILD_PROFILE>

Name of the build profile to use.

If unspecified, forc will use debug build profile.

--dca-graph

Print the computed Sway DCA graph

--error-on-warnings

Treat warnings as errors

--finalized-asm

Print the finalized ASM.

This is the state of the ASM with registers allocated and optimisations applied.

-g, --output-debug <DEBUG_FILE>

If set, outputs source file mapping in JSON format

--gas-limit <LIMIT>

Gas limit for the transaction

[default: 100000000]

--gas-price <PRICE>

Gas price for the transaction

[default: 0]

-h, --help

Print help information

--intermediate-asm

Print the generated ASM.

This is the state of the ASM prior to performing register allocation and other ASM optimisations.

--ir

Print the generated Sway IR (Intermediate Representation)

--json-abi

By default the JSON for ABIs is formatted for human readability. By using this option JSON output will be "minified", i.e. all on one line without whitespace

--json-abi-with-callpaths

Outputs json abi with callpaths instead of names for struct and enums

--json-storage-slots

By default the JSON for initial storage slots is formatted for human readability. By using this option JSON output will be "minified", i.e. all on one line without whitespace

--locked

Requires that the Forc.lock file is up-to-date. If the lock file is missing, or it needs to be updated, Forc will exit with an error

--maturity <MATURITY>

Block height until which tx cannot be included

[default: 0]

--node-url <NODE_URL>

The URL of the Fuel node to which we're submitting the transaction. If unspecified, checks the manifest's network table, then falls back to [crate::default::NODE_URL]

[env: FUEL_NODE_URL=]

-o, --output-bin <BIN_FILE>

If set, outputs a binary file representing the script bytes

--offline

Offline mode, prevents Forc from using the network when managing dependencies. Meaning it will only try to use previously downloaded dependencies

--output-directory <OUTPUT_DIRECTORY>

The directory in which the sway compiler output artifacts are placed.

By default, this is <project-root>/out.

-p, --path <PATH>

Path to the project, if not specified, current working directory will be used

--random-salt

Generate a random salt for the contract. Useful for testing or deploying examples to a shared network

--release

Use release build plan. If a custom release plan is not specified, it is implicitly added to the manifest file.

If --build-profile is also provided, forc omits this flag and uses provided build-profile.

--salt <SALT>

Optional 256-bit hexadecimal literal(s) to redeploy contracts.

For a single contract, use --salt <SALT>, eg.: forc deploy --salt 0x0000000000000000000000000000000000000000000000000000000000000001

For a workspace with multiple contracts, use --salt <CONTRACT_NAME>:<SALT> to specify a salt for each contract, eg.:

forc deploy --salt contract_a:0x0000000000000000000000000000000000000000000000000000000000000001 --salt contract_b:0x0000000000000000000000000000000000000000000000000000000000000002

-t, --terse

Terse mode. Limited warning and error output

--time-phases

Output the time elapsed over each part of the compilation process

--unsigned

Do not sign the transaction

-V, --version

Print version information

EXAMPLE

You can use forc deploy, which triggers a contract deployment transaction and sends it to a running node.

Alternatively, you can deploy your Sway contract programmatically using fuels-rs, our Rust SDK.

You can find an example within our fuels-rs book.

forc-run

Run script project. Crafts a script transaction then sends it to a running node

USAGE:

forc run [OPTIONS] [SIGNING_KEY]

ARGS:

<SIGNING_KEY> Set the key to be used for signing

OPTIONS:

--ast

Print the generated Sway AST (Abstract Syntax Tree)

--build-profile <BUILD_PROFILE>

Name of the build profile to use.

If unspecified, forc will use debug build profile.

--contract <CONTRACT>

32-byte contract ID that will be called during the transaction

-d, --data <DATA>

Hex string of data to input to script

--dca-graph

Print the computed Sway DCA graph

--dry-run

Only craft transaction and print it out

--error-on-warnings

Treat warnings as errors

--finalized-asm

Print the finalized ASM.

This is the state of the ASM with registers allocated and optimisations applied.

-g, --output-debug <DEBUG_FILE>

If set, outputs source file mapping in JSON format

--gas-limit <LIMIT>

Gas limit for the transaction

[default: 100000000]

--gas-price <PRICE>

Gas price for the transaction

[default: 0]

-h, --help

Print help information

--intermediate-asm

Print the generated ASM.

This is the state of the ASM prior to performing register allocation and other ASM optimisations.

--ir

Print the generated Sway IR (Intermediate Representation)

--json-abi

By default the JSON for ABIs is formatted for human readability. By using this option JSON output will be "minified", i.e. all on one line without whitespace

--json-abi-with-callpaths

Outputs json abi with callpaths instead of names for struct and enums

--json-storage-slots

By default the JSON for initial storage slots is formatted for human readability. By using this option JSON output will be "minified", i.e. all on one line without whitespace

--locked

Requires that the Forc.lock file is up-to-date. If the lock file is missing, or it needs to be updated, Forc will exit with an error

--maturity <MATURITY>

Block height until which tx cannot be included

[default: 0]

--node-url <NODE_URL>

The URL of the Fuel node to which we're submitting the transaction. If unspecified, checks the manifest's network table, then falls back to [crate::default::NODE_URL]

[env: FUEL_NODE_URL=]

-o, --output-bin <BIN_FILE>

If set, outputs a binary file representing the script bytes

--offline

Offline mode, prevents Forc from using the network when managing dependencies. Meaning it will only try to use previously downloaded dependencies

--output-directory <OUTPUT_DIRECTORY>

The directory in which the sway compiler output artifacts are placed.

By default, this is <project-root>/out.

-p, --path <PATH>

Path to the project, if not specified, current working directory will be used

-r, --pretty-print

Pretty-print the outputs from the node

--release

Use release build plan. If a custom release plan is not specified, it is implicitly added to the manifest file.

If --build-profile is also provided, forc omits this flag and uses provided build-profile.

--simulate

Execute the transaction and return the final mutated transaction along with receipts (which includes whether the transaction reverted or not). The transaction is not inserted in the node's view of the blockchain, (i.e. it does not affect the chain state)

-t, --terse

Terse mode. Limited warning and error output

--time-phases

Output the time elapsed over each part of the compilation process

--unsigned

Do not sign the transaction

-V, --version

Print version information

forc-client

A forc plugin for interacting with a Fuel node.

USAGE:

forc-submit [OPTIONS] <TX_PATH>

ARGS:

<TX_PATH> Path to the Transaction that is to be submitted to the Fuel node.

Paths to files ending with .json will be deserialized from JSON. Paths to files ending with .bin will be deserialized from bytes using the fuel_tx::Transaction::try_from_bytes constructor.

OPTIONS:

--await <await>

Whether or not to await confirmation that the transaction has been committed.

When true, await commitment and output the transaction status. When false, do not await confirmation and simply output the transaction ID.

[default: true]

-h, --help

Print help information

--node-url <NODE_URL>

The URL of the Fuel node to which we're submitting the transaction

[env: FUEL_NODE_URL=] [default: http://127.0.0.1:4000]

--tx-status-json <json>

Output the resulting transaction status as JSON rather than the default output

[default: false]

-V, --version

Print version information

Forc

Usage: forc-doc [OPTIONS]

Options:

--manifest-path <MANIFEST_PATH>

Path to the Forc.toml file. By default, forc-doc searches for the Forc.toml file in the current directory or any parent directory

--document-private-items

Include non-public items in the documentation

--open

Open the docs in a browser after building them

--offline

Offline mode, prevents Forc from using the network when managing dependencies. Meaning it will only try to use previously downloaded dependencies

-s, --silent

Silent mode. Don't output any warnings or errors to the command line

--locked

Requires that the Forc.lock file is up-to-date. If the lock file is missing, or it needs to be updated, Forc will exit with an error

--no-deps

Do not build documentation for dependencies

-h, --help

Print help

-V, --version

Print version

forc-explore

Forc plugin for running the Fuel Block Explorer.

USAGE:

forc-explore [OPTIONS] [SUBCOMMAND]

OPTIONS:

-h, --help

Print help information

-p, --port <PORT>

The port number at which the explorer will run on localhost [default: 3030]

-V, --version

Print version information

SUBCOMMANDS:

clean

Cleans up any existing state associated with the fuel block explorer

help

Print this message or the help of the given subcommand(s)

forc-fmt

Forc plugin for running the Sway code formatter.

USAGE:

forc-fmt [OPTIONS] [FILE]

ARGS:

<FILE> Formats a single .sw file with the default settings. If not specified, current working directory will be formatted using a Forc.toml configuration

OPTIONS:

-c, --check

Run in 'check' mode.

  • Exits with 0 if input is formatted correctly. - Exits with 1 and prints a diff if formatting is required.

-h, --help

Print help information

-p, --path <PATH>

Path to the project, if not specified, current working directory will be used

-V, --version

Print version information

forc-lsp

Forc plugin for the Sway LSP (Language Server Protocol) implementation.

USAGE:

forc-lsp

OPTIONS:

-h, --help

Print help information

-V, --version

Print version information