{ skip to content }

User Defined Value Types in Solidity

Posted by Solidity Team on September 27, 2021

Explainers

Solidity v0.8.8 introduces user defined value types as a means to create zero-cost abstractions over an elementary value type that also increases type safety and improves readability.

Motivation

A problem with primitive value types is that they are not very descriptive: they only specify how the data is stored and not how it should be interpreted. For example, one may want to use uint128 to store the price of some object as well as the quantity available. It is quite useful to have stricter type rules to avoid intermingling of the two different concepts. For example, one may want to disallow assigning a quantity to a price or vice versa.

One option for solving this issue is by using structs. For example, price and quantity can be abstracted as structs as follows:

struct Price { uint128 price; }
struct Quantity { uint128 quantity; }

function toPrice(uint128 price) returns(Price memory) {
    return Price(price);
}
function fromPrice(Price memory price) returns(uint128) {
   return price.price;
}
function toQuantity(uint128 quantity) returns(Quantity memory) {
    return Quantity(quantity);
}
function fromQuantity(Quantity memory quantity) returns(uint128) {
    return quantity.quantity;
}

However, a struct is a reference type and therefore always points to a value in memory, calldata or storage. This means that the above abstraction has a runtime overhead, i.e., additional gas when compared to using just uint128 to represent the underlying value. In particular, the functions toPrice and toQuantity involve storing the value in memory. Similarly, the functions fromPrice and fromQuantity read the respective value from memory. Together, these functions pass the value from stack -> memory -> stack which wastes memory and incurs a runtime cost. This issue is solved by user defined value types, which are abstractions of elementary value types (such as uint8 or address), without any additional runtime overhead.

Syntax for User Defined Value Types

A user defined value type is defined using type C is V;, where C is the name of the newly introduced type and V has to be a built-in value type (the "underlying type"). They can be defined inside or outside contracts (including libraries and interfaces). The function C.wrap is used to convert from the underlying type to the custom type. Similarly, the function C.unwrap is used to convert from the custom type to the underlying type.

Going back to the problem from the motivation section, one can replace the structs by:

pragma solidity ^0.8.8;

type Price is uint128;
type Quantity is uint128;

The functions toPrice and toQuantity can be replaced by Price.wrap and Quantity.wrap respectively. Similarly, the functions fromPrice and fromQuantity can be replaced by Price.unwrap and Quantity.unwrap respectively.

The data-representation of the values of such types are inherited from the underlying type and the underlying type is also used in the ABI. This means that the following two transfer functions would be identical, i.e., they have the same function selector as well as the same ABI encoding and decoding. This allows using user defined value types in a backwards compatible way.

pragma solidity ^0.8.8;

type Decimal18 is uint256;

interface MinimalERC20 {
    function transfer(address to, Decimal18 value) external;
}

interface AnotherMinimalERC20 {
    function transfer(address to, uint256 value) external;
}

Notice in the above example, how the user defined type Decimal18 makes it clear that a value is supposed to represent a number with 18 decimals.

Example

The following example illustrates a custom type UFixed representing a decimal fixed point type with 18 decimals and a minimal library to do arithmetic operations on the type.

// SPDX-License-Identifier: GPL-3.0
pragma solidity ^0.8.8;

// Represent a 18 decimal, 256 bit wide fixed point type
// using a user defined value type.
type UFixed is uint256;

/// A minimal library to do fixed point operations on UFixed.
library FixedMath {
    uint constant multiplier = 10**18;

    /// Adds two UFixed numbers. Reverts on overflow,
    /// relying on checked arithmetic on uint256.
    function add(UFixed a, UFixed b) internal pure returns (UFixed) {
        return UFixed.wrap(UFixed.unwrap(a) + UFixed.unwrap(b));
    }
    /// Multiplies UFixed and uint256. Reverts on overflow,
    /// relying on checked arithmetic on uint256.
    function mul(UFixed a, uint256 b) internal pure returns (UFixed) {
        return UFixed.wrap(UFixed.unwrap(a) * b);
    }
    /// Take the floor of a UFixed number.
    /// @return the largest integer that does not exceed `a`.
    function floor(UFixed a) internal pure returns (uint256) {
        return UFixed.unwrap(a) / multiplier;
    }
    /// Turns a uint256 into a UFixed of the same value.
    /// Reverts if the integer is too large.
    function toUFixed(uint256 a) internal pure returns (UFixed) {
        return UFixed.wrap(a * multiplier);
    }
}

Notice how UFixed.wrap and FixedMath.toUFixed have the same signature but perform two very different operations: The UFixed.wrap function returns a UFixed that has the same data representation as the input, whereas toUFixed returns a UFixed that has the same numerical value. One can allow some form of type-encapsulation by only using the wrap and unwrap functions in the file that defines the type.

Operators and Type Rules

Explicit and implicit conversions to and from other types are disallowed.

Currently, no operators are defined for user defined value types. In particular, even the operator == is not defined. However, allowing operators is currently being discussed. To give a short outlook on the applications, one may want to introduce a new integer type that always does wrapping arithmetic as follows:

/// Proposal on defining operators on user defined value types
/// Note: this does not fully compile on Solidity 0.8.8; only a concept.

type UncheckedInt8 is int8;

function add(UncheckedInt8 a, UncheckedInt8 b) pure returns(UncheckedInt8) {
    unchecked {
        return UncheckedInt8.wrap(UncheckedInt8.unwrap(a) + UncheckedInt8.unwrap(b));
    }
}
function addInt(UncheckedInt8 a, uint b) pure returns(UncheckedInt8) {
    unchecked {
        return UncheckedInt8.wrap(UncheckedInt8.unwrap(a) + b);
    }
}

using {add as +, addInt as +} for UncheckedInt8;

contract MockOperator {
    UncheckedInt8 x;
    function increment() external {
        // This would not revert on overflow when x = 127
        x = x + 1;
    }
    function add(UncheckedInt8 y) external {
        // Similarly, this would also not revert on overflow.
        x = x + y;
    }
}

You can join or follow this discussion in the Solidity forum and issue #11969. Also, you can join or follow the discussion about allowing the constructor syntax for user defined value types in the issue #11953.

Previous post

Next post

Get involved

GitHub

Twitter

Mastodon

Matrix

Discover more

BlogDocumentationUse casesContributeAboutForum

2024 Solidity Team

Security Policy

Code of Conduct