How to convert a uint to bytes in Solidity?

Question

Is there an easy way to convert a uint to bytes in Solidity?

Answer 1

It seems there now is, since solidity version 0.4.24 you can use abi.encodePacked

E.G:

uint i = 0;
i_bytes = abi.encodePacked(i);

Answer 2

The alternative to @eth's answer is to use assembly:

function toBytes(uint256 x) returns (bytes b) {
    b = new bytes(32);
    assembly { mstore(add(b, 32), x) }
}

This is significantly more gas-efficient, but depends on the internal memory layout used by the Solidity compiler. In theory, this can change in future, but in practice it should be fairly stable.

Answer 3

Here's a comparison of the gas used in the three methods by @NickJohnson, @Eth and @k26dr. I've added a constant to the function modifiers as these functions do not alter the blockchain:

pragma solidity ^0.4.2;

contract Test {
    function toBytesNickJohnson(uint256 x) constant returns (bytes b) {
        b = new bytes(32);
        assembly { mstore(add(b, 32), x) }
    }

    function toBytesEth(uint256 x) constant returns (bytes b) {
        b = new bytes(32);
        for (uint i = 0; i < 32; i++) {
            b[i] = byte(uint8(x / (2**(8*(31 - i))))); 
        }
    }

    function toBytesNicolasMassart(uint256 x) constant returns (bytes c) {
        bytes32 b = bytes32(x);
        c = new bytes(32);
        for (uint i=0; i < 32; i++) {
            c[i] = b[i];
        }
    }    
}

You can see the gas cost of running these functions in Remix (Solidity Browser):

Answer 4

There are no easy ways to convert anything to bytes. Here's a function:

function toBytes(uint256 x) returns (bytes b) {
    b = new bytes(32);
    for (uint i = 0; i < 32; i++) {
        b[i] = byte(uint8(x / (2**(8*(31 - i))))); 
    }
}

Based on Solidity Gitter chat.

Answer 5

If you're overly concerned about gas we can improve on @NickJohnson answer too.

function toBytes(uint _num) returns (bytes _ret) {
    assembly {
        _ret := mload(0x10)
        mstore(_ret, 0x20)
        mstore(add(_ret, 0x20), _num)
    }
}

This will shave approximately a further 15% off the gas cost, just be careful that you're not using that memory for something else as 0x10 is a direct reference to memory.

Also, if you want to copy an 8bit int to byte here's an alternative which only uses half the gas cost:

function toByte(uint8 _num) returns (byte _ret) {
    assembly {
        mstore8(0x20, _num)
        _ret := mload(0x20)
    }
}

Again, be mindful that 0x20 is another direct memory reference and in all honesty with this one I would stick with

function toByte(uint8 _num) returns (byte _ret) {
    return byte(_num);
}

The gas price between the two was almost identical, but the functional ASM was undercutting it by about 70 wei

Edit: If you're concerned about overwriting memory you could use:

let m_alloc := add(msize(),0x1)

Instead of referencing memory yourself

Answer 6

You can convert to bytes32 then convert to bytes:

uint u = 200;
bytes32 b = bytes32(u);
bytes memory c = new bytes(32);
for (uint i=0; i < 32; i++) {
    c[i] = b[i];
}

Answer 7

You can avoid padding the bytes array with 0's with the toBytes implementations by determining the scriptNumSize

function scriptNumSize(uint256 i) public view returns (uint256) {
    if      (i > 0x7fffffff) { return 5; }
    else if (i > 0x7fffff  ) { return 4; }
    else if (i > 0x7fff    ) { return 3; }
    else if (i > 0x7f      ) { return 2; }
    else if (i > 0x00      ) { return 1; }
    else                     { return 0; }
}

function toBytes(uint256 x) public view returns (bytes memory b) {
    uint a = scriptNumSize(x);
    b = new bytes(a);
    for (uint i = 0; i < a; i++) {
        b[i] = byte(uint8(x / (2**(8*(a - 1 - i))))); 
    }
}

For example uint 1563384765 would return 0x5d2f5bbd not 0x000000000000000000000000000000000000000000000000000000005d2f5bbd

This was written specifically for timestamps. If you want larger numbers you can add more if cases.