Skip to main content

Vyper ERC-721 Contract Walkthrough

vyper
erc-721
python
Beginner
Ori Pomerantz
April 1, 2021
20 minute read

Introduction

The ERC-721 standard is used to hold the ownership of Non-Fungible Tokens (NFT). ERC-20 tokens behave as a commodity, because there is no difference between individual tokens. In contrast to that, ERC-721 tokens are designed for assets that are similar but not identical, such as different cat cartoons (opens in a new tab) or titles to different pieces of real estate.

In this article we will analyze Ryuya Nakamura's ERC-721 contract (opens in a new tab). This contract is written in Vyper (opens in a new tab), a Python-like contract language designed to make it harder to write insecure code than it is in Solidity.

The Contract

#pragma version >0.3.10

The #pragma version line tells the compiler which Vyper versions the contract is written for. >0.3.10 accepts any release newer than 0.3.10, but this contract in fact needs Vyper 0.4.1 or later: it logs events with keyword arguments (for example log IERC721.Transfer(...)), a feature added in Vyper 0.4.1, so it does not compile on 0.4.0.

# @dev Implementation of ERC-721 non-fungible token standard.
# @author Ryuya Nakamura (@nrryuya)
# Modified from: https://github.com/vyperlang/vyper/blob/de74722bf2d8718cca46902be165f9fe0e3641dd/examples/tokens/ERC721.vy

Comments in Vyper, as in Python, start with a hash (#) and continue to the end of the line. Comments that include @<keyword> are used by NatSpec (opens in a new tab) to produce human-readable documentation.

from ethereum.ercs import IERC165
from ethereum.ercs import IERC721

implements: IERC721
implements: IERC165

The ERC-165 and ERC-721 interfaces ship with the Vyper compiler, so we import them from ethereum.ercs rather than writing them ourselves. You can see the ERC-721 interface definition here (opens in a new tab). The implements: lines declare that this contract provides both interfaces, and the compiler checks that every function they require is defined below.

The ERC721Receiver Interface

# Interface for the contract called by safeTransferFrom()
interface ERC721Receiver:
    def onERC721Received(

ERC-721 supports two types of transfer:

  • transferFrom, which lets the sender specify any destination address and places the responsibility for the transfer on the sender. This means that you can transfer to an invalid address, in which case the NFT is lost for good.
  • safeTransferFrom, which checks if the destination address is a contract. If so, the ERC-721 contract asks the receiving contract if it wants to receive the NFT.

To answer safeTransferFrom requests a receiving contract has to implement ERC721Receiver.

            _operator: address,
            _from: address,

The _from address is the current owner of the token. The _operator address is the one that requested the transfer (those two may not be the same, because of allowances). By convention, most function parameters in this contract start with an underscore (_).

            _tokenId: uint256,

ERC-721 token IDs are 256 bits. Typically they are created by hashing a description of whatever the token represents.

            _data: Bytes[1024]

The request can have up to 1024 bytes of user data.

        ) -> bytes4: nonpayable

To prevent cases in which a contract accidentally accepts a transfer the return value is not a boolean, but a specific four-byte value, the function selector of onERC721Received. The function is nonpayable because a receiving contract may change its own state when it accepts a token.

Events

Events are emitted to inform users and servers outside of the blockchain of events. Note that the content of events is not available to contracts on the blockchain. The three ERC-721 events are defined by the IERC721 interface we imported, so this contract does not declare them itself; it emits them with log IERC721.<Event>(...), as we will see in the transfer functions below.

Transfer (sender, receiver, token_id) reports a change in the ownership of an NFT. This is similar to the ERC-20 Transfer event, except that we report a token_id instead of an amount. Nobody owns address zero, so by convention we use it to report creation and destruction of tokens. The one exception is contract creation, during which any number of NFTs may be created and assigned without emitting Transfer.

An ERC-721 approval is similar to an ERC-20 allowance: a specific address is allowed to transfer a specific token, and Approval (owner, approved, token_id) is emitted whenever that approved address is set or reaffirmed. This gives a mechanism for contracts to respond when they accept a token. Contracts cannot listen for events, so if you just transfer the token to them they don't "know" about it. This way the owner first submits an approval and then sends a request to the contract: "I approved for you to transfer token X, please do ...". This is a design choice to make the ERC-721 standard similar to the ERC-20 standard. Because ERC-721 tokens are not fungible, a contract can also identify that it got a specific token by looking at the token's ownership.

Finally, ApprovalForAll (owner, operator, approved) is emitted when an operator is enabled or disabled for an owner. It is sometimes useful to have an operator that can manage all of an account's tokens of a specific type (those that are managed by a specific contract), similar to a power of attorney. For example, I might want to give such a power to a contract that checks if I haven't contacted it for six months, and if so distributes my assets to my heirs (if one of them asks for it, contracts can't do anything without being called by a transaction). In ERC-20 we can just give a high allowance to an inheritance contract, but that doesn't work for ERC-721 because the tokens are not fungible. This is the equivalent. The approved value tells us whether the event is for an approval, or the withdrawal of an approval.

State Variables

These variables contain the current state of the tokens: which ones are available and who owns them. Most of these are HashMap objects, unidirectional mappings that exist between two types (opens in a new tab).

# @dev Mapping from NFT ID to the address that owns it.
idToOwner: HashMap[uint256, address]

# @dev Mapping from NFT ID to approved address.
idToApprovals: HashMap[uint256, address]

User and contract identities in Ethereum are represented by 160-bit addresses. These two variables map from token IDs to their owners and those approved to transfer them (at a maximum of one for each). In Ethereum, uninitialized data is always zero, so if there is no owner or approved transferor the value for that token is zero.

# @dev Mapping from owner address to count of his tokens.
ownerToNFTokenCount: HashMap[address, uint256]

This variable holds the count of tokens for each owner. There is no mapping from owners to tokens, so the only way to identify the tokens that a specific owner owns is to look back in the blockchain's event history and see the appropriate Transfer events. We can use this variable to know when we have all the NFTs and don't need to look even further in time.

Note that this algorithm only works for user interfaces and external servers. Code running on the blockchain itself cannot read past events.

# @dev Mapping from owner address to mapping of operator addresses.
ownerToOperators: HashMap[address, HashMap[address, bool]]

An account may have more than a single operator. A simple HashMap is insufficient to keep track of them, because each key leads to a single value. Instead, you can use HashMap[address, bool] as the value. By default the value for each address is False, which means it is not an operator. You can set values to True as needed.

# @dev Address of minter, who can mint a token
minter: address

New tokens have to be created somehow. In this contract there is a single entity that is allowed to do so, the minter. This is likely to be sufficient for a game, for example. For other purposes, it might be necessary to create a more complicated business logic.

# @dev Static list of supported ERC165 interface ids
SUPPORTED_INTERFACES: constant(bytes4[2]) = [
    # ERC165 interface ID of ERC165
    0x01ffc9a7,
    # ERC165 interface ID of ERC721
    0x80ac58cd,
]

ERC-165 (opens in a new tab) specifies a mechanism for a contract to disclose how applications can communicate with it, to which ERCs it conforms. SUPPORTED_INTERFACES is a constant list of the two four-byte interface IDs this contract conforms to: ERC-165 itself and ERC-721.

Functions

These are the functions that actually implement ERC-721.

Constructor

@deploy
def __init__():

In Vyper, as in Python, the constructor function is called __init__. It is marked with the @deploy decoration, which means it runs once, when the contract is deployed.

    """
    @dev Contract constructor.
    """

In Python, and in Vyper, you can also create a comment by specifying a multi-line string (which starts and ends with """), and not using it in any way. These comments can also include NatSpec (opens in a new tab).

    self.minter = msg.sender

To access state variables you use self.<variable name> (again, same as in Python). The constructor records the account that deployed the contract as the minter.

View Functions

These are functions that do not modify the state of the blockchain, and therefore can be executed for free if they are called externally. If the view functions are called by a contract they still have to be executed on every node and therefore cost gas.

@view
@external

These keywords prior to a function definition that start with an at sign (@) are called decorations. They specify the circumstances in which a function can be called.

  • @view specifies that this function is a view.
  • @external specifies that this particular function can be called by transactions and by other contracts.
def supportsInterface(interface_id: bytes4) -> bool:

In contrast to Python, Vyper is a static typed language (opens in a new tab). You can't declare a variable, or a function parameter, without identifying the data type (opens in a new tab). In this case the input parameter is bytes4, a four-byte value, and the output is a boolean value.

    """
    @dev Interface identification is specified in ERC-165.
    @param interface_id Id of the interface
    """
    return interface_id in SUPPORTED_INTERFACES

Return True if interface_id is one of the interface IDs in the SUPPORTED_INTERFACES list.

### VIEW FUNCTIONS ###

These are the view functions that make information about the tokens available to users and other contracts.

This line asserts (opens in a new tab) that _owner is not the zero address, written as empty(address). If it is, there is an error and the operation is reverted.

In the Ethereum Virtual Machine (evm) any storage that does not have a value stored in it is zero. If there is no token at _tokenId then the value of self.idToOwner[_tokenId] is zero. In that case the function reverts.

Note that getApproved can return zero. If the token is valid it returns self.idToApprovals[_tokenId]. If there is no approver that value is zero.

This function checks if _operator is allowed to manage all of _owner's tokens in this contract. Because there can be multiple operators, this is a two level HashMap.

Transfer Helper Functions

These functions implement operations that are part of transferring or managing tokens.


### TRANSFER FUNCTION HELPERS ###

@view
@internal

This decoration, @internal, means that the function is only accessible from other functions within the same contract. By convention, these function names also start with an underscore (_).

There are three ways in which an address can be allowed to transfer a token:

  1. The address is the owner of the token
  2. The address is approved to spend that token
  3. The address is an operator for the owner of the token

The function above can be a view because it doesn't change the state. To reduce operating costs, any function that can be a view should be a view.

When there's a problem with a transfer we revert the call.

Only change the value if necessary. State variables live in storage. Writing to storage is one of the most expensive operations the EVM (Ethereum Virtual Machine) does (in terms of gas). Therefore, it is a good idea to minimize it, even writing the existing value has a high cost.

We have this internal function because there are two ways to transfer tokens (regular and safe), but we want only a single location in the code where we do it to make auditing easier.

To emit an event in Vyper you use a log statement (see here for more details (opens in a new tab)). Because the events belong to the imported interface, we refer to them as IERC721.Transfer and pass their fields by keyword.

Transfer Functions

This function lets you transfer to an arbitrary address. Unless the address is a user, or a contract that knows how to transfer tokens, any token you transfer will be stuck in that address and useless.

The @payable decoration is here because the IERC721 interface declares transferFrom, safeTransferFrom, and approve as payable, so a contract that implements the interface has to match those signatures.

It is OK to do the transfer first because if there's a problem we are going to revert anyway, so everything done in the call will be cancelled.

    if _to.is_contract: # check if `_to` is a contract address

First check to see if the address is a contract (if it has code). If not, assume it is a user address and the user will be able to use the token or transfer it. But don't let it lull you into a false sense of security. You can lose tokens, even with safeTransferFrom, if you transfer them to an address for which nobody knows the private key.

        returnValue: bytes4 = extcall ERC721Receiver(_to).onERC721Received(msg.sender, _from, _tokenId, _data)

Call the target contract to see if it can receive ERC-721 tokens. Vyper 0.4 requires calls to other contracts to be marked, so the call is prefixed with extcall.

        # Throws if transfer destination is a contract which does not implement 'onERC721Received'
        assert returnValue == method_id("onERC721Received(address,address,uint256,bytes)", output_type=bytes4)

If the destination is a contract, but one that doesn't accept ERC-721 tokens (or that decided not to accept this particular transfer), revert.

By convention if you want not to have an approver you appoint the zero address, not yourself.

    # Check requirements
    senderIsOwner: bool = self.idToOwner[_tokenId] == msg.sender
    senderIsApprovedForAll: bool = (self.ownerToOperators[owner])[msg.sender]
    assert (senderIsOwner or senderIsApprovedForAll)

To set an approval you can either be the owner, or an operator authorized by the owner.

Mint New Tokens and Destroy Existing Ones

The account that created the contract is the minter, the super user that is authorized to mint new NFTs. However, even it is not allowed to burn existing tokens. Only the owner, or an entity authorized by the owner, can do that.

### MINT & BURN FUNCTIONS ###

@external
def mint(_to: address, _tokenId: uint256) -> bool:

This function always returns True, because if the operation fails it is reverted.

Only the minter (the account that created the ERC-721 contract) can mint new tokens. This can be a problem in the future if we want to change the minter's identity. In a production contract you would probably want a function that allows the minter to transfer minter privileges to somebody else.

    # Throws if `_to` is zero address
    assert _to != empty(address)
    # Add NFT. Throws if `_tokenId` is owned by someone
    self._addTokenTo(_to, _tokenId)
    log IERC721.Transfer(sender=empty(address), receiver=_to, token_id=_tokenId)
    return True

By convention, the minting of new tokens counts as a transfer from address zero.

Anybody who is allowed to transfer a token is allowed to burn it. While a burn appears equivalent to transfer to the zero address, the zero address does not actually receives the token. This allows us to free up all the storage that was used for the token, which can reduce the gas cost of the transaction.

Using this Contract

In contrast to Solidity, Vyper does not have inheritance. This is a deliberate design choice to make the code clearer and therefore easier to secure. So to create your own Vyper ERC-721 contract you take this contract (opens in a new tab) and modify it to implement the business logic you want.

Conclusion

For review, here are some of the most important ideas in this contract:

  • To receive ERC-721 tokens with a safe transfer, contracts have to implement the ERC721Receiver interface.
  • Even if you use safe transfer, tokens can still get stuck if you send them to an address whose private key is unknown.
  • When there is a problem with an operation it is a good idea to revert the call, rather than just return a failure value.
  • ERC-721 tokens exist when they have an owner.
  • There are three ways to be authorized to transfer an NFT. You can be the owner, be approved for a specific token, or be an operator for all of the owner's tokens.
  • Past events are visible only outside the blockchain. Code running inside the blockchain cannot view them.

Now go and implement secure Vyper contracts.

See here for more of my work (opens in a new tab).