dmccartney / micro-erc721

experimenting w/ more gas efficient erc721

Micro ERC721

This contains my WIP experiments creating a smaller footprint ERC721.

TODO package/publish for use, doc install

SizedERC721 - size-optimized ERC721

This is a drop-in replacement for the OpenZeppelin ERC721 that uses storage optimizations for a specified maximum size. This lets it keep gas costs low for mint/transfer/burn while also adding extra functionality (e.g. IERC721Enumerable-tokenOfOwnerByIndex).

Whereas ERC721 supports any uint256 ID and up to 2^256 tokens, this SizedERC721 requires a specified maximum token ID (e.g. 10,000) and token IDs can then only be minted within that range (e.g. 0 - 10,000). This is a typical assumption many projects enforce anyways.

Benchmarks

• ExampleSized runs are instances of SizedERC721.
• OZVanilla runs are instances of OpenZeppelin's ERC721.
• OZEnumerable runs are instances of OpenZeppelin's ERC721Enumerable.
• ChiruERC721 runs are instances of Chiru Lab's ERC721a.

Right now the SizedERC721 has the same gas prices as OpenZeppelin's vanilla ERC721 for mint/burn/transfer. But it does this while also implementing IERC721Enumerable-tokenOfOwnerByIndex. By contrast, the OpenZeppelin edition of ERC721Enumerable (which includes tokenOfOwnerByIndex) dramatically increases the mint/burn/transfers costs by 2x-3x. So this means your SizedERC721 can list a person's tokens on-chain w/o a big spike in the cost of a mint.

Multi-minting

Chiru Lab's ERC721a does some clever stuff to make multi-minting (minting more than one at a time) cheaper up-front. It offloads some of those minting costs on later transfers and requires tokens mint use sequential IDs starting from 0.

I've added multiMinting to the benchmark to track this. Comparing multi-minting costs is a little tricky. Where Chiru is a hair costlier for n = 1 it quickly becomes the cheapest for minting multiple. And if you look at transferFrom, you can see where it offloads that onto later transfers.

TODO doc data structure cleverness / next explorations

• TODO explain EnumerableBitMaps
• TODO explore incorporating Chiru's null-run technique for SizedERC721's _owners (and/or consider contributing to ERC721a to speed up enumeration)
• TODO explore static indexed 8192 (8x32 index (fits into 1 uint256) for a 8x256 bitmap of holdings)

development setup

• $ npm install
• $ npm test
• $ npm run pretty
• $ REPORT_GAS=true npm test