/mangrove-core

🌴 Core smart contracts for Mangrove with deployment scripts and example offer logics

Primary LanguageSolidityOtherNOASSERTION

CI

This package contains the Solidity implementation of Mangrove as well as deployment and governance scripts.

Documentation

If you are looking for the Mangrove developer documentation, the site to go to is docs.mangrove.exchange.

Use as a foundry dependency

Just forge install mangrovedao/mangrove-core.

⚠️ You will not get the usual remapping mangrove-core/=lib/mangrove-core/src/ (because forge's remapping generation heuristic sees the preprocessing/lib/ directory and decides to remap to the parent dir). Instead, you will get:

@mgv/src/=lib/mangrove-core/src/
@mgv/lib/=lib/mangrove-core/lib/
@mgv/test/=lib/mangrove-core/test/
@mgv/script/=lib/mangrove-core/script/

Use this likely-unique prefix even internally so projects that depend on Mangrove don't mess with Mangrove's internal dependencies.

Installing prerequisites

For Linux or macOS everything should work out of the box, if you are using Windows, then we recommend installing everything from within WSL2 and expect some quirks.

  1. Node.js 14.14+, we recommend installation through nvm, e.g.:

    $ curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.39.1/install.sh | bash
    # Reopen shell
    $ nvm install --lts
  2. Yarn 2, with Node.js >= 16.10:

    $ corepack enable
  3. Foundry:

    $ curl -L https://foundry.paradigm.xyz | bash
    # Reopen shell
    $ foundryup
  4. Clone the git repo with sub-modules

    $ git clone --recurse-submodules https://github.com/mangrovedao/mangrove-core.git
    # Or set the global git config once: git config --global submodule.recurse true

Usage

The following sections describe the most common use cases in this repo.

Initial setup

After cloning the repo, you should run yarn install in the root folder.

$ yarn install

Then you need to setup the local environment (still in the root folder). Start by copying the test file provided:

$ cp .env.example .env

And then open .env in your favorite editor and put in settings for, e.g., node urls, for instance pointing to Alchemy. (The discussion around setting up an environment for testing out the strat library on a local chain on docs.mangrove.exchange might be helpful.)

Build

To build, run

$ yarn build

Addresses and Configuration

The repo supports the use of named addresses. The addresses and names are read from JSON files and, on deployment, written to the same JSON files.

By default, these addresses are located in ./addresses/ and are populated as part of the build process with the addresses in the npm packages mangrove-deployments (Mangrove, periphery, and strat addresses) and context-addresses (eg token addresses).

You can configure where the scripts will look for addresses files and it's possible to point to multiple directories. By default, the scripts will use the paths specified in the {projectRoot}/mgvConfig.json file:

{
  "addresses_paths": ["addresses/"],
  "deployment_addresses_path": "addresses/"
}

The deployment_addresses_path is the path to the folder where the addresses are written to when deploying. The addresses_paths is an array of paths to folders where the addresses are read from. The paths are relative to the project root.

If you want to read addresses from other folders, then you can add paths to the json file. If you do not want to change the mgvConfig.json file, but still want to read some extra addresses, then you you can set MGV_ADDRESSES_PATHS to the addresses paths that should be read from. And if you don't want to read the mgvConfig.json addresses at all, then you can set MGV_READ_ADDRESSES_PATHS to false. The MGV_ADDRESSES_PATHS variable has same structure as the mgvConfig.json. Here is an example:

export MGV_ADDRESSES_PATHS='{ "addresses_paths": ["/addresses/"] }'
export MGV_READ_ADDRESSES_PATHS=false

In this example we disable the default paths and set the path to /addresses/. Remember the path is relative to the project root. This way you can easily read addresses from multiple sources.

When adding paths, then you have to remember to add the path in the foundry.toml file as well, in order for foundry to be able to read from that path.

Tests

To run all tests in the package, just run yarn test.

This package contains a comprehensive test suite for Mangrove, implemented in Solidity using Foundry.

The tests are located in ./test.

Refer to the documentation of Foundry for details on how tests are structured and options for running it.

Foundry and its use in this package

This package relies heavily on the Foundry development framework for Ethereum. It includes an EVM interpreter with special hooks for

  • interpreting console.log-type statements
  • displaying Solidity stack traces on reverts.

For example, you can use console.log in contracts for debugging; those logs survive transaction revert. More in Foundry's. Example:

string memory s = "Hello";
uint n = 31;
console.log("Message %s number %d",s,d);

Deploying on Mangrove

Mangrove uses forge script

forge script <scriptName> executes an arbitrary smart contract function locally. Then, any CALLs executed therein and preceded by the cheatcode vm.broadcast() can be broadcast to a remote node by reading a forge-generated run-*.json field.

Generating Mangrove address files

The log of transactions generated by forge script gets written to broadcast/<scriptName>/<chainId>/run-latest.json. It is an array of low-level transactions info with some additions like newly created contract names.

Mangrove needs to:

  • Name its contract (multiple instances of the same contract may be deployed)
  • Ignore script names (different scripts are used for different networks)
  • Have all its deployed contracts in one place

To do the above, Mangrove adds a layer to forge script deployment.

  • Deployment scripts should inherit the Deployer contract.
  • You should call outputDeployment() at the end of your scripts.
    • When outputDeployment() gets called, a file with all known deployed contracts are written to addresses/deployed.backup/<network>-latest.json
  • You should set WRITE_DEPLOY=true when running scripts that you want to broadcast.
    • When WRITE_DEPLOY=true, the contract set is also written to addresses/deployed/<network>.json.

(Note that for mumbai, network=maticmum)

Foundry keywords for rpc and verification

We use foundry's [rpc_endpoints] and [etherscan] config sections. If the same key exists in both, you can drop the --etherscan-api-key from the commandline arguments. For instance if mumbai is defined in both sections, you can say forge script --fork-url mumbai ... --verify and any deployed contracts will get verified through etherscan using the mumbai API key of the [etherscan] section.

(Note: in this context, etherscan can mean "polygonscan" or any block explorer)

Chain-dependent broadcast with private keys in .env

The vm.broadcast() cheatcode implicitly selects a sender for the broadcast transactions: either the default sender, or the address associated to the --private-key given in the CLI, or to --mnemonic information, etc.

It is tiring to always add --private-key 0x.. to scripts, especially since the key may be different for each network. The Deployer contract has a broadcast() function that reads the <NAME>_PRIVATE_KEY var off the environment, where NAME is the name of the chain you're talking to.

  • You should have vars such as MUMBAI_PRIVATE_KEY , POLYGON_PRIVATE_KEY in your .env file. You can use .env.example as a template.
  • You should use broadcast() instead of vm.broadcast() in scripts. The deployer contract will look for the correct private key, and fallback to the CLI-provided key if none was found.

(Note that for Mumbai, name=”mumbai”)

Generate documentation

The Mangrove Solidity files contain documentation that can be extracted to a nicely formatted and navigable HTML file by running yarn doc which will generate a doc/MgvDoc.html.