apollo-ethereum

What is apollo-ethereum?

apollo-ethereum is the missing link between on-chain contract data and your dApp. It allows you to express your data requirements in declarative GraphQL, while handling data retrieval and common behaviours for you. The result is simplified access to the Ethereum network, and significantly reduced frontend application logic.

Installation | Usage | Features

Example

Given the following contract:

contract SimpleStorage {
  uint256 storedData = 42;

  function set(uint256 x) public {
    storedData = x;
  }

  function get() public view returns (uint256) {
    return storedData;
  }
}

At build-time, apollo-ethereum will generate the GraphQL Schema:

type Mutation {
  SimpleStorage(address: Address!): SimpleStorageMutative
}

type Query {
  SimpleStorage(address: Address!): SimpleStorage
}

type SimpleStorage {
  _address: Address
  get: BigNumber
}

type SimpleStorageMutative {
  set(x: BigNumber): Boolean
}

You may then import the generated schema (and associated contract ABIs) into your app, to run queries and mutations against Ethereum using GraphQL syntax. For example, you might call the set method using the following GraphQL query:

mutation($value: BigNumber!) {
  SimpleStorage(address: "0x123...") {
    set(x: $value)
  }
}

Calls to Ethereum (via the configured web3 provider) will occur automatically, and you may code against loading and error states as per standard Apollo GraphQL client behaviour.

Installation

apollo-ethereum includes both build-time and runtime modules, therefore it must be installed as a production dependency using either:

yarn add apollo-ethereum

or

npm install apollo-ethereum --save

Usage

Building the GraphQL schema

The apollo-ethereum binary is responsible for converting a folder of compiled Solidity contracts into the final GraphQL schema. First, ensure your contracts have been compiled into a build folder, then:

1. Create a configuration file

In the root of your project, create a file named eth.config.yaml. Add an entry for each contract you wish to include in the generated schema. A minimal configuration file might look like the folllowing:

contracts:
  SimpleStorage: true

NB: See further below for Features which allow you to configure the schema output and behaviour for each contract

2. Run the schema generator

The included binary accepts input in the form of:

apollo-ethereum [contracts_dir] [output_dir]

This script will place two new files in the output dir:

1. index.js - configuration module to be included in your runtime build
2. ethereum.graphql - a superfluous schema file which is not required, but may be used by various GraphQL tools to implement additional GraphQL-related tooling. For example, query linting or generating types for use with your application. Note: Recipes for these tasks will be available shortly.

Configuring the Apollo GraphQL client

Apollo Client is agnostic to the frontend framework in-use, therefore only the base client configuration instructions are provided. For integrating your framework with Apollo Client, see the documentation for your relevant binding library e.g. react-apollo

In the following example code, graph is the name of the output directory.

import { ApolloClient } from 'apollo-client'
import { createEthereumLink } from 'apollo-ethereum'
import { InMemoryCache } from 'apollo-cache-inmemory'
import ethereumConfig from './graph'

const link = createEthereumLink(ethereumConfig, {
  provider: window.ethereum,
  // additional feature configuration (see below)
})
const cache = new InMemoryCache()
const client = new ApolloClient({ link, cache })

Features

To see these features in action, refer to the tests directory.

ERC1820 interface address registry

If you notice in all examples of this readme, the generated schema requires an address to be provided for the requested contract. However, if the contract is registered with an ERC1820 registry, and you know the interface name of the contract, you are able to skip the provision of a hard-coded address and allow apollo-ethereum to lookup the contract's address automatically. You can update the config YAML with the contracts interface name as follows:

contracts:
  MyERC20Token:
    interfaceName: ERC20

This will make the address argument optional for the given contract, allowing you to perform a query without hard-coding an address:

query {
  MyERC20Token {
    symbol
    decimals
  }
}

To enable this piece of functionality, you must provide the registry address and lookup function method name when instantiating the Apollo Link, e.g.

const link = createEthereumLink(ethereumConfig, {
  provider: ethereum,
  erc1820: {
    lookupAddress: process.env.ERC1820_ADDR,
    lookupMethod: 'policyFor',
  }
})

Type conversion

The schema generation process makes use of both the contract ABI and AST to determine input and output types, however there may be some scenarios where a more suitable type can be selected. For example, a date stored as uint256, or perhaps the contract is storing an int256 value that can be safely converted to a JavaScript-native number without requiring the use of the BigNumber library.

The config YAML allows you to specify type overrides for output fields. You may choose from built-in GraphQL types (String, Int, etc) or scalar types that come bundled with apollo-ethereum, listed as follows:

• Bytes
• Address
• BigNumber
• Timestamp

BigNumber and Timestamp will return deserialized data in the appropriate data structure (BigNumber and Date respectively).

Contract source

contract Coercion {
  /* uint to Int */
  uint256 public smallInteger = 2048;

  /* uint to Timestamp */
  uint256 public createdAt = now;
}

eth.config.yaml

contracts:
  Coercion:
    fields:
      smallInteger: Int
      createdAt: Timestamp

Generated schema

type Coercion {
  _address: Address
  createdAt: Timestamp
  smallInteger: Int
}

Example query

query {
  Coercion(address: "0x123...") {
    createdAt # returns an instance of JS-native Date
    smallInteger # converts the BigNumber into a JS-native Number
  }
}

Mappings and arrays

When mapping and array state variables are compiled, each entry in the mapping or array must be accessed individually, using either a key (mapping) or index (array). For arrays, you may be able to create a function that returns an array of all members, as long as the array members are a primitive type.

Contract source

contract Types {
  uint8[] public ints = [1, 2, 3, 4, 5];

  mapping(bytes3 => uint256) public tlaScore;

  function allInts() public view returns (uint8[] memory) {
    return ints;
  }
}

eth.config.yaml

The schema generation process handles this automatically - no additional configuration is required.

Generated schema

type Types {
  _address: Address
  ints(index: BigNumber!): Int
  tlaScore(key: Bytes!): BigNumber
  allInts: [Int]
}

Example query

query {
  Types(address: "0x123...") {
    ints(index: 1) # returns second member of array
    tlaScore(key: "tla") # returns BigNumber located at given key of mapping
    allInts # returns an array of all integers
  }
}

Mapping index pattern

A common task in querying a contract mapping is attempting to return all known members of a mapping data structure. To attempt to retrieve each individually would likely result in first retrieving the known keys for a mapping, then asynchronously iterating through each known key to retrieve the associated member. Thankfully apollo-ethereum provides a quick & convenient way to retrieve many members of a mapping, using the "mapping index" pattern. A mapping index is an array containing known keys for a mapping, which can be used to retrieve the associated mapping entry for each key.

Contract source

contract Index {
  struct Profile {
    string name;
    string email;
  }

  /* the mapping */
  mapping(address => Profile) public profiles;

  /* the mapping index */
  address[] _profilesIndex;

  function addProfile(address _addr, string memory _name, string memory _email) public {
    profiles[_addr] = Profile(_name, _email);
    _profilesIndex.push(_addr);
  }
}

eth.config.yaml

contracts:
  Index:
    fields:
      allProfiles:
        mappingIndex: profiles

Generated schema

type Index {
  _address: Address
  profiles(key: Address!): Index_Profile
  allProfiles: [Index_allProfiles] @mappingIndex(mapping: "profiles")
}

type Index_allProfiles {
  key: Address
  value: Index_Profile
}

type Index_Profile {
  name: String
  email: String
}

Example query

To retrieve all profiles in the mapping:

query {
  Index(address: "0x123...") {
    allProfiles {
      key
      value {
        name
        email
      }
    }
  }
}

Contracts using selfdestruct

Contracts which make use of selfdestruct cannot have state variables queried after they have been destroyed. The only pieces of contract data which remain available after their destruction are the events produced during active state on the Ethereum network. apollo-ethereum allows you to configure a selfdestruct contract and produces two types during generation - one representing a contract "active" state, and another representing the same contract after it has been destroyed, otherwise known as the "complete" state.

Contract source

contract SelfDestruct {
  string public message = "only available on active contracts";

  event Log(string message);

  constructor() public {
    emit Log("Creating the contract");
  }

  function destroy() public {
    emit Log("Destroying the contract");
    selfdestruct(msg.sender);
  }
}

eth.config.yaml

contracts:
  SelfDestruct:
    canSelfDestruct: true

Generated schema

type Mutation {
  SelfDestruct(address: Address!): SelfDestructMutative
}

type Query {
  SelfDestruct(address: Address!): SelfDestruct
}

interface SelfDestruct {
  _address: Address
  Log: [SelfDestruct_Log]
}

type SelfDestruct_Log {
  message: String
  _timestamp: Timestamp
}

type SelfDestructActive implements SelfDestruct {
  _address: Address
  Log: [SelfDestruct_Log]
  integer: Int
  message: String
}

type SelfDestructComplete implements SelfDestruct {
  _address: Address
  Log: [SelfDestruct_Log]
}

type SelfDestructMutative {
  destroy: Boolean
}

Example query

Notice the use of a fragment to specify which fields should be queried, only when the contract is still active on the chain. Using this syntax ensures that the requested data is only fetched from the chain if the contract has not yet self-destructed.

Additionally, you may want to make use of the GraphQL-level field __typename to determine whether the contract at the given address is active or not. It will return a string set to either '<CONTRACT_NAME>Active' or '<CONTRACT_NAME>Complete'.

query {
  SelfDestruct(address: "0x123...") {
    __typename
    _address
    Log {
      _timestamp
      message
    }
    ... on SelfDestructActive {
      message
    }
  }
}

Contract relationships

By far, the most helpful feature of apollo-ethereum is the ability to model relationships between contracts. A common pattern seen in Ethereum is storing the address of Contract B in the state data of Contract A. In this scenario, we have a relationship between both contracts that is well-suited for GraphQL modeling.

Contract source

Car.sol

contract Car {
  string public name;

  address[] public reviews;

  constructor(string memory _name) public {
    name = _name;
  }

  function addReview(address _address) public {
    reviews.push(_address);
  }

  function allReviews() public view returns (address[] memory) {
    return reviews;
  }
}

CarReview.sol

contract CarReview {
  enum Rating { Bad, Good }

  Rating public rating;
  string public review;

  constructor(Rating _rating, string memory _review) public {
    rating = _rating;
    review = _review;
  }
}

eth.config.yaml

We can create a one-to-many link from Car to CarReview, using the following configuration in eth.config.yaml (making use of type overrides, as demonstrated above):

contracts:
  CarReview: true
  Car:
    fields:
      reviews: CarReview
      allReviews: CarReview

Generated schema

This will produce the following GraphQL schema, giving you the ability to reach through contract relationships, and fetch data from related contracts in a single query.

type Car {
  _address: Address
  name: String
  reviews(index: BigNumber!): CarReview @contract
  allReviews: [CarReview] @contract
}

type CarMutative {
  addReview(address: Address): Boolean
}

type CarReview {
  _address: Address
  rating: CarReview_Rating
  review: String
}

enum CarReview_Rating {
  Bad
  Good
}

type Mutation {
  Car(address: Address!): CarMutative
}

type Query {
  Car(address: Address!): Car
  CarReview(address: Address!): CarReview
}

Example query

query {
  Car(address: "0x123...") {
    _address
    name
    allReviews {
      _address
      rating
      review
    }
  }
}

Contributing

PRs accepted.

License

MIT © Eco