Frequently Asked Questions
Frequently asked questions related to viem.
TL;DR: viem tries to avoid creating unnecessary abstractions on top of existing systems.Feel free to add to this document if you notice frequently asked questions that are not covered here.
Why use the terms "Wallet" & "Account" instead of "Signer"
viem attempts to align to the "Wallet" and "Account" terminology on Ethereum.org. The term "Signer" was adapted from ethers.js.
Let's clear up on some terms before we dive in.
- Wallet: An application or interface that holds Account(s).
- Account: An object that represents an address, balance, nonce, and optional storage and code.
- Private Key: Proves ownership of an Account, and can sign messages & transactions.
In the context of viem, a Wallet Client is an interface that can hold an Account. The Account may or may not hold a Private Key.
In viem, there are two types of Accounts:
- Local Account: can synchronously & directly sign messages and transactions using its Private Key. A signature is guaranteed.
- JSON-RPC Account: asynchronously requests signing of messages and transactions from the target Wallet over JSON-RPC (e.g. Browser Extension or WalletConnect). The target Wallet holds the Account & Private Key. A signature is not guaranteed (the target Wallet may not have permitted the Account, or the Wallet may have rejected the request).
We do not use the term "Signer" because there are noticeable behavioral differences between signing locally and signing over JSON-RPC.
Why are contract function args
with fully-named inputs represented as unnamed tuple types instead of object types?
Let's look at an example! Suppose I have the following function in my contract:
function transferFrom(address sender, address recipient, uint256 amount) returns (bool)
All the inputs are named (sender
, recipient
, and amount
) so I might be tempted to represent the parameters as the following TypeScript type:
type Args = {
sender: `0x${string}`;
recipient: `0x${string}`;
amount: bigint;
}
This improves developer experience a bit because now I can see the names of the parameters in my editor.
import { createWalletClient, parseAbi } from 'viem'
const client = createWalletClient(…)
client.writeContract({
address: '0x…',
abi: parseAbi([
'function transferFrom(address sender, address recipient, uint256 amount) returns (bool)',
]),
functionName: 'transferFrom',
args: {
sender: '0x…',
recipient: '0x…',
amount: 100n,
},
})
However, this only works if all the inputs are named (some compilers will strip names from inputs). If any of the inputs are unnamed, then you'll have to use a tuple instead:
client.writeContract({
address: '0x…',
abi: parseAbi([
'function transferFrom(address, address, uint256) returns (bool)',
]),
functionName: 'transferFrom',
args: ['0x…', '0x…', 100n],
})
This can get even more complicated when a function has overrides:
function safeTransferFrom(address, address, uint256) {}
function safeTransferFrom(address from, address to, uint256 tokenId, bytes data) {}
In this case, the type of the overload parameters start to diverge from each other:
type Args =
| [`0x${string}`, `0x${string}`, bigint]
| {
from: `0x${string}`;
to: `0x${string}`;
tokenId: bigint;
data: string;
}
If you want to switch between the two overloads in your code, you'll need to completely change the type instead of just adding or removing a single positional argument from the end. (Objects also don't enforce type-level ordering so you can put them in whatever order you want. This would also mean that viem would also need to internally validate order during runtime, adding some extra overhead.)
client.writeContract({
address: '0x…',
abi: parseAbi([
'function safeTransferFrom(address, address, uint256)',
'function safeTransferFrom(address from, address to, uint256 tokenId, bytes data)',
]),
functionName: 'safeTransferFrom',
- args: ['0x…', '0x…', 100n],
+ args: {
+ from: '0x…',
+ to: '0x…',
+ tokenId: 100n,
+ data: '0x…',
+ },
})
Even though overloads are an edge case, it would be sufficiently astonishing to come across this behavior. So what's the best way to represent args
? Well, they are positional at the contract-level so it makes sense to represent them that way in viem too.
Not all is lost when it comes to developer experience though! Tuple types in TypeScript can have names attached to them:
type Args = [from: `0x${string}`, to: `0x${string}`, tokenId: bigint]
These names show up in your editor so you get nice developer experience when using autocomplete, etc. Unfortunately, TypeScript doesn't support dynamic named tuples right now, but we are watching this issue closely and once it is implemented, we will add it to viem. In the meantime, hang tight!
Why is a contract function return type returning an array instead of an object?
Suppose your ABI looks like this:
[
{
inputs: [],
name: "latestRoundData",
outputs: [
{ name: "roundId", type: "uint80" },
{ name: "answer", type: "int256" },
{ name: "startedAt", type: "uint256" },
{ name: "updatedAt", type: "uint256" },
{ name: "answeredInRound", type: "uint80" },
],
stateMutability: "view",
type: "function",
}
]
You might be confused why the following does not return an object:
import { createPublicClient, parseAbi } from 'viem'
const client = createPublicClient(…)
const res = await client.readContract({
address: '0x…',
abi: […], // abi from above
functionName: 'latestRoundData',
})
res
// ^? const res: [bigint, bigint, bigint, bigint, bigint]
This is expected. "latestRoundData"
outputs
is an array of types, so you get an array of decoded values as the return type. viem only maps explicitly typed tuples as objects
Why does viem follow this approach? Here is the contract function definition for latestRoundData
with two different return types:
function latestRoundData() external view
returns (
uint80 roundId,
int256 answer,
uint256 startedAt,
uint256 updatedAt,
uint80 answeredInRound
);
struct Data {
uint80 roundId;
uint256 answer;
uint256 startedAt;
uint256 updatedAt;
uint80 answeredInRound
}
function latestRoundData() external view returns (Data data);
The first function returns a set of five items, so viem maps it to an array. The reason why we don't convert it to an object is because things get ambiguous when we come to decode structs. How do you determine the difference between a "return" tuple (first function) and a "struct" tuple (second function).
Another reason is that folks might expect it to be an array (because it is a set of return items). Other libraries, like ethers, mitigate this by returning a hybrid Array/Object type, but that kind of type is not serializable in JavaScript, and viem prefers to not try and "hack" JavaScript types.
Why doesn't Wallet Client support public actions?
Wallet Client doesn't support public actions because wallet providers (Injected window.ethereum
, WalletConnect v2, etc.) may not provide a large majority of "node"/"public" RPC methods like eth_call
, eth_newFilter
, eth_getLogs
, etc. This is because these methods are not required for a wallet provider to function properly. For example, a wallet provider may only support eth_sendTransaction
and eth_sign
and nothing else.