Dynamic primitives
This primitive allows registering new extensions in runtime, without changing the configuration files. These are triggered by a generic smart contract event. A use-case for this would be a factory contract.
Currently the dynamic extensions can be one of:
Configuration
ERC721 example
# extensions.yaml
extensions:
- name: "Dynamic erc721"
type: "dynamic-primitive"
startBlockHeight: 100
contractAddress: "0x5FbDB2315678afecb367f032d93F642f64180aa3"
eventSignature: "ERC721Created(address)"
abiPath: "./FactoryERC721.json"
targetConfig:
type: "erc721"
scheduledPrefix: "nft"
burnScheduledPrefix: "nftburn"
dynamicFields:
contractAddress: nftAddress
Meaning
There are two type of fields in this configuration. The top level fields are used for the extension that have the role of monitoring the network for the trigger event. These are the following:
contractAdress
is the address of the smart contract that will emit the event that will trigger the erc721 primitive creation.abiPath
, specifying a path to a .json file describing the compiled contract – the only field required in this file is theabi
field in the top-level object;eventSignature
, specifying the signature of the event consisting only of the event name followed by parameter types (without names) in order, enclosed in parentheses and separated by commas.
The nested fields instead are used to construct the configuration for the
dynamically generated extensions. It's divided into a static part, which is in
targetConfig
. And a dynamic part which is in dynamicFields
. Dynamic in this
context means that it depends on the data included on the emitted event.
Static configuration
-
scheduledPrefix
is the prefix used for the events emitted by the dynamic extensions. This primitive won't emit events by itself. -
For the erc721 case, there is also
burnScheduledPrefix
, for which the comment above also applies. -
For the generic primitive the
targetConfig
field would look like this instead:
targetConfig:
type: "generic"
abiPath: "./abis/MyCustomContract.json"
eventSignature: "MyEvent(address,uint256)"
scheduledPrefix: "cst"
Dynamic configuration
contractAddress
has the name of the parameter in the emitted event that contains the address for the new primitive.
Factory contracts
Solidity has no standardized way to create a factory contract. Instead, it's up to you to create your own factory contract whatever way works best, and to emit events that trigger the dynamic primitive inside it. The example configuration assumes that the event (such as in your factory contract) has the following signature:
// note: the name of the argument (nftAddress) has to match the name specified in dynamicFields
event ERC721Created(address indexed nftAddress);
Which could be part of the following implementation.
// SPDX-License-Identifier: MIT
pragma solidity ^0.8.20;
import {IERC721} from "@openzeppelin/contracts/token/ERC721/IERC721.sol";
import {ERC721} from "@openzeppelin/contracts/token/ERC721/ERC721.sol";
contract CustomERC721 is ERC721 {
constructor(string memory name, string memory symbol) ERC721(name, symbol) {}
}
contract FactoryERC721 {
event ERC721Created(address indexed nftAddress); // emitted when ERC721 token is deployed
/// @dev deploys an ERC721 token with given parameters
/// @return address deployed
function deployERC721(string memory _name, string memory symbol) public returns (address) {
CustomERC721 t = new CustomERC721(_name, symbol);
emit ERC721Created(address(t));
return address(t);
}
}
NOTE: Extra parameters in the event will be ignored, and the position does not matter.
Utility functions
Generated extensions have a name, that is derived from the name of the trigger extension, and the order of the dynamic primitive. The following function can be used to generate this name.
export function generateDynamicPrimitiveName(parentName: string, id: number): string;
For example, the first dynamic primitive that gets created will have a name of
generateDynamicPrimitiveName(parentCdeName, 0)
, and so on. This name can be
then used as an argument to the other utility functions.
There is also a public utility function to get the list of all of the
dynamically generated extensions. The config
field has a JSON with the
concrete parameters for that particular extension. These are the ones that would
be in the configuration file if
the primitive was not dynamic. The only exception is the name
field, which is
instead provided as a separated field.
export async function getDynamicExtensions(
readonlyDBConn: Pool,
parent: string
): Promise<{ name: string; config: string }[]>;
There is also the following function, which can be used to get the parameters
for a specific primitive by name. This can be used to access the contract
address in the state transition
function, by first getting
the name from the inputData.extensionName
field.
export async function getDynamicExtensionByName(
readonlyDBConn: Pool,
name: string
): Promise<{ contractAddress: string; startBlockHeight: number }[]>;
Performance implications
During a typical execution, the steps are as follows:
- Fetch all primitives for a block range concurrently (depending on
funnelBlockGroupSize
orpresyncStepSize
). - Construct the blocks from the fetched data to feed to the state machine.
However, dynamic primitive can change the set of primitives the rollup monitors (and therefore change what has to be fetched for a given block range). Therefore, if there are dynamic primitives configured, funnels will instead
- Request all of the events for dynamic primitives and update the set of primitives tracked by the rollup
- Fetch primitives (both the static ones and the new ones after the dynamic primitive update)
- Send blocks to the state machine as before
This extra step introduces a performance reduction (whose impact depends how long it takes to fetch data from the node your game is connecting to)