Tables

Each piece of data in Store is stored as a record in a table. You can think of tables in two ways, either as a relational database or as a key-value store.

• Each table is identified by a unique ResourceId tableId.
• Each record in a table is identified by a unique bytes32[] keyTuple. You can think of the key tuple as a composite key in a relational database, or as a nested mapping in a key-value store.
• Each table has a value schema (all the schema fields that aren't part of the key) that defines the types of data stored in the table. You can think of the value schema as the column types in a table in a relational database, or the type of structs stored in a key-value store.

Tables are registered in the Store contract at runtime. When a table is registered, its ID, key and value types, as well as key and field names are stored in the internal Tables table. This emits an event that can be used by offchain indexers to start replicating the state of the new table.

The recommended way of reading from tables and writing to tables is via the typed table libraries. However, it is also possible to use the low-level IStore (external) or StoreCore (internal) API directly.

Types of tables

There are two types of tables in Store: Onchain tables and offchain tables. We often omit the prefix from onchain tables and just call them tables.

As the name suggests, onchain tables store their state onchain, in the Store contract. In addition, an event is emitted on every write operation, to allow offchain indexers to replicate the onchain state.

Offchain tables on the other hand don't store any state onchain, but only emit the events for offchain indexers. This makes them suitable for use cases where data doesn't need to be retrieved onchain, but should still be synchronized from the Store contract to offchain indexers.

Onchain you can write to offchain tables with the same methods as onchain tables, except for reading data and modifying partial data in a dynamic field.

Advanced

Table ID

The table ID encodes the table's type and name in a 32-byte ResourceId value and uniquely identifies a table.

import { ResourceId, ResourceIdLib } from "@latticexyz/store/src/ResourceId.sol";
import { RESOURCE_TABLE, RESOURCE_OFFCHAIN_TABLE } from "@latticexyz/store/src/storeResourceTypes.sol";
 
...
 
// Using the `RESOURCE_TABLE` type makes it an onchain table
ResourceId tableId = ResourceIdLib.encode({
  typeId: RESOURCE_TABLE,
  name: "Balance"
});
 
// Using the `RESOURCE_OFFCHAIN_TABLE` type makes it an offchain table
ResourceId offchainTableId = ResourceIdLib.encode({
  typeId: RESOURCE_OFFCHAIN_TABLE,
  name: "Balance"
});

Key schema

The key schema represents the data types of the (composite) key of the table. Internally the composite key is always handled as bytes32[], but the key schema allows offchain indexers to create properly typed composite key columns, and allows MUD to generate typed table libraries.

The key schema can contain up to 28 static (fixed length) data types. Note that no dynamic (variable length) data types (e.g. arrays, bytes or string) are supported in the key schema.

import { Schema, SchemaLib } from "@latticexyz/store/src/Schema.sol";
import { SchemaType } from "@latticexyz/schema-type/src/solidity/SchemaType.sol";
 
// NOTE: the key schema can only include static types
SchemaType[] memory keySchemaTypes = new SchemaType[](1);
keySchemaTypes[0] = SchemaType.ADDRESS;
Schema keySchema = SchemaLib.encode(keySchemaTypes);

Value schema

The value schema represents the data types of the value fields of the table. Similar to key schemas, it allows offchain indexers to create properly typed value columns, and allows MUD to generate typed table libraries.

The value schema can contain up to 28 fields in total, of which up to 5 can be dynamic (variable length) data types (e.g. arrays, bytes, string). All static (fixed length) data types must come before the dynamic data types.

import { Schema, SchemaLib } from "@latticexyz/store/src/Schema.sol";
import { SchemaType } from "@latticexyz/schema-type/src/solidity/SchemaType.sol";
 
// NOTE: static types must come before dynamic types
SchemaType[] memory valueSchemaTypes = new SchemaType[](1);
valueSchemaTypes[0] = SchemaType.UINT256;
Schema valueSchema = SchemaLib.encode(valueSchemaTypes);

Key and field names

Key and (value) field names are simple string arrays, and must have the same length as the key and value schemas respectively.

string[] memory keyNames = new string[](1);
keyNames[0] = "owner";
 
string[] memory fieldNames = new string[](1);
fieldNames[0] = "balance";

Field layout

The field layout encodes the byte length of each static data type in the table. This is mostly a way to save gas in onchain operations that require the length of static fields, as it avoids additional steps to translate the schema into the corresponding byte length.

import { FieldLayout, FieldLayoutLib } from "@latticexyz/store/src/FieldLayout.sol";
 
uint256[] memory staticFields = new uint256[](1);
staticFields[0] = 32; // The byte length of the first field
FieldLayout fieldLayout = FieldLayoutLib.encode(staticFields, 0);

Manually register a table

Registering a table is a prerequisite for writing to the table, and allows offchain indexers to replicate the onchain state.

import { IStore } from "@latticexyz/store/src/IStore.sol";
 
IStore store = IStore(STORE_ADDRESS);
store.registerTable({
  tableId: tableId,
  fieldLayout: fieldLayout,
  keySchema: keySchema,
  valueSchema: valueSchema,
  keyNames: keyNames,
  fieldNames: fieldNames
});