@magmalayer/bedrocksdk v0.0.29-alpha

@magmalayer/bedrocksdk

BedRockDB SDK to create, manage and query open data.

!WARNING
This SDK is a work-in-progress and is being developed in parallel with the BedRockDB node.\ Things will change, however, the core components have been ported over from the Bedrock Social SDK and should have a stable-enough interface.\ Find any notable or breaking changes in CHANGELOG.md.

Installation

The SDK is available publicly on NPM. You can install it using your preferred package manager.

yarn add @magmalayer/bedrocksdk

Description

BedRockDB SDK is a client-side complement to BedRockDB - a decentralized database built on top of Ceramic.\ It inherits the DX of our TS SDK which enables simple user authentication while providing new (more generic) methods to manipulate data.

Initialize the SDK

Initializing the SDK requires 2 gateways - one for your Ceramic node and another one for your BedRockDB.

import { BedRockDB } from "@magmalayer/bedrocksdk"

const db = new BedRockDB({
    ceramic: {
        gateway: "YOUR_CERAMIC_NODE_URL"
    },
    nodes: [
        {
            gateway: "YOUR_BEDROCK_NODE_URL"
        }
    ]
})

Why is nodes argument an array?

We have plans to support connecting to multiple BedRockDB instances for fallback, load-balancing as well as automatic query rerouting. Currently, only the first node will be used and no node rotation will happen.

Ceramic gateways might be inferred from the BedRockDB node's metadata in the future, however, we want to make sure exposing the Ceramic node is optional to ensure privacy and security of your infrastructure.

Handling errors

try / catch

Standard try/catch practices apply.

let document
try{
    document = await bedrock.insert(...).run()
}catch(error){
    console.log("Error", error)
}

console.log("Result", document)

catchError

This is a utility method provided by Bedrock, originally implemented in Radash. We've modified the call signature to make it more convenient for our use case.

import { catchError } from "@magmalayer/bedrocksdk"

const [document, error] = await catchError(
    () => bedrock.insert(...).run()
)

if(error){
    console.warn("Error", error)
}

console.log("Result", document)

User authentication

Authentication is handled by BedrockAuthenticators which generate the DID session in did:pkh (BedrockEVMAuth, BedrockSolanaAuth, BedrockTezosAuth) and did:key (BedrockKeyDidAuth) formats.

By default, sessions are persisted in localStorage and are valid for up to 3 months.\ In order to bypass this behavior, pass { saveSession: false } to the connectUser method.

EVM (did:pkh)

import { BedRockDB } from "@magmalayer/bedrocksdk"
import { BedrockEVMAuth } from "@magmalayer/bedrocksdk/auth"

// Browser provider
const provider = window.ethereum

// Ethers provider
const provider = new Wallet(...)

// Bedrock Authenticator
const auth = new BedrockEVMAuth(provider)

// Authenticate the user and persist the session in localStorage
const authResult: BedrockConnectResult = await bedrock.connectUser({ auth })

// Authenticate, but don't persist the session in localStorage
const authResult: BedrockConnectResult = await bedrock.connectUser({ auth, saveSession: false })

// Log the result
console.log({ authResult })

KeyDid (did:key)

import { BedRockDB } from "@magmalayer/bedrocksdk"
import { BedrockKeyDidAuth } from "@magmalayer/bedrocksdk/auth"

// Generate the seed
const seed = await BedrockKeyDidAuth.generateSeed()

// Initiate the authneticator using the generated (or persisted) seed
const auth = await BedrockKeyDidAuth.fromSeed(seed)

// Authenticate the user and persist the session in localStorage
const authResult: BedrockConnectResult = await bedrock.connectUser({ auth })

// Authenticate, but don't persist the session in localStorage
const authResult: BedrockConnectResult = await bedrock.connectUser({ auth, saveSession: false })

// Log the result
console.log({ authResult })

Check if a user is connected

This method always returns true/false.

// Check if any user is connected
const connected = await bedrock.isUserConnected()

// Check if a user with the specified wallet address is connected
const connected = await bedrock.isUserConnected("0x00...")

Get the currently connected user

This method either returns the currently connected user (BedrockConnectResult) or false.

// Get the currently connected user
const currentUser = await bedrock.getConnectedUser()
if(!currentUser){
  // Notify the user or reconnect
  throw "There is no active user session."
}

console.log({ currentUser })

Managing data

BedRockDB SDK makes creating, updating and reading data simple and consistent.\ We took inspiration from Web2 SDKs from solutions like Supabase/PostgREST, Knex, MongoDB, etc.

Operations are divided into insert, update and select.

All methods allow you to use friendly model names if you have them set up in the connected BedRockDB node.\ Contexts are also a Ceramic-native feature and are exposed in all data management methods.

Method chaining is being used to construct queries with all methods and a .run() method executes the chain.

DELETE statement-equivalent is WIP as we're looking to solve this at the core protocol layer.

INSERT

Inserts execute Ceramic MID writes. This has been abstracted using a query-builder interface to simplify execution and allow optimizations of the underlying calls in the future, without modifying the original interface.

Insert a single row

const insertStatement = await bedrock
    .insert("MODEL_ID" | "TABLE_NAME")
    .value(
        {
            column: value,
            column2: value2,
        }
    )
    // optionally, you can scope this insert to a specific context
    .context("CONTEXT_ID")

// Perform local JSON Schema validation before running the query
const validation = await insertStatement.validate()
if(!validation.valid){
    throw "Error during validation: " + validation.error
}

const [result, error] = await catchError(() => insertStatement.run())

// All runs of a statement are stored within the statement, in case you want to reuse the same statmenet
console.log(insertStatement.runs)

Insert multiple rows

const insertStatement = await bedrock
    .bulkInsert("MODEL_ID" | "TABLE_NAME")
    .values(
        {
            column: value,
            column2: value2,
        },
        {
            column: value,
            column2: value2,
        },
        ...
    )
    .value(
        {
            column: value,
            column2: value2,
        }
    )

// Perform local JSON Schema validation before running the query
const validation = await insertStatement.validate()
if(!validation.valid){
    console.error("Errors during validation", validation.errors)
    throw "Errors during validation"
}

// bulkStatements DO NOT throw in case a run partially fails
// As each insert is handled as an isolated case, you may have partial-success
const { success, errors } = await insertStatement.run()

if(errors.length){
    console.error("Errors occurred during execution", errors)
}

console.log(success)

// All runs of a statement are stored within the statement, in case you want to reuse the same statmenet
console.log(insertStatement.runs)

UPDATE

Updates can replace the entire row or perform shallow merging with existing data.

Replace a row

// This will replace the provided row with provided values
const updateStatement = await bedrock
    .update("DOCUMENT_ID")
    .replace(
        {
            column: value,
            column2: value2,
        }
    )

const [result, error] = await catchError(() => updateStatement.run())

// All runs of a statement are stored within the statement, in case you want to reuse the same statmenet
console.log(updateStatement.runs)

Update a row partially

// This will perform a shallow merge before updating the document 
// { ...oldContent, ...newContent }
const updateStatement = await bedrock
    .update("DOCUMENT_ID")
    .set(
        {
            column: value,
        }
    )

const [result, error] = await catchError(() => updateStatement.run())

// All runs of a statement are stored within the statement, in case you want to reuse the same statmenet
console.log(updateStatement.runs)

SELECT

Querying data is done using a custom-built query builder.\ The interface has been kept simple and familiar, as it mimics popular QB solutions such as Knex.js.

Query is being sent to the BedRockDB node in JSON format where it gets parsed and executed.

You can preview the final query by using .build().

Why a custom query builder?

Our initial POCs were using existing QB solutions such as Knex.js and waterfall/JSON SQL builders.\ However, these libraries are built with backend environments in mind and made our query interface more complex, as we aren't executing queries against a DB engine directly.

Building a custom QB gave us the option to separate query building, serializing and final SQL outputs.\ It also allows us to expose custom options such as .context() and .contexts(), further abstracting the underlying data model and making future optimizations and changes in the node easier.

We also did not require multiple engine support and we kept our dependencies to the minimum.

We will keep expanding QB functionality with simple joins, new operators and other features that will make interacting with BedRockDB simpler and more efficient.

Building a SELECT query

const selectStatement = await bedrock
    // SELECT column1, column2
    // if no columns are passed, all columns (*) will be returned
    .select("column1", "column2")
    // FROM model_id | table_name | view_id
    .from("MODEL_ID" | "TABLE_NAME" | "VIEW_ID")
    // WHERE ...conditions
    // unless specified, all conditions will be treated as logical AND
    .where(
        {
            // column = "value"
            column: "value",
            // columns2 in (value1, value2)
            column2 = ["value1", "value2"]
        }
    )
    // you can scope this query to a specific context
    .context("CONTEXT_ID")
    // or multiple contexts
    .contexts("CONTEXT_ID", "CONTEXT_ID", ...)
    // ORDER BY
    .orderBy(
        // orderBy a single column
        ["column", "asc" | "desc"],
        // or multiple columns by providing additional
        // column arrays
        ["column2", "asc" | "desc"]
    )
    // LIMIT
    .limit(number)
    // OFFSET
    .offset(number)

const query = selectStatement.build()
console.log("Query that will be run", query)

const [result, error] = await catchError(() => selectStatement.run())
if(error){
    throw error
}

// columns: Array<string>
// rows: Array<T | Record<string, any>>
const { columns, rows } = result

console.log({ columns, rows })

Using operators

Operator helpers are exposed to provide query flexibility.\ These include logical, comparison and aggregation operators.

You can find the entire list of operators and resulting queries here.

import { count, sum, contains, ilike, or, gte } from "@magmalayer/bedrocksdk/operators"

const selectStatement = await bedrock
    // if no columns are passed, all columns (*) will be returned
    .select(
        "column1", 
        "column2", 
        sum("column3"), 
        count("column4", "count_column4")
    )
    .from("MODEL_ID" | "TABLE_NAME" | "VIEW_ID")
    // unless specified, all conditions will be treated as logical AND
    .where(
        {
            // column = "value"
            column: "value",
            // columns2 in ("value1", "value2")
            column2 = ["value1", "value2"],
            // column3 ILIKE "%value"
            column3: ilike("%value"),
            // column4 LIKE "%value%"
            column4: contains("value"),
            // column5 >= 5
            column5: gte(5),
            // column = "value" OR column2 = "value2"
            ...or(
                {
                    column: "value"
                },
                {
                    column2: "value2"
                }
            )
        }
    )

const query = selectStatement.build()
console.log("Query that will be run", query)

const [result, error] = await catchError(() => selectStatement.run())
if(error){
    throw error
}

// columns: Array<string>
// rows: Array<T | Record<string, any>>
const { columns, rows } = result

console.log({ columns, rows })