@naturalcycles/db-lib

Lowest Common Denominator API to supported Databases

Defines 3 things:

• CommonDB interface
• CommonDao class
• DBQuery class

CommonDB serves as a Lowest Commond Denominator between different DB implementations (see further). So you can use same syntax, e.g getById<DBM>(id: string): Promise<DBM | undefined> across different DBs.

DBQuery allows to use the same query syntax across different DBs! E.g:

const q = DBQuery.create('table1')
  .filterEq('type', 'cat')
  .filter('updated', '>', '2019-01-17')
  .order('name', true)

await db.runQuery(q)

So, you can run it against Datastore, Firestore, Redis, MongoDB, Airtable, etc. Different DBs, same syntax!

You can swap DB implementations without changing your application code. Migrate Datastore to Firestore? Easy.

You can test your code against InMemoryDB (that implements full CommonDB interface, even with querying, streaming, etc). So, your unit tests can use exactly same querying syntax, or even exactly same services, DAOs. Just swap real DB with InMemoryDB in your setupJest.ts (for example).

Supported databases

• InMemoryDB (with optional file persistence, Redis-like; implemented in this package)
• datastore-lib (GCP Datastore, or Firestore in Datastore mode)
• firestore-lib (Firestore in Native mode)
• mysql-lib (MySQL)
• redis-lib (Redis)
• mongo-lib (MongoDB)
• airtable-lib (Airtable)
• HttpDB (CommonDB exposed via REST API, implemented in backend-lib)
• spreadsheet-lib "Google Spreadsheets as a Database"
• github-db "github branch as a Database"
• sqlite-lib SqliteDB (in progress), SqliteKeyValueDB (done)

Features

• CommonDB, CommonDao, DBQuery
• Streaming (Node.js streams with backpressure)
• DBM / BM, validation, conversion (Joi-powered)
• Conventions
  • String ids
  • created, updated (unix timestamps)
  • Dates as ISO strings, e.g 2019-06-21
  • Timestamps as unixtimestamps (seconds, not milliseconds; UTC)
  • Complex objects as JSON serialized to string (DBM), converted to object (BM)

Concept

CommonDB is a low-level API (no high-level sugar-syntax). CommonDao is the opposite - a high-level API (with convenience methods), built on top of CommonDB.

Concerns of CommonDB:

• Access to DB (all tables): CRUD (create, read, update, delete)
• Batch methods (cause they can be more optimal if implemented "natively")
• Querying
• Streaming

Concerns of CommonDao:

• Access to one DB Table ("kind")
• Transformation between DBM and BM, validation/conversion
• Auto-generating id, created, updated fields
• Anonymization hook to be able to plug your implementation (privacy by design)

CommonDB API

• ping
• getByIds
• runQuery
• runQueryCount
• streamQuery
• saveBatch
• deleteByIds
• deleteByQuery
• getTables
• getTableSchema
• createTable

ping

ping(): Promise<void>

Call this to check that DB connection, credentials, configuration is working. Should throw an error if any of above is invalid.

getByIds

getByIds<DBM>(table: string, ids: string[]): Promise<DBM[]>

await db.getByIds('table1', ['id1, 'id2'])
// [ { id: 'id1', ... }, { id: 'id2', ... } ]

Should return items in the same order as ids in the input.

Only returns items that are found, does not return undefined (absent) items.

runQuery

runQuery<DBM>(q: DBQuery<DBM>): Promise<RunQueryResult<DBM>>

const q = DBQuery.create('table1').filterEq('type', 'cat').order('name', true) // desc

await db.runQuery(q)
// { records: [ { ... }, { ... }, ... ] }

runQueryCount

runQueryCount(q: DBQuery): Promise<number>

await db.runQuery(DBQuery.create('table1'))
// 5

streamQuery

streamQuery<DBM>(q: DBQuery<DBM>): ReadableTyped<DBM>

Returns ReadableTyped (typed wrapper of Node.js Readable).

Streams in Node.js support back-pressure by default (if piped properly by the consumer).

const q = DBQuery.create('table1') // "return all items" query

await _pipeline([
  db.streamQuery(q),
  writableForEach(item => {
    console.log(item)
  }),
])

// { item1 }
// { item2 }
// ...

Alternative:

await db.streamQuery(q).forEach(item => {
  console.log(item)
})

saveBatch

saveBatch<DBM>(table: string, dbms: DBM[]): Promise<void>

Since CommonDB is a "minimal API", there's no save method for a single item, only for multiple. Pass an array with single item to save just one item.

const items = [
  { item1 },
  { item2 },
]

await db.saveBatch('table1', items) // returns void
await db.runQuery(DBQuery.create('table1') // "get all" query
// [ { item1 }, { item2 } ]

deleteByIds

deleteByIds(table: string, ids: string[]): Promise<number>

Returns number of deleted items (not all CommonDB implementations support that).

await db.deleteByIds('table1', ['id1', 'id2'])
// 2

deleteByQuery

deleteByQuery(q: DBQuery): Promise<number>

Returns number of deleted items.

await db.deleteByQuery(DBQuery.create('table1'))
// 2

getTables

getTables(): Promise<string[]>

await db.getTables()
// [ 'table1', 'table2' ]

getTableSchema

getTableSchema(table: string): Promise<JsonSchemaObject>

await db.getTableSchema('table1')

Returns a JsonSchema, generated from the table.

createTable

createTable(table: string, schema: JsonSchemaObject): Promise<void>

Applicable to Relational DBs, like MySQL. Will invoke smth like create table Table1 ... ;. Takes a JsonSchema as an argument.

DBQuery

Object that defines "DB Query".

// Simplest query - "get all" query
DBQuery.create('table1')

// where type = "cat"
DBQuery.create('table1').filter('type', '==', 'cat')

// OR
DBQuery.create('table1').filterEq('type', 'cat')

// Where updated > 2019-01-17
DBQuery.create('table1').filter('updated', '>', '2019-01-17')

// order by 'name'
DBQuery.create('table1').filter('updated', '>', '2019-01-17').order('name')

// order by 'name' in descending order
DBQuery.create('table1').filter('updated', '>', '2019-01-17').order('name', true)

Features:

.filter(key: string, operator: Operator, value: any)

.filter('updatedDate', '>', '2019-01-17')

.filterEq(key: string, value: any)

.filterEq('updated', true)

.order(key: string, descending: boolean = false)

.order('updated') // asc
.order('updated', true) // desc

.limit(lim: number)

.limit(1000)
.limit(0) // no limit

.select(fields: string[])

Allows "projection queries" - queries that return subset of fields. Like select a,b,c from Table in SQL, as opposed to select * from Table.

Passing empty array will actually return an array of empty objects (documented edge case).

.select([]) // returns [ {}, {}, {} ]
.select(['id']) //=> [ { id: 'id1' }, { id: 'id2' }, ... ]

Exports

• / root
• /adapter/file
• /adapter/cachedb
• /testing
  • dbTest
  • daoTest
  • Test models, utils, etc
• /validation
  • Joi validation schemas for DBQuery, CommonDBOptions, CommonSchema, etc

Packaging

• engines.node >= LTS
• main: dist/index.js: commonjs, es2020
• types: dist/index.d.ts: typescript types
• /src folder with source *.ts files included