Skip to content
This repository has been archived by the owner on Dec 11, 2022. It is now read-only.

oflynned/Mongoize-ORM

Repository files navigation

Mongoize ORM

npm CI codecov

Setup

$ npm install -S mongoize-orm
$ yarn add mongoize-orm

Disclaimer

Data generally becomes relational after a while, you should switch to a relational database at some point when your POC becomes mature enough instead of using a pseudo-relational wrapper on documents.

Seriously, use Postgres or MySQL when you get to this point.

Show me some actual examples

Have a look here for some simple examples

https://github.com/oflynned/Mongoize-ORM/tree/master/src/example

Have a look here for a full TypeScript Express web server that gets transpiled to js

https://github.com/oflynned/Mongoize-ORM-Example/

Quick start

You just need to implement the two abstract classes Schema and BaseDocument to get started with using Mongoize-ORM. You'll also need to implement your own model type extending BaseModelType for the TypeScript layer of things.

import { Schema, BaseDocument, BaseModelType } from 'mongoize-orm'

// define a ts type interface for strong typing
export interface AnimalType extends BaseModelType {
  name: string;
  legs?: number;
}

// define a db schema for validating data passed
export class AnimalSchema extends Schema<AnimalType> {
  joiBaseSchema(): object {
    return {
      name: Joi.string().required(),
      legs: Joi.number().min(0)
    };
  }

  joiUpdateSchema(): object {
    return {
      name: Joi.string(),
      legs: Joi.number().min(0)
    };
  }
}

// hey presto you now have an instance with an abstract method to implement
// with a strong interface type
class Animal extends BaseDocument<AnimalType, AnimalSchema> {
  joiSchema(): AnimalSchema {
    return new AnimalSchema();
  }
}

To persist records, you need a database client (either in-memory or mongodb) to connect in order to use Mongoize ORM.

import { InMemoryClient, MongoClient } from 'mongoize-orm';

const client = await new InMemoryClient().connect();

Now you have a fully-usable model that you can perform actions on or commit to a database.

const animal: Animal = new Animal().build({name: "Doggo", legs: 4}).save(client);
console.log(animal.toJson());

// don't forget to close the connection when you're done
await client.close();

{
  name: 'Doggo',
  legs: 4,
  _id: '0b498ead-4915-4076-adf8-eb49ac72d12c',
  createdAt: 2020-03-16T14:22:35.277Z,
  updatedAt: null,
  deletedAt: null,
  deleted: false,
}

Non-Relational Data

For any records that correspond to being documents where you either want to nest everything as a property, or don't care about relationships (yet), then your model needs to extend BaseDocument and you follow your own schema that you set. No redundant abstract methods, no hackiness.

Type

import { BaseModelType } from "mongoize-orm";

export interface AnimalType extends BaseModelType {
  name: string;
  legs?: number;
}

Schema

import { Schema, Joi } from "mongoize-orm";
import { AnimalType } from "./type

export class AnimalSchema extends Schema<AnimalType> {
  joiBaseSchema(): object {
    return {
      name: Joi.string().required(),
      legs: Joi.number().min(0)
    };
  }

  joiUpdateSchema(): object {
    return {
      name: Joi.string(),
      legs: Joi.number().min(0)
    };
  }
}

Model

import { MongoClient, BaseDocument, Repository } from "mongoize-orm";
import { AnimalType } from "./type";
import { AnimalSchema } from "./schema";

class Animal extends BaseDocument<AnimalType, AnimalSchema> {
  joiSchema(): AnimalSchema {
    return new AnimalSchema();
  }
}

export default Animal;

Relational Data

Models generally have relationships to one another, the RelationalDocument type is a subtype of BaseDocument that allows you to specify how documents are related to each other. The .populate method on a relational document fetches the document's layer of relationships, and does not populate relationships of relationships to prevent infinite loops.

Type

import { BaseModelType } from "mongoize-orm";

export interface AnimalType extends BaseModelType {
  name: string;
  legs?: number;
  ownerId?: string;
}

Relationship

import { BaseRelationshipType } from "mongoize-orm";

export interface AnimalRelationships extends BaseRelationshipType {
  owner?: Person;
}

Schema

import { Schema, Joi } from "mongoize-orm";
import { AnimalType } from "./type

export class AnimalSchema extends Schema<AnimalType> {
  joiBaseSchema(): object {
    return {
      name: Joi.string().required(),
      legs: Joi.number().min(0)
    };
  }

  joiUpdateSchema(): object {
    return {
      name: Joi.string(),
      legs: Joi.number().min(0)
    };
  }
}

Model

import { MongoClient, RelationalDocument, Repository } from "mongoize-orm";
import { AnimalType } from "./type" }
import { AnimalSchema } from "./schema";
import { AnimalRelationships } from "./relationships";

import Person from "../person";

class Animal extends RelationalDocument<
  AnimalType,
  AnimalSchema,
  AnimalRelationships
> {
  joiSchema(): AnimalSchema {
    return new AnimalSchema();
  }

  async relationalFields(client: MongoClient): Promise<AnimalRelationships> {
    return {
      owner: await this.owner(client)
    };
  }

  private async owner(client: MongoClient): Promise<Person> {
    return Repository.with(Person).findById(client, this.toJson().ownerId);
  }
}

export default Animal;

You need to call .populate on the instance to populate that level of relationships. Otherwise it will be undefined!

Database Clients

All clients extend from the abstract class DatabaseClient which has no functionality directly. To interact with a database, you should use MongoClient or InMemoryClient depending on your use case.

MongoClient

MongoClient extends DatabaseClient and implements the MongoDB driver so it acts as a wrapper for it.

Using the client is simple, it will default to localhost:27017. You need to pass the database as an option or through the URI.

const client = await new MongoClient().connect({ database: 'mongoize' });

Passing options is done through the .connect method. Keep in mind that a URI will be prioritised over a raw config with individual options:

type UriConnectionOptions = {
  uri: string;
};

type AuthConnectionOptions = {
  username?: string;
  password?: string;
  host: string;
  port: number;
  database: string;
};

There is also an option to append the NODE_ENV environment value to the db when you enable the option appendDatabaseEnvironment. For a database named mongoize on the development environment, the database is automatically set to mongoize-development. This value defaults to false, it must be set to true through options on connect. It is not available on the in-memory client.

Mongo client options can be customised by overriding the typed .mongoOptions method. It defaults to:

mongoOptions(): MongoClientOptions {
    return {
      useNewUrlParser: true,
      useUnifiedTopology: true
    };
  }

InMemoryClient

This is great for use in ephemeral testing scenarios, or short-lived servers (would not recommend that). No config is needed as it does all the setup & teardown itself as part of the lifecycle of the in-memory server.

const client = await new InMemoryClient().connect();

// your db client is now ready for use

Model Definition

Models comprise of three distinct parts:

  • TypeScript property type
    • Allows auto-prediction of types and records
    • Pertains to any data (including volatile) that needs to be shown in a context menu
  • Joi validation schema
    • Used by Joi to validate any data provided for creation/update functions
    • For validation checks against a schema before persisting to a db
  • Model document
    • Centralised point for extending a MongoDB document and override any of the lifecycle hooks
.build()

Construct a new instance with the type and base type schema in mind.

.collection()

Defaults to the name of the child class's constructor with an appended s

User -> users

This can be overridden in the derived model class if the name has an irregular plural

collection(): string {
  return "people";
}
.from()*

Don't use this method unless you want to copy a record directly into another from the schema definition-level. This is used internally to cast database records into model instances. Use .build() instead to obey schemas.

On a model instance however, the type is still inferred but no base record content will be populated (_id, createdAt, updatedAt, deletedAt, deleted)

.save()

Saves the instance to the persistence layer, only allowed to be called once as _id has a unique constraint.

Use .update() or Repository#updateOne() if you want to update data stored in a record.

.update()

Updates a record's fields as per its definition type interface.

.delete()

Defaults to soft-delete which sets .deleted and .deletedAt fields. Keeps record in database without purging it.

There is also hard-delete which purges the record from the store and sets the instance to undefined.

.toJSON()

Returns an untyped object of all the fields on the record inherited from IBaseRecord (timestamps, id, etc) and the fields set on the interface type.

Model Instance Lifecycle Methods

Create

When await new User().build({...}).save() is called:

  • onPreValidate
  • validate*
  • onPostValidate
  • onPreSave
  • save*
  • onPostSave
Update

When await user.update({...}) or await Repository.with(User).updateOne(client, "id", {...}) is called:

  • validateOnUpdate*
  • onPreUpdate
  • update*
  • onPostUpdate
Delete

When await user.delete({...}) or await Repository.with(User).deleteOne(client, "id", {...}) is called:

  • onPreDelete
  • delete*
  • onPostDelete

Repositories

The repository is for directly interacting with the db via a facade without having a model to work with.

Fetch an instance of the repository with Repository.with(User). You can use the methods below once you acquire it.

.count

Returns the count of records matching the query.

.deleteCollection

Purges the collection by name.

.deleteMany

Deletes records (hard or soft) with a query.

.deleteOne

Deletes a single record by id.

.findOne

Returns a single record if it exists (first in the array). Returns undefined if not.

.findById

Returns a record if the id exists. Returns undefined if not.

.findMany

Returns an array of records by a query. Returns [] if nothing matches.

.exists

Returns true if the query contains at least 1 record. Returns false if not.

.updateOne

Dispatches the update when validated to the db. Returns an updated record if successfully validated. Throws an error if the record does not exist.

For auto-completion on the param types, the full list of generics needs to be passed to the static repository instance:

Repository
    .with<AnimalType, Animal, AnimalSchema>(Animal)
    .updateOne(client, animal.toJson()._id, { name: "Doggo" });

If you know the params, then you can just pass them untyped.

Repository
    .with(Animal)
    .updateOne(client, animal.toJson()._id, { name: "Doggo" });

Beware - the validator will cut out any unknown keys unless you manually turn it off in the options parameter.

Repository
    .with(Animal)
    .updateOne(client, animal.toJson()._id, { newParameter: "Cool" }, { validateUpdate: false });

About

Simplify MongoDB model persistence with a smart layer in between you and your database

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published