Mongoose Patch History is a mongoose plugin that saves a history of JSON Patch operations for all documents belonging to a schema in an associated "patches" collection.
$ npm install mongoose-patch-history
To use mongoose-patch-history for an existing mongoose schema you can simply plug it in. As an example, the following schema definition defines a Post schema, and uses mongoose-patch-history with default options:
import mongoose, { Schema } from 'mongoose'
import patchHistory from 'mongoose-patch-history'
/* or the following if not running your app with babel:
const patchHistory = require('mongoose-patch-history').default;
const mongoose = require('mongoose');
const Schema = mongoose.Schema;
*/
const PostSchema = new Schema({
title: { type: String, required: true },
comments: Array,
})
PostSchema.plugin(patchHistory, { mongoose, name: 'postPatches' })
const Post = mongoose.model('Post', PostSchema)mongoose-patch-history will define a schema that has a ref field containing the ObjectId of the original document, a ops array containing all json patch operations and a date field storing the date where the patch was applied.
Continuing the previous example, a new patch is added to the associated patch collection whenever a new post is added to the posts collection:
Post.create({ title: 'JSON patches' })
.then(post => post.patches.findOne({ ref: post.id }))
.then(console.log)
// {
// _id: ObjectId('4edd40c86762e0fb12000003'),
// ref: ObjectId('4edd40c86762e0fb12000004'),
// ops: [
// { value: 'JSON patches', path: '/title', op: 'add' },
// { value: [], path: '/comments', op: 'add' }
// ],
// date: new Date(1462360838107),
// __v: 0
// }mongoose-patch-history also adds a static field Patches to the model that can be used to access the patch model associated with the model, for example to query all patches of a document. Whenever a post is edited, a new patch that reflects the update operation is added to the associated patch collection:
const data = {
title: 'JSON patches with mongoose',
comments: [{ message: 'Wow! Such Mongoose! Very NoSQL!' }],
}
Post.create({ title: 'JSON patches' })
.then(post => post.set(data).save())
.then(post => post.patches.find({ ref: post.id }))
.then(console.log)
// [{
// _id: ObjectId('4edd40c86762e0fb12000003'),
// ref: ObjectId('4edd40c86762e0fb12000004'),
// ops: [
// { value: 'JSON patches', path: '/title', op: 'add' },
// { value: [], path: '/comments', op: 'add' }
// ],
// date: new Date(1462360838107),
// __v: 0
// }, {
// _id: ObjectId('4edd40c86762e0fb12000005'),
// ref: ObjectId('4edd40c86762e0fb12000004'),
// ops: [
// { value: { message: 'Wow! Such Mongoose! Very NoSQL!' }, path: '/comments/0', op: 'add' },
// { value: 'JSON patches with mongoose', path: '/title', op: 'replace' }
// ],
// "date": new Date(1462361848742),
// "__v": 0
// }]rollback(ObjectId, data, save)Documents have a rollback method that accepts the ObjectId of a patch doc and sets the document to the state of that patch, adding a new patch to the history.
Post.create({ title: 'First version' })
.then(post => post.set({ title: 'Second version' }).save())
.then(post => post.set({ title: 'Third version' }).save())
.then(post => {
return post.patches
.find({ ref: post.id })
.then(patches => post.rollback(patches[1].id))
})
.then(console.log)
// {
// _id: ObjectId('4edd40c86762e0fb12000006'),
// title: 'Second version',
// __v: 0
// }Further the rollback method accepts a data object which is injected into the document.
post.rollback(patches[1].id, { name: 'merged' })
// {
// _id: ObjectId('4edd40c86762e0fb12000006'),
// title: 'Second version',
// name: 'merged',
// __v: 0
// }To rollback the document to a specific patch but without saving it back to the database call the method with an empty data object and the save flag set to false.
post.rollback(patches[1].id, {}, false)
// Returns the document without saving it back to the db.
// {
// _id: ObjectId('4edd40c86762e0fb12000006'),
// title: 'Second version',
// __v: 0
// }The rollback method will throw an Error when invoked with an ObjectId that is
- not a patch of the document
- the latest patch of the document
PostSchema.plugin(patchHistory, {
mongoose,
name: 'postPatches',
})mongoose📌 required
The mongoose instance to work withname📌 required
String where the names of both patch model and patch collection are generated from. By default, model name is the pascalized version and collection name is an undercore separated versionremovePatches
Removes patches when origin document is removed. Default:truetransforms
An array of two functions that generate model and collection name based on thenameoption. Default: An array of humps.pascalize and humps.decamelizeincludes
Property definitions that will be included in the patch schema. Read more about includes in the next chapter of the documentation. Default:{}excludes
Property paths that will be excluded in patches. Read more about excludes in the excludes chapter of the documentation. Default:[]trackOriginalValue
If enabled, the original value will be stored in the change patches under the attributeoriginalValue. Default:false
PostSchema.plugin(patchHistory, {
mongoose,
name: 'postPatches',
includes: {
title: { type: String, required: true },
},
})This will add a title property to the patch schema. All options that are available in mongoose's schema property definitions such as required, default or index can be used.
Post.create({ title: 'Included in every patch' })
.then((post) => post.patches.findOne({ ref: post.id })
.then((patch) => {
console.log(patch.title) // 'Included in every patch'
})The value of the patch documents properties is read from the versioned documents property of the same name.
There is an additional option that allows storing information in the patch documents that is not stored in the versioned documents. To do so, you can use a combination of virtual type setters on the versioned document and an additional from property in the include options of mongoose-patch-history:
// save user as _user in versioned documents
PostSchema.virtual('user').set(function (user) {
this._user = user
})
// read user from _user in patch documents
PostSchema.plugin(patchHistory, {
mongoose,
name: 'postPatches',
includes: {
user: { type: Schema.Types.ObjectId, required: true, from: '_user' },
},
})
// create post, pass in user information
Post.create({
title: 'Why is hiring broken?',
user: mongoose.Types.ObjectId(),
})
.then(post => {
console.log(post.user) // undefined
return post.patches.findOne({ ref: post.id })
})
.then(patch => {
console.log(patch.user) // 4edd40c86762e0fb12000012
})In case of a rollback in this scenario, the rollback method accepts an object as its second parameter where additional data can be injected:
Post.create({ title: 'v1', user: mongoose.Types.ObjectId() })
.then(post =>
post
.set({
title: 'v2',
user: mongoose.Types.ObjectId(),
})
.save()
)
.then(post => {
return post.patches.find({ ref: post.id }).then(patches =>
post.rollback(patches[0].id, {
user: mongoose.Types.ObjectId(),
})
)
})In situations where you are running Mongoose queries directly instead of via a document, you can specify the extra fields in the query options:
Post.findOneAndUpdate(
{ _id: '4edd40c86762e0fb12000012' },
{ title: 'Why is hiring broken? (updated)' },
{ _user: mongoose.Types.ObjectId() }
)PostSchema.plugin(patchHistory, {
mongoose,
name: 'postPatches',
excludes: [
'/path/to/hidden/property',
'/path/into/array/*/property',
'/path/to/one/array/1/element',
],
})
// Properties
// /path/to/hidden: included
// /path/to/hidden/property: excluded
// /path/to/hidden/property/nesting: excluded
// Array element properties
// /path/into/array/0: included
// /path/into/array/345345/property: excluded
// /path/to/one/array/0/element: included
// /path/to/one/array/1/element: excludedThis will exclude the given properties and all nested paths. Excluding / however will not work, since then you can just disable the plugin.
- If a property is
{}orundefinedafter processing all excludes statements, it will not be included in the patch. - Arrays work a little different. Since json-patch-operations work on the array index, array elements that are
{}orundefinedare still added to the patch. This brings support for laterremoveorreplaceoperations to work as intended.
TheARRAY_WILDCARD*matches every array element.
If there are any bugs experienced with the excludes feature please write an issue so we can fix it!