Minimongo

A client-side MongoDB implementation which supports basic queries, including some geospatial ones.

Uses code from Meteor.js minimongo package, reworked to support more geospatial queries and made npm+browserify friendly. It was forked in January 2014.

It is either IndexedDb backed (IndexedDb), WebSQL backed (WebSQLDb), Local storage backed (LocalStorageDb) or in memory only (MemoryDb).

Autoselection is possible with utils.autoselectLocalDb(options, success, error). success is called with the selected database.

sqlite plugin is also supported when available, activate sqlite plugin with option {storage: 'sqlite'} in utils.autoselectLocalDb

Usage

Minimongo is designed to be used with browserify or Webpack

// Require minimongo
var minimongo = require("minimongo");

var LocalDb = minimongo.MemoryDb;

// Create local db (in memory database with no backing)
db = new LocalDb();

// Add a collection to the database
db.addCollection("animals");

doc = { species: "dog", name: "Bingo" };

// Always use upsert for both inserts and modifies
db.animals.upsert(doc, function() {
	// Success:

	// Query dog (with no query options beyond a selector)
	db.animals.findOne({ species:"dog" }, {}, function(res) {
		console.log("Dog's name is: " + res.name);
	});
});

// Access collections via collection for Typescript
await db.collection["animals"].upsert(doc)

Upserting

db.collection["sometable"].upsert(docs, bases, success, error) can take either a single document or multiple documents (array) for the first and second parameter.

docs is the document(s) to upsert. If bases is present, it is the base version on which the update is based. It can be omitted to use the current cached value as the base, or put as null to force an overwrite (a true upsert, not a patch)

Promise interface is also available:

await db.collection["sometable"].upsert(docs, bases) can take either a single document or multiple documents (array) for the first and second parameter.

Removing

db.collection["sometable"].remove(docId, success, error) to remove a document.

Promise interface is also available:

db.collection["sometable"].remove(docId)

Finding

db.collection["sometable"].find(selector, options).fetch(success, error)

selector is a standard MongoDB selector, e.g. { x: 5, y: { $in: ['a', 'b'] } }

options are MongoDB find options: e.g. { limit: 10 }, { sort: ["x"] }

Promise interface is also available:

await db.collection["sometable"].find(selector, options).fetch()

Caching

A set of rows can be cached in a local database. Call

db.collection["sometable"].cache(docs, selector, options, success, error)

selector and options are the selector and options that were used to perform the find that produced docs. The local database will add/remove/update its local copy appropriately.

Seeding

A set of rows can be seeded in a local database. Seeding does not overwrite an existing row; it only makes sure that a row with that _id exists.

db.collection["sometable"].seed(docs, success, error)

Un-caching

Cached rows matching a selector can be removed with:

db.collection["sometable"].uncache(selector, success, error)

It will not affect upserted rows.

Resolving upserts

Upserts are stored in local databases in a special state to record that they are upserts, not cached rows. The base document on which the upsert is based is also stored. For example, if a row starts in cached state with { x:1 } and is upserted to { x: 2 }, both the upserted and the original state are stored. This allows the server to do 3-way merging and apply only the changes.

To resolve the upsert (for example once sent to central db), use resolveUpserts on collection

db.collection["sometable"].resolveUpserts(upserts, success, error) takes the list of upserts to resolve

resolveUpserts does not resolve an upsert if another upsert on the same row has taken place. Instead, the base value is updated (since the change has been accepted by the server) but the new upserted value is left alone.

Resolving removes

Removed rows are still stored locally until they are resolved. This is so they can be later sent to the server.

To resolve all removes, first get a list of all ids to be removed, then resolve them one by one:

const idsToRemove = await new Promise((resolve, reject) => collection.pendingRemoves(resolve, reject))

for (const id of idsToRemove) {
	await new Promise((resolve, reject) => collection.resolveRemove(id, resolve, reject))
}

ReplicatingDb

Keeps two local databases in sync. Finds go only to master.

IndexedDb

To make a database backed by IndexedDb:

// Require minimongo
var minimongo = require("minimongo");

var IndexedDb = minimongo.IndexedDb;

// Create IndexedDb
db = new IndexedDb({namespace: "mydb"}, function() {
	// Add a collection to the database
	db.addCollection("animals", function() {
		doc = { species: "dog", name: "Bingo" };

		// Always use upsert for both inserts and modifies
		db.animals.upsert(doc, function() {
			// Success:

			// Query dog (with no query options beyond a selector)
			db.animals.findOne({ species:"dog" }, {}, function(res) {
				console.log("Dog's name is: " + res.name);
			});
		});
	});
}, function() { alert("some error!"); });

Caching

Rows can be cached without creating a pending upsert. This is done automatically when HybridDb uploads to a remote database with the returned upserted rows. It is also done when a query is performed on HybridDb: the results are cached in the local db and the query is re-performed on the local database.

The field _rev, if present is used to prevent overwriting with older versions. This is the odd scenario where an updated version of a row is present, but an older query to the server is delayed in returning. To prevent this race condition from giving stale data, the _rev field is used.

HybridDb

Combines results from the local database with remote data. Multiple options can be specified at the collection level and then overriden at the find/findOne level:

interim: (default true) true to return interim results from the local database before the (slower) remote database has returned. If the remote database gives different results, the callback will be called a second time. This approach allows fast responses but with subsequent correction if the server has differing information.

cacheFind: (default true) true to cache the find results from the remote database in the local database

cacheFindOne: (default true) true to cache the findOne results from the remote database in the local database

shortcut: (default false) true to return findOne results if any matching result is found in the local database. Useful for documents that change rarely.

useLocalOnRemoteError: (default true) true to use local results if the remote find fails. Only applies if interim is false.

To keep a local database and a remote database in sync, create a HybridDb:

hybridDb = new HybridDb(localDb, remoteDb)

Be sure to add the same collections to all three databases (local, hybrid and remote).

Then query the hybridDb (find and findOne) to have it get results and correctly combine them with any pending local results. If you are not interested in caching results, add { cacheFind: false, cacheFindOne: false } to the options of find or findOne or to the addCollection options.

When upserts and removes are done on the HybridDb, they are queued up in the LocalDb until hybridDb.upload(success, error) is called.

upload will go through each collection and send any upserts or removes to the remoteDb. You must call this to have the results go to the server! Calling periodically (e.g every 5 seconds) is safe as long as you wait for one upload call to complete before calling again.

findOne will not return an interim null result, but will only return interim results when one is present.

RemoteDb

Uses AJAX-JSON calls to an API to query a real Mongo database. API is simple and contains only query, upsert, patch and remove commands.

If the client field is passed to the constructor, it is appended as a query parameters (e.g. ?client=1234) to each request made.

Example code:

remoteDb = new minimongo.RemoteDb("http://someserver.com/api/", "myclientid123")

This would create a remote db that would make the following call to the api for a find to collection abc:

GET http://someserver.com/api/abc?client=myclientid123

The client is optional and is a string that is passed in each call only to make authentication easier.

The API that RemoteDb should support four HTTP methods for each collection:

GET /<collection>

Performs a query, returning an array of results. GET query parameters are:

selector (optional) : JSON of query, in MongoDB format. e.g. {"a": 1} to find records with field a having value 1 fields (optional) : JSON object indicating which fields to return in MongoDB format. e.g. {"a": 1} to return only field a and _id sort (optional) : JSON of MongoDB sort field. e.g. ["a"] to sort ascending by a, or [["a","desc"]] to sort descending by a limit (optional) : Maximum records to return e.g. 100

Possible HTTP response codes:

200 : normal response 401 : client was invalid

POST /<collection>/find (optionally implemented)

Performs a query, returning an array of results. POST body parameters are:

selector (optional) : JSON of query, in MongoDB format. e.g. {"a": 1} to find records with field a having value 1 fields (optional) : JSON object indicating which fields to return in MongoDB format. e.g. {"a": 1} to return only field a and _id sort (optional) : JSON of MongoDB sort field. e.g. ["a"] to sort ascending by a, or [["a","desc"]] to sort descending by a limit (optional) : Maximum records to return e.g. 100

Possible HTTP response codes:

200 : normal response 401 : client was invalid

POST /<collection>

Performs a single upsert, returning the upserted row. POST value is the document to upsert. Possible HTTP response codes:

200 : document was upserted. Returns the upserted object (see notes below on merging) 400 : document did not pass validation 401 : client was invalid or not present 403 : permission denied to upsert 409 : another client was upserting same document. Try again. 410 : document was already removed and cannot be upserted

On 403 or 410, the change is automatically discarded in the HybridDb.

If array is POSTed, upsert each one and return array of upserted documents

PATCH /<collection>

Performs a patch, returning the upserted row. PATCH value is the following structure:

{
	doc: <the document in its new form. Can also be array of documents>
	base: <base document on which the changes were made. Can also be array of base documents, which match length of doc array>
}

For example, to change { x:1, y:1 } to set x to be 2, PATCH would send

{
	doc: { x:2, y: 1 }
	base: { x:1, y: 1 }
}

Possible HTTP response codes:

200 : document was upserted. Returns the upserted object 400 : document did not pass validation 401 : client was invalid or not present 403 : permission denied to upsert 409 : another client was upserting same document. Try again. 410 : document was already removed and cannot be upserted

On 403 or 410, the change is automatically discarded in the HybridDb.

DELETE /<collection>/<_id>

Note: the RemoteDb does not support remove({ filter }), but only removing a single document by _id!!

Removes to the local collection are converted into a series of _ids to be removed when sent to the server.

Removes a document. _id of the document to remove

200 : document was removed 401 : client was invalid or not present 403 : permission denied to remove 410 : document was already removed and cannot be removed again

On 403 or 410, the change is automatically discarded in the HybridDb.

Merging

Minimongo is designed to work with a server that performs three-way merging of documents that are being upserted by multiple users.

It can also be used with a simple server that just overwrites documents completely on upsert, just taking the doc value of PATCH, though this is not recommended.

Testing

To test, run testem in the main directory.

To test a RemoteDb implementation, use test/LiveRemoteDbTests.ts. Server must have a collection called scratch with fields as specified at top of tests file.

Quickfind

Finds can be very wasteful when the client has large collections already cached. The quickfind protocol shards the existing docs on the client by id and then sends a hash of them to the server, which just responds with the changed ones. See src/quickfind.ts for more details. It needs to be enabled and is off by default.