sharedb/README.md

JSON OT database backend

_This README is for `sharedb@1.x`. For `sharedb@1.x-beta`, see [the 1.x-beta branch](https://github.com/share/sharedb/tree/1.x-beta). To upgrade, see [the upgrade guide](https://github.com/share/sharedb/wiki/Upgrading-to-sharedb@1.0.0-from-1.0.0-beta)._

# ShareDB

  [![NPM Version](https://img.shields.io/npm/v/sharedb.svg)](https://npmjs.org/package/sharedb)
  [![Build Status](https://travis-ci.org/share/sharedb.svg?branch=master)](https://travis-ci.org/share/sharedb)
  [![Coverage Status](https://coveralls.io/repos/github/share/sharedb/badge.svg?branch=master)](https://coveralls.io/github/share/sharedb?branch=master)

ShareDB is a realtime database backend based on [Operational Transformation
(OT)](https://en.wikipedia.org/wiki/Operational_transformation) of JSON
documents. It is the realtime backend for the [DerbyJS web application
framework](http://derbyjs.com/).

For questions, discussion and announcements, join the [ShareJS mailing
list](https://groups.google.com/forum/?fromgroups#!forum/sharejs) or [check the FAQ](./docs/faq.md).

Please report any bugs you find to the [issue
tracker](https://github.com/share/sharedb/issues).

## Features

- Realtime synchronization of any JSON document
- Concurrent multi-user collaboration
- Synchronous editing API with asynchronous eventual consistency
- Realtime query subscriptions
- Simple integration with any database - [MongoDB](https://github.com/share/sharedb-mongo), [PostgresQL](https://github.com/share/sharedb-postgres) (experimental)
- Horizontally scalable with [pub/sub integration](#pubsub-adapters)
- Projections to select desired fields from documents and operations
- Middleware for implementing access control and custom extensions
- Ideal for use in browsers or on the server
- Offline change syncing upon reconnection
- In-memory implementations of database and pub/sub for unit testing

### Reconnection

**TLDR**
```javascript
const WebSocket = require('reconnecting-websocket');
var socket = new WebSocket('ws://' + window.location.host);
var connection = new sharedb.Connection(socket);
```

The native Websocket object that you feed to ShareDB's `Connection` constructor **does not** handle reconnections.

The easiest way is to give it a WebSocket object that does reconnect. There are plenty of example on the web. The most important thing is that the custom reconnecting websocket, must have the same API as the native rfc6455 version.

In the "textarea" example we show this off using a Reconnecting Websocket implementation from [reconnecting-websocket](https://github.com/pladaria/reconnecting-websocket).



## Example apps

[<img src="examples/counter/demo.gif" height="300">
Simple app demonstrating realtime sync](examples/counter)

[<img src="examples/leaderboard/demo.gif" height="436">
Leaderboard app demonstrating live queries](examples/leaderboard)

## Data model

In ShareDB's view of the world, every document has 3 properties:

- **version** - An incrementing number starting at 0
- **type** - An OT type. OT types are defined in
[share/ottypes](https://github.com/share/ottypes). Documents
which don't exist implicitly have a type of `null`.
- **data** - The actual data that the document contains. This must be pure
acyclic JSON. Its also type-specific. (JSON type uses raw JSON, text documents
use a string, etc).

ShareDB implicitly has a record for every document you can access. New documents
have version 0, a null type and no data. To use a document, you must first
submit a *create operation*, which will set the document's type and give it
initial data. Then you can submit editing operations on the document (using
OT). Finally you can delete the document with a delete operation. By
default, ShareDB stores all operations forever - nothing is truly deleted.

## Server API

### Initialization

First, create a ShareDB server instance:

```js
var ShareDB = require('sharedb');
var share = new ShareDB(options);
```

__Options__

* `options.db` _(instance of `ShareDB.DB`)_
  Store documents and ops with this database adapter. Defaults to `ShareDB.MemoryDB()`.
* `options.pubsub` _(instance of `ShareDB.PubSub`)_
  Notify other ShareDB processes when data changes
  through this pub/sub adapter. Defaults to `ShareDB.MemoryPubSub()`.
* `options.milestoneDb` _(instance of ShareDB.MilestoneDB`)_
  Store snapshots of documents at a specified interval of versions

#### Database Adapters
* `ShareDB.MemoryDB`, backed by a non-persistent database with no queries
* [`ShareDBMongo`](https://github.com/share/sharedb-mongo), backed by a real Mongo database
  and full query support
* [`ShareDBMingoMemory`](https://github.com/share/sharedb-mingo-memory), backed by
  a non-persistent database supporting most Mongo queries. Useful for faster
  testing of a Mongo-based app.
* [`ShareDBPostgres`](https://github.com/share/sharedb-postgres), backed by PostgresQL. No query support.

#### Pub/Sub Adapters
* `ShareDB.MemoryPubSub` can be used with a single process
* [`ShareDBRedisPubSub`](https://github.com/share/sharedb-redis-pubsub) can be used
  with multiple processes using Redis' pub/sub mechanism

Community Provided Pub/Sub Adapters
* [wsbus](https://github.com/dmapper/sharedb-wsbus-pubsub)

#### Milestone Adapters

* [`sharedb-milestone-mongo`](https://github.com/share/sharedb-milestone-mongo), backed by Mongo

### Listening to WebSocket connections

```js
var WebSocketJSONStream = require('@teamwork/websocket-json-stream');

// 'ws' is a websocket server connection, as passed into
// new (require('ws').Server).on('connection', ...)
var stream = new WebSocketJSONStream(ws);
share.listen(stream);
```

For transports other than WebSockets, expose a duplex
stream that writes and reads JavaScript objects. Then
pass that stream directly into `share.listen`.

### Middlewares

Middlewares let you hook into the ShareDB server pipeline. In
middleware code you can read and also modify objects as they
flow through ShareDB. For example,
[sharedb-access](https://github.com/dmapper/sharedb-access) uses middlewares
to implement access control.

`share.use(action, fn)`
Register a new middleware.

* `action` _(String)_
  One of:
  * `'connect'`: A new client connected to the server.
  * `'op'`: An operation was loaded from the database.
  * `'readSnapshots'`: Snapshot(s) were loaded from the database for a fetch or subscribe of a query or document
  * `'query'`: A query is about to be sent to the database
  * `'submit'`: An operation is about to be submitted to the database
  * `'apply'`: An operation is about to be applied to a snapshot
    before being committed to the database
  * `'commit'`: An operation was applied to a snapshot; The operation
    and new snapshot are about to be written to the database.
  * `'afterWrite'`: An operation was successfully written to
    the database.
  * `'receive'`: Received a message from a client
  * `'reply'`: About to send a non-error reply to a client message
* `fn` _(Function(context, callback))_
  Call this function at the time specified by `action`.
  * `context` will always have the following properties:
    * `action`: The action this middleware is hanlding
    * `agent`: A reference to the server agent handling this client
    * `backend`: A reference to this ShareDB backend instance
  * `context` can also have additional properties, as relevant for the action:
    * `collection`: The collection name being handled
    * `id`: The document id being handled
    * `op`: The op being handled
    * `req`: HTTP request being handled, if provided to `share.listen` (for 'connect')
    * `stream`: The duplex Stream provided to `share.listen` (for 'connect')
    * `query`: The query object being handled (for 'query')
    * `snapshots`: Array of retrieved snapshots (for 'readSnapshots')
    * `rejectSnapshotRead(snapshot, error)`: Reject a specific snapshot read (for 'readSnapshots')
      - `rejectSnapshotReadSilent(snapshot, errorMessage)`: As above, but causes the ShareDB client to treat it as a silent rejection, not passing the error back to user code.
    * `data`: Received client message (for 'receive')
    * `request`: Client message being replied to (for 'reply')
    * `reply`: Reply to be sent to the client (for 'reply')

### Projections

ShareDB supports exposing a *projection* of a real collection, with a specified (limited) set of allowed fields. Once configured, the projected collection looks just like a real collection - except documents only have the fields you've requested. Operations (gets, queries, sets, etc) on the fake collection work, but you only see a small portion of the data.

`addProjection(name, collection, fields)`
Configure a projection.

 * `name` The name of the projected collection.
 * `collection` The name of the existing collection.
 * `fields` A map (object) of the allowed fields in documents.
   * Keys are field names.
   * Values should be `true`.

For example, you could make a `users_limited` projection which lets users view each other's names and profile pictures, but not password hashes. You would configure this by calling:

```js
share.addProjection('users_limited', 'users', { name:true, profileUrl:true });
```

Note that only the [JSON0 OT type](https://github.com/ottypes/json0) is supported for projections.

### Logging

By default, ShareDB logs to `console`. This can be overridden if you wish to silence logs, or to log to your own logging driver or alert service.

Methods can be overridden by passing a [`console`-like object](https://developer.mozilla.org/en-US/docs/Web/API/console) to `logger.setMethods`:

```javascript
var ShareDB = require('sharedb');
ShareDB.logger.setMethods({
  info: () => {},                         // Silence info
  warn: () => alerts.warn(arguments),     // Forward warnings to alerting service
  error: () => alerts.critical(arguments) // Remap errors to critical alerts
});
```

ShareDB only supports the following logger methods:

  - `info`
  - `warn`
  - `error`

### Shutdown

`share.close(callback)`
Closes connections to the database and pub/sub adapters.

## Client API

The client API can be used from either Node or a browser. First, get a `ShareDB.Connection` object by connecting to the ShareDB server instance:

From Node:
```js
// `share` should be a ShareDB server instance
var connection = share.connect();
```

To use ShareDB from a browser, use a client bundler like Browserify or Webpack. The following
code connects to the ShareDB server instance over WebSockets:
```js
var ShareDB = require('sharedb/lib/client');
var socket = new WebSocket('ws://localhost:8080');
var connection = new ShareDB.Connection(socket);
```

For transports other than WebSockets, create an object implementing
the WebSocket specification and pass it into the `ShareDB.Connection` constructor.

### Class: `ShareDB.Connection`

`connection.get(collectionName, documentId)`
Get a [`ShareDB.Doc`](#class-sharedbdoc) instance on a given collection and document ID.

`connection.createFetchQuery(collectionName, query, options, callback)`
`connection.createSubscribeQuery(collectionName, query, options, callback)`
Get query results from the server. `createSubscribeQuery` also subscribes to
changes. Returns a [`ShareDB.Query`](#class-sharedbquery) instance.

* `query` _(Object)_
  A descriptor of a database query with structure defined by the database adapter.
* `callback` _(Function)_
  Called with `(err, results)` when server responds, or on error.
* `options.results` _(Array)_
  Prior query results if available, such as from server rendering.
* `options.*`
  All other options are passed through to the database adapter.

`connection.fetchSnapshot(collection, id, version, callback): void;`
Get a read-only snapshot of a document at the requested version.

* `collection` _(String)_
  Collection name of the snapshot
* `id` _(String)_
  ID of the snapshot
* `version` _(number) [optional]_
  The version number of the desired snapshot. If `null`, the latest version is fetched.
* `callback` _(Function)_
  Called with `(error, snapshot)`, where `snapshot` takes the following form:

  ```javascript
  {
    id: string;         // ID of the snapshot
    v: number;          // version number of the snapshot
    type: string;       // the OT type of the snapshot, or null if it doesn't exist or is deleted
    data: any;          // the snapshot
  }
  ```

`connection.fetchSnapshotByTimestamp(collection, id, timestamp, callback): void;`
Get a read-only snapshot of a document at the requested version.

* `collection` _(String)_
  Collection name of the snapshot
* `id` _(String)_
  ID of the snapshot
* `timestamp` _(number) [optional]_
  The timestamp of the desired snapshot. The returned snapshot will be the latest snapshot before the provided timestamp. If `null`, the latest version is fetched.
* `callback` _(Function)_
  Called with `(error, snapshot)`, where `snapshot` takes the following form:

  ```javascript
  {
    id: string;         // ID of the snapshot
    v: number;          // version number of the snapshot
    type: string;       // the OT type of the snapshot, or null if it doesn't exist or is deleted
    data: any;          // the snapshot
  }
  ```

### Class: `ShareDB.Doc`

`doc.type` _(String_)
The [OT type](https://github.com/ottypes/docs) of this document

`doc.id` _(String)_
Unique document ID

`doc.data` _(Object)_
Document contents. Available after document is fetched or subscribed to.

`doc.fetch(function(err) {...})`
Populate the fields on `doc` with a snapshot of the document from the server.

`doc.subscribe(function(err) {...})`
Populate the fields on `doc` with a snapshot of the document from the server, and
fire events on subsequent changes.

`doc.unsubscribe(function (err) {...})`
Stop listening for document updates. The document data at the time of unsubscribing remains in memory, but no longer stays up-to-date. Resubscribe with `doc.subscribe`.

`doc.ingestSnapshot(snapshot, callback)`
Ingest snapshot data. The `snapshot` param must include the fields `v` (doc version), `data`, and `type` (OT type). This method is generally called interally as a result of fetch or subscribe and not directly from user code. However, it may still be called directly from user code to pass data that was transferred to the client external to the client's ShareDB connection, such as snapshot data sent along with server rendering of a webpage.

`doc.destroy()`
Unsubscribe and stop firing events.

`doc.on('load', function() {...})`
The initial snapshot of the document was loaded from the server. Fires at the
same time as callbacks to `fetch` and `subscribe`.

`doc.on('create', function(source) {...})`
The document was created. Technically, this means it has a type. `source` will be `false` for ops received from the server and defaults to `true` for ops generated locally.

`doc.on('before op'), function(op, source) {...})`
An operation is about to be applied to the data. `source` will be `false` for ops received from the server and defaults to `true` for ops generated locally.

`doc.on('op', function(op, source) {...})`
An operation was applied to the data. `source` will be `false` for ops received from the server and defaults to `true` for ops generated locally.

`doc.on('del', function(data, source) {...})`
The document was deleted. Document contents before deletion are passed in as an argument. `source` will be `false` for ops received from the server and defaults to `true` for ops generated locally.

`doc.on('error', function(err) {...})`
There was an error fetching the document or applying an operation.

`doc.removeListener(eventName, listener)`
Removes any listener you added with `doc.on`. `eventName` should be one of `'load'`, `'create'`, `'before op'`, `'op'`, `'del'`, or `'error'`. `listener` should be the function you passed in as the second argument to `on`. Note that both `on` and `removeListener` are inherited from [EventEmitter](https://nodejs.org/api/events.html#events_class_eventemitter).

`doc.create(data[, type][, options][, function(err) {...}])`
Create the document locally and send create operation to the server.
* `data` Initial document contents
* `type` _([OT type](https://github.com/ottypes/docs))_
  Defaults to `'ot-json0'`, for which `data` is an Object
* `options.source` Argument passed to the `'create'` event locally. This is not sent to the server or other clients. Defaults to `true`.

`doc.submitOp(op, [, options][, function(err) {...}])`
Apply operation to document and send it to the server.
`op` structure depends on the document type. See the
[operations for the default `'ot-json0'` type](https://github.com/ottypes/json0#summary-of-operations).
Call this after you've either fetched or subscribed to the document.
* `options.source` Argument passed to the `'op'` event locally. This is not sent to the server or other clients. Defaults to `true`.

`doc.del([options][, function(err) {...}])`
Delete the document locally and send delete operation to the server.
Call this after you've either fetched or subscribed to the document.
* `options.source` Argument passed to the `'del'` event locally. This is not sent to the server or other clients. Defaults to `true`.

`doc.whenNothingPending(function(err) {...})`
Invokes the given callback function after

 * all ops submitted via `doc.submitOp` have been sent to the server, and
 * all pending fetch, subscribe, and unsubscribe requests have been resolved.

Note that `whenNothingPending` does NOT wait for pending `model.query()` calls.

### Class: `ShareDB.Query`

`query.ready` _(Boolean)_
True if query results are ready and available on `query.results`

`query.results` _(Array)_
Query results, as an array of [`ShareDB.Doc`](#class-sharedbdoc) instances.

`query.extra` _(Type depends on database adapter and query)_
Extra query results that aren't an array of documents. Available for certain database adapters and queries.

`query.on('ready', function() {...}))`
The initial query results were loaded from the server. Fires at the same time as
the callbacks to `createFetchQuery` and `createSubscribeQuery`.

`query.on('error', function(err) {...}))`
There was an error receiving updates to a subscription.

`query.destroy()`
Unsubscribe and stop firing events.

`query.on('changed', function(results) {...}))`
(Only fires on subscription queries) The query results changed. Fires only once
after a sequence of diffs are handled.

`query.on('insert', function(docs, atIndex) {...}))`
(Only fires on subscription queries) A contiguous sequence of documents were added to the query result array.

`query.on('move', function(docs, from, to) {...}))`
(Only fires on subscription queries) A contiguous sequence of documents moved position in the query result array.

`query.on('remove', function(docs, atIndex) {...}))`
(Only fires on subscription queries) A contiguous sequence of documents were removed from the query result array.

`query.on('extra', function() {...}))`
(Only fires on subscription queries) `query.extra` changed.

### Class: `ShareDB.Backend`

`Backend` represents the server-side instance of ShareDB. It is primarily responsible for connecting to clients, and sending requests to the database adapters. It is also responsible for some configuration, such as setting up [middleware](#middlewares) and [projections](#projections).

#### `constructor`

```javascript
var Backend = require('sharedb');
var backend = new Backend(options);
```

Constructs a new `Backend` instance, with the provided options:

* `db` _DB (optional)_: an instance of a ShareDB [database adapter](#database-adapters) that provides the data store for ShareDB. If omitted, a new, non-persistent, in-memory adapter will be created, which should _not_ be used in production, but may be useful for testing
* `pubsub` _PubSub (optional)_: an instance of a ShareDB [Pub/Sub adapter](#pubsub-adapters) that provides a channel for notifying other ShareDB instances of changes to data. If omitted, a new, in-memory adapter will be created. Unlike the database adapter, the in-memory instance _may_ be used in a production environment where pub/sub state need only persist across a single, stand-alone server
* `milestoneDb` _MilestoneDB (optional)_: an instance of a ShareDB [milestone adapter](#milestone-adapters) that provides the data store for milestone snapshots, which are historical snapshots of documents stored at a specified version interval. If omitted, this functionality will not be enabled
* `extraDbs` _Object (optional)_: an object whose values are extra `DB` instances which can be [queried](#class-sharedbquery). The keys are the names that can be passed into the query options `db` field
* `suppressPublish` _boolean (optional)_: if set to `true`, any changes committed will _not_ be published on `pubsub`
* `maxSubmitRetries` _number (optional)_: the number of times to allow a submit to be retried. If omitted, the request will retry an unlimited number of times

#### `connect`

```javascript
var connection = backend.connect();
```

Connects to ShareDB and returns an instance of a [`Connection`](#class-sharedbconnection). This is the server-side equivalent of `new ShareDBClient.Connection(socket)` in the browser.

This method also supports infrequently used optional arguments:

```javascript
var connection = backend.connect(connection, req);
```

* `connection` _Connection (optional)_: a [`Connection`](#class-sharedbconnection) instance to bind to the `Backend`
* `req` _Object (optional)_: a connection context object that can contain information such as cookies or session data that will be available in the [middleware](#middlewares)

Returns a [`Connection`](#class-sharedbconnection).

#### `listen`

```javascript
var agent = backend.listen(stream, req);
```

Registers a `Stream` with the backend. This should be called when the server receives a new connection from a client.

* `stream` _Stream_: a [`Stream`](https://nodejs.org/api/stream.html) (or `Stream`-like object) that will be used to communicate between the new `Agent` and the `Backend`
* `req` _Object (optional)_: a connection context object that can contain information such as cookies or session data that will be available in the [middleware](#middlewares)

Returns an [`Agent`](#class-agent), which is also available in the [middleware](#middlewares).

#### `close`

```javascript
backend.close(callback);
```

Disconnects ShareDB and all of its underlying services (database, pubsub, etc.).

* `callback` _Function_: a callback with the signature `function (error: Error): void` that will be called once the services have stopped, or with an `error` if at least one of them could not be stopped

#### `use`

```javascript
backend.use(action, middleware);
```

Adds [middleware](#middlewares) to the `Backend`.

* `action` _string | string[]_: an action, or array of action names defining when to apply the middleware
* `middleware` _Function_: a middleware function with the signature `function (context: Object, callback: Function): void;`. See [middleware](#middlewares) for more details

Returns the `Backend` instance, which allows for multiple chained calls.

#### `addProjection`

```javascript
backend.addProjection(name, collection, fields);
```

Adds a [projection](#projections).

* `name` _string_: the name of the projection
* `collection` _string_: the name of the collection on which to apply the projection
* `fields` _Object_: a declaration of which fields to include in the projection, such as `{ field1: true }`. Defining sub-field projections is not supported.

#### `submit`

```javascript
backend.submit(agent, index, id, op, options, callback);
```

Submits an operation to the `Backend`.

* `agent` _[`Agent`](#class-agent)_: connection agent to pass to the middleware
* `index` _string_: the name of the target collection or projection
* `id` _string_: the document ID
* `op` _Object_: the operation to submit
* `options` _Object_: these options are passed through to the database adapter's `commit` method, so any options that are valid there can be used here
* `callback` _Function_: a callback with the signature `function (error: Error, ops: Object[]): void;`, where `ops` are the ops committed by other clients between the submitted `op` being submitted and committed

#### `getOps`

```javascript
backend.getOps(agent, index, id, from, to, options, callback);
```

Fetches the ops for a document between the requested version numbers, where the `from` value is inclusive, but the `to` value is non-inclusive.

* `agent` _[`Agent`](#class-agent)_: connection agent to pass to the middleware
* `index` _string_: the name of the target collection or projection
* `id` _string_: the document ID
* `from` _number_: the first op version to fetch. If set to `null`, then ops will be fetched from the earliest version
* `to` _number_: The last op version. This version will _not_ be fetched (ie `to` is non-inclusive). If set to `null`, then ops will be fetched up to the latest version
* `options`: _Object (optional)_: options can be passed directly to the database driver's `getOps` inside the `opsOptions` property: `{opsOptions: {metadata: true}}`
* `callback`: _Function_: a callback with the signature `function (error: Error, ops: Object[]): void;`, where `ops` is an array of the requested ops

#### `getOpsBulk`

```javascript
backend.getOpsBulk(agent, index, fromMap, toMap, options, callback);
```

Fetches the ops for multiple documents in a collection between the requested version numbers, where the `from` value is inclusive, but the `to` value is non-inclusive.

* `agent` _[`Agent`](#class-agent)_: connection agent to pass to the middleware
* `index` _string_: the name of the target collection or projection
* `id` _string_: the document ID
* `fromMap` _Object_: an object whose keys are the IDs of the target documents. The values are the first versions requested of each document. For example, `{abc: 3}` will fetch ops for document with ID `abc` from version `3` (inclusive)
* `toMap` _Object_: an object whose keys are the IDs of the target documents. The values are the last versions requested of each document (non-inclusive). For example, `{abc: 3}` will fetch ops for document with ID `abc` up to version `3` (_not_ inclusive)
* `options`: _Object (optional)_: options can be passed directly to the database driver's `getOpsBulk` inside the `opsOptions` property: `{opsOptions: {metadata: true}}`
* `callback`: _Function_: a callback with the signature `function (error: Error, opsMap: Object): void;`, where `opsMap` is an object whose keys are the IDs of the requested documents, and their values are the arrays of requested ops, eg `{abc: []}`

#### `fetch`

```javascript
backend.fetch(agent, index, id, options, callback);
```

Fetch the current snapshot of a document.

* `agent` _[`Agent`](#class-agent)_: connection agent to pass to the middleware
* `index` _string_: the name of the target collection or projection
* `id` _string_: the document ID
* `options`: _Object (optional)_: options can be passed directly to the database driver's `fetch` inside the `snapshotOptions` property: `{snapshotOptions: {metadata: true}}`
* `callback`: _Function_: a callback with the signature `function (error: Error, snapshot: Snapshot): void;`, where `snapshot` is the requested snapshot

#### `fetchBulk`

```javascript
backend.fetchBulk(agent, index, ids, options, callback);
```

Fetch multiple document snapshots from a collection.

* `agent` _[`Agent`](#class-agent)_: connection agent to pass to the middleware
* `index` _string_: the name of the target collection or projection
* `ids` _string[]_: array of document IDs
* `options`: _Object (optional)_: options can be passed directly to the database driver's `fetchBulk` inside the `snapshotOptions` property: `{snapshotOptions: {metadata: true}}`
* `callback`: _Function_: a callback with the signature `function (error: Error, snapshotMap: Object): void;`, where `snapshotMap` is an object whose keys are the requested IDs, and the values are the requested `Snapshot`s

#### `queryFetch`

```javascript
backend.queryFetch(agent, index, query, options, callback);
```

Fetch snapshots that match the provided query. In most cases, querying the backing database directly should be preferred, but `queryFetch` can be used in order to apply middleware, whilst avoiding the overheads associated with using a `Doc` instance.

* `agent` _[`Agent`](#class-agent)_: connection agent to pass to the middleware
* `index` _string_: the name of the target collection or projection
* `query` _Object_: a query object, whose format will depend on the database adapter being used
* `options` _Object_: an object that may contain a `db` property, which specifies which database to run the query against. These extra databases can be attached via the `extraDbs` option in the `Backend` constructor
* `callback` _Function_: a callback with the signature `function (error: Error, snapshots: Snapshot[], extra: Object): void;`, where `snapshots` is an array of the snapshots matching the query, and `extra` is an (optional) object that the database adapter might return with more information about the results (such as counts)

### Class: `ShareDB.Agent`

An `Agent` is the representation of a client's `Connection` state on the server. If the `Connection` was created through `backend.connect` (ie the client is running on the server), then the `Agent` associated with a `Connection` can be accessed through a direct reference: `connection.agent`.

The `Agent` will be made available in all [middleware](#middlewares) requests. The `agent.custom` field is an object that can be used for storing arbitrary information for use in middleware. For example:

```javascript
backend.useMiddleware('connect', (request, callback) => {
  // Best practice to clone to prevent mutating the object after connection.
  // You may also want to consider a deep clone, depending on the shape of request.req.
  Object.assign(request.agent.custom, request.req);
  callback();
});

backend.useMiddleware('readSnapshots', (request, callback) => {
  const connectionInfo = request.agent.custom;
  const snapshots = request.snapshots;

  // Use the information provided at connection to determine if a user can access the snapshots.
  // This should also be checked when fetching and submitting ops.
  if (!userCanAccessCollection(connectionInfo, request.collection)) {
    return callback(new Error('Not allowed to access collection ' + request.collection));
  }
  // Check each snapshot individually.
  for (const snapshot of snapshots) {
    if (!userCanAccessSnapshot(connectionInfo, request.collection, snapshot)) {
      request.rejectSnapshotRead(snapshot,
        new Error('Not allowed to access snapshot in ' request.collection));
    }
  }

  callback();
});

// Here you should determine what permissions a user has, probably by reading a cookie and
// potentially making some database request to check which documents they can access, or which
// roles they have, etc. If doing this asynchronously, make sure you call backend.connect
// after the permissions have been fetched.
const connectionInfo = getUserPermissions();
// Pass info in as the second argument. This will be made available as request.req in the
// 'connection' middleware.
const connection = backend.connect(null, connectionInfo);
```

### Logging

By default, ShareDB logs to `console`. This can be overridden if you wish to silence logs, or to log to your own logging driver or alert service.

Methods can be overridden by passing a [`console`-like object](https://developer.mozilla.org/en-US/docs/Web/API/console) to `logger.setMethods`

```javascript
var ShareDB = require('sharedb/lib/client');
ShareDB.logger.setMethods({
  info: () => {},                         // Silence info
  warn: () => alerts.warn(arguments),     // Forward warnings to alerting service
  error: () => alerts.critical(arguments) // Remap errors to critical alerts
});
```

ShareDB only supports the following logger methods:

  - `info`
  - `warn`
  - `error`


## Errors

ShareDB returns errors as an instance of `ShareDBError`, with a machine-parsable `code`, as well as more details in the human-readable `message`.

### Common error codes

#### `ERR_OP_SUBMIT_REJECTED`

The op submitted by the client has been rejected by the server for a non-critical reason.

When the client receives this code, it will attempt to roll back the rejected op, leaving the client in a usable state.

This error might be used as part of standard control flow. For example, consumers may define a middleware that validates document structure, and rejects operations that do not conform to this schema using this error code to reset the client to a valid state.

#### `ERR_OP_ALREADY_SUBMITTED`

The same op has been received by the server twice.

This is non-critical, and part of normal control flow, and is sent as an error in order to short-circuit the op processing. It is eventually swallowed by the server, and shouldn't need further handling.

#### `ERR_SUBMIT_TRANSFORM_OPS_NOT_FOUND`

The ops needed to transform the submitted op up to the current version of the snapshot could not be found.

If a client on an old version of a document submits an op, that op needs to be transformed by all the ops that have been applied to the document in the meantime. If the server cannot fetch these ops from the database, then this error is returned.

The most common case of this would be ops being deleted from the database. For example, let's assume we have a TTL set up on the ops in our database. Let's also say we have a client that is so old that the op corresponding to its version has been deleted by the TTL policy. If this client then attempts to submit an op, the server will not be able to find the ops required to transform the op to apply to the current version of the snapshot.

Other causes of this error may be dropping the ops collection all together, or having the database corrupted in some other way.

#### `ERR_MAX_SUBMIT_RETRIES_EXCEEDED`

The number of retries defined by the `maxSubmitRetries` option has been exceeded by a submission.

#### `ERR_DOC_ALREADY_CREATED`

The creation request has failed, because the document was already created by another client.

This can happen when two clients happen to simultaneously try to create the same document, and is potentially recoverable by simply fetching the already-created document.

#### `ERR_DOC_WAS_DELETED`

The deletion request has failed, because the document was already deleted by another client.

This can happen when two clients happen to simultaneously try to delete the same document. Given that the end result is the same, this error can potentially just be ignored.

#### `ERR_DOC_TYPE_NOT_RECOGNIZED`

The specified document type has not been registered with ShareDB.

This error can usually be remedied by remembering to register any types you need:

```javascript
var ShareDB = require('sharedb');
var richText = require('rich-text');

ShareDB.types.register(richText.type);
```

#### `ERR_DEFAULT_TYPE_MISMATCH`

The default type being used by the client does not match the default type expected by the server.

This will typically only happen when using a different default type to the built-in `json0` used by ShareDB by default (eg if using a fork). The exact same type must be used by both the client and the server, and should be registered as the default type:

```javascript
var ShareDB = require('sharedb');
var forkedJson0 = require('forked-json0');

// Make sure to also do this on your client
ShareDB.types.defaultType = forkedJson0.type;
```

#### `ERR_OP_NOT_ALLOWED_IN_PROJECTION`

The submitted op is not valid when applied to the projection.

This may happen if the op targets some property that is not included in the projection.

#### `ERR_TYPE_CANNOT_BE_PROJECTED`

The document's type cannot be projected. `json0` is currently the only type that supports projections.