Skip to content

Latest commit

 

History

History
158 lines (120 loc) · 17.6 KB

assignment-4.md

File metadata and controls

158 lines (120 loc) · 17.6 KB

<< Back to Main

Part 4: Data Persistence using MongoDB

This assignment builds on the work you have done for Assignment 3.

You will have noticed in the previous assignment that when you restart the server, the chatroom list is reset and the chat messages are gone. In this assignment, you will be implementing mechanisms to persist data in the server. Here is a high-level overview of what you will be doing:

  • Set up a MongoDB database for the application
  • Write a "driver" for interacting with the MongoDB service.
  • Create REST endpoints for reading objects from the database.

We continue to prohibit the use of third-party JavaScript frameworks, except those specifically mentioned, such as Express.js on the server side..

Directory Structure

Same as previous assignments. If you still have client/chat.html and client/profile.html, you can delete them.

/client/
    /assets/
    /index.html
    /style.css
    /app.js
/Database.js    (ADDED in this assignment)
/server.js
/package.json

All your server-side code for this assignment goes in server.js. In addition, you will be implementing Database.js - we provide the initial implementation here.

Tasks

We use the following object schemas to describe the structure of objects:

Room = {
    "_id": String | ObjectId(String),
    "name": String,
    "image" : String
}

Message = {
    "username": String,
    "text": String
}

Conversation = {
    "_id": ObjectId(String),
    "room_id": String,
    "timestamp": Number,
    "messages": [ Message, Message, Message, ... ]
}

For example, when we refer to a "Room object" in the instructions below, we mean an object with fields "_id", "name", and "image" as shown in the schema above.

  1. [Dependency] As the first step, you will need to install MongoDB. You can download the latest MongoDB Community Server, provided as free and open source software at mongodb.com. Once you have installed MongoDB, also install the MongoDB JavaScript driver from NPM: npm install mongodb. You can refer to the official "Quick Start" guide.

    • We provide a MongoDB script to initialize your database with some initial data. You are not required to use this script, but you will be responsible for managing the structure of all the database objects, and configuring the test script accordingly. To run the MongoDB script, start the MongoDB Shell by entering mongo in the command line (make sure the MongoDB service is running). Once in the MongoDB Shell, load the script by typing: load("initdb.mongo"). This script will set up 2 Collections - chatrooms and conversations - with some initial data. At any point during development, you can reset the database to the initial state by re-running this script.
  2. (1 Points) [JS] We can use the mongodb library to interact with the database. However, the API provided by the module are quite low-level and we want to abstract out these details so we don't have to call the low-level methods all the time. In this assignment you will write a Database object that will carry out the low-level queries. We have provided an initial implementation of Database, whose methods you will complete. In server.js, declare a variable db, assigning a Database instance initialized with the appropriate arguments (mongoUrl is the address of the MongoDB service you want to connect to -- e.g., mongodb://localhost:27017 by default. Look in initdb.mongo to find out the dbName argument). If the Database instance successfully connects to the MongoDB service, you should see in the NodeJS console: [MongoClient] Connected to ${mongoUrl}/${dbName}.

  3. (5 Points) [JS] In this task, we want to make the application read the list of Rooms from the database.

    • A) In Database.js, complete the implementation of Database.prototype.getRooms.
      • It does not accept any arguments.
      • It should find documents in the MongoDB chatrooms collection, and return a Promise that resolves to an array of Room objects.
    • B) In server.js, initialize the messages object as you have done in the previous assignment, but this time based on the results from db.getRooms (initialize an empty array for each Room, using the Room ID as the key). Note that MongoDB API will use the key _id for the database objects and not id. You will have to update your client side app accordingly to use _id instead of id -- if you have followed the instructions closely up to this point, you will only have to modify refreshLobby and the "click" handler in the LobbyView.
    • C) In server.js, update the /chat endpoint's GET request handler to get the list of chat rooms using db.getRooms method you just implemented. You can now remove the chatrooms variable you have defined in Assignment 3. It should still return an array built from combining the messages object with the result of db.getRooms (consistent with Assignment 3).
    • D) In Database.js, complete the implementation of Database.prototype.getRoom.
      • It should accept a single argument room_id, which can either be a String or a mongodb.ObjectID. In case the given argument is ambiguous, mongodb.ObjectID type should get priority -- i.e., if there are 2 documents: A with ID = "11111111" and B with ID = ObjectID("11111111"), B should be returned.
      • It should find a single document in the MongoDB chatrooms collection that has the _id equal to the given room_id, and return a Promise that resolves to the document found.
      • If the document does not exist, the Promise should resolve to null.
    • E) In server.js, create a new GET endpoint /chat/:room_id where room_id is a dynamic route parameter (see Express.js documentation). In the request handler, search for the Room in the database using the db.getRoom method you just implemented in Task 1.D. If the Room was found, send it back to the client. If the Room was not found, return HTTP 404 with an error message (e.g., "Room X was not found")
  4. (3 Points) [JS] In this task, we want to make the application write a Room to the database.

    • A) In Database.js, complete the implementation of Database.prototype.addRoom.
      • It should accept a single argument room, which is a Room object with or without the _id field.
      • It should insert a Room document in the MongoDB chatrooms collection.
      • It should return a Promise that resolves to the newly inserted Room object, including the _id as assigned by the MongoDB service.
      • If the name field in the room object is not provided, the Promise should reject with an Error.
    • B) In server.js, update the /chat endpoint's POST request handler to use the db.addRoom method you have just defined in Task 3. The error handling behaviour (e.g., sending HTTP 400 Bad Request upon malformed payload) should remain unchanged and the messages object should still be updated as it were in the previous assignment. By the end of this task, the list of chatrooms should be persistent even after you restart the server.
  5. (7 Points) [JS] The list of chat rooms are now persistent, but all the chat message are ephemeral. We need to be clever about saving the chat messages, because they are created very frequently - writing to the database for each message will create a lot of overhead, potentially increasing the risk of a server crash. To handle this more efficiently, we will deal with chat messages in terms of "conversation" blocks - you could think of this as a buffering mechanism. A "conversation" will contain a fixed-length array of messages, and whenever the server accumulates the predefined number of text messages, it will save the block in the database. If the server crashes, it would lose messages in the current block, but any messages before that should be persistent.

    • A) In Database.js, complete the implementation of Database.prototype.addConversation.
      • It accepts a single argument conversation, which is a Conversation object.
      • It should insert a Conversation document in the MongoDB conversations collection.
      • It should return a Promise that resolves to the newly inserted Mongo Document.
      • If any of the fields (room_id, timestamp, messages) is not provided, the Promise should reject with an Error.
    • B) In Database.js, complete the implementation of Database.prototype.getLastConversation.
      • It accepts 2 arguments: room_id and optional before.
      • If before is not given, use the current UNIX time in milliseconds given by Date.now.
      • It should find the Conversation object with room_id field equal to the given room_id argument, and timestamp less than the given before argument. If multiple Conversation objects are found, it should select the one whose timestamp is closest to before.
      • It should return a Promise that resolves to the Conversation object found.
      • The Promise should resolve to null if no Conversation was found.
    • C) In server.js declare a global constant messageBlockSize and assign a number (set a small number like 10 during development, so you can test it easily). This number will indicate how many messages to include in a conversation.
    • D) In server.js, update the message handler for the broker client WebSocket, performing the following in addition to what it was doing in Assignment 3:
      • If the length of the corresponding messages[message.roomId] array is equal to messageBlockSize after pushing the new message, create a new Conversation object and add to the database by calling db.addConversation. The timestamp of the new Conversation should be UNIX time in milliseconds. After successfully saving the Conversation into the database, empty the corresponding messages array to collect new messages.
    • E) In server.js, define a new GET endpoint /chat/:room_id/messages, where room_id is the room ID. This endpoint should return a Conversation based on the room_id parameter and the before parameter in the query string. For example, if a GET request is made to /chat/room-1/messages?before=1604188800000, the server should return a Conversation with room_id equal to "room-1" and timestamp less than 1604188800000.
    • F) In the Service object in app.js (client-side), define a new function getLastConversation, which accepts 2 arguments: roomId and before. It should simply make an AJAX GET request to the /chat/:room_id/messages endpoint, encoding the before parameter as a query string in the URL. It should return a Promise that resolves to the Conversation object returned by the server.
  6. (8 Points) [JS] By now, messages should persist at least up to the last Conversation block. In this task, we will create a mechanism to continuously read the Conversation blocks as we scroll up in the chat view - a UI mechanism commonly refered to as "infinite scrolling". To achieve this, we will make some changes to the Room and ChatView classes, and use a generator function to load the messages from the server.

    • A) In the Room class in app.js, define a new method named addConversation(conversation). This method is similar to addRoom, except that here you want to insert the given messages at the beginning of the Room.messages array. Make sure the order of messages is chronological. After inserting the messages to the Room object, call the onFetchConversation(conversation) event listener callback, which will be assigned later by the ChatView. This callback is used to notify the ChatView that new data is available (similar to how onNewMessage works).
    • B) In app.js, define a generator function named makeConversationLoader(room). The generator function will "remember" the last conversation fetched, and incrementally fetch the conversations as the user scrolls to the top of the chat view, until there is no more conversation blocks to be read.
      • i. In this function, you should use a local variable to keep track of the last conversation block fetched (or just the timestamp of the last conversation).
      • ii. Then, using a conditional loop, use the Service.getLastConversation method you implemented in Task 4.F to fetch the conversation block from the server. The last conversation read has information needed for determining the next query parameters.
      • iii. Just before you make a request to the server, set the room.canLoadConversation property to false. This property has not been defined yet (you will do that next), but you will use this boolean flag to indicate whether more data can be loaded and if the request is in progress.
      • iv. Upon fetching a result, update the local variable (the one you're using to maintain the generator state), and set the canLoadConversation flag to true. If a Conversation object was returned, call the room.addConversation method. If there is no result, leave the canLoadConversation flag as-is (false), and the function should terminate.
      • v. This generator function should yield a Promise object that resolves to the conversation data if exists, and resolves to null otherwise.
    • C) In the Room object's constructor, create a Generator instance by calling makeConversationLoader, passing in the Room instance itself as the argument. Assign this Generator to the Room instance's getLastConversation property. Also in the constructor, define a boolean variable named canLoadConversation and initialize it to true. You will use in ChatView to determine whether to allow making further requests or not.
    • D) In ChatView, you will need to trigger the entire fetch-update-render cycle when the mouse is scrolled up in the chat view (we assume that this.chatElem has a minimum and maximum height, and it has a scrollbar in the y direction). Attach a wheel event listener in the this.chatElem element. Now that each room has a generator for loading conversations, this event listener should be incredibly simple. Invoke the generator's next function, only if the following conditions are met:
      • The scroll is at the top of the view
      • Mouse scroll direction is "up"
      • this.room.canLoadConversation is true
    • E) In the setRoom method of ChatView, create an anonymous function and attach it to the this.room instance's onFetchConversation property -- this is the callback you use to receive new conversation blocks. The behaviour is almost the same as onNewMessage, except that you will be attaching multiple messages at once, at the beginning of the DOM. After you render the messages, you will have to adjust the scroll position, so you may want to keep track of the scroll height before you render the messages. More specifically, if the scroll height is hb pixels before rendering, the scroll height will increase to ha pixels after adding the messages. The scroll's vertical position should be at ha - hb after rendering.

Testing

The testing infrastructure is identical to the one used in the previous assignment, except for the URL of the scripts used. Therefore, we will not elaborate on how to use the test script's API; you can refer to the "Testing" section in the previous assignment, and update the URLs of the test scripts accordingly.

  • Client-side:

    • URL: http://99.79.42.146/cpen322/test-a4.js (this goes in index.html)
    • Default Values (can be customized with cpen322.setDefault(key, val)):
      • testRoomId: 'room-1',
      • image: 'assets/everyone-icon.png',
      • webSocketServer: 'ws://localhost:8000'
    • Exported closure variables: lobby, chatView
  • Server-side:

    • URL: http://99.79.42.146/cpen322/test-a4-server.js (this goes in the cpen322.connect function inside server.js. You still need to use the same cpen322-tester.js module from Assignment 3)
    • Exported global variables: app, db, messages, messageBlockSize
  • NOTE: Unlike the test scripts in the previous assignments, this test script will not clean up the test objects it creates during the test. Previously, we could safely delete the test objects because we know that the objects were all in-memory and definitely belong to the application. In this assignment, the application connects to a database, which we consider as "external" to the application. Since we cannot make assumptions about the database service you're connecting to, we refrain from performing destructive operations. (Most of you would probably be running your own fresh database service, but we don't neglect the chance that some of you might be using an existing/shared database or connecting to a cloud-hosted service).

Marking

There are 5 tasks for this assignment (Total 24 Points):

  • Task 1: 1 Points
  • Task 2: 5 Points
  • Task 3: 3 Points
  • Task 4: 7 Points
  • Task 5: 8 Points

Submission instructions

Copy the commit hash from Github and enter it in Canvas.

For step-by-step instructions, refer to the tutorial.

Deadline

These deadlines will be strictly enforced by the assignment submission system (Canvas).

  • Sunday, Nov 21, 2021 23:59:59 PST