CS 448B Visualization

Stanford CS course on data visualization techniques (Fall 2024)

Location: Skilling Aud. (Rm 80)
Time: MW 10:30am-12:20pm


Assignment 3: Creating Interactive Visualization Software

Due: Monday Oct 28, 2024 by 10:30am

banner image

In this assignment, you will explore the issues involved in implementing interactive visualization software. Specifically we would like you to implement the interactive technique of dynamic queries – first explored in the HomeFinder application. However, instead of housing prices you will build an interactive visualizations for a dataset containing information about Bay Area Restaurants. Note that this data is fetched from the Yelp Fusion API in early October 2024, and it has been filtered to focus on restaurants on the Peninsula.

Requirements

The data includes lat/lon locations for each restaurant along with other descriptive fields as noted below. There are almost 2000 restaurants all together in the dataset. Your goal is to show these restaurants as data points on a map of the Bay Area and provide the following dynamic query functionality:

  • You must allow users to specify two draggable locations A (e.g. point of interest 1) and B (e.g. point of interest 2), as well as an adjustable radius for each one and filter the restaurants to those that lie within the intersection of the circles around A and B. (Other restaurants should be grayed out.) The locations of the circles should be draggable, and the circles can already be on the page upon loading. You may use either sliders or a draggable edge on the circles to adjust their radii. If you want to be fancy you may try letting users adjust the radii using either sliders or a draggable edge, but be careful, implementing both so that they update properly (i.e. the sliders update when you drag the edge and vice versa) is somewhat tricky.

  • When the user hovers over a restaurant location point, the name of the restaurant and other relevant information e.g. average rating (if not null) should be visible. These should not be visible otherwise.

  • You must provide at least 1 additional filtering controls that allow users to filter out specific aspects of the data (e.g. limit the average rating or price to a range of values, limit to restaurants with names starting with a specific letter, limit the location to a certain street name, etc.)

The application should update at interactive rates (0.1s update rate) and part of this assignment is to write the code so that the filters operate quickly.

You can work by yourself or with a partner for this assignment. Groups of three or more are not allowed. Your group must write code for this assignment. You are free to write the code in any programming language/environment you prefer, including JavaScript, C++, Java, etc. In addition you may use any software toolkit to help you build the code. However, we strongly recommend using JavaScript and D3 for this assignment. The teaching staff will provide support for coding the assignment using these tools either in Observable or as a standalone D3 application. We cannot guarantee that we can help you in other coding environments. We expect you to write the code from scratch, but if you use any pre-existing resources (e.g. Stack Overflow, extensively peruse related code on GitHub, chatGPT etc.) we expect you to list them as part of your submission and explain how you used them.

No matter what language/libraries you use we would like you to submit either an Observable notebook or a final executable program that we can execute on our own on either a Mac OS X or a Windows machine. Ideally, you should submit the work as a link to a website (e.g. an Observable notebook or a standalone webpage, such as one hosted through GitHub pages) where we can run your code along with the source code. If this is a problem for you, please talk to us right away.

Deliverables

Your final submission should include:

  • A link to a published version of your Observable notebook or the bundled source code for your application uploaded as file (either a .zip or .tar.gz archive) to Canvas. If your code is designed to be run in a browser you must also provide a link to a live version on the web, Please ensure that the software submitted is in working order. If any special instructions are needed for building or running your software, please provide them.
  • For submissions by groups of two, please also include a breakdown of how the work was split among the group members.
  • We expect you to write the code from scratch, but if you use any pre-existing resources (e.g. Stack Overflow, extensively peruse related code on GitHub, chatGPT, etc.) we expect you to list them as part of your submission and explain how you used them.
  • Finally, please include a commentary on the development process, including answers to the following questions: Roughly how much time did you spend developing your application? What aspects took the most time?

Upload a link to a published version of your Observable notebook with an embedded write-up, or upload the bundled code and your write-up as a PDF, to Canvas. If you’re working in a group of two, please just have one person submit to Canvas, but make sure to include both group members’ names in your write up.

Your assignment must be posted to Canvas by Monday Oct 28, 2024 by 10:30am.

Upload your work to Canvas.

Bay Area Restaurants

We have set up the Bay Area Restaurants dataset to contain about 2000 rows with the following fields:

  • id: A string describing the unique ID of the restaurant given by the Yelp Fusion API.
  • name: A string describing the name of the restaurant location.
  • url: A string representing the URL for business page on Yelp.
  • image_url: A string pointing to the URL of photo for the restaurant.
  • phone: A string representing the phone number of the restaurant.
  • rating: A string representing the average rating for the business
  • review_count: A string representing the integer number of reviews for the business.
  • address: A string representing the address of the restaurant.
  • latitude: A string representing the floating point latitude coordinate.
  • longitude: A string representing the floating point longitude coordinate.
  • price: A string with one of the following values: “$, \(,\)$ or \(\)”
  • categories: A string of categories associated with the restaurant.
  • transactions: A string representing a list of Yelp transactions that the business is registered for. (pickup, delivery, and restaurant_reservation)
  • business_hours: A JSON string representing the business hours.
  • is_closed: A boolean that represents whether the business has been permanently closed

Resources

Map

You can use this PNG map of the region covered in the dataset as the base for your visualization. (You are also welcome to use a different strategy for mapping, but make sure we can see geographic detail comparable to the provided map.)

If you use the provided map, here’s a sample D3 snippet to set it up.

// Assumes you've included D3 version 7 somewhere above:
// e.g. <script src="https://d3js.org/d3.v7.min.js"></script>

const mapWidth = 948
const mapHeight = 844

// These are the extent of our datapoints, coordinates-wise
const longitudeRange = ["-121.781739849809", "-122.50685"]
const latitudeRange = ["37.22070801115405", "37.820673"]

mapFrameGeoJSON = {
  const coords = `[[${longitudeRange[0]}, ${latitudeRange[0]}],[${longitudeRange[1]}, ${latitudeRange[1]}]]`
return JSON.parse(`{"type":"Feature","geometry":{"type":"LineString","coordinates": ${coords}}}`)
}

// This projects a given [longitude, latitude] pair into [x, y] pixel positions onto our image. 
projection = d3.geoConicConformal()
  .parallels([37 + 4 / 60, 38 + 26 / 60])
  .rotate([120 + 30 / 60], 0)
  .fitSize([mapWidth, mapHeight], mapFrameGeoJSON)

var container = d3.create("div");
var svg = container
  .append("svg")
  .attr("width", mapWidth)
  .attr("height", mapHeight)
  .style("border", "1px solid black");

  // Add map image to our container
svg.append("image")
  .attr("width", mapWidth)
  .attr("height", mapHeight)
  .attr("xlink:href", "data/map.png"); // Replace with the image url

The projection function returns a mapping from the coordinates ([longitude, latitude]) to the [x, y] pixel on your static map backdrop that matches to the longitude/latitude pair. The projection is an instance of a D3 projection. Feel free to read the documentation to see more information on what you can do with the projection function.

If you have a datapoint d with a given latitude and longitude, you can add a circle to the map in the following way:

var projectedLocation = projection([business_longitude, business_latitude]);
var circle = svg.append('circle')
  .attr('cx', projectedLocation[0])
  .attr('cy', projectedLocation[1])
  .attr('r', 1);

To load your data, you can use the following code:

const data = d3.csv(
  "CSV_FILE_URL",
  (row) => {
    // Here you can add any pre-processing to each row, if needed 
    return row;
  }
)

You are welcome to fork this starter Observable notebook: https://observablehq.com/d/18d09ac17e65db4b.

FAQ

How to respond to DOM events e.g. clicks?

Similar to jQuery, D3 provides a simple interface to add even listeners: use the on method on any selection. For example, to listen to click events on circles and print out the associated data object:

d3.selectAll('circle')
  .on('click', function(d) { console.log(d); });

Why is my data undefined?

You are most likely trying to use your data before it is ready/loaded. In JavaScript, HTTP requests are handled asynchronously. When you call d3.csv, the browser starts makes an HTTP request to that resource, and it immediately continues to execute the following code:

// In D3 v7, the csv function uses Promises instead of asynchronous callbacks to load data
d3.csv("file.csv").then(function(data) {
	console.log(data);
});

// This code is going to run before data is loaded, and you cannot use the data here
console.log('We don't have the data yet.');
nonDataRelatedStuff();

// This will print:
// => We don't have the data yet.
// => We have the data now!

How should I be doing my D3 development?

If you submit an Observable notebook link, we will be testing your visualizations by running the cells directly within the notebook itself, using the most recent stable version of Google Chrome.

If you submit bundled code, we’ll be testing your visualizations in the most recent stable version of Google Chrome (unless you come talk to us with a really good reason to do something differently for you), so use Chrome to develop. Chrome’s DevTools can be quite helpful as you work. Chrome also supports many ES6 and beyond features (const and let, arrow functions, async and await, etc.) so you’re welcome to use these if you’re familiar with them, but there’s definitely no need.

You might also need to run a local web server while you’re developing, because Chrome may fail to load data through d3.csv (or other XMLHttp Requests) for security reasons if you don’t. To do this, you can run python -m http.server from the directory where your code lives. (Your command line should give you a localhost link.) If that doesn’t work for you, come talk to us.

You can use GitHub Pages to host your web-based visualization software at no cost. As a student, you can upgrade your GitHub account to the Pro version for free by signing up for the Student Developer Pack. The Pro version allows you to keep private repos.

Other resources for learning D3 (and other web programming)

  • Review the slides for week 4 of the class (These have some links to other resources in them too), and other links on the homepage of the website.

  • There are lots of D3 code samples on Observable, a website run by the creator of D3, Mike Bostock. You can definitely take a look at examples here (and on Stack Overflow, etc.) for learning techniques, but please be very transparent by citing any external code snippets that you adapt, or even ones that simply inspire how you do something. We expect your design choices and your implementation to be original.

  • When in doubt, refer to the D3 API documentation. It is dense in places, but very thorough.