This is the codebase for a website of maps to accompany a research project on datafication and restaurants in several cities across the United States. Data from Yelp on Black-owned restaurants in Detroit comes from @amychiv. Google Maps data and the code in this repository come from myself, Nathan Kim. The project as a whole is led by Matthew Bui.
Data from Yelp is scraped by @amychiv and can be viewed here.
Data from Google Maps was obtained through the Places Text Search API. For each city in the scope of this website, I split the city into a 50x50 grid and queried Google Maps' Text Search API with the query "restaurants" once for each unit area. I then repeated the process with the query "Black-owned restaurants" and a 10x10 grid. Once a list was established, I used shapefiles for each city to clip the result set to only include restaurants within city boundaries, and exported them for mapping.
To replicate or extend:
- Create a project, link a billing account, and enable the Places API on Google Cloud Platform.
Once you enable the Places API, you should receive an API key.
- You must link a valid credit card, but as long as you only run the scraper linked in here once, it will not be charged. Google gives $200 worth of Maps API access for free every month, which comes out to around 6,500 queries. Our scraper consumes about 5,200 queries.
- It is a good idea to restrict the API key to only the Places API. If you start a new repository
following this methodology, please make sure to store files like
.envor.Renvironin your.gitignorebefore you commit anything.
- Locate a shapefile representing the city/cities you'd like to obtain restaurant data for. These
can be in any format, so long as it can be imported into R and as a simple features object via
st_read.- A custom shapefile for each city must be used because, to my knowledge, no dataset exists that comprehensively tracks city boundaries. I saw that the Census Bureau keeps track of municipal areas, but these can be a fair amount larger than cities.
- I recommend that you place the data in
static/to keep things tidy, but they can be anywhere as long as they are within the repository.
- Adjust the
configobject at the beginning ofdata/get_googlemap.Rto include your new city, following the format of the existing entries. - (Optional) Adjust logic for the scraper. The scraper is split into roughly three functions, one
to perform a single query, one to scrape data for an entire city using the first function, and
one to combine and clean results from that function into one result object. Whatever you do, you
should export two
FeatureCollections of points for each city into thestatic/folder, one dealing with specificalyl Black-owned restaurants and one covering non-Black-owned restaurants.
The maps here were created through Mapbox's toolkit, firstly through generating a custom theme on Mapbox Studio and then using the Mapbox GL JS library to manipulate and present data on the basemap. I used Mapbox's clustering functions to group data together. The restaurant icon comes from Material Design.
The website itself was scaffolded through the Svelte framework for building UIs in JavaScript. I acknowledge it's a bit niche and not the best choice for collaborative projects, but it's by far the fastest for me in terms of development time.
Each page on the site passes an object of options to the renderMap function located at
src/components/renderMap.ts. You can manipulate values here, as shown in
src/routes/yelp/+page.svelte, to override defaults and insert custom presentation. For Yelp data,
I only pass in one city (Detroit), and change the popup rendering function to account for properties
named differently.
To add a new city, you should place GeoJSON files labeled {city}_non_black_owned.json and
{city}_black_owned.json in a subdirectory in static/. Then, create a new page in routes/ by
creating a folder, and follow the Yelp example to pass as the path_prefix parameter in the
Options object the name of the folder you created in static/.
Of course, you can always just ask me to do it or to assist. The logic is a bit idiosyncratic and can be finnicky to figure out.