Bikemap Integration

Integration-round task built around real HTTP API calls. Read JSON describing geometric/landmark data, hit a remote API to download tiles or static map images, then progressively annotate the images with markers and connecting lines. Speed-coding round — most candidates finish 3 of 5 parts. Requirements Task 1: Read a JSON file that contains geometric data (coordinates / track points). Load from the given path, take the first N coordinates (usually 10), and print them. Inspect the layout and the coordinate order — latitude-first vs longitude-first — because the later tasks depend on it. Guard against a missing or malformed file. Task 2: Make an authenticated HTTP request to a tile/image API and save the returned image to disk. Know in advance how to write binary content from a requests response to a file. The base map is centered on the Stripe office. Task 3: Same as Task 2 but plot the Task 1 points on the image and connect them with line segments. Task 4: Read a second JSON file with landmarks; for each landmark find the nearest point on the track and annotate both on the image. Task 5: A speed/timing computation; rarely reached in 60 min. Examples # Task 1 shape: load, slice, print import json with open("coordinates.json") as f: data = json.load(f) coords = data["coordinates"][:10] # confirm lat/lon order before using downstream for c in coords: print(c) Notes Most reports get through 3 of 5 parts; 3.5/5 has been seen as a passing signal. Pillow / PIL is the standard imaging library. Know Image.open, ImageDraw, ImageDraw.line, ImageDraw.ellipse. Practice saving images from requests.Response.content. One report notes the interviewer is light on feedback during this round; assume you will be working solo and budget time accordingly. The setup is typically: clone a starter repo (data + task list provided), code on your own machine, share your screen. You may solve it in the language of your choice — Python, JavaScript/Node.js, Java, or others are all accepted — so warm up the HTTP / file-IO / JSON idioms in whichever you pick. Do not over-engineer auth setup — most candidates lose 5+ minutes fighting authorization headers and JSON parsing they could have warmed up beforehand. Time budget is 45–60 minutes. Alternate canonical variant — server-rendered route via POST payload A common shape draws the route server-side rather than locally with Pillow: instead of fetching a base tile and drawing on it, you send the coordinates inside the request payload and the map API returns the finished image with the route already rendered. Step 2 (download base map): send an HTTP POST to the map URL using the JSON payload provided in the starter code; the response body is binary image data — save it with "wb" (write bytes, not a string), then open the file to confirm a map renders. Step 3 (draw the route): add the Task 1 coordinates to the same JSON payload and re-POST; the returned image now has the path drawn on it. You may also need to set line color / thickness in the payload. Coordinate-order corner case: this API typically expects [longitude, latitude] (GeoJSON order), which is the reverse of the latitude-first order familiar from Google Maps. Getting this backwards is a frequent silent bug — the route renders in the wrong place. This is why Task 1 asks you to nail down the order up front. Optional extras (if time): change line color/thickness/style, draw multiple routes, add markers, or split the route into segments based on the data. Discussion follow-ups After coding, expect open-ended design questions: How would you handle thousands of coordinates? How would you design this as a service used by many people? What happens if the API is slow or returns an error? How would you write tests for this code? Preparation Run a full practice round end-to-end against a public API such as https://pokeapi.co/: read JSON, save responses, paginate, draw on an image with Pillow. Pre-write helper snippets for: GET/POST with bearer token, write bytes to file, open image + draw circle/line, parse Unix epoch into datetime, and load JSON + slice the first N items safely. Practice reading long prompts under time pressure — multiple reports cite "read time > code time" as a failure mode.