These are instructions for your coding exercise. Please read them carefully and make sure to ask whatever question pops into your mind.
We would like to create a system for scheduling appointments with providers (doctors), which would able to:
- Search for providers to set up an appointment with.
- Select an appointment date from the search results & notifying the provider of the new appointment.
- Receive requests for adding, updating or deleting providers
Our system supports both synchronous queries and async queries. We use synchronous queries where possible, but sometimes when we rely on slow/unreliable external systems or when we need an event-based workflow we use an async system. The synchronous queries use REST APIs - one of which you will build. The asynchronous queries use a publish-subscribe (pub-sub) system based on “channels”. You will connect to this system as both a subscriber and a publisher.
- There is no “right” solution! As long as your code works, is readable and extensible- it’s great!
- You may use whatever external library you need.
- We’re looking for big picture stuff:
- Is your code architecture clean?
- Is it easy to add new functionality?
- Is your code testable?
- Is your code readable?
- Does your code handle errors well?
- Does your code have good logging?
- We’re not looking to find unimportant mistakes:
- Don’t get stuck if you don’t understand the instructions. We will help you!
- Don’t get stuck on silly syntax errors in libraries you have no experience with. We will help you!
- Don’t get stuck on coding environment / bootstrapping issues. We will help you!
- We want you to be able to finish this exercise so we can talk about your solution.
- You are not expected to finish any bonus from part C. They are there if you happen to have enough time and to open up discussion points.
- You are not expected to use any specific framework or even any framework at all - work with what you like.
- We encourage you to execute each part of this exercise independently. For each part - read the instructions, make sure you understand them with your interviewer, implement the part, verify the tests pass and call your interviewer before moving on to the next part.
A provider is a doctor. Its Object model looks like this:
{
“name”: “Roland Deschain”, //Provider’s name
“specialties”: [“Neurologist”, “Cardiologist”], //The provider’s specialties
“availableDates”: [ //Available time slots for appointments
{“from”:1548932400000,”to”:1548943200000},
{“from”:1549000800000,”to”:1549031400000}
]
“score”: 9.3 //Vim’s “secret sauce” - a provider’s score
}
We take the provider's name as unique, which we can differentiate between providers - you can rely on that when you think about data modeling.
Dates: The dates (all dates in our system) are formatted as Milliseconds since epoch (Which means that a date like this - 31/11/1990 16:30
is written like this - 660061800000
).
The dates are inclusive, meaning that if a provider has an availability of {"from":100, "to":200}
then acceptable dates for an appointment would be: 150
, 100
, 200
but not 99
, 201
.
Datasource: For the purpose of this exercise, you will get a Provider list from a .json file and all reading/writing from/to it should be done in memory. Design your system in a way that supports millions of Providers, each with thousands of Specialties and AvailableDates, all in memory.
- Make sure you have node.js, npm and git installed.
- Clone this repository and create a new branch.
- The relevant folders for Part A:
./providers
contains mock data./test
contains the tests for your code
- Run
npm install
inside./test
. - To run the tests for this part, you will need to have the server that you will implement running in the background.
- Your server should be running on port 3500. You can change this in
./test/config.js
. - In
./test
executenpm run test-a
.
- Your server should be running on port 3500. You can change this in
The goal of this part is to create a REST endpoint to allow users to set up appointments. Users look for a provider with a specific specialty (e.g ‘Neurologist’, ‘Cardiologist’) and with availability for a certain date. They should receive a list of providers ordered by relevance, and should be able to select one and set up an appointment with them.
Use the mock info under /providers/providers.json
as your data source, but write your code such that it will be easy to switch this mock for an actual data source like a database / another HTTP endpoint.
-
Create a REST server with the following endpoint:
GET /appointments?specialty=<SPECIALTY>&date=<DATE>&minScore=<SCORE>
- The results should take into acount:
- Threshold: They should only get providers whose scores are matching that threshold (inclusive, if
minScore=9.0
then a provider with score 9.0 should be valid) - Specialty: They should only get providers that specialize in that specific specialty. Specialty is not case sensitive.
- Availability: They should only get providers that are available in the specific time requested
- The providers should be ordered by score
- Threshold: They should only get providers whose scores are matching that threshold (inclusive, if
- The endpoint should return an array of provider names according to the order defined above.
- If there are no suitable providers the endpoint should return an empty array.
- If the user gave bad parameters, like a missing specialty or a bad date format (should be milliseconds since epoch), the server should return a
400 (BAD REQUEST)
code.
- The results should take into acount:
-
Create an endpoint to set up an appointment:
POST /appointments
BODY: { “name”: string, “date”: date }
:- The server should validate that such an availability exists. If it doesn’t, the server should return
400 (BAD REQUEST)
. - If such an availability does exist, the server should return a
200 (OK)
response to the client.
- The server should validate that such an availability exists. If it doesn’t, the server should return
Before moving forward, let's just make sure - did you finish part A? Did you go over it with your interviewer? Good job :)
Our pubsub system is an HTTP server that listens on two endpoints: publish
, subscribe
.
The system is based on channels, which according to their name you can either publish messages and subscribe to. It is up to the publishers and subscribers to decide which channels to define and how to use and name them.
When a message is sent to a specific channel, the pubsub system sends that message to all of the listeners that subscribed to that specific channel. The pubsub system sends a message to a listener by executing a POST
request to an endpoint that was given to it by the subscriber.
Publishing a message:
POST /publish
{ “channel”: string,
“payload”: object,
“metadata”: object (optional) }
- Channel (Required): The channel name you’re publishing a message to.
- Payload (Required): The message content. Must be a JSON object.
- Metadata (Optional): Any metadata you want to add. Must be a JSON object. This could include, for example, the publisher’s name, or the date of the published message, or a random id. It is it up to the publisher to decide which metadata is relevant.
The server will return
200 (OK)
on successful publish,400 (BAD REQUEST)
if it received a bad parameter and5XX
on failed publish.
Subscribing to a channel:
POST /subscribe
{ “channel”: string,
“address”: string }
- Channel (Required): The channel name you’re publishing a message to.
- Address (Required): The address to which the pubsub system will send messages that are published on the channel. For example, if you’re interested in the channel “providerUpdates”, you might want to create a REST endpoint at
localhost:port/providerUpdates
and send that address to the pubsub system when subscribing. The server will return200 (OK)
on successful subscription,400 (BAD REQUEST)
if it received a bad parameter and5XX
on failed subscription.
Receiving messages:
Once you’re subscribed to a specific channel, when a message is published on this channel your defined endpoint will receive a POST
request with the following body:
{ “payload”: object,
“metadata”: object (optional) }
Both parameters are JSON objects and are completely defined by the message’s publisher. Metadata is optional.
Cancelling subscriptions:
To delete the listeners you can call GET /reset
. This will delete all subscriptions from all channels.
- Make sure you've completed part A of this exercise.
- Run
npm install
inside./pubsub
. - Running the pubsub server:
- To run the pubsub server, run
npm start
inside./pubsub
. By default it runs on port 3535. Change it using thePORT
env variable. For example:PORT=6700 npm start
. - The tests expect the pubsub server and your server on default ports 3535 and 3500 respectively. They also need port 3838 to be available. You can change all of these in
./test/config.js
. - In
./test
executenpm run test-a
,npm run test-b
ornpm run test
to test part A, part B or both.
- To run the pubsub server, run
The goal of this part is to support sending/receiving messages to/from the providers' applications by using the pubsub system. Your server should support:
- Setting a new appointment
- Adding / Updating providers' info
- Removing providers
These changes are given to you asynchronously using the pubsub system. For this part, as we said before, assume providers are identified by their name, there are no duplicates in name. You should support 'upserting', which means that if a name doesn't exist - add as new provider, if it does - update it.
All the changes for the providers’ info should happen in memory, not on disk. Restarting your server should revert them.
- Whenever a user sets a new appointment using the
POST /appointments
endpoint you wrote in Part A, use the pubsub system to publish a new message to a channel called ‘newAppointments’. The message should contain a payload:{“name”: string, “date”: date}
- Subscribe to the channel called ‘addProvider’. The messages on this channel have a payload that’s a provider according to the object model described above. Your server should receive updates on this channel and add/update your data model according to the changes received.
- Subscribe to the channel called ‘deleteProvider’. The messages on this channel have a payload that looks like this:
{“name”: string}
. Your server should delete the provider according to the deletions received.
- Create a simple web interface that allows searching for providers and setting appointments according to part A.
- Create a simple web interface that allows creating/updating/deleting providers’ information according to part B.
- We’re interested in analytics - your CTO wrote a service that subscribes to the pubsub system on the channel “analytics”. Design and implement messages on this channel that would support:
- Understanding when and which service received/published messages on certain channel
- Performance - How much time each subscriber/publisher worked on each message?
- Understanding a workflow - If a specific request from the client creates a chain of messages moving from channel to channel, how can we easily tell what the workflow is?
- Create a wrapper library for the pubsub system. It should be used as a JS client for people who want to use the pubsub system without executing HTTP requests by themselves. It should find an available port, spin up a server to listen on that port and use that port to listen to published messages.