Creating A Information API Utilizing Kafka, Rockset & Postman


On this submit I’m going to point out you ways I tracked the placement of my Tesla Mannequin 3 in actual time and plotted it on a map. I stroll via an finish to finish integration of requesting knowledge from the automobile, streaming it right into a Kafka Subject and utilizing Rockset to show the information by way of its API to create actual time visualisations in D3.

jp-valery-Qm n6aoYzDs-unsplash

Getting began with Kafka

When beginning with any new instrument I discover it finest to go searching and see the artwork of the potential. Inside the Rockset console there’s a catalog of out of the field integrations that will let you connect Rockset to any variety of present functions you’ll have. The one which instantly caught my eye was the Apache Kafka integration.

This integration lets you take knowledge that’s being streamed right into a Kafka matter and make it instantly out there for analytics. Rockset does this by consuming the information from Kafka and storing it inside its analytics platform virtually immediately, so you possibly can start querying this knowledge straight away.

There are a selection of nice posts that define intimately how the Rockset and Kafka integration works and how you can set it up however I’ll give a fast overview of the steps I took to get this up and operating.

Organising a Kafka Producer

To get began we’ll want a Kafka producer so as to add our actual time knowledge onto a subject. The dataset I’ll be utilizing is an actual time location tracker for my Tesla Mannequin 3. In Python I wrote a easy Kafka producer that each 5 seconds requests the true time location from my Tesla and sends it to a Kafka matter. Right here’s the way it works.

Firstly we have to setup the connection to the Tesla. To do that I used the Sensible Automobile API and adopted their getting began information. You’ll be able to strive it free of charge and make as much as 20 requests a month. When you want to make extra calls than this there’s a paid choice.

As soon as authorised and you’ve got all of your entry tokens, we are able to use the Sensible Automobile API to fetch our car information.

vehicle_ids = smartcar.get_vehicle_ids(entry['access_token'])['vehicles']
# instantiate the primary car within the car id checklist
car = smartcar.Automobile(vehicle_ids[0], entry['access_token'])

# Get car information to check the connection
information = car.information()

For me, this returns a JSON object with the next properties.

        "id": "XXXX",
        "make": "TESLA",
        "mannequin": "Mannequin 3",
        "yr": 2019

Now we’ve efficiently linked to the automobile, we have to write some code to request the automobile’s location each 5 seconds and ship that to our Kafka matter.

from kafka import KafkaProducer
# initialise a kafka producer
producer = KafkaProducer(bootstrap_servers=['localhost:1234'])

whereas True:
      # get the autos location utilizing SmartCar API
    location = car.location()
      # ship the placement as a byte string to the tesla-location matter
    producer.ship('tesla-location', location.encode())

As soon as that is operating we are able to double verify it’s working through the use of the Kafka console client to show the messages as they’re being despatched in actual time. The output ought to look just like Fig 1. As soon as confirmed it’s now time to hook this into Rockset.


Fig 1. Kafka console client output

Streaming a Kafka Subject into Rockset

The crew at Rockset have made connecting to an present Kafka matter fast and simple by way of the Rockset console.

  1. Create Assortment
  2. Then choose Apache Kafka
  3. Create Integration – Give it a reputation, select a format (JSON for this instance) and enter the subject title (tesla-location)
  4. Observe the 4 step course of supplied by Rockset to put in Kafka Join and get your Rockset Sink operating

It’s actually so simple as that. To confirm knowledge is being despatched to Rockset you possibly can merely question your new assortment. The gathering title would be the title you gave in step 3 above. So throughout the Rockset console simply head to the Question tab and do a easy choose out of your assortment.

choose * from commons."tesla-integration"

You’ll discover within the outcomes that not solely will you see the lat and lengthy you despatched to the Kafka matter however some metadata that Rockset has added too together with an ID, a timestamp and a few Kafka metadata, this may be seen in Fig 2. These will likely be helpful for understanding the order of the information when plotting the placement of the car over time.


Fig 2. Rockset console outcomes output

Connecting to the REST API

From right here, my subsequent pure thought was how you can expose the information that I’ve in Rockset to a entrance finish net software. Whether or not it’s the true time location knowledge from my automobile, weblogs or some other knowledge, having this knowledge in Rockset now offers me the ability to analyse it in actual time. Slightly than utilizing the in-built SQL question editor, I used to be searching for a method to permit an internet software to request the information. This was once I got here throughout the REST API connector within the Rockset Catalog.


Fig 3. Relaxation API Integration

From right here I discovered hyperlinks to the API docs with all the data required to authorise and ship requests to the in-built API (API Keys could be generated throughout the Handle menu, underneath API Keys).

Utilizing Postman to Take a look at the API

After getting your API key generated, it’s time to check the API. For testing I used an software referred to as Postman. Postman gives a pleasant GUI for API testing permitting us to rapidly rise up and operating with the Rockset API.

Open a brand new tab in Postman and also you’ll see it’s going to create a window for us to generate a request. The very first thing we have to do is locate the URL we wish to ship our request to. The Rockset API docs state that the bottom deal with is and to question a set it’s essential to append /v1/orgs/self/queries – so add this into the request URL field. The docs additionally say the request kind must be POST, so change that within the drop down too as proven in Fig 4.


Fig 4. Postman setup

We are able to hit ship now and check the URL we have now supplied works. If that’s the case it’s best to get a 401 response from the Rockset API saying that authorization is required within the header as proven in Fig 5.


Fig 5. Auth error

To resolve this, we want the API Key generated earlier. When you’ve misplaced it, don’t fear because it’s out there within the Rockset Console underneath Handle > API Keys. Copy the important thing after which again in Postman underneath the “Headers” tab we have to add our key as proven in Fig 6. We’re basically including a key worth pair to the Header of the request. It’s vital so as to add ApiKey to the worth field earlier than pasting in your key (mine has been obfuscated in Fig 6.) While there, we are able to additionally add the Content material-Sort and set it to software/json.


Fig 6. Postman authorization

Once more, at this level we are able to hit Ship and we must always get a special response asking us to supply a SQL question within the request. That is the place we are able to begin to see the advantages of utilizing Rockset as on the fly, we are able to ship SQL requests to our assortment that can quickly return our outcomes to allow them to be utilized by a entrance finish software.

So as to add a SQL question to the request, use the Physique tab inside Postman. Click on the Physique tab, ensure that ‘uncooked’ is chosen and make sure the kind is ready to JSON, see Fig 7 for an instance. Inside the physique area we now want to supply a JSON object within the format required by the API, that gives the API with our SQL assertion.


Fig 7. Postman uncooked physique

As you possibly can see in Fig 7 I’ve began with a easy SELECT assertion to only seize 10 rows of information.

    "sql": {
       "question": "choose * from commons."tesla-location" LIMIT 10",
       "parameters": []

It’s vital you utilize the gathering title that you simply created earlier and if it comprises particular characters, like mine does, that you simply put it in quotes and escape the quote characters.

Now we actually are able to hit ship and see how rapidly Rockset can return our knowledge.


Fig 8. Rockset outcomes

Fig 8 exhibits the outcomes returned by the Rockset API. It gives a collections object so we all know which collections have been queried after which an array of outcomes, every containing some Kafka metadata, an occasion ID and timestamp, and the lat lengthy coordinates that our producer was capturing from the Tesla in actual time. In accordance with Postman that returned in 0.2 seconds which is completely acceptable for any entrance finish system.

In fact, the probabilities don’t cease right here, you’ll usually wish to carry out extra advanced SQL queries and check them to view the response. Now we’re all arrange in Postman this turns into a trivial process. We are able to simply change the SQL and preserve hitting ship till we get it proper.

Visualising Information utilizing D3.js

Now we’re capable of efficiently name the API to return knowledge, we wish to utilise this API to serve knowledge to a entrance finish. I’m going to make use of D3.js to visualise our location knowledge and plot it in actual time because the automobile is being pushed.

The stream will likely be as follows. Our Kafka producer will likely be fetching location knowledge from the Tesla each 3 seconds and including it to the subject. Rockset will likely be consuming this knowledge right into a Rockset assortment and exposing it by way of the API. Our D3.js visualisation will likely be polling the Rockset API for brand spanking new knowledge each 3 seconds and plotting the newest coordinates on a map of the UK.

Step one is to get D3 to render a UK map. I used a pre-existing instance to construct the HTML file. Save the html file in a folder and title the file index.html. To create an internet server for this so it may be considered within the browser I used Python. In case you have python put in in your machine you possibly can merely run the next to begin an internet server within the present listing.

python -m SimpleHTTPServer

By default it’s going to run the server on port 8000. You’ll be able to then go to in your browser and in case your index.html file is setup accurately it’s best to now see a map of the UK as proven in Fig 9. This map would be the base for us to plot our factors.


Fig 9. UK Map drawn by D3.js

Now we have now a map rendering, we want some code to fetch our factors from Rockset. To do that we’re going to write down a operate that can fetch the final 10 rows from our Rockset assortment by calling the Rockset API.

operate fetchPoints(){
  // initialise SQL request physique utilizing postman instance
  var sql="{ "sql": { "question": "choose * from commons."tesla-location" order by _event_time LIMIT 10","parameters": [] }}"
  // ask D3 to parse JSON from a request.
    // setting headers the identical means we did in Postman
    .header('Authorization','ApiKey AAAABBBBCCCCDDDDEEEEFFFFGGGGG1234567')
    .header('Content material-Sort','software/json')
    // Making our request a POST request and passing the SQL assertion
      // now we have now the response from Rockset, lets print and examine it
      var response = JSON.parse(d.response)
      // parse out the checklist of outcomes (rows from our rockset assortment) and print
      var newPoints = response.outcomes

When calling this operate and operating our HTTP server we are able to view the console to take a look at the logs. Load the webpage after which in your browser discover the console. In Chrome this implies opening the developer settings and clicking the console tab.

It is best to see a printout of the response from Rockset exhibiting the entire response object just like that in Fig 10.


Fig 10. Rockset response output

Beneath this ought to be our different log exhibiting the outcomes set as proven in Fig 11. The console tells us that it is an Array of objects. Every of the objects ought to characterize a row of information from our assortment as seen within the Rockset console. Every row contains our Kafka meta, rockset ID and timestamp and our lat lengthy pair.


Fig 11. Rockset outcomes log

It’s all coming collectively properly. We now simply must parse the lat lengthy pair from the outcomes and get them drawn on the map. To do that in D3 we have to retailer every lat lengthy inside their array with the longitude in array index 0 and the latitude in array index 1. Every array of pairs ought to be contained inside one other array.

[ [long,lat], [long,lat], [long,lat]... ]

D3 can then use this as the information and mission these factors onto the map. When you adopted the instance earlier within the article to attract the UK map then it’s best to have all of the boilerplate code required to plot these factors. We simply must create a operate to name it ourselves.

I’ve initialised a javascript object for use as a dictionary to retailer my lat lengthy pairs. The important thing for every coordinate pair would be the row ID given to every outcome by Rockset. This can imply that once I’m polling Rockset for brand spanking new coordinates, if I obtain the identical set of factors once more, it gained’t be duplicated in my array.

    _id : [long,lat],
    _id : [long,lat],

With this in thoughts, I created a operate referred to as updateData that can take this object and all of the factors and draw them on the map, every time asking D3 to solely draw the factors it hasn’t seen earlier than.

operate updateData(coords){
    // seize solely the values (our arrays of factors) and cross to D3  
    var mapPoints = svg.selectAll("circle").knowledge(Object.values(coords))
    // inform D3 to attract the factors and the place to mission them on the map
    .attr("cx", operate (d) { return projection(d)[0]; })
    .attr("cy", operate (d) { return projection(d)[1]; })
    .attr("r", "2px")
    .attr("fill", "crimson")


All that’s left is to alter how we deal with the response from Rockset in order that we are able to constantly add new factors to our dictionary. We are able to then preserve passing this dictionary to our updateData operate in order that the brand new factors get drawn on the map.

//initialise dictionary
var factors = {}

operate fetchPoints(){
  // initialise SQL request physique utilizing postman instance
  var sql="{ "sql": { "question": "choose * from commons."tesla-location" order by _event_time LIMIT 10","parameters": [] }}"
  // ask D3 to parse JSON from a request.
    // setting headers the identical means we did in Postman
    .header('Authorization','ApiKey AAAABBBBCCCCDDDDEEEEFFFFGGGGG1234567')
    .header('Content material-Sort','software/json')
    // Making our request a POST request and passing the SQL assertion
      // now we have now the response from Rockset, lets print and examine it
      var response = JSON.parse(d.response)
      // parse out the checklist of outcomes (rows from our rockset assortment) and print
      var newPoints = response.outcomes

      for (var coords of newPoints){
          // add lat lengthy pair to dictionary utilizing ID as key
          factors[coords._id] = [coords.long,]
          console.log('updating factors on map ' + factors)
          // name our replace operate to attract factors on th

That’s the bottom of the appliance accomplished. We merely must loop and constantly name the fetchPoints operate each 5 seconds to seize the newest 10 information from Rockset to allow them to be added to the map.

The completed software ought to then carry out as seen in Fig 12. (sped up so you possibly can see the entire journey being plotted)


Fig 12. GIF of factors being plotted in actual time

Wrap up

By way of this submit we’ve learnt how you can efficiently request actual time location knowledge from a Tesla Mannequin 3 and add it to a Kafka matter. We’ve then used Rockset to eat this knowledge so we are able to expose it by way of the in-built Rockset API in actual time. Lastly, we referred to as this API to plot the placement knowledge in actual time on a map utilizing D3.js.

This offers you an thought of the entire again finish to entrance finish journey required to have the ability to visualise knowledge in actual time. The benefit of utilizing Rockset for that is that we couldn’t solely use the placement knowledge to plot on a map but additionally carry out analytics for a dashboard that might for instance present journey size or avg time spent not transferring. You’ll be able to see examples of extra advanced queries on linked automobile knowledge from Kafka on this weblog, and you may strive Rockset with your personal knowledge right here.

Lewis Gavin has been a knowledge engineer for 5 years and has additionally been running a blog about expertise throughout the Information neighborhood for 4 years on a private weblog and Medium. Throughout his pc science diploma, he labored for the Airbus Helicopter crew in Munich enhancing simulator software program for army helicopters. He then went on to work for Capgemini the place he helped the UK authorities transfer into the world of Huge Information. He’s presently utilizing this expertise to assist rework the information panorama at, a web-based charity cashback web site, the place he’s serving to to form their knowledge warehousing and reporting functionality from the bottom up.

Picture by Jp Valery on Unsplash


Leave a Reply