An exercise in Scala/Akka recording clicks/impressions for 'net advertising.
Specifically, this is probably what most people would call a microservice too (it has virtually no external dependencies).
Internet advertising is sold in two forms: by the click and by the impression.
The first is referred to as PPC (Pay Per Click) and the second is PPM (Pay Per Mille, where Mille by convention means a thousand impressions). The general gist is that more eyeballs on an advertisement are generally better than fewer. And of course it's really important to get those advertisements in front of the right eyeballs. So demographic and user profiles are valuable information (just ask Facebook or Google).
As should be clear by now, this is a business that is based upon volume. And of course the goal of all this advertising is name brand recognition, but perhaps more importantly: clickthrough (or conversion). Or in other words, how many people were actually motivated to part with their cash and actually buy something by the advertisement. And all of this needs to be tracked, so that the advertising agency or Amazon/Facebook/Google can get paid.
For some reason, Scala has a reputation for being difficult. I'm not sure why, compared with C++ or Lisp or Scheme or Ops5 it's actually pretty simple. Sure, purely functional or reactive or reactive/functional programming can be very hard (conceptually at least it can be hard to break old habits) - but these are not imposed by Scala. It's entirely possible to write Scala code that looks almost exactly like the Java code that it is supposed to replace. In fact it is exactly possible, since Scala runs on a JVM (Java Virtual Machine).
Be that as it may, Akka is definitely a pretty heavy-duty framework.
Here I used Akka (specifically akka-http) simply due to the very high-order
transaction volume.
[I had previously used Spray - which is a framework much like node.js or Python's Flask or even Golang's Gorilla/mux. However, Spray is now deprecated so I decided to use akka-http which is its replacement (another option might have been Play).]
I am including the specification I was given.
Run the following command to build the service:
$ sbt test
This will also run a test suite:
< ... >
[info] Done compiling.
[info] AnalyticsServiceSuite:
[info] - should return 0 records in response to get Statistics (9-10 PM Pre-New Year)
[info] - should return 1 record in response to get Statistics (10-11 PM Pre-New Year)
[info] - should respond to get Statistics (New Year)
[info] - should respond to get Statistics pre-New Year
[info] - should respond to get Statistics (1 AM)
[info] - should respond to post Statistic
[info] - should use cached Statistics (First 1 AM)
[info] - should NOT use cached Statistics (First)
[info] - should use cached Statistics (Second 1 AM)
[info] - should NOT use cached Statistics (Second)
[info] Run completed in 6 seconds, 389 milliseconds.
[info] Total number of tests run: 10
[info] Suites: completed 1, aborted 0
[info] Tests: succeeded 10, failed 0, canceled 0, ignored 0, pending 0
[info] All tests passed.
[success] Total time: 12 s, completed 22-Mar-2018 9:24:30 PM
$
Run the following command:
$ sbt run
The service runs on port 8080 by default and should be available at:
http://localhost:8080/analytics
It may also be tested with curl.
[Press Enter or Return to terminate.]
The procedure will be to create an analytic, and then query for that analytic.
To create an analytic for the given timestamp:
$ curl -v -X POST 127.0.0.1:8080/analytics?timestamp=1514793600000\&user=Fred\&event=impression
* Trying 127.0.0.1...
* Connected to 127.0.0.1 (127.0.0.1) port 8080 (#0)
> POST /analytics?timestamp=1514793600000&user=Fred&event=impression HTTP/1.1
> Host: 127.0.0.1:8080
> User-Agent: curl/7.47.0
> Accept: */*
>
< HTTP/1.1 204 No Content
< Server: akka-http/10.1.0
< Date: Fri, 23 Mar 2018 04:10:07 GMT
<
* Connection #0 to host 127.0.0.1 left intact
$
To get analytics for the current hour of the timestamp, up to and including the actual millisecond of the timestamp:
$ curl -v 127.0.0.1:8080/analytics?timestamp=1514793600000
* Trying 127.0.0.1...
* Connected to 127.0.0.1 (127.0.0.1) port 8080 (#0)
> GET /analytics?timestamp=1514793600000 HTTP/1.1
> Host: 127.0.0.1:8080
> User-Agent: curl/7.47.0
> Accept: */*
>
< HTTP/1.1 200 OK
< Server: akka-http/10.1.0
< Date: Fri, 23 Mar 2018 04:10:55 GMT
< Content-Type: text/plain; charset=UTF-8
< Content-Length: 38
<
unique_users,1
clicks,0
impressions,1
* Connection #0 to host 127.0.0.1 left intact
$
As instructed, this project was meant to be an MVP (Minimum Viable Product), meaning a proof-of-concept, spike or demo suitable for showing to a client. It was never meant to be a complete and finished product. Given the vagueness of the specification, it was meant to be a jumping-off point for clarifying the actual requirements.
Even so, there are still a few things that - given time - I would like to rewrite and/or clean up. A partial list follows.
- Write the same functionality in Golang and carry out a head-to-head benchmark
- Rewrite the persistence portion (possibly in ScalikeJDBC)
- Rewrite the code which makes use of mutable DateTime objects to be more functional in style
- Rewrite the test suite
AnalyticsServiceSuite.scalato be more functional in style