The WattDepot Katas Project: Thoughts on an Open Source API

This week, I worked on a series of code katas for the WattDepot API, an open-source interface used to retrieve data on daily power and energy generation and use from remote sensors. The WattDepot API is intended to give users the tools to process technical, obtuse, automatically generated energy data and put it into a form that means something to the people who will use it.

I was able to complete the first Kata, SourceListing, in one or two hours; it asked for an alphabetical listing of all the Sources on the WattDepot server and their descriptions. Most of the code for this can be attributed to the SimpleApplication.java example program included in the distribution, as a Source object includes a name and a description.

The second Kata, SourceLatency, which required a program that calculated latency for each Source and printed the sources in ascending order based on latency, took two days. I had somehow managed to finish ICS 211 without learning how to use Hashtable, or else I had just forgotten that it existed. As a result, I spend the next two days trying to figure out how to sort one array according to the ordering of another array, without any success whatsoever. The solution I eventually found, and the way in which I found it, demonstrate why it is better to work on complex projects in groups. The reason I finally gave up on my insane attempt to sort two arrays simultaneously was because one of my classmates suggested during peer-programming exercises that I use a Hashtable. It turned out that a Hashtable with array-list values to store Sources with the same latency values worked just fine using the latencies as keys. As in some of my other projects, I have found that solving programming problems is a strange mix of happiness that something works, and frustration that it took me several sleep-deprived hours to figure out a simple mistake.

The third Kata, SourceHierarchy, required a program which printed Sources and their SubSources with increasing indentation for each level of sub-source. This kata also took two days, mostly because I sought clarification on how to parse the SubSources. The major issue I had at first, which turned out to be a non-issue when I figured out the actual format of the assignment, was that the SubSources returned by Source.getSubSources are not a recursive data structure (that is, a SubSource does not contain the Source data for its SubSources, only their URIs), and it was not possible to construct the hierarchy recursively. The trick, as Professor Johnson said, was to make a single pass through the list to figure out which sources were SubSources, and then use that to create the hierarchy. The solution I eventually implemented parsed the URIs to retrieve the SubSource names and did not print any Source which was listed as another source's SubSource.

The fourth Kata, EnergyYesterday, required a program which totaled up the energy used by each source on the previous day and printed each Source and its energy use in ascending order by energy use. This was similar in design to SourceLatency; in both cases, I used Hashtables to store Lists of Sources with identical energy or latency values, using the energy or latency values as the keys to retrieve them, then iterating through the lists and printing each Source with its associated key. Once I had fixed my SourceLatency issues, it did not take much editing to create EnergyYesterday; the main additional problem was the simple task of creating a timestamp which was rolled back by one day.

The fifth Kata, HighestRecordedPowerYesterday, required a program to print the highest recorded power for each source for the previous day, sorted in ascending order. It took up almost all of Monday, and several hours of Tuesday morning; this blog entry is being written at 6:00 AM on Tuesday morning, November 8, and the last kata still isn't finished. I had to write several utility methods for this class, mostly to generate the timestamps for power data requests and to parse the timestamps back into AM/PM format. The kata appears to be working correctly, except for one thing: I was never able to figure out whether it was a bug, or just a reflection of college study habits, but the peak power measured for every single sensor was at 12:00 AM midnight:

// My output for Kata 5 is reproduced below:
Connected successfully to: http://server.wattdepot.org:8190/wattdepot/
Server: http://server.wattdepot.org:8190/wattdepot/

Source              Highest recorded power in watts ( 07-Nov-2011 )
Lokelani-12-telco   1,203  12:00am
Lehua-10-telco      1,263  12:00am
Mokihana-08-telco   1,390  12:00am
Mokihana-04-telco   1,612  12:00am
Ilima-12-telco      1,818  12:00am
Lehua-04-telco      1,868  12:00am
Ilima-10-telco      2,227  12:00am
Lokelani-04-telco   2,432  12:00am
Mokihana-10-telco   2,662  12:00am
Lokelani-06-telco   2,890  12:00am
Ilima-08-telco      3,276  12:00am
Ilima-08-lounge     4,036  12:00am
Lokelani-10-lounge  4,204  12:00am
Ilima-10-lounge     4,285  12:00am
Lokelani-06-lounge  4,667  12:00am
Ilima-06-lounge     5,208  12:00am
Ilima-A             5,237  12:00am
Lokelani-04-lounge  5,458  12:00am
Mokihana-12-lounge  5,665  12:00am
Lokelani-D          5,845  12:00am
Mokihana-A          5,862  12:00am
Lehua-08-lounge     6,233  12:00am
Ilima-D             6,512  12:00am
Mokihana-10-lounge  6,612  12:00am
Mokihana-C          6,877  12:00am
Lehua-B             7,258  12:00am
Mokihana-E          7,484  12:00am
Lehua-C             7,760  12:00am
Ilima-B             8,022  12:00am
Lokelani-08-lounge  9,564  12:00am
Ilima               33,961  12:00am
Mokihana            35,322  12:00am

The sixth kata, MondayAverageEnergy, required me to write out the average energy consumed by a Source over the previous two Mondays. It is probable that I did not finish this kata in the hours remaining before the deadline. The main problem is in figuring out if today is Monday, then subtracting the number of days since the last Monday (to generate the first data set), then rolling back the date seven days to the previous Monday (to generate the second data set), and if I had had more time I would most likely have been able to reuse some of the code from EnergyYesterday to do this. I would have to put most of the blame for my poor scheduling on the two days of time I used to try to sort two arrays simultaneously for the second and fourth katas.

Though I may have frequently been frustrated with my own lack of knowledge or the difficulties of certain parts of the assignment, the WattDepot katas generally served as another example of how code katas can force developers to learn about a language or API's capabilities very quickly. Parsing WattDepot energy data into human-readable form is mostly a matter of timestamp generation, sorting among multiple sets of data elements, and parsing strings, and I was forced to learn new capabilities of the Java language and the WattDepot API for all of these tasks. The difficulties of making energy conservation relevant to consumers and enterprise users begin with putting information about their energy use into a form they can understand, and the WattDepot API provides a flexible platform for this purpose.