DayPath Journal

Google Interview CUMSUM Challenge

I am told that there are more than a few former Microsoft employees working fulltime at Google. I talked to one in fact. He was prepared to share a Google Doc with me and test me out but I had to reschedule and the second guy was clearly non-Microsoft. He works on a Google-customized version of Eclipse every day (I historically preferred NetBeans). His team officially does not recognize git culture as he proudly pronounced the word Perforce. Most importantly, when I attempted to work on his coding challenge he stopped me. He did not want me to use LINQ (from my point of view he wanted me to re-implement LINQ in addition to working on the challenge he presented).

He presented me with the challenge of implementing the CUMSUM function from the R programming language. Long story short: after his rejection of LINQ, I refused to participate further in the interview. (It might help to mention that Google contacted me—I would not volunteer to run this experiment on them.)

Microsoft products like Typescript are being used at Google but the penetration of Microsoft cultural awareness is not as deep as I would prefer (optimism on my part). Unless I am applying for a job to work on the .NET Framework itself, I do not expect an interviewer to stop me from using LINQ. Should I have been able to solve the problem within the time pressure allotted, I still would have been taken aback by the Google-cultural distance from Microsoft technologies. Additionally, this was the second live-coding interview under my belt (my first was actually with Microsoft) and I am even more convinced that this approach is too adversarial and ultimately meaningless (I would give a take-home test and follow it up with in-depth questions about the code—and then resort to live coding a variation based on the take-home original—probably too time-consuming for busy, expensive engineers—and, yes, I have conducted developer interviews).

As is customary, after my failure during the interview, I took some time with LINQPad to solve the problem alone. The interviewer supplied me with the signature of this function and two examples of output:

//cumsum(int[] array, int startPos, int count);
var set = new[] { 4, 1, 0, 3, 2 };
cumsum(set, 3, 4) == new[] { 3, 5, 9, 10 };
cumsum(set, 4, 2) == new[] { 2, 6 };

He also mentioned that count can extend beyond the upper bound of the array, leading me to add that the logic should wrap around the end of the array, pulling items from the beginning. I stated within the allotted time of the interview that I would work on developing a generic function that can perform this ‘wrap’ operation. What I failed to do was produce this function seconds after my statement (which means that I am not worthy of Google). Anyway, this is my wrapping extension method using LINQ:

public static IEnumerable<T> Wrap<T>(this IEnumerable<T> enumerable, int startPos, int count)
    if (enumerable == null) return Enumerable.Empty<T>();
    var length = enumerable.Count();
    if(startPos > length) throw new ArgumentException("The start position is larger than the length of the enumerable.", "startPos");
    var wrappedSet = enumerable
        .Union(enumerable.Take(length - startPos));
    return wrappedSet;

In LINQPad I can run this:

var set = new[] { 4, 1, 0, 3, 2 };
set.Wrap(3, 4).Dump();

and Dump() out this:

new[] { 3, 2, 4, 1 }

Now, I have no idea why anyone would want to do this because I am not an R programmer or a serious student of statistics (and it has been decades since I wrote the words Eigen values). My ignorance does not stop me from moving to the next step in the form of another LINQ extension method:

public static IEnumerable<int> ToCulmulativeSum(this IEnumerable<int> enumerable, int startPos, int count)
    if (enumerable == null) return Enumerable.Empty<int>();
    var wrappedSet = enumerable.Wrap(startPos, count);
    wrappedSet.Dump("wrapped set");
    return wrappedSet
        .Select((x, i) => x + wrappedSet.Take(i).Sum());

This ToCulmulativeSum() method is the correct solution by my standards—but, as whined about earlier, it is not sufficient for Google. In order to be a “smart” developer, I would have to implement every single LINQ extension method used in the examples above. Money cannot buy everything and this is intellectually (and aesthetically) satisfying enough for me:

var set = new[] { 4, 1, 0, 3, 2 };
set.ToCulmulativeSum(3, 4).Dump();
set.ToCulmulativeSum(4, 2).Dump();

As soon as Google owns a programming language that has LINQ-like syntax that most Google employees use every day, all of this stuff here will suddenly be OK. I doubt that Google will use C# in spite of its open-source status.

It may be of interest to mention that I would likely not be able to solve this problem again in another interview. I tend to fail to memorize solutions to problems that I have already written down. Programmers in the real world are experts at looking stuff up—even the stuff we’ve written down in the past. The luxury of memorizing even 5% of what we have learned comes from socializing with—and even dearly befriending—others who truly love to hear about these little adventures. Some of us do not have this luxury—and we literally pay for it in terms of velocity of career growth.

Related Links