It's Time for Programming Puzzles

Published on 17 Oct 2021 by Susam Pal


In the beginning of this month, we concluded our previous reading sessions on analytic number theory. It took about 79 hours spread across 120 meetings and 7 months to complete reading a 300-page textbook on analytic number theory. After a short break of two weeks, we are now going to resume our club activities. We have chosen the CSES Problem Set and the associated book written by Antti Laaksonen for this new series of reading sessions.

Puzzles, Programming, and Mathematics

I have been fond of puzzles (mathematical, programming, or otherwise) since my childhood days. Despite being a considerable time sink and having arguably very little utility in life, the activity of solving puzzles is something I find very enjoyable. I enjoy looking at a problem from multiple perspectives and gaining a deeper appreciation of the underlying patterns that lead to interesting mathematical properties related to the problem.

As a result of my personal liking for puzzles, I am eagerly looking forward to hosting these reading sessions. The problems from the CSES problem set we will be looking at are also known as competitive programming problems. They are called so because often problems of this nature are used in programming contests. The problems are usually contrived and solving them generally require skilled application of mathematics, logic, data structures, and algorithms.

To be honest, I do not have a lot of experience in competitive programming. Twelve years ago, a friend insisted that I must try my hand at an upcoming competitive programming contest. He perhaps suggested that after observing my fondness for mathematical puzzles. I took his advice, participated in the CodeChef July 09 contest, and was pleasantly surprised to see that I managed to obtain a decent score. Not bad for the first try!

Apart from that brief encounter, I have not done much competitive programming. However, I have always been involved in mathematics as well as algorithms in a non-competitive manner both as hobby as well as part of my professional work. I still remain fond of puzzles. I am hoping that this club activity will help me to sharpen my skills for solving programming puzzles further!


In the first couple of weeks we are going to look at and discuss the introductory problems from the CSES problem set. We will then move on to sorting and searching, dynamic programming, graph algorithms, and other themes laid out in the problem set. Often seemingly convoluted problems turn out to have simple and elegant solutions. Let us see a couple of examples of the introductory problems.

Note that there are spoilers in the two subsections below. If you want to avoid looking at the spoilers, skip ahead to the section named Community.

Number Spiral

Consider the introductory problem named Number Spiral where we need to predict the integer at row \( y \) and column \( x \) in a spiral like this: \begin{array}{r r r r r} 1 & 2 & 9 & 10 & 25 & \dots \\ 4 & 3 & 8 & 11 & 24 & \dots \\ 5 & 6 & 7 & 12 & 23 & \dots \\ 16 & 15 & 14 & 13 & 22 & \dots \\ 17 & 18 & 19 & 20 & 21 & \dots \\ \vdots & \vdots & \vdots & \vdots & \vdots \end{array} If we represent the integer at row \( y \) and column \( x \) as \( f(y, x) \), where both \( y \) and \( x \) are \( 1 \)-based indices, it turns out that \( f(y, x) \) can be expressed as \[ f(y, x) = 1 + k(k - 1) + (-1)^k (y - x) \] where \( k = \max(y, x) \). Here is a quick illustration of this formula. Say, we want to find the integer at row \( 3 \) and column \( 5 \). We first set \( k = \max(3, 5) = 5 \). Then we compute \( f(3, 5) \) as \[ f(3, 5) = 1 + 5(5 - 1) + (-1)^5 (3 - 5) = 23. \] Indeed the integer at the third row and fifth column is \( 23 \). It takes only a couple of lines to implement this solution as code:

long long f(long long y, long long x)
    long long k = std::max(x, y);
    return (1 + k * (k - 1)) + (k % 2 == 0 ? 1 : -1) * (y - x);

It is this kind of interesting results that make the activity of solving programming puzzles quite fun! How a clumsy looking problem turns into a neat formula like this is something we are going to discuss in our club meetings. In case you are curious, take a look at the relevant meeting boards for this club.

Coin Piles

Here is another interesting introductory problem named Coin Piles. It involves two coin piles containing \( a \) and \( b \) number of coins. On each move, we can remove one coin from one pile and two coins from the other pile. Given \( a \) and \( b \), we need to find out if there is a way to empty both the files after a finite number of moves.

Since we remove exactly \( 3 \) coins in each move, it is easy to see that a necessary condition for emptying both piles successfully is that \( a + b \) must be a multiple of \( 3 \). Further, assuming \( a \le b \), it is also necessary that \( a \) is at least half of \( b \), i.e., the inequality \( a \ge b/2 \) must be satisfied. If \( a < b/2 \), we can never make both piles empty. Even if we were to remove only \( 1 \) coin from the first pile in every move, after \( a \) moves, the second pile would still contain \( b - 2a \) coins left but the first file would be empty thus preventing us from making any subsequent move.

In the previous paragraph, we obtained two necessary conditions with a couple of quick observations. If either of the necessary conditions is violated, then there is no way to make both piles empty. Are the necessary conditions also sufficient? If \( a + b \) is a multiple of \( 3 \) and \( a \ge b/2 \), must we certainly be able to make both piles empty after a finite number of moves? Yes, it turns out that the necessary conditions are also sufficient. It takes a tiny bit of more work to show that it is really so. As a result, assuming that we have already ensured \( a \le b \), the essence of the solution can be expressed as a single line of code:

return ((a + b) % 3 == 0 && 2 * a >= b) ? "YES" : "NO";

Again, we will discuss these things in our meetings. But if you are curious, you can take a look at the relevant meeting boards.


This topic of CSES problem set was chosen by popular demand from the current members of the club. Most existing members of this club have come from ##math and #algorithms channel of Libera IRC. Many members of this club have also come from the Hacker News community. I know from prior meetings on analytic number theory that some members of our club have a strong background in mathematics as well as competitive programming. Some members are far more skilled at these areas than I am. A few members have also made significant progress with the problem set already.

I am hoping that we can learn a lot from the collective experience and expertise of various club members who join our meetings. That has been my experience in the prior book club meetings on analytic number theory too. I learnt a lot more from the insightful comments from other very bright members than I would have learnt if I were reading the book alone. It was a very rewarding experience and I expect the learning experience in the upcoming meetings to be similar too.

Meeting Format

Our meeting style so far has been to go through every page of the book. Additionally, for the upcoming series of meetings, we also plan to go through the programming problems in the CSES problem set. We plan to spend about 20% time on reading the book and 80% time on discussing the problems.

Unlike other book clubs, we do not expect the participants to read the book in advance and we do not limit our discussion to only an overall summary of the sections and chapters from the book. Instead we go through the entire book together in our meetings, page by page. The most serious participants do read the book in advance. As a meeting host, I too read the book in advance. But we read the book together once again line by line during the meetings. We pause at the end of every section to give everyone a chance to add their comments and insights about the section.

In a nutshell, the meeting host and the serious participants do a second pass over the chapters of the book during the meetings. The first pass is typically done privately in advance. However, many participants do their first pass over the material only in our meetings. We consider that fine.

For the subject of analytic number theory in our previous meetings, where the textbook was very dense, a two-pass detailed reading like this was very helpful in thoroughly understanding the material. I do not know yet if the same approach would work well for an easier material like the CSES problem set and the associated book. We will see what works well in our meetings and what does not as we make progress with our meetings. We will then fine tune the meeting format accordingly.

We normally do not record the meetings. However, we are exploring if we can make the meeting recordings available on a temporary basis (say, for a week or two). But do not count on it. Details about this will be shared in our IRC channel mentioned in the next section.

Join Us

If solving programming puzzles sounds like fun to you, consider joining us! Visit CSES Problem Solving Club to get the meeting link, schedule, plan, etc. The first meeting is going to be on Tue, 19 Oct 2021 at 17:00 UTC. See the meeting boards being prepared now to get a glimpse of the kind of stuff we will discuss in the upcoming meetings.

These meetings are open to anyone who is willing to join and study the reference book and the problem set with us. Note that silently lurking is considered completely fine in our meetings. In fact, most members of the club join in and stay silent throughout the meetings. Only a few members talk via audio/video or chat. This is considered absolutely normal in our meetings, so please do not hesitate to join our meetings!

Also, consider hanging out with us at our channel #offbeat on Libera IRC to be a part of our club activities. Alternatively you can also join our channel via its Matrix bridge at Both the channel link and the bridge link point to the same channel, so you need to join only one of them, not both. If you are not an active IRC user, prefer joining the Matrix bridge because it is more convenient for someone unfamiliar with IRC. See you there!