Math, culture and programming languages

Can a programmer’s background determine whether he’ll be a great programmer? By background, I mean his upbringing, the values learnt, his primary (and perhaps secondary) spoken/written language and so on.

I don’t know. However, I have arguments for and against the proposition. Let’s start with…

Learning to count

You think counting is easy? Apparently not. Recently I read a book, Outliers by Malcolm Gladwell. According to Gladwell, American children the age of four can, on average, count up to fifteen. Chinese children at that age can, on average, count up to forty.

His reasoning is that, the system of naming numbers is different in English and in Chinese.

For example, think about counting from twenty to thirty (I’m deliberately using the English form instead of the Arabic numerals to highlight the difference). You have twenty, twenty one, twenty two and so on till twenty nine and thirty. How about thirty to forty? Thirty, thirty one, thirty two and so on till thirty nine and forty.

Consider counting from ten to twenty. Ten, eleven, twelve, thirteen and so on till eighteen, nineteen and twenty. The pattern is different and irregular.

Before I show you the Chinese method of counting, let me show you the first 10 numbers, shown by the Arabic numeral, the English name and the (closest) Chinese pronunciation I can give:

1 one (yi, like in “yeast”)
2 two (er, like in “brighter”)
3 three (san, “sahn”)
4 four (si, do a short hissing sound with the s)
5 five (wu, “woo” and keep it short)
6 six (liu, “li” and “ou” as in “shoulder”, and string li-ou together quickly)
7 seven (qi, “cheese” without the s, and shorter in length)
8 eight (ba, as in “barter”)
9 nine (jiu, “gi” as in “gin” and “ou” as in “shoulder”. Like that of six.)
10 ten (shi, like that of four, with the h)

There are actually 4 tonal inflections for a Chinese character pronunciation, and we’ll ignore that for this discussion.

So to count in Chinese from ten to twenty, we have shi, shi yi, shi er, shi san, shi si, shi wu, shi liu, shi qi, shi ba, shi jiu, er shi. They are literally “ten”, “ten one”, “ten two”, “ten three” and so on till “ten nine” and “two ten”. There’s an implicit “one” in front of the “ten”, so it’s “one ten one” for eleven. For twenty three, it’s “er shi san”, or literally “two tens and three”.

Gladwell says this gives structure to the counting system, so children are able to grasp larger numbers quicker. The faster you can count to larger numbers, the more operations you can do on them. Additions, subtractions, summations and so on.

Ok, I’m not saying the Chinese number naming system is better than the English system. It’s just different. Gladwell says this difference also makes memorising short number sequences easier. For example, I can remember my Identification Card Number (equivalent to the Social Security Number in America) easier in Chinese than in English.

It explains why when someone asks for my phone number in English, I have a problem. Because I’m mentally translating my memory of the phone number from Chinese to English. Did you know it’s kinda hard to say out 8 digits in English while translating them from a Chinese memory? I can even mentally picture the numbers. It’s the speaking out that’s taking up mental processing time.

Gladwell also made a point about Cantonese (a Chinese dialect) pronunciation of numbers. So I tried saying out numbers in Cantonese (yes, I’m multilingual), and wow, it is easier to say and memorise! The Cantonese pronunciation of numbers are short in length, which makes it easier to spit them out *smile*

And the relation to programming? Programming is made up largely of counting and solutions formed from abstract ideas. When I first learned C, I was surprised that many fellow students had difficulty counting the number of iterations in say,

for (i=0; i<10; ++i)
	if (i>7) break;
	// do something

Counting and iterating leads to lists of data, or sets of data. In SQL, you can manipulate sets of data as if it’s one unit, abstracting away the fact that the data is actually iterated one by one. For example, you can select information from another data set, or a subquery as it’s known.

Of course, there are still some people who have difficulty visualising SQL data sets as one unit, hence their need to iterate over that one record by one record at a time, even when there’s no need. What does that tell you about these people?

Hard work is valued

Gladwell also made a point about culture. That successful people seem to grow up with a culture of valuing hard work. The ability to think on a problem long enough to come up with a solution.

He said something worth thinking about. There’s a educational researcher by the name of Erling Boe at the University of Pennsylvania. Boe says that one can know if a child will do well in math without asking that child a single math question.

The example in the book was a fictional Math Olympics. Before the test, there was a questionnaire to be filled in. There were tons of questions inside, none related to math. Boe asserts that a child who finishes that questionnaire will also do correspondingly just as well in the math test.

It’s a question of perseverance, the willingness to put one’s mind to work, even if one doesn’t feel like it.

Hard work is something valued in the countries of “wet-rice agriculture and meaningful work”, as Gladwell puts it. Based on the research of Boe, the top countries are Singapore (yay!), South Korea, Taiwan, Hong Kong and Japan. Though I’m not so sure of my fellow countryman’s motivations… The parents and children could be driven more by the race for top scores rather than the culture passed on by our forefathers.

Well, I haven’t read much on Boe’s research, but the hardworking nature of my countrymen is fairly accurate. Students are willing to study for long periods of time (most of them anyway…). Adults are willing to work just a little longer, a little harder at work. For example, I recently knew that my friend worked overtime till 4 am at the office. Personally, I think that’s crazy, but to each his (in this case, her) own.

The native language

There was a question in StackOverflow about coding in other spoken languages, which was highlighted in Jeff and Joel’s recent podcast.

The reserved words in a programming language are fixed. Usually they are in English, though there are programming languages in say Chinese. I didn’t know there was a Chinese version of BASIC!

Once, I took up an SQL reference book, written in Chinese. I want to mention that I cannot read a programming reference book written in Chinese. That is to say, I can certainly read the Chinese characters, but I can’t understand the heck what it means.

I need a reference book written in English for SQL, because the native (human) language for SQL is English! Unless there’s a variant I don’t know about…

I’ve seen some code written in Spanish before, I think. Can’t remember. Anyway, the native language for the programming language (C, I think) was English, so the code reads fine. The variable names look different, but I didn’t have too much difficulty.

My guess is that my math background prepared me for abstract notions and symbols, and still be able to work with them. So I treated variables named in a foreign language as just another symbol. And continued to read the code based on that.

And this brings me full circle to…

So does background really matter?

I am unfamiliar with how an American (or English, or French) grows up. I don’t really know the values valued, or the culture surrounding the upbringing of a child.

I do know mine. I’m brought up learning two languages (English and Chinese), two Chinese dialects (Hokkien and Cantonese). I taught myself to read Japanese characters. I’m brought up around people who wake up before dawn to work, and work long hours, regularly and consistently and over long periods of time.

Personal values, personality, genes and luck. I agree they play a part in the makings of a great programmer. In particular, I believe that one’s background influences personal values and personality, so in that sense, background does matter.

And specifically, I think my math background makes grasping programming concepts easier for me.

Of course, everything you’ve just read could be hogwash, because I’m still telling (interesting, I hope) anecdotes to illustrate points as Joel points out vehemently. I haven’t read a lot about computer science and its history in America (or pretty much anywhere in the world). I research just enough so that what I write is as true as I understand it (sometimes I don’t research at all!).

I admit that I’m still naive and easily impressed. I still pretty much trust what I read as true. It’s only when I start internalising the information that I really think about them.

So what do you think?

Solving the Resident Evil Zero game math puzzle


On the screen is shown a “00/81”. A number pad with number keys 1 to 9 is on the right. When a number key is entered, a knob is lit in red and the value of the number is added to the numerator of the displayed ratio. There are 10 unlit knobs.

The puzzle should be obvious. Match the numerator and the denominator with 10 entries.

So how would I solve it? My hint as stated, was:

the last entry is the most important

The first 9 entries are to get you as close to the required number as possible. This is actually similar to blackjack, a card game where you try to get as close to 21 points as possible without going over. The difference is that you don’t get to choose your next number (or card value) in blackjack. You do here.

[image by pavlen]

So my first instinctual thought was to finish up the first 9 entries to get to the last entry. My second instinctual thought was to make these 9 entries all the same number value. It’s a timed effort. I don’t have time to go figure out different combinations of sums.

Based on those thoughts, my next (instinctual) step was to divide the required number by 9, rounding down to nearest integer. Don’t ask me why at this point, because I’ll know the reason only after thinking it through, and I’m not thinking it through at this point.

So we have 81 divide by 9, and rounded down, we get 9. We have a problem. It’s not just 9. It’s exactly 9. There’s no number value 0 for the final entry. So we use 8 (1 less than 9).

Then we multiply 8 by 9 (entries) to get 72. And then 81 (required number) – 72 is 9, the final entry. So the answer is 8,8,8,8,8,8,8,8,8,9.

This method gives the simplest of combinations (only 2 distinct numbers used) and is the fastest to get you to the end point, the final entry. It’s the final entry that “corrects” the sum to the required number.

I have to say, the changing of 9 to 8 is (I’m starting to use the word generously) instinctive. I don’t know how changing 9 to 8 would solve the problem. I just know.

Let’s use this method on the second part of the puzzle: 67.

67 divide by 9 rounded down gives 7. 7 multiplied by 9 is 63. 67 – 63 is 4. So the answer is 7,7,7,7,7,7,7,7,7,4.

There you have it, the algorithm to solving the math puzzle. The second puzzle uses the “normal” mode of the algorithm. The first puzzle tested the edge case of the algorithm.

Hmm… can we write a function that spits out the combination using the above algorithm? Can we write a function that spits out all possible combinations?

Can you solve this game math puzzle in seconds?

The game in question is Resident Evil Zero. Didn’t think there’d be a math puzzle in a survival horror game? Neither did I.

The game video footage (video link) is over 10 minutes long, so I’ll highlight the relevant parts. Warning: the player is uh, slightly liberal in his use of words. This video’s considered ok, comparing with his other videos. His commentary is colourful and funny, which is why I watch his videos. He makes surviving in a horror story, fun. *smile*

The player mentions a difficult math puzzle at about 1 minute 44 seconds (henceforth noted 1:44) into the video. That got my attention. A math puzzle in a game?

He even tagged it as such:

Stupid math puzzle

I watched the video till about 7:44, where the math puzzle finally made its entrance.

Resident Evil Zero math puzzle part 1

On the screen is shown a “00/81”. A number pad with number keys 1 to 9 is on the right. When a number key is entered, a knob is lit in red and the value of the number is added to the numerator of the displayed ratio. There are 10 unlit knobs.

The puzzle should be obvious. Match the numerator and the denominator with 10 entries.

This is the math puzzle?! Still, in an action game, any kind of math puzzle is probably fatal. I can fumble on simple additions too. Guess I need to practise more.

Anyway, his answer was 9,9,9,9,9,9,9,8,7,3 for a total of 81. Puzzle solved. Wait, there’s more.

Resident Evil Zero math puzzle part 2

In the second part of the puzzle (at about 8:13), the numerator was obscured. The denominator was 67. This time, the player had to really know what he should enter before entering anything. He fumbled for the right combination for about 2 minutes. And he hadn’t saved the game since the beginning of the game…

With about 20 seconds left on the clock, he finally solved it. His answer was 5,5,5,5,9,9,9,9,3,8 for a total of 67.

So my challenge to you is, can you quickly come up with a solution in a timed situation? How would you solve the problem? What goes through your mind while you’re coming up with an algorithm for it?

I’d love to hear your thoughts and comments. I’ll publish what my thoughts are too (some days later). Hint: the last entry is the most important.

You’re most welcome to submit well thought solutions in pseudocode too, eliminating the limited time factor.

Game programmer? Need math

I read this article on video game degrees and I felt like “Finally! Someone who got what I’m trying to do here!”. I didn’t know the United Kingdom is so supportive of video game productions. They need mathematicians and computer scientists huh?

What happened to great programming skills, you ask? “I can tell the difference between an ‘is a’ and ‘has a’ relationship in object oriented programming!”

Programming skills are still needed. We’re just talking about complementary skills to that coding ability. Math course too difficult? Well, it’s difficult for a reason.

Math is good

According to this article,

mathematics help students to develop a quantitative sense as well as a strong analytical and logical reasoning skills, which will assist them in acquiring new knowledge, be innovative and take calculated risks with confidence.

Strong analytical and logical reasoning skills? Isn’t that what you need when programming?

Then in another article,

most of the major problems we face are math and science problems

Time to brush up on some algebra!

Tackling problems with indefinite answers

In your programming career, have you ever faced a problem where there’s more than one answer? Or that you don’t even know if there is a right answer? Let me tell you one of mine.

Recently, one of my colleagues came to me for advice on a validation condition he faced. It turned out to be mathematical in nature. Wow! Anyway, in his database he stored number_of_days, month and prorated_fee. The stored data contains records of the results of prorating calculations.

What is prorating? Say a customer was charged a monthly fee for a service, and the customer had just signed up for the monthly service near the end of June. And he had the service for only 5 days, June has 30 days, and the monthly fee is $20. Prorating means he’d be charged 5 / 30 * $20 = $3.33. You will have noticed that prorating often has inexact results because the calculated values are rounded to 2 decimal places.

Back to my colleague’s problem. Suppose that the records from the database looked like this

days  month  prorated
5     30     3.33
8     30     5.33

I’ve translated the month into the number of days in that month for easier visualisation.

The problem was, given those records, could he prove that the prorated fee was calculated from the same monthly fee?

For this example, the monthly fee used was $20, so the results match. What he needed from me was, was there a way to programatically determine if there was a correct original monthly fee?

My initial instinctive answer was no, because all precision of the calculation was lost when the final value was rounded. Then he said, he’s not looking for precision. He just wanted to know if there was a possibility that the records were calculated using the same monthly fee. Aaahhhh, now we’re talking. So I told him to give me some time to work out an equation or something. Then I went to work.

Mathematically formulating the problem, I got:
Given d, m and p, where
d is the number of days in service
m is the number of days of that month in service
p is the calculated prorated value
find f, the original monthly fee, such that abs(fd/m - p) < 0.005

The equation is fd/m = p', where p' is the exact calculated value. Then p' is rounded to p.

The abs(fd/m - p) < 0.005 condition was because of the rounding. The difference between the calculated value and the stored value must be less than 0.005 for the rounding to be correct. And the error margin is 0.005 because 2 decimal place rounding means the value to be rounded is less than half of 0.01.

Rewriting the condition, we get
-0.005 < fd/m - p < 0.005
=> -0.005m < fd - pm < 0.005m
=> -0.005m + pm < fd < 0.005m + pm
=> (p - 0.005)m/d < f < (p + 0.005)m/d [d is always positive, so division is fine]

Substituting our first record's values into the inequality, we get
(3.33 - 0.005)(30)/(5) < f < (3.33 + 0.005)(30)/(5) => 19.95 < f < 20.01 Substituting our second record's values into the inequality, we get (5.33 - 0.005)(30)/(8) < f < (5.33 + 0.005)(30)/(8) => 19.96875 < f < 20.00625 Now if there were more records, we'd continue substituting, and we'll end up with a lot of inequality ranges. So what's the solution to the problem? If there was an f such that it satisfies all the inequalities, then the prorated values could have been calculated from the same monthly fee.

I went back to my colleague and told him that even if the inequalities could be solved, he might still have a range of values that could have contained the original monthly fee. So there's still no exact answer. He said that was fine. He rephrased his objective, that he wanted to know if the records from the database could have been calculated from the same monthly fee within reasonable limits.

Ohhh. Ok. I told him if he could find a range of values of f satisfying the set of inequalities from above, and that there was a value in that range with at most 2 decimal places, then he's done.

Example, a range of (19.956, 20.00) would be valid, because the original value could be 19.96 or 19.97.
But a range of (19.956, 19.959) would be invalid, because there were no values that had at most 2 decimal places. The reason for this was that the original monthly fee was at most 2 decimal places.

So, how do we find that range of values of f satisfying all the inequalities?

Let L be the set of the left hand values of the inequalities. Let R be the set of right hand values of the inequalities. Then there's a range of values of f satisfying the inequalities if and only if the maximum value in L is less than or equal to the minimum value in R. Stated mathematically,

max(L) <= min(R)

I gave him the piece of paper where I scribbled my notes, and he copied some of the important formulae. What about determining the solution programmatically? What, are you kidding? I wracked my brain for a mathematical solution, distilled it into understandable logic steps. How to implement this was up to him... *grin*