Good Language Choice For School Programming Test?

An anonymous reader writes "The Australian Informatics Olympiad programming test is being run in a couple of months. I'm an experienced programmer and I'm thinking of volunteering to tutor interested kids at my children's school to get them ready. There will be children of all levels in the group, from those that can't write 'hello world' in any language, to somewhat experienced programmers. For those starting from scratch, I'm wondering what language to teach them to code in. Accepted languages are C, C++, Pascal, Java, PHP, Python and Visual Basic. I'm leaning towards Python, because it is a powerful language with a simple syntax. However, the test has a run-time CPU seconds limit, so using an interpreted language like Python could put the students at a disadvantage compared to using C. Is it better to teach them something in 2 months that they're likely to be able to code in but possibly run foul of the CPU time limit, or struggle to teach them to code in a more complicated syntax like C/C++ which would however give them the best chance of having a fast solution?"

Comments

Many good choices

[DoofusOfDeath]

You don't want a test that favors kids who have studied that particular language in the past. I suggest the Turing-complete language LaTeX. It's the only way to be sure.

I'm guessing the CPU limits are generous.

[Jason Pollock]

I'm guessing that the CPU limits are generous and are more about filtering out bad algorithms than bad languages.

For example, someone using stooge sort instead of quicksort...

While the language used would increase the budget, the algorithms used will very quickly swamp any language gains.

When I did programming contests, they were more bound on thought (how quickly you can come up with an algorithm) and then implementation time. Rarely did compute time come into it.

Re:I'm guessing the CPU limits are generous.

[bill_mcgonigle]

Right, the OP says that Python is an accepted language, so it's only sensible to assume that this isn't a sneaky trap set by the Ruby Mafia intending to disqualify testers for using it.

The Navy research showed Python is a good language for learning OOP and if programming contests today are anything like what they were when I was in ACSL, boy would having a scripting language be handy. I wonder how anybody could compete in c (but perhaps that's a built-in test - picking the right language for the job is an important skill).

Psyco

[headkase]

For Python, can they use: Psyco as a library? That would help being practically a Just In Time compiler. It's x86 architectures only but that should be what they're running. As a side point I find it irritating that a language that is designed to be friendly and powerful is disadvantaged by counting CPU cycles: especially since in the real world those are plentiful compared to the scarce resources available for the hard work of debugging. And in Python if the CPU is your constraint - which it isn't in most programs - then you write that little bit of CPU code in C or C++ and call that one part from Python. This keeps the rest of the program easy to debug and portable.

Language shouldn't push you past the limit

[EEBaum]

Having competed in a handful of collegiate programming contests about 10 years ago, the CPU time limit was never even a passing concern. Granted, we were coding in C++, but even in Python, any solution that hits the CPU limit on these contests is quite likely an unnecessarily complex algorithm. I always considered the CPU limit to be a safeguard against programs with infinite loops or REALLY slow solutions, so that the teams wouldn't claim "no, really, it works!" and drag the contest on for hours insisting that the right answer will reveal itself shortly. If your solution works, but has complexity of O(n!), I'd have a hard time calling it acceptable.

If one of our entries was rejected due to exceeding the CPU limit, it was always due to a problem in our logic that the sample data hadn't triggered, but that the actual test data had.

Re:There's C then there's C written by newbies

[betterunixthanunix]

That is not universally true, at least not in the college level contests I participated in as an undergrad. They used to test with very large inputs, to deliberately throw off solutions that leaked memory (they set a ulimit before running the code) or inefficient solutions (except in cases where they gave us NP complete problems)...

Go With Your Gut

[Bob(TM)]

Unless these kids already have a programming proficiency, go with your gut.

The exercise is as much about allowing them to test the programming waters as it is about them winning. If you are starting with a blank slate, that means you need to create an environment that is intriguing. If YOU think Python is the thing, you'll be passionate ... and that is a lot of what makes a good coach/tutor in an olympiad.

Further, I think you could do a lot worse than Python. It is really a great language and is popping up in a lot of disciplines as the scripting tool of choice. It will perform well and has great characteristics that make it well suited for someone learning the ropes. Plus, the language is modern enough to be relevant should they desire to pursue IT further.

I would stay away from C/C++. In the hands of novices in a timed activity, I would wager it would be more trouble than it's worth.

Re:Faster than you think

[WrongSizeGlass]

I think you should pick something that would be the easiest for them to learn. Python is probably the best choice out of those languages (I'm not a Python programer and my exposure to Python has been limited - though I do code in C/C++, PHP & VB and I've coded and taught Java & PASCAL).

You can't really worry about the runtime limit since it should be rather liberal for a student's competition and you'll never know what the system's config will be (so Python may be fast or slow).

If you're going to be teaching new programmers, get them started on something they can use and something they can expand upon. Of the languages available to you I would say Python is the best choice.

If Python is an allowed choice then...

[Frequency Domain]

Any CPU limits should be generous enough to accommodate correct solutions in any of the permissible languages.

Re:Too easy... can't resist...

[WrongSizeGlass]

It's the whole "4 out of 5 doctors ..." scenario. Any kid who picks Java will be taken away to be deprogrammed and then reintroduced to programming society once they're ready to be released into the wild.

Born free, free as the wind blows ...

Python will give you th least problems

[gweihir]

C, C++ and Java are not an option. They all need years to master and have numerous non-obvious pitfalls. Pascal is pretty limitad and definitly a historic design. PHP is obscure in palaces. Visual Basic still is a bad joke, confined to just one platform and wioth numerous design problems.

On the other hand Python is fine, with the only thing to master the indention. Not on the list, but Ruby would also be nice. And for a minimalistic, yet powerful language, loot at (again not on the list) Lua.

Re:Python will give you th least problems

[dunkelfalke]

Pray tell me how a modern Pascal dialect (as used by Free Pascal Compiler for example) is "limitad".

Re:Strange limitation

[godrik]

The contest has a cpu time limit which seems to be reasonnable for a programming contest.

Re:Faster than you think

[gd2shoe]

You can't really worry about the runtime limit since it should be rather liberal for a student's competition...

As the summary says "Accepted languages" (presumably the competition rules), I would tend to agree. They're not going to ask your students to complete something that those languages cannot reasonably accomplish. Bad C++ code will be slower than halfway-decent Python code. Teach a good foundation, and let the chips fall where they may.

(And I too vote Python, by the way.)

Perl !!

[QX-Mat]

Perl is very simple to learn (check out "Robert's Perl Tutorial"). Perl allows the programmer to do what he wants - regardless of the style. There's even English.pm. Perl has many upsides ie: it's a powerful parser and indispensable toolkit... and highly embeddable. If there was ever a seed language to learn, it's Perl.

There's also the fact that Perl programmers are real unix men - and let's face it, you'd rather have real unix graduates than Ruby fannies?

Re:Perl !!

[LodCrappo]

I'm a big fan of Perl myself but I wouldn't recommend it for new programmers. The phrase "enough rope to hang yourself" doesn't begin to describe it. It's also probably a bad choice in this scenario because it isn't on the list of accepted languages, although if you were decently skilled I'm sure you could write one liner to translate your perl source into any one of the accepted list.

Hello World

[gd2shoe]

On the plus side, you could make a really spiffy "Hello World".

Python or Java

[M. Baranczak]

I assume you want something that'll give the kids an advantage in the competition, and be useful if they want to start programming seriously.

Pascal - it was nice back when Reagan was in the White House and Michael Jackson still had his nose, but it's obsolete.

PHP is for writing dynamic web pages. It's good at what it does, but it isn't really a general-purpose language.

C and C++ are faster than interpreted languages. That is, sometimes they're slightly faster. And they're almost always harder to write in.

VB? Go fuck yourself.

So that leaves Java and Python.

Re:Strange limitation

[Charan]

I'm sure the CPU time limit would be generous enough that it won't matter if your programming language is interpreted 10x slower than hand-tuned assembly. They want to make sure you aren't using a brute-force O(n^3) algorithm when a linear one would work well enough.

Plus, the judges need a rule to allow them to terminate programs that may be stuck in infinite loops. Otherwise, a contestant could delay the results of a competition indefinitely.

(Imagine: "This competition was rigged! The judges killed my program before it had a chance to finish. It was working fine, and I was the first one to submit answers to all the problems. What? So it has a long start up time. You don't have a rule against 100-hour programs.")

surely not; Pascal was meant for this

[r00t]

We can sort of classify the languages into "slow" and "not slow" in various ways.

• Learning: Pascal is fast. C++ is horribly slow. The rest are moderate.
• Compiling: While it can't beat "no compile", Pascal is shockingly fast for people used to something like C or Java. C++ is horribly slow.
• Start-up: Pascal again does fine. Here it's joined by C and C++. The interpreted stuff is typicallly an embarassment.
• Run-time: Pascal again does fine. Here it's joined by C. C++ can be fast if you treat it like C or if you're a God-like expert in compiler/library details that normal mortals don't understand. The interpreted stuff is typically an embarassment.

This should come as no surprise. Pascal was intended as a teaching language. None of the other languages was designed for teaching, unless you somehow equate Visual Basic with BASIC.

The second choice is probably C, but you'll have some issues teaching about declarations/definitions (inside out) and string-related stuff. The meaning of '=' is also not compatible with normal math notation.

The third choice is probably Java. You get most of the trouble of C, less speed, and greater need for boilerplate sourcecode bloat.

Re:surely not; Pascal was meant for this

[cibyr]

Oh come on, Python was designed as a teaching language and in my experience students find it much easier to learn than Pascal (and it's much less limiting once you get past the basics).

As far as speed is concerned, according to the Programming Language Game Pascal is at best 60x faster than Python, and these sorts of competitions usually give you a few orders of magnitude in margin - the idea is to make sure your solution is in the right complexity class, not to try and enforce the most efficient possible solution.

Re:surely not; Pascal was meant for this

[Glonoinha]

I took a few minutes to RTFA and the software equivalent of scholastic olympics today are pretty much the same as they were in 1984 when my team took second place at the state level. It is a medium low level software engineering test demonstrated via a few medium low level implementations.

They are looking for two things:
Can you read the requirements and come up with an appropriate algorithm?
Can you implement the algorithm in their favorite language (pretty much any language in common use)?

The CPU time limit is to keep people from using bogo-sort (ie, randomize the data and then check if it is sorted, repeat until it's actually sorted.) If the students understand how to break a problem down and implement it using a standard approach, they will beat the CPU time limit. Even if they use bubble sort.

The real thing they are up against is the design and implementation time limit - I'm guessing they have three or four problems and four hours per team. The libraries they are allowed to use are basically nonexistent, meaning they software they are expected to write is going to be a software implementation of some algorithm, and they aren't going to be able to Google for syntax assistance so they need to be fluent in their language. They will have access to hardcopy manuals and manuals in electronic format, but that will just slow them down.

Given the limit on imported packages and code complexity (not to mention source code size - no more than 40,000 characters (which is ultra-overkill - it would take a week to hand write 40k worth of source code for an application that actually needs 40k worth of source code)) compile time will be negligible. A few seconds tops. Don't worry about compile times for this competition, worry about the time to actually come up with the algorithm and then to implement it in the correct syntax.

They aren't doing any serious object oriented coding because everything is in one class, and because they're allowing Borland's Pascal (well technically Delphi but they are using is as a super powerful Turbo Pascal. This means implementing simple algorithms.

If you want your kids to succeed in this arena, here's the best toolset to train them. Two piece of paper, a pencil, and a few crayons. Verbally describe a simple problem for them. Explain it only once. As you are explaining it to them, have them take notes on one piece of paper and capture the requirements. On the second piece of paper, using the crayons diagram their solution. A sort can be diagrammed by drawing boxes and arrows showing how things move around. String manipulation or array manipulation can be diagrammed by drawing a word with numbers below it and arrows above it showing how the elements are moved around. Teach them to understand the problem, come up with a practical approach, and then code it in the basic syntax in any language.

Because I expect the problems to have solutions that would be simpler if the language has iterators, and since they are disallowing pretty much all of the good Java external packages to do those things - I'm going to suggest a language with simple pointers - either Pascal or C/C++. Rearranging arrays (of characters, structures, or whatever) is trivial in either of those languages, and rearranging or iterating over an array is the perfect example of algorithm design and implementation typical of this type of challenge.

Re:There's C then there's C written by newbies

[oldhack]

Actually, I take this back.

Go with C++. If not, in 20 years, some of these kids will come after you for ruining their lives by seducing them into the nasty painful world of programmers. It's a nasty way of life. Scare them shitless off it - teach them Perl and make them debug each other's code. Tell them all the grownups use Perl.

Re:Faster than you think

[Larryish]

LOLCODE would be easiest to learn for the cellphone-texting rugrats of today.

Example of Hello World program:

HAI
CAN HAS STDIO?
VISIBLE "HAI WORLD!"
KTHXBYE

Things you might want to consider

[sydneyfong]

I've participated in local and regional Informatics Olympiads, and went to the IOI once. I was involved in training local students a few years ago, so I know quite a bit of the ins and outs of these competitions.

All the languages have pros and cons, but PHP and VB obviously aren't suitable (if only for their encouragement of crappy coding practices). Java offers little advantage over C/C++, and it forces OOP onto you so it adds an unnecessary layer of complexity for the students.

The "industry standard" of competitive programming is C++, since it offers near-C speed with the power of various algorithms (eg. sort) and data structures ( maps, sets, priority queues ). Those who intend to take the competition seriously should be using C++ as their primary language. But then judging from your requirements the vast majority of your students won't fall into this category, so I wouldn't recommend C++, at least not at this stage.

Python generally is a good first language for its simplicity and power. There are a few problems with using python for competitive programming though. First, speed can become a problem for *some* contests, which have rather tight runtime constraints. The contests that cater for a wide range of languages are usually better in this respect, but a lot out there are primarily C/C++/Pascal/Java. Secondly, the fact that python supports a range of built in advanced data structures and algorithms means that you may lose the chance to teach them how to implement the basic stuff, eg. using a binary search tree to implement a map (typing `` mymap = {}; mymap[foo] = bar; '' is surely easier than implementing a BST yourself). You might ask why learn to reinvent the wheel when most modern languages provide these features, but these data structures and algorithms is the core of informatics olympiads, and one needs to learn from the basics. That being said, if the timeframe is just 2 months, I think teaching python might be most rewarding for the students.

C and Pascal are basically on the same league. C is a bit more "archaic" than Pascal in terms of the way it does things, but feature-wise they are roughly equivalent. The languages are simple, fast, and bare bones enough to force the user to implement the basic stuff. In the long term they are good languages for teaching data structures and algorithms, but require a bit of patience on the part of the student since you need to know quite a bit before you can do anything "flashy" with them.

My experience with most average students is that they usually struggle to form precise ideas on what to tell the program to do, and then when they do have rough ideas they then fail to write a syntactically correct program, and if they really do write a syntactically correct program the program usually fails on correctness for most inputs, or simply do the wrong thing. Running time shouldn't be a concern before the students actually get a correct program, and my experience is, unless you have really really bright students, most of them probably won't be able to come close to writing a correct program within competition constraints, so don't worry too much about running time.

In short, I recommend python, but in the long term you might want to think about using C/Pascal. And if you restrain yourself to a subset of C++, it might work too.

Re:Too easy... can't resist...

[binarylarry]

C++ is definitely the best language in the world.

My god

Frameworks are more important than language

[u.hertlein]

* Learning: Pascal is fast. C++ is horribly slow. The rest are moderate.
        * Compiling: While it can't beat "no compile", Pascal is shockingly fast for people used to something like C or Java. C++ is horribly slow.
        * Start-up: Pascal again does fine. Here it's joined by C and C++. The interpreted stuff is typicallly an embarassment.
        * Run-time: Pascal again does fine. Here it's joined by C. C++ can be fast if you treat it like C or if you're a God-like expert in compiler/library details that normal mortals don't understand. The interpreted stuff is typically an embarassment.

What I find interesting is that no one has mentioned that what's actually more important than the language itself is the available libraries/toolkit/frameworks that it comes with.

Let's face it, after you've learned a few languages it pretty much comes down to "what is printf called this week?" when you pick up a new language (functional languages aside). Getting familiar with frameworks is actually what takes the most time.

C++ and C are pretty bad in this department. STL is nice but only gets you so far. Threads? Nope (not yet). Sockets? No. XML/encryption/whatnot? Sorry. You have to write an awful lot of code to come up with this or find (and learn!) a support lib that does this. (I do C++ for a living.)

So I'd say: Python. Or C#/Mono (but that's not on the list, why?)

Re:Frameworks are more important than language

[cbhacking]

Python, Java, and VB (one must assume it's VB.NET, and thus has the same libraries as C#) are definitely what I would suggest on the basis of libraries. OK, you can do C++ with the CLR and CLR libraries, but C++ is *not* a good choice for a first language.

There's a lot of hate on /. for VB, perhaps because not many Slashdotters have tried any version of the language from the last 8 years, but it *is* a good first language - static and strong typing (by default now, I believe), either procedural or OOP (full support for inheritance, unlike legacy VB), BASIC-style syntax (intuitive and quick to pick up), fully memory managed (real GC, not just ref-counting), structured exception handling (no, it's not On Error Goto anymore), and all the .NET/Mono libraries available (before you ask, yes Mono has a VB compiler/runtime, and once you compile it all down to CIL the libraries are usable by any language).

Java has about half those advantages plus a lot of silly restrictions (although at least Scanner makes getting keyboard input a lot less retarded than it used to be), but it's still a popular choice for learning a first language, and a practical choice for those who want to make a career of programming (though they should learn multiple languages anyhow). Python is a better choice, and possibly quicker to learn, although I'm not a fan of dynamically typed languages with significant whitespace.

assume they don't get it right the first time

[r00t]

This is a student's programming contest, right? Why are you even concerned about compile time or "startup" time?

Normally there is a time limit during which the competition runs. The faster your tools are, the more time you have to write code.

A kid might need 50 tries to get his program right.

Java != slow

[Vapula]

Well, it's true that the JVM may take some time to start... but once it's running, JIT and other make it run quite fast... You see game servers hosting hundred of players on a "normal" machine (not 8-CPU 4-core or similar), web sites using JSP, ...

And if you don't start to use the latest API (reflexion and similar) and stick to "standard" java, it'll make you create clean code...

But I agree that starting the JVM is rather slow... Even slower if you start it in "server" mode (but then, you'll get a speed increase while running the code)

It's just me

[Exitar]

or teaching a language to non programmers children and sending them to Informatics Olympiad after two months is pretending too much?
In addition to the language, you must first teach them very basic concepts like variables, loops, recursion...
As soon as they start to digest them and write their first "Hello World"... BANG! You send them to Informatics Olympiad!

Re:Teach them C++

[dunkelfalke]

forget turbo pascal! that's an educational language, which is hardly used in the real world (esp. because of the lack of object orientation)

How does it feel living at least 22 years behind the rest of the world? I am asking because Turbo Pascal supports OOP for that long.