Why My Teenage Code Was Terrible: Sorting Algorithms and Big O Notation 

Dashlane continue to be a great sponsor for these videos! You can get their password manager for free on your first device at www.dashlane.com/tomscott - there's a 30-day free trial, and 10% off with my code, "tomscott"!

I usually clean my room with bogosort

Bogosort has the possibility of sorting any list in 1 try, it's the most op.

>Select all >FInd and replace every name with "content" >Presto, everything has a technically correct description

I personally prefer quantum bogosort: 1. Randomize the list. 2. If it isn't sorted, destroy the universe. In some parallel universe, it got sorted instantly, and that's good enough.

My favourite sorting algorithm still has to be miracle sort, check if the data is sorted, if yes, great, if not wait for a bit and check again...repeat. Eventually due to bit flipping or divine intervention the data will eventually be sorted...

If it makes you feel any better my adult code is terrible

I prefer Schrodinger's sort: the list is sorted and unsorted, as long as if you never find out what the list actually is. No work required.

Got a funny feeling my brain is running on bogosort

I actually implemented bogosort when I was bored at work once. I think it took about 6 hours to sort a list of 12 numbers.

I recently read a paper on a new linear sorting algorithm: It's called Stalin-Sort It achieves this by simply eliminating any element that isn't in order.

As someone who develops for a living, I can tell everyone to never ever ever try to get clever for the sake of being clever. The end result is you solving a puzzle and a load of frustrated users that probably didn't get their needs met, and team members that now need to jump through hoops to support your code.

I can not imagine Tom as a teenager, my brain refuses to.

"They're not disenchanted enough with the world yet". The most British sentence ever uttered, folks

I'm always amazed how Tom can do these videos in one take without a single stutter, pause or 'umm'

3:03 "each block bubbling up" I have a compsci degree but I've just now learned why it's called bubble sort...

"Randomise the list. Is it sorted?" "How do I know if it's sorted?" "Check it against some kind of sorting algorithm I guess"

"It's the job you give the 'work experience' kid, because they're not disenchanted enough with the world yet." - might be the most cuttingly true thing Tom has ever said. Much love, Tom.

Loved the end. Hit the issue on the head. Solutions are only good if they solve the problem that was asked.

2:02 “Complete with BLEEP BLOOP sound effects when I was, tiny.” This made me really laugh. I will now forever refer to my childhood as, “The Time I was Tiny”

1. Calculate a combination of elements & speed required for bogosort to complete a sort in approximately 1 year 2. Plug it in to a visualization with sound and begin sorting 3. Start a 24/7 Livestream 4. ??? 5. Profit

Also, another limitation to remember with Big O notation is that some algorithms scale really well but are slower on smaller samples; and if you know you only need to deal with under a hundred items every operation but you need to do this a lot of times per minute, it might be better to use the algorithm that scales poorly but runs fast at that sample size. Pick the algorithm that's right for your use case.

I love how Tom asserts dominance by doing these videos in one take

Quicksort may be O(n^2) in the worst case, but that's only if your list is already sorted (the wrong way). Interestingly, in the real world, this is highly likely to be the case, because any time a system touches data, it usually imposes order on it. So this is why it's recommended that every quicksort implementation be started with a shuffle. Once shuffled, it's nearly certain to be O(n*log n). And the shuffle itself is O(n) so it adds effectively nothing to the total time as long as your list isn't stupidly short.

If you're interviewing candidates for an IT-related job, the most important question should be what the worst mistake they've ever made was. If they don't have an answer then they're either lying or inexperienced. If they give an answer then the level of detail they provide about why it went wrong and how it could have been prevented will give you more than enough of an idea of how capable they are.

Something else to keep in mind is that just because something has a worse complexity class, it doesn't mean it is slower. Or that an algorithm can have multiple complexity classes. Bubble sort is O(n) in the best case, while Quick sort maintains O(n log n). If your list is already sorted, Bubble sort will figure that out faster than Quick sort. A single Quick sort operation is also much worse than a single Bubble sort operation. While Quick sort scales better, Bubble sort is actually better for small lists. The default sort function in Java, for example, uses Bubble sort for lists of length 5 or lower, and Quick sort for anything else. Sometimes O(n²) is better than O(n log n) - the complexity does not tell the whole story, it just focuses on how it scales.

...or, as Donald Knuth had it: "the fastest code is the code that never runs".

A good developer can write an efficient sort algorithm. A great developer just uses the system libraries.

I came up with a sorting algorithm that runs in O(1) Stalin Sort: 1. Declare input list to be sorted 2. If user complain, send to gulag.

Disappointed not to hear the “ONE TAKE” screech at the end of the video

Back in my undergrad days I earned a bit of extra money working in the library. Part of this was sorting books prior to putting them back on the shelves. My colleagues used insertion sort. I used quicksort. Happy new year!

Bogosort is like a move in a video game that has a 10% chance to insta-kill an enemy.

The best sorting algorithm is obviously the intelligent design sort, which is instantaneous as given any random list of items, the algorithm simply returns the unaltered list, saying it must have been put in that order by a higher power which is the correct order and no further examination is required.

Is school: Why do I need to learn this? Now: It’s kind of nice knowing if this will take 10 minutes or if the expected time is a week and I should fix the code instead of waiting.

One Cut. Very impressed!

The best solution isn't always the fastest or the smartest, its the one that works for everyone. BRILLIANT!

It's sometimes counterintuitive, but the most common lesson I have to teach graduates coming into our industry (Game development) is that a lot of those nice fancy algorithms that are Nlog(N) and similar - frequently are not the best choice in practice. Most of the datasets we deal with are actually tiny, you'd be amazed how frequently replacing a smart hashmap with a dumb array and some brute force, or a quicksort with an insertion sort (or ideally, Introsort) saves a huge amount of time when you're not dealing with thousands of elements.

"So that the teachers have a break". You sunshine are having a laugh. What it actually meant was driving round the locale, checking up on the kids, finding things to do with the ones that got fired, filling in forms with feedback from the companies, fending off calls from parents complaining that their kids hadn't got suitable placements, chasing up the ones that just went missing. And all that was at the "good" grammar school I was teaching at.

I took a highschool class on web design once, even got certified for it. all I remember now is how to make your background pink and your text comic sans.

The biggest issue i struggle with as a CS student is trying to come up with the " perfect " solution , while there is none , and ignoring the simplest solutions , just because they seemed " Too easy "

When I was 14 my website stored passwords, comments, and profile names/descriptions in plain text

I’m 39 years old, and I miss “utterly unjustified confidence”.... sigh, the nostalgia.

If Tom Scott explained every programming concept I ever struggled with, I would no longer struggle.

Bogos sorted?

When I was in high school, I was in charge of sorting in the largest online image gallery of a certain TV show that was popular, but also slightly obscure enough that we could be completely sure that it was in fact the largest gallery of images from that show and it's corresponding merchandise on the web at that time. I created folders and subfolders, probably the most folders. Nothing had tags, so I had to organize one by one into the correct folders and subfolders that I came up with. I could move files that were sent to my section and create and delete folders, but not change any base code. I was excellent at it and thought about what would be most efficient as I sorted away. This wasn't for school or anything and I'm sure it'll help everyone sleep at night knowing that the gallery was last forever after it was put on temporary hold while a Google search consultant was organizing a site revamp and the owner died. I spent two or three years of my free time on that.

As a software developer in the 70's I realised most of our problems - AND we had a lot, were caused because we didn't really understand what the client really wanted. After several meetings and arguments we decided to implement a "system" to overcome this. To most peoples astonishment our software quality made an order of magnitude improvement over night. We produced software much quicker, we developed it much cheaper, and more importantly it worked and did what it was supposed to do. Not rocket science but even today an important factor all too often missed by "smart" programmers, (and development managers). Nice video Thanks.

Tom, this might actually be the last recorded time that an end user actually knew exactly what they wanted!

Years of programming later my technique is now 1. Does my this need a tree? 2. Does my thing need a linked list? 3. If no to the above, screw it, quicksort. I aint got time for the rest.

As a software engineer who takes the quality of my work seriously, I have mixed feelings about this. There are times when I _have_ had to tell clients to suck it up and learn how to use the thing I made for them, because it was in fact better than what they asked for _and would scale with the growth of their workload over time._ Clients know what they want, and on rare occasion they might have even figured out what they need _today_ -- but they almost never know what they _will_ need _next year._ Just because the world operates on a "no time to do it correctly, we'll just re-do it later" mentality, that doesn't mean professionals have to enable that mentality.

"Should have done what the client asked for" if they were actually clients sure. Giving that job to a work experience kid would be illegal in Australia because its a waste of work experience

You have my gratitude for removing the inevitable CRT whine that was in this video

I LOVED the conclusion ! "Do what's NEEDED, not what you think would make you look cook". Sure, your software might be cool or whatever, but the client just wants something that works for them !

When I saw Tom Scott and “My Teenage Code” and “Bogosort” I cried

Bubble Sort, although yes has an exponential algorithm of O(n^2), you can make more efficient. Since you're "bubbling the largest to the top", or shifting your elements right, for each subsequent pass you don't need to compare the last elements because they're already sorted. For each "and then you do it again", you subtract 1 from the length of the list. Because you keep subtracting one, each subsequent "sort" is performed on a smaller list, therefore making it quicker to progress through. In short: There's no point in trying to sort things you've already sorted. You know the largest elements are at the end of the list, so don't bother with them on subsequent passes.

Knowing about Big O and runtime and so on is one of those things in computer science that you hardly ever really "use" (and it gets fuzzy really quickly if you think about it too hard) but knowing about it makes you think about problems in a totaly different way. And I like the remark at the end about premature optimization :-)

Why is it that “big O” is all anyone talks about on the internet but in college we didn’t even use it at all. We used theta notation exclusively.

Adult me: Eh, at least you could do a bit of coding at that age! Teenage me: **chortles every single time Tom said "Big O"**

That is one of the most important engineering and development lessons: do what is the best solution for the user.

2:50 Bit to add here, in computer science class we learnt that once you have sorted the biggest and put it at the end (here anyway), you don't need to check it again, so you stop checking pre-sorted sectors. I've never actually PROGRAMMED a bubble sort, but that's the theory we learned anyway, this would be much more efficient. Still though, correct me if i'm actually wrong.

"Do what the client asks for" I have to tell people that so often! So many developers think they can do the job faster by themselves and turn in something that doesn't do what the client wanted

You are incredibly humble. You use your mistakes as a learning experience and have no problem using your own mistakes to teach others. Thanks, Tom.

I recently watched Ted-ed's video about the fastest way to sort books, and I was happy to realize that the same concept is used in programming! Thanks for the great video!

Two days work experience? Our teachers got rid of us for a whole fortnight.

Stalin Sort is always the best due to it being Big O of N. Always. It goes something like this: for n in arr, if arr[n] < arr[n-1] pass, else drop arr[n]. By the end, the list will be ordered, give or take a few dropped items that didn’t make the cut.

The fact you do these videos in one single shot is why I enjoy watching them so much. That, and the content is always interesting!

I remember trying to code a 'quick' sort at computer science a level which took ages and didnt really work for a lot of different tests.

I misheard that as, "When I was 50" and was going, "Ok wait HOW old are you for that to be a teen!? Are you secretly the Norse God of Red Shirts!?"

0:10 American schools have “work experience” as well for their older students. It’s called different things in different districts. Mine was called CWE (Continuing Work Experience). It wasn’t an internship because it was just an entry level job in retail or food service for most of the kids. But it got them out of the school and making some money for a few hours work a day.

"The big problem wasn't that I used a bad algorithm, the big problem was that I was ignoring what my users actually needed because I wanted to show off how clever I thought I was" I'm seriously printing this quote and putting it up on my desk as a jab to some assholes I have to deal with.

When I was about that age, I came up with a sort algorithm, it had a run time of n! x 2^n .

In my heart, the best sorting Algo still is Sleep Sort ;)

3:13 That animation with that hand movement was so satisfying!!!

I Love this series Tom! Please keep going. As a Junior Software engineer almost finished with university i Love going Back to the basics and love your personal touch.

I feel like I finally understand what my teacher's trying to say 2 years ago.

This is an incredibly well timed video for someone going into a second semester programming course starting yesterday

Tom you never cease to teach me something new, thank you.

Tom's facial expressions while storytelling are just one of a kind! ❤️

It's so weird i was thinking about sorting algorithms the whole day and you uploaded a video about it now

at 15 when you were doing all that I was still figuring out how to eat ice cream by not dropping it on my shirt. Absolute Genius

Wow. Hit me right in the premature optimization.

I tend to imagine that the person who created the Bogosort didn't know that it would become the joke sorting algorithm.

As a former teacher I can confirm the two weeks when an entire year group of snotty kids, I mean delightful teenagers, was out on work experience was like a mini-holiday. And yes, I know, as a teacher I had plenty of holiday anyway. The payback was every teacher had to go and visit at least two of the pupils at their placement to see how they were getting on, but that was also a rare trip out of school and meant we could get some shopping in or have a coffee.

as a gcse computer science student, i cried watching this

I like how the graph for bogosort goes nearly vertical almost instantly

Before this Derek told to use lastpas and now you're saying dashlane.. what to do?? Idea: I'll use dashlane to protect my lastpas master password 👍

There is an oversight/improvement in the animation of Quicksort. You don't need to look at any pivot element ever again after splitting the list because you know they are in the correct position

I love the message at the end of the video about what you should have done. It highlights the importance of the goal. If your goal is to show-off, you are going to think very differently than if your goal was meeting client needs. There is a time and place for most things, so make sure you're setting the right goal for the job and it's constraints. At work and in real life.

I'm a fan of heat-death sort. It's O(1). 1. Wait until the heat-death of the universe 2. There's no one left to care if the list is sorted now or not

Besides just being a generally amazing explanation of Big O, thank you so much for that ending. I can imagine many CS students stumbling upon this video, and they should really hear this

"the best solution isn't always the fastest or the smartest. it's the one that works for everyone, long-term." - Tom Scott

My favourite is the trust search algorithm: it returns you an unchanged array in hopes that it was sorted before you passed it

"Do what the client asked you to do" Tom is growing old.

"When I was tiny" Tom Scott 2020

You do such a outstanding job explaining these concepts compared to some popularized books. 👏🏻

Big O was one of those lessons I remember not thinking that much of in school, but I've found it to be incredibly helpful when I try to plan out whatever projects I'm working on. One mistake that I see programmers make very frequently though, is not taking into account the Big O performance of built in functions. A very common example is string operations in Javascript. They seem to just think that whatever strings they're working with will always be very small, and so they're code ends up scaling horribly. The sad thing is the modifications needed to make it scale better usually aren't that complicated.

7:54 thank you for saying this, not enough people do :D Also, the way you describe "work experience" is not very different from the concept of an internship in America. You are given busy work to do so they can get as much work for free out of you as possible and you learn virtually nothing.

This episode seemed to touch on User Experience towards the end... make a "the basics" video about UX!

This video came with perfect timing for me. We are just now starting to learn different sorting algorithms in my AP Computer Science A class.

The reason quantum mechanics is complicated (on computers) is that the best algorithm we have, in terms of accuracy, is O(n^n) where n is number of particles. That one isn't used, however. Instead, there are series of approximations, each one a litte more inaccurate and much faster. The common ones for chemistry are, in order from most accurate to least accurate: CCSD (n^7), CCS(n^6), CC2 *and* MP2 (n^5), and finally HF (n^4). This means that when my computer uses 4 minutes to calculate the lowest-energy state of water (3 atoms), calculating the same for alcohol (8 atoms) takes roughly 3 hours using the fastest QM method we have. Meanwhile, a typical medicinal drug (~50 atoms) would by the same calculation use 50 days. Hence why we use supercomputers.

You just made shorting algorithms interesting and I consider that to be a small miracle.