[jammygit]

As a busy self-learner, when is it time to learn data structures in depth? I feel like there are always 10 other technologies I need to know more urgently (eg, more bash, Linux, testing frameworks, deep learning, c++ libraries, linear algebra, common security mistakes/attacks, OpenGL, etc)

[espeed]

All time. Everyday. You'll know you understand it when you can build it, when you can see its structure.

"Smart data structures and dumb code works a lot better than the other way around." -- Guy Steele

"The Representation Principle: Once a problem is described using an appropriate representation, the problem is almost solved." -- P. H. Winston

"Show me your [code] and conceal your [data structures], and I shall continue to be mystified. Show me your [data structures], and I won't usually need your [code]; it'll be obvious." -- Fred Brooks

"I will, in fact, claim that the difference between a bad programmer and a good one is whether he considers his code or his data structures more important. Bad programmers worry about the code. Good programmers worry about data structures and their relationships." -- Linus Torvalds

"Rule 5: Data dominates. If you've chosen the right data structures and organized things well, the algorithms will almost always be self-evident. Data structures, not algorithms, are central to programming." -- Rob Pike

"What are tensors? The facts of the universe." -- Lillian Lieber

[zphds]

Same here. I had the same problem especially when dealing with problems that required me to go back and (un)learn stuff from scratch.

I found that the greatest barrier to deal with seemingly hard concepts was getting the right intuition on what "might" be happening under the hood. For me, this is absolutely crucial even if my intuition about a "thing" might not be completely correct.

As someone who didn't go to a classroom to be taught a lot of the basics, I dive in head-first to read papers, I mean a lot of papers, to help build this mental model EVEN though I don't understand everything in it, but it lets you pick up patterns for solving similar problems you might be having.

For things you don't understand, you can always fit in these gaps in your mental model as you dig deeper into the abstraction cake and repeat the whole "intuition" building loop.

I find that the same also applies to Math and other domains which build on a bunch of primitives that has to be really well known to be able to reason in higher abstractions.

I feel software is no different... dig deep enough, there will be a hashmap of some sort.

[knicholes]

It depends if you're learning for pleasure or for your career. ;) When I bust out my algorithms/data structures books, I find myself engrossed much like I imagine people get with a good novel. So many times I say to myself, "Whoah! That's so cool!" because of how elegantly/efficiently a problem can be solved that I had never even considered.

My favorite so far is Skiena's "The Algorithm Design Manual". I especially love the computational geometry section.

[gnufied]

How does one read a algorithm or data structure book like an novel? most books require busting out keyboard or pen and paper and solving one thing or another.

Just genuinely curious because I love reading technical books but I can never finish them from page to page.

[knicholes]

This book has "war stories," where he describes the problem he was faced with and his different approaches before finding his awesome solution, so that's probably what gives me that feeling. But additionally, maybe I'm a freak.

[verall]

I think this is just a personal thing. kinda like some people immediately spill their brain to some cache (whiteboard, pen/paper) and others stare at a wall for minutes then write when they have a fuller thought.

If it's a technical book I probably read and enjoy it and forget most of it. If I see a similar problem it might jog my memory and then I can refer back. Only works for good books. Its a different experience than directly interacting with the examples by coding/working out yourself.

[caconym_]

If you're learning these topics for purposes of career advancement, probably there is not much reason to go much beyond the fundamentals that you might study for a FAANG-type interview. If you know that much, you can probably recognize a situation where you're missing something and go look for it.

Even that is probably overkill for a lot of people, practically speaking.

On the other hand, if you're learning for fun, you should study the thing you're most excited about because you will do the most interesting things with knowledge and skills that you're passionate about.

Ultimately, in my personal experience, the latter course has also led to career advancement.

[dcolkitt]

> As a busy self-learner, when is it time to learn data structures in depth?

From a practical perspective... The best sign is if you're spending a lot of time trying to wrangle performance or scalability and most of the bottleneck is inside the standard data structures.

E.g. if you're spending too much time or memory doing dictionary lookups, then it may be worth looking into Bloom filters. But if dictionary lookups are a trivial component of your overall performance, then there are no real gains to be had from replacing with a fancier data structure.

There's a follow-up point I'd like to make too. Skill with data structures is more about knowing the right questions to ask, rather than memorizing an encyclopedic list of a whole bunch of exotic structures. When you understand the major tradeoffs and design considerations underlying data structure design, then it's usually easy to find a solution in the literature even if you were completely unaware of it prior.

Effectively that means something like being able to translate between business requirements and academic language. "Oh, I need something that works on directed graphs, handles cycles, logarithmic in space, linear in average case time, resistant to adversaries, respects cache locality during lookup, and behaves deterministically."

The best way to get learn this stuff is to read a bunch of famous papers in data structure. Again, the point isn't to memorize a zoo. It's to get comfortable with the already well-developed framework that previous researchers have used. It's very unlikely that your business problem is truly unique at a theoretical level, so most likely someone before you has already figured out the tradeoffs and considerations that you are facing.

[chrisweekly]

> "Effectively that means something like being able to translate between business requirements and academic language. "Oh, I need something that works on directed graphs, handles cycles, logarithmic in space, linear in average case time, resistant to adversaries, respects cache locality during lookup, and behaves deterministically."

Excellent comment! I'd love to see a catalog of data structures that documented those characteristics; do you know of a good resource like that?

[dcolkitt]

Unfortunately, I don't have any great suggestions for a single book or resource. (Doesn't mean something good isn't out there, just not anything I've come across.)

For me personally, the best resource has been simply to work through famous papers in the space. Blogs like the Morning Paper, centered around breaking down papers, are also helpful.

[elevenoh]

Would also like to know if there's a resource like this out there.

[snarfy]

You can have twenty years experience, but if all you learn is testing frameworks, opengl, c++ libraries, etc you never really master your profession. CS fundamentals like algorithms and data structures are much more important to learn in depth. The knowledge stays with you and is always relevant, the latest testing framework, not so much.

[joebergeron]

Given the bulk of the things you've listed here, which seem to serve very practical purposes, maybe never. I wouldn't call data structures a "technology" so much as I would label them as a theoretical underpinning to much of computer science. If your goal is to learn practical technologies (which is a good goal!) you don't really _need_ to learn data structures deeply -- they're just not a skill you need in day-to-day software life (for most people). I'm a software engineer, and the idea of implementing so much as a binary heap (which is nearly one of the first data structures you learn) at work is a preposterous idea. That said, if you want to learn about data structures, go for it! They're incredibly interesting and will give you a way deeper appreciation for computer science and the work that goes into developing complex algorithms. There's just not necessarily a "pedagogical" place to stick in learning data structures with the sorts of technologies you've listed.

[ascar]

>"they're just not a skill you need in day-to-day software life (for most people)."

I mostly disagree. You won't need to implement them, but a working knowledge of the basics is the basis of any performance analysis and can make a huge difference in system performance in day to day programming. At some point all the small inefficiencies also add up.

I would go so far as to say that data structures and basic algorithms (e.g. Dijkstra) are more important to become a competent engineer than any of the practical tools he listed. It's like the difference between a mechanic and a mechanical engineer.

There is also a reason they are taught very early in a computer science degree.

[TrackerFF]

I think we should remember that the vast majority of computer developers and engineers will not, and probably never, work on stuff that requires deep understanding of things like data structures and algorithms, even those with advanced degrees. Most are just, how to put this, IT janitors.

To use your mechanic and mechanical engineer analogy: They're not even mechanics, they simply change the oil and light bulbs from time to time, along with cleaning the interior.

Not saying that it's not useful - you'd get a much better understanding of things...but this obsession with learning everything under the sun just to be "ready" for the job is overkill

[ascar]

I guess it depends what you want to work on and what your career goals are. I think my mechanic vs mechanical engineer example is actually a pretty good analogy of actual conditions. You have (many) people learning how to use tools and fix known problems (like mechanics, or e.g. CRUD developers) versus the others that actually need to solve problems with unknown solutions or optimize existing systems (the engineers). Maybe also with known solutions, but with a far deeper theoretical knowledge as a base for doing that.

Yes, the majority of work is on solved problems and we need far more mechanics than engineers. But it's not such a clear cut and a lot of actual software engineers work on solved problems most of their time and only spend a small amount of time on the engineering part (which is probably similar to other engineering disciplines). But when you spend the small amount of time on engineering work these basics are important and I realized multiple times that I could make better decisions because of it. And I'm kind of certain that's the case for most engineers with an academic background or self-taught basics though they might not even realize it.

[jlevers]

From this perspective, which people/roles would you consider to be the software engineering equivalent of a mechanic? How about the equivalent a mechanical engineer?

[Swizec]

> the idea of implementing so much as a binary heap

In my experience knowing how to implement such things isn't very useful in daily practical programming. Knowing that they exist, what problems they solve, the pros and cons that is incredibly useful day to day. Especially if you can recognize when you've accidentally strayed into implementing one of them under the guise of solving a business problem. You can then apply known patterns or have the foresight to say "Yo this is a solved problem, let's find a library"

[Koshkin]

Not to disagree, but even if one visits restaurants for their "day to day" meals, one might begin wondering, Hmm.. How would I cook that?

[mrbill29]

The only time this information is really helpful is if you care about optimization. If most of the data you deal with is within 1000 items, or there's no time/hardware constraint, then this isn't super helpful. But if you're trying to take a function that costs you tons of CPU time down, there may be some value in knowing how it all works under the hood, and how to improve it.

[mettamage]

If I were you I'd send an email to prof. Erik Demaine about this. If you write your email thoughtfully enough, then I think he's the kind of person to answer. I would in his position send at least a quick reply.

I have the same question, except I already have a partial answer (i.e. less interested in going after it, I sometimes do these things if I'm interested enough). I'm currently working my way up 6.006, but I wonder why I'd have to do advanced data structures. My guess is: I think I don't, because it's more of a research area than anything else.

I would need to take a quick look to see if there are some interesting data structures solving some interesting problems.

Personally, I'd be interested in time travel and memory hierarchy (caching) as topics. I think they might have some immediate practical utility in those areas.

[jolmg]

As someone in the same situation as you, I can advice that it's better to hold off on learning such things until you have a good moment to make use of them.

I studied a lot of Intel's 64 and IA-32 Architectures Software Developer Manuals, particularly the first and third volumes, and have forgotten a lot of details of what I read. I should have waited until I was ready to invest time in writing an OS or a first-stage bootloader or something. The broader things were useful, but I learned a lot of details that I haven't used and have forgotten. I'm sure it's still useful in the sense that I can quickly pick it back up if I ever needed to, but still. Same for all those things I read from the OSDev Wiki.

[didibus]

What do we mean by in-depth? Basic knowledge is pretty fundamental. Beyond that, I'd say assume you know them, and when you're stuck because you actually don't then go learn only as much about them to get you unstuck.