Which Computer Science Concept Do you Value the Most?

Question

For some reason, I notice that I end up using a lot of finite state machines at work. In particular, when I'm implementing a custom TCP/serial protocol, they are very helpful and produce a very robust output (in my opinion).

My days in CS classes are long behind me. As such my recollection of the stuff I learned there is fuzzy. I was curious if there are other concepts people are leveraging that I've forgotten about.

There is no "right" answer. Vote up the answers containing the concept you use this most. We'll simply end up with the most used concepts on top. For me, it'll be a list of stuff to study up on.

-Robert

Answer 1

The "Google" concept ;)

Answer 2

Complexity.

Answer 3

Time/Space complexity.

Answer 4

Model View Controller pattern is the one I use more so than any other.

Answer 5

Breaking down a problem into smaller sub-problems, I think, is something I quite often do, even if not really thinking about it :

• it helps getting to the solution
• and it also help getting cleaner code (smaller functions / methods, that do "unit stuff", for instance)

Still, maybe it's not really a "concept"... Event if I remember some algorithm lessons where we were taught "divide to conquer" ^^

If you want something more concrete, I'd go with :

• testing ; it's something we don't do when we are just out of school... And we learn the hard way that it's definitly something we have to do more, and better !
• Some Design Patterns, probably
• Thinking before coding -- maybe the most important thing in our jobs ^^

Answer 6

Object Oriented Programming

Answer 7

Graph theory

Answer 8

Keep it simple. If possible, make it simpler.

Answer 9

Concurrency and parallel computing. I didn't touch it for many years, but it's become more relevant with each passing year (and each core count doubling).

Answer 10

Object Oriented Programming and Data Structure

Answer 11

Copying and Modifying Existing Code.

Answer 12

Big O notation

Answer 13

Requirements analysis and relational databases.

Answer 14

It's not CS-specific, but just remember that producing results that are simple or concise are both good goals. If you can produce something simple and concise then you're likely producing high-quality work.

Answer 15

Understanding and utilizing the data structures and algorithms provided for me by language libraries (either from the standard or third parties, like Boost). Don't reinvent the wheel, and learn what wheels are out there that are better than your own.

Answer 16

Analysis of Algorithms

Introduction to Algorithms: Cormen, et al.

The Art of Computer Programming: Knuth (the entire series)

Answer 17

I believe we all do FSM's day in and out. OOP models a FSM , so does a MVC.

Infact, OOP/MVC etc are patterns for expressing a FSM.

Are there ANY applications where we DO NOT build a FSM?

Consider a simple application - sorting. Well, that's a FSM too!

This should be the question in fact : Are there ANY applications where we DO NOT build a FSM?

Answer 18

Problem solving...

Answer 19

Estimating space/time complexity and using appropriate data structures to get much simpler/faster code. Modeling certain problems as graphs also came useful once in a while.

Answer 20

Encapsulation or information hidding

Answer 21

Functional decomposition.

Answer 22

"Debugging code is twice as hard as writing it. Therefore, if you write code as cleverly as you can, you are by definition not clever enough to debug it."

Answer 23

Singleton, template and strategy patterns.

Also: YAGNI - You ain't gonna need it
KISS - Keep it simple, stupid

Answer 24

Don't repeat yourself.

Answer 25

Avoiding premature optimizations, as Mr Knuth said:

"We should forget about small efficiencies, say about 97% of the time: premature optimization is the root of all evil." (from wikipedia)

Answer 26

Data Binding, you gotta use it all the time... in numerous different ways....

Answer 27

Modelling

Formal verification for small but hard functions.

Lambda expression (C# :p)

Concurrency "theory"

Design patterns

Writing specs before writing code.

Answer 28

The fundamental Theorem of software engineering: any software engineering problem can be solved by adding a level of indirection.

(I think that it has been formulated in this form by A. Koenig)

But then, too many levels of indirection can become hard to follow :-)

Answer 29

"Any problem in computer science can be solved with another layer of indirection." — David Wheeler, chief programmer for the EDSAC project in the early 1950s

When well-applied, this leads to reasonable generalization as seen in examples such as abstract data types, reusable classes with virtual methods, etc.

When poorly-applied, it leads to overly-indirect implementations with lots of runtime overhead due to over-generalization, e.g. the Intel 432 architecture.

Answer 30

Coupling and cohesion.

It's essentially the divide-and-conquer paradigm the basis of all software.

You are looking for orthogonal concepts and orthogonal software entities, those that exhibit loose coupling and high cohesion.

Used a gosub in Basic? You're using C&C.

Answer 31

Relational Model for data management and normalization.

Answer 32

These are the university courses/concepts I found most useful for my professional career

• Introduction to Databases
• DBMS - how they're working
• Algorithms and Data structures
• Object Oriented Programming concepts
• Design patterns (mostly MVC, application layering)
• Requirements engineering
• Software Quality Management
• Software Metrics

guess they're all...I did not mention specific technologies here but just the concepts.

Answer 33

"First make it work, then make it work /fast/."

Of course, you must not lock yourself in to a slow design, but a lot of time can be wasted trying to optimize routines that would never become the bottleneck of the entire solution - also, if you later discover the need to redesign that optimized module, your optimization effort was arguably wasted.

Answer 34

What data structures exist, their space/time characteristics, and what situations each should be used for. There is no better way to write fast, maintainable code than to use the appropriate data structure for your use case.

Btw, even though you could make the case that data structures are design patterns, I intend for this to be considered as distinct from design patterns.

Answer 35

Abstraction

Aho and Ullman write in the introduction to Foundations of Computer Science

But fundamentally, computer science is a science of abstraction — creating the right model for thinking about a problem and devising the appropriate mechanizable techniques to solve it.

Answer 36

The longer a fault exists in software the more costly it is to detect and correct the less likely it is to be properly corrected

Answer 37

I'm surprised no one has mentioned encapsulation.

Answer 38

The most important phrase that pops into my head a lot was from my operating systems professor -- and I didn't do very well in operating systems. Nevertheless...

"There is no magic."

He meant that if a computer can do it, a computer programmer can figure out how it is done. When someone waves the magic "high technology" wand, look closer, and you'll see a heuristic.

When I have to accomplish something hard, it gives me the courage to find a way to do it.

When I am presented with some "magical"-seeming piece of technology in the media or in marketing material, it makes me skeptical and dig for the truth.

Answer 39

I frequently apply Computer Science concepts every day to write my own custom splay trees and ray-tracing algorithms for the Java web applications that I develop.

Answer 40

Vectorization. I use MATLAB at work for some of my programming and am regularly porting scalar operations to a vector implementation. Sometimes it's a lot of fun - sometimes it feels like picturing a hypercube flatlanding a Mobius strip while chatting with Escher.

Answer 41

Time and space estimation. How long will this thing take to run? Not just big-O notation, but also some idea of whether something will take seconds or hours. Also how much space will something take up? Is it 1 meg of data I can load into RAM, a gig of data I can work on in RAM with a bit of work, or 100 gigs of data that will take disk and/or a distributed system?

I interviewed a lot of fresh-out-of-undergrad engineer candidates at Google. I'd say about 50% of them couldn't tell me how much RAM was in a computer, or estimate how long it'd take to, say, parse a million web pages. Without some basic intuition of how big a problem is you're not equipped to solve it.

Answer 42

Garbage in, garbage out.

Answer 43

Appropriate levels of Abstraction