Avoiding Unnecessary Computations

Loading...

Do curso por University of Washington

Programming Languages, Part B

366 classificações

University of Washington

Programming Languages, Part B

366 classificações

[As described below, this is Part B of a 3-part course. Participants should complete Part A first -- Part B "dives right in" and refers often to material from Part A.] This course is an introduction to the basic concepts of programming languages, with a strong emphasis on functional programming. The course uses the languages ML, Racket, and Ruby as vehicles for teaching the concepts, but the real intent is to teach enough about how any language “fits together” to make you more effective programming in any language -- and in learning new ones. This course is neither particularly theoretical nor just about programming specifics -- it will give you a framework for understanding how to use language constructs effectively and how to design correct and elegant programs. By using different languages, you will learn to think more deeply than in terms of the particular syntax of one language. The emphasis on functional programming is essential for learning how to write robust, reusable, composable, and elegant programs. Indeed, many of the most important ideas in modern languages have their roots in functional programming. Get ready to learn a fresh and beautiful way to look at software and how to have fun building it. The course assumes some prior experience with programming, as described in more detail in the first module of Part A. Part B assumes successful completion of Part A. The course is divided into three Coursera courses: Part A, Part B, and Part C. As explained in more detail in the first module of Part A, the overall course is a substantial amount of challenging material, so the three-part format provides two intermediate milestones and opportunities for a pause before continuing. The three parts are designed to be completed in order and set up to motivate you to continue through to the end of Part C. Week 1 of Part A has a more detailed list of topics for all three parts of the course, but it is expected that most course participants will not (yet!) know what all these topics mean.

Na lição

Section 5 and Homework 4 (First Module with Racket)

Let's get started programming with Racket and then learning idioms related to delaying evaluation. The welcome message has a few additional comments about picking up a new language and how to approach the homework assignment, so let's get started...

Conheça os instrutores

Dan Grossman
Professor
Computer Science & Engineering

, So let's now continue our study of using thunks to delay computations we might not

need. See how it's, that's a good idea and how that is a less good idea. So, thunks

let you skip an expensive computation if you're not going to need it. So this is

great in this sort of first code skeleton you see here. If I have some function that

takes in a thunk, and I have some if that ends up taking the true-branch, the branch

that doesn't use the thunk, this is much better than passing in the result of some

expensive computation. If I didn't use a thunk here and I had to pass in the result

of the expensive computation, I would have done a lot of unnecessary work. So, that's

fine, and in those situations, using thunk is straightforward. But what if you're

more in the situation like here at the bottom where I have a bunch of separate

conditionals? I don't know if any of them are going to evealuate to true or evaluate

to false. I don't know how many. But they do all need, if they're false as in this

case, doesn't matter whether they're true or false, they do all need the same

result. So should I precompute the results of the expensive computations for all of

them? That would be wasteful if none of them need it, but if multiple of them need

it, this situation with the thunk is actually worse, because I'm going to

reevaluate the thunk every time one of these ifs ends up being false. So that's

the trade-off, we're going to end up getting the best of both worlds in a few

minutes. But first, I want to show you an actual example it's a silly example, but

it's actual code instead of something where I've left out all the interesting

parts. So just to make this interesting, this first function is ridiculous, it adds

its arguements, but I've put in enough extra code that it takes a long time to

evaluate. So if I do something like slow-add 3 4, you see it actually takes a

second before it produces 7. that way we can actually see the difference in, in the

example I'm about to show you. Okay? Then I have this functi on that does

multiplication, but in sort of a strange way. It takes in a number x, and it takes

in a thunk y that when you call it, returns a number. And here is how it then

multiplies x and the result of calling y. If x is 0, it returns zero without ever

executing the thunk, because that's how multiplication works. We don't care what

the thunk is. If x is 1, then, we evaluate the thunk and that's our answer.

Otherwise, we evaluate the thunk and we recursively call my-mult with x minus 1

and not the result of calling the thunk, but with the thunk itself, because that is

what the my-mult expects, it expects a thunk for its second argument. So if you

look at this, this is going to end up calling y-thunk once every time for x. So,

if, if x is seven, its going to call the thunk seven times, alright?

So, indeed, therefore, even though, something like slow-add of 3 and 4 is very

slow. Right? If I call my-mult with 0 and a thunk of slow-add 3 and 4, that's very

fast. want to see it again? Very fast. But if I were instead multiplying by 1, well,

this is sort of unavoidable. I need to add three and four, but that's okay. I mean,

what, what else am I going to do? But if I called it with 2, it's actually going to

take twice as long, because it's calling that thunk twice. and I don't even want to

sit here while I did something like multiply by 20. Okay? So thunking was

great in the zero case. It was fine in the one case, we did have to add three and

four. but it was terrible if for anything greater than one, because it was actually

a net loss compared to the version of multiplication that just took in an x and

a y, and therefore, the caller would have to precomputed the two numbers and

therefore, it would have already done slow-add. In fact let me show you that

version as well. I can actually do it with my-mult. So suppose I passed in zero, but

then for the second argument, I precomputed slow-add, put that in a let,

and then in my lambda, just look up x. Alright? So, all I'm doing is my second a

rgument to my-mult, which is evaluated right away, it evaluates x to slow-add of

3 and 4, and then, creates a thunk that when you call it, will look up x. So, this

is going to call slow-add 3 and 4 once before you ever call my-mult. So now, if

you call it with 0, it does just take a little bit because we did call slow-add.

But two doesn't take any longer, and in fact, 20 doesn't take any longer. They all

take about the same amount of time. Okay? So that's all fine, but I gave up zero

being fast. It's not fast anymore. Okay? So that's sort of our motivation. So, what

if we could get the best of both worlds? What if we could take some computation

that always return the same result, had no side effects, didn't matter when it was

executed, and we did not compute it until we needed it? But then, once we did need

it we remembered the answer, so in the future we, didn't have to compute it

again. We would just return that remembered result immediately. Well, this

has a name, it's called lazy evaluation, and languages where most constructs work

this way, were in fact all function work this way, are called lazy languages and

Haskell is the most well-known, successful example today. But what I'm going to do in

the next segment in racket, is show you how we could just code this up. Now,

racket does not work this way for function argument. Function arguments are all

evaluated at the call site. We do not have lazy evaluation in Racket just like we

didn't in ML. But we will be able to code it up, just ourselves. Racket does have

some built-in support that is very slightly different syntactically in what

it's doing. I'd rather show you the implementation, so we will use our own,

and then, we will come back and revisit this multiplication example. So, in our

implementation, all we're going to need are thunks and mutable pairs, and the

cons, two things we've seen from previous segments. So we're going to be able to put

together some previous ideas and get these best of both worlds.

Explore nosso catálogo

Registre-se gratuitamente e obtenha recomendações, atualizações e ofertas personalizadas.

O Coursera proporciona acesso universal à melhor educação do mundo fazendo parcerias com as melhores universidades e organizações para oferecer cursos on-line.

© 2018 Coursera Inc. Todos os direitos reservados.

Baixar na App Store

Baixar no Google Play