Rust Programming Language Rust Malaysia, 14 Jul 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: RustLang Meetup Malaysia - July 2021

Description

Do join us at discord https://discord.gg/9Xj8H2EXTD

Rust malaysia meetup july 2021 discussing zero-cost abstractions, iterators and live coding.

A

Okay, hello welcome to meetup july today.

B

I will be talking.

A

About uh I'll be talking about zero cost abstraction in rust and iterators and then at the end there will be like a live coding session, and this time there will be more interactions and.

B

uh Before before we start, I would like to introduce the speaker for today we have ivan, which is on my right side and then um co-organizer for the meetup. So if you are interested in joining us, please do join the discord discord channel. uh If otherwise, uh you have you, you are free, uh feel free to leave a comment on youtube streaming.

B

B

We will try our best to respond to any questions you have throughout the presentation. So with that.

A

Are you ready? Okay, welcome, welcome to rasmita july and today I'll be talking about zero cost abstraction in rust, and then it will be a lot of the most of the parts will be talking about iterators, and uh today there will be more. You know more discussion, more interactions, so y'all may need to ask questions in in the in the group, discord, group and and then at the end there will be a live coding session and just wondering have you ever heard of zero cost extraction?

A

Yeah the talk will be about zero cost action and on the youtube search right. I found a few talks related, but the content will be different. Anyone heard of it all right.

B

C

A

A

Okay, anyone, if you, if you want to say case, just stay in the discord, okay, it will be.

A

Oh, no one, okay, then I'll start zero consumption. Why is zero cost?

A

We abstraction see this more often the old days right now we see this last and this is you see this in the last book in the introduction it mentioned this zero cost. It mentioned zero, cost abstraction and also in the old website.

A

It also mentioned the zero cost abstraction here, but the new website, the new website, doesn't have these.

A

Yeah, the new website, you, don't you, don't see this, they change a lot of stuff. So always I have more more technical.

B

You please show the website.

A

This one you can, can you see.

B

A

uh I think I'm I'm streaming, I'm showing the website right now right.

B

No, you are not sharing a screen.

A

I'm not sharing a screen nope.

A

I think I'm showing you the screen right.

A

Look. I can see the screen. Look! Okay! You can see right just now. Do you see the slide.

B

A

Okay, can you see this.

A

Can you see the slide here right there, all right got it yeah and then this is the, and this is the can you can you see the uh website now on and this is the o1.

A

So this one is the book. It's called the raspberry, and here you mentioned zero cost abstraction here and the old one also measures zero cost expression. When first I got into this.

A

Okay, so uh at first I was confused. Why is zero cost expression then? Now I I get it so I'll explain or if anyone want to take a guess what what? What is zero cost expression means anyone want to guess.

A

Okay, maybe I'll just continue so expression is like a high level common high level concepts. Rust can have high level concepts, but it compiles to low level code.

A

As in you, don't you don't just interface with directly interface with to do indexing, changing the characters on on original, using for loop everywhere right so with expressions. You can write something like high level high level you can. You can have an iterator, which is high level. You don't need to write every time. You don't need to do like a for loop and there's like classes interfaces trades that are all considered abstractions and, for example, you see here is you can have right?

A

So you can you can do this? Okay, here's an.

A

Example, okay, so so now you see it starts right.

C

A

Okay- and so you see here, you can collect what this does is. It adds the rack for each item, this map, similar to what you see in javascript and for each item. It adds x by one and then it collects it back to a rack yeah. We will talk about this later. This is like.

A

Because we don't know this collect right, it's like there's some types there. So a lot of this abstraction eras is.

A

Built with types and at first, I thought zero course is like a thing that is like magically make make the code faster because they say zero cost. I didn't know that it is referring to the abstraction that means zero cost.

A

As in like, uh for example, you write, like you, write some iterators in other language. You have a course like if you want to do some classes or something in most languages with uh in most high-level languages, they have cost like using using. Sometimes you might hear using for loop is faster than using while loop, so using follow is faster than using the iterator api and or something like that.

A

So you need to fall back to the for loop, the lower level stuff that it can optimize easily and and then, when you use the iterator api, it might be a bit slower.

A

That's what I see in other comparisons like I.

A

A

No, I I don't think I have example right now, but I've seen what I see in the like other youtube channels, saying that, with the similar talks right, they do a comparison like, for example, in c or java.

A

It is likely that the for loop will be faster than the iterator and there will be like uh the iterator will be a bit. It will be slower than the follow, but in rush right the iterator will be faster than in most cases it will be faster than what your have right for those that doesn't know is iterator. Iterator is just a synthetic synthetic trigger for like wow, I iterator next and then do something with it.

A

It's something like that and then, when you see the chips right, so there are a few type of modulator areas. So essentially it's just it's just like synthetic sugar for for loop and then and then, if you see for loop, is just c dot into ether.

A

So it means like, for example, like this. If you have 4 x in c, so this means wow.

B

Can you enlarge it a little bit.

B

A

A

It will be to be something like this, that c equals to c dot into ether, and then, while that sum x.

A

It will be this. You see this will be similar to this. It's essentially just a synthetic sugar to use the iterator and c dot. Next is the iterator api. Iterator is just if you don't know what iterator iterator is like an expression to let you iterate on something like an array yeah, that's why it's an iterator and then expressions about zero cost like how. How does it do that? Because we thought like how come how come like, for example, you see in this comparison right.

A

This is the comparison using for loop and this comparison using your iterator, and you see that the iterator is faster than the for loop, like how so the thing is usually in iterator, it might be faster than what you can right in for loop. If it's a simple one and even a complicated one, most of the time it can be faster, it is faster than the for loop using like manually unloop it you see something like this iterator.

A

And but like how? So the thing is the iterator right have a lot of types behind it and abstractions usually have runtime calls in other languages, but in rust the extractions the iterator performs as fast as the hero look faster.

A

So the zero cost abstractions essentially means that the runtime cost of the iterator is actually being a compound time cost in which that the the component is slow right. So it means that it is compiled, is probably one the reason. One of the reasons why a couple times low because of too many attractions like this and but yeah there's, no, that's like almost no. They don't cause anything during runtime like the types will optimize out into like without without types and the enums doesn't cost anything even in runtime.

A

And how they do it is okay, I'll show a few types instead activator. So this is how how it works.

A

So the iterator module right- you can see here, there's an iterator module.

A

So the base iterator phase looks like this. Iterator is a trade a trade inside? If you don't, if you don't know us, a trade is something that defines the behavior of something it's a bit like uh abstract interface, it's sort of like abstract interface. You just say like what is the behavior or something you can? You can add methods for you can implement a trade for like a strut or an enum.

A

And you in iterator days there's a type item this this type is called license temple and then right, there's a titan, so iterator contains a type here, as in, like you, see like something like.

A

This something like this, so there will be item type and then next, if you call the function next right, it will mutable mutably, take itself and then return the next item in the iterator. So basically this like the main iterator. Usually if you you just you just do this, the call, usually you just implement this and it's fine- you don't need to implement other items in iterator and then there are three forms of iterator. One is eater. The mute and intuitor ether is just iterate over a reference of part.

A

T and interval is just iterates over multiple iterator, integral with the reference of type t and intuitor converts the t into iterator iterate away.

A

Okay here here it mentioned how to create an iterator yourself. If you look into this, so we have counter structure which contains count of yield, size, unsigned integer, and you implement the methods for the counter and fnu. You create a new counter.

A

Then you write, you implement the trade iterator for counter. So what this does is the counter we will have. We will will contain the iterator or what items do size, which is like uh the counting and then next. Next is the only required method right, so you can only just implement next and then you get the iterator.

A

So you can just self count your increment account by one and then, if, if it's less than six you'll return, the self count and then you'll get like it's, not your return, none. So the option is like uh office. Whether something is there or not. So this is to prevent, if you don't know why the option is just just like algebraic data type that lets you cover. None, for example, it could be, it could be something or it could be none.

A

So you can see a return one, two, three, four five. So this is like roughly how an iterator works and then there's four loops and iterator. How it works is roughly the sugar into this in mesh and it will keep looping and get the next value, and this is roughly how it works in iterator.

A

So the thing is in iterator runs you see you. Can you can see that there's? There are many types of sharp here, because there are all of them have like some of them have like specialized function.

A

For example, map is just created from map the map method and then some of these some of these structures right, have implement certain trade. Some of them doesn't implement certain trade, so it's like they can optimize this to a very, very to a very deep level. They can optimize it and there are a few of the trades that helps in improving the speed of the iterator and to identify the pattern of the iterator so that there are multiple traits in iterator.

A

For example, double double-ended iterator means that you can. You can iterate from the start or you can iterate from the back, so you can iterate from 2n like a rack, you can iterate from the from the start and from the back as well, but sometimes you can't iterate from both sides.

A

Then there are x, there's exact size, iterator exercise iterator is like you know the you know the size of the iterator, whether it's empty or or you get the length. So these are some of the methods that let like, for example, there are. There are some specializations that let's say if they have this kind of trade, they'll use some specialized function to speed it up and then there's also like fused iterator.

A

These are some mechanism like uh once the iterator is fused, it will stop producing value, and so these are some tricks, for example, and then speeding up the iterator is usually done by using the iterator size in right. I always go there iterator, okay, let's look at the iterator trade.

A

So you see that so it will give you a hint on the lower bound and upper bound of the element of the iterator like how how many items you there will be left so iterator, like all of the iterator right, then make use of this size hint to do the optimization and, for example, you can, because you can know that how many items is left then they can directly allocate just just the just enough capacity.

A

They can just allocate once just enough capacity for the whole whole thing, which is why these are. These are one trick they do and then there's also trusted land trusted is not here. I don't think it's here. Oh he's here so translate is just another trade, but this is a likely trick. So this lets the user know that the length is exciting but exact, except that this knows that users know that you have exactly like that size.

A

Exactly that's that size for the iterator. So you can. You can have that many items.

A

And then there's also in place iterable another trick, but this is not here. This.

A

A

Now this is this is to let you know that you can. You can directly modify it in place. You can you can you can take the it you can take the internal of the iterator like in the iterator there's like enter the buffer right.

A

You can, you can directly take the original buffer and you can directly modify that buffer to take the buffer and take it for another iterator and optimization step, but this is uh not not a good for everyone to use, so this is only they only implement this for the internal brass iterable structs, but most of the time you don't need to do this. Usually it's fast enough.

A

If you just use the center iterator yeah and something new since rust, right, you can use the cons generic air indexing, like you, can do one like one, two, three decent array. You see this. This is an array, but you don't need to do this either. Last time you need to do one dot, eater and chain two, but right now you don't need to do this. uh It's a lot more!

A

It's a lot more organic! You can just do like one chain two, this. What you should do is chain iterator of array containing one with two and then this will produce like okay. If you do next you'll get one and you do next again, you get two.

A

And other benefits of iterator includes expressiveness, your camera is very expensive. You can collect the iterator into multiple types like you can collect it into hash set or you can collect it into hash map. For example.

A

Okay, let's see.

A

A

C

A

Wait this generic yeah, so it needs two item.

A

Okay, so it doesn't know the size.

A

Yeah there will be some errors. Sometimes I just fix it. What this does, is you just okay? Now you see now you can get a map from this iterator.

A

See now you can get a map map is like a dictionary in python or in rust. They call it hash map, there's also b3 map, so the key is two value is three key: is one value is two you see this collect can can be collected into this, and also, if you, if you use the same, collect right.

A

Let's say you want to call it back to a web.

A

uh Intuitor into eater is just covered forward that forward the uh original thing into an iterator which iterates over the value, rather than rather than the reference it iterates directly over the original value and collect, is just collect into a different struct, but this doesn't work. You see.

A

This won't work, it's because.

B

Why why why, why did I take it and why it doesn't work, but it intuitive works.

A

Okay, so it either works over the value, but the intuitor works over the reference.

A

Okay, let's see.

A

You mean this right, yep.

A

So the issue is because it is a reference.

A

You see, the item is iterating over a reference of a superstruct containing yield size and usas, but iterator is not implemented for this. Okay, let's see the.

A

Iterator, okay for iterator right, we need to look into the.

A

For the iterator, we need to look into the collect method.

A

So the collect what this collect does is collect to any item that implements the trade from iterator, so anything that that implements from iterator can have the correct method. So I just go into this to see what it implements and, let's look at hashmap.

A

You know you see that there's only one implementation for from iterator, which is using the original value. You see you see this.

A

This is using the original value, but here we are using this.

A

A

A

What we need is because this wasn't implemented, because this wasn't implemented in.

C

A

B

They are not implemented.

A

By what not implemented for hashmap.

A

You'll notice the difference, because here right, the type is the type. Is this it's a reference: iterator, okay, look iterator.

A

A

A

Okay, I don't know what I don't know why I can't find a wet eater.

A

I can't find to it by an issue uh I can't find it shouldn't be like this, but now mine are open. True, so, okay, you look into this right. You can see at it right for this implemented, not for what, but for slice.

A

No, it's right. I don't know why? Isn't here, okay, open issue? Okay, so you see that the you see that the wet here returns, an ether of p.

A

So essentially, this is just a reference. Obviously it's just a reference. I didn't check.

A

Yeah, you see it's just a reference.

A

You notice this. The original value is one but you'll get a reference to the one.

A

Where did this show in reference.

A

You can't you can see from this signature.

A

Okay, you see the iterator. The item for this iterator right is, is a reference of t. So in this case you see that the wet here.

A

And then, with the ether right, it will convert it into the eater.

A

With the item he will convey it into eater. With the item of t. The p is this and then the iterator is called when the fn ether.

A

This is the iterator right, so iterator for this is item is.

A

So the item is this, but what the component expected is, what is expected is this.

A

So but this is being expected, but then we got the reference, which means you can't construct it from the original value. Do you get.

A

It and then that's why the errors say so this means the trade is not implemented for the hashmap. As you can see here,.

A

Yeah you can see from iterator here right.

A

A

A

Okay, so now understand me yeah, I understand what I'm saying here. You can't you can't use. You can construct a hash map using just reference. You need to construct.

A

You need to construct it by using the original value, but I don't know I don't know why they didn't implement from iterator uh the if the reference for origination I didn't, I didn't know why maybe there's a reason, but I'm not sure why it's the reason yeah. I don't know why yeah they will implement, extend with reference.

A

Oh yeah, so I think I think I understand now right or are you still.

A

B

Yes, yes, yes, continue.

A

Okay, okay, so yeah. This doesn't work.

A

You see this, this doesn't work. Why? Because plus plus supports type nation, but in this case it doesn't know what type it is going to call it into. You see that the from iterator right.

A

From iterator, it's like how like how it can, how the how a struct might be, how a structure or an enum will be created from an iterator.

A

So you'll see that hashmap from iterator is in print for hashmap and from iterator is also implemented for set. You know for web.

A

You see from iterator type t is also implemented for wack, so here just now, just now we were using from iterator, that's mostly just white okay. So just now we were using from iterator yield size view size for hashmap right for hashmap, your size, your size.

A

You can see map, and now we want this from iterator t for red t.

A

In our case, it is from iterator view size your size for rack, your size design. So you see just one iterator. You can call it into multiple types. So here that's covalent is confused. What type does what type does it needs to convert into?

A

So I just like this time. You should know what time is it yeah? So now it works. So it's connected into a rack that is like a vector, and so this is the expressiveness for the collect.

A

Italics are readable.

A

Okay, I I think yeah I increase the size should should work right now. I think.

A

Okay, any question on this collect thing is: basically, you want to collect an iterator into different type into another type.

A

So you can, you can collect into multiple different types as long as e implements from iterator, for example, let's say the common one. You can call it into string, you can call it into vector.

A

You can collect into all these different types.

A

Yeah, the most useful you also want is, you can use the same, collect for hashmap and hashtag. Headset is just basically set and hashtag is just map any question on this. So this is the expressiveness, but there are too many types behind the scenes scene. So sometimes you might get confused a bit.

A

And any question on this expressiveness like how how it is able to collect into different types. If I go too fast, maybe you want to slow me down.

A

Yeah any question at least like collecting into different types like you have an iterator. What it does is you have an iterator this one. You have an iterator and you want to collect this iterator into another type yeah. So this question just asked any.

A

A

Okay, so this is the collect how it works, so there are multiple trades behind there's into iterator from iterator and the collect uses both of them.

A

Then, okay, now, let's look look to auto vectorization in a lot a lot of time right, russ authorities, the code as in autobacterization means that you use simd, which stands for single instruction. Multiple.

A

A

Yeah single instruction, multiple data, so it uses one cpu instruction, but it processes like multiple data like four like four times the data or eight times the data, then what this does is it just you auto automatically use the faster function like it will automatically iterate over the four four items together.

A

You'll go through four item eight item in one shot, rather than going through one by one. So like it's like it's a free optimization, but sometimes you need to tweak it a bit the iterator to eat a bit. But it's not always it's not always, okay, so only sometimes you might be able to make use of smd. But if you, if it makes use of smd, it will be a lot faster.

A

Okay and then this is. These are mostly because, due to the types in place, the types other like, for example, yeah the types they have in place and a lot. It depends a lot on the compiler. But if you still want to manually vectorize manually the right smd code, you might want to like use the smd, which is a lot harder to write.

A

The code is a lot more troublesome to write.

A

Example, an example.

A

A

Not like this either, maybe this is a very high level one, but yeah not like this. Something like uh here's an example srmd.

A

I think maybe they improve it. A lot already.

A

Yeah something like this, you see uh you see like mm set one epi, eight yeah. You need to manually write this to get to get the feature and yeah. This is about automatization. So a lot of time you get the free performance from us.

A

Then later you can also easily switch to rayon to parallelize stuff.

A

So, for example,.

A

Let's look at okay.

A

So, usually you just you, have an iterator, don't don't say this? Originally it was eater right. Then they have a map. You have a sum, but you want to parallelize it. It's just like you, you just add par underscore.

A

Then it will automatically change that to use all your cpu calls. Now. You just add like these four characters, then you change the whole whole iterator to use all of your all of your cpu cores, which is very easy, yeah, just just one step.

A

So when you are already using iterator. If you want to convert it cover that your application to use that all of the cost right, you can structure it in an iterator format.

A

It doesn't cover all the cases, but a lot of the cases you can just par underscore at this. Then you can make use of the parallel parallelism that your computer already has yeah. This is rayon, but it's only useful for cpu. If your bound tasks really needs a lot of cpu.

A

Okay, now any questions here about rayon or auto vectorization say any question on.

A

A

Yeah any question: no, then I'll continue.

A

So next there will be guns, like last I face when using the iterator api clickable. If you use speakable, which is like.

A

Clickable is like this: you can pick the next value in iterator. You can like take the you, can look at the next value without taking the item you. If you next write the iterator, you go. One item next go one item forward, but pig allows you to look into the next item without taking it out.

A

So there's a difference.

A

So if you use pickable and you use.

A

A

Iterator, so you borrow an iterator, then you have multiple iterator there and you use the together with the together applicable and you use this take or something sometimes you might take something.

A

Sometimes you might pick something, and sometimes you don't take something which will make you confused, so try trying not to merge these two together and yeah, but other case I don't see any issue. If you don't use pickable, uh it will be very safe most of the time otherwise.

A

But it's not oh, it's not pyoxidizer, but yeah. So there are quite a lot of iterators I'll. Just look through a few first last iterator you can have. You can have all of these. If you just implement the next method in iterator, you just implement this one required method. Then you get all of this for free.

A

You have like first last mph fine position. First, let's you give you the first item. Let's give you the last item nth, give you the mph item.

A

Fine, for example, will you find you the item based on the predicate like, for example, you want to find like something that matches this like five equals to two yeah matches these two five? Five, you cannot find anything position. Position is like fine.

A

Other, besides fine, rather than giving you the original item, it gives you the position of the item. Yes, you just give me the position of the item and you have step by max min.

A

Step by side just keep the skip certain step, no step by you go through like every nth step, like, for example, you have to step by two. It will be zero, two four, it will go two step and then max is just get. The max item in the iterator mean get the minimum item in the iterator zip. You can zip multiple iterator together, you can zip another iterator, so you'll get two items in one shot, so that's it and enumerate.

A

In numerator is like inside the python enumerate, it will give you the number at the side, but you see it doesn't take any. It doesn't take any method.

A

Okay, let's show.

A

A

Okay, let's say f2l one: two.

A

A

To animate it and then it should have like.

A

Your print and and x, so now you see zero. The first line will be zero one. Oh, you need a.

A

A

So this will be the type and this will be the.

A

A

Okay, it doesn't allow you to use the length.

A

B

A

Yeah, so you get zero one one two two three, but what if you want to start from the run to innovate from the item, one you want to start from one rather than start from zero in your brain, like people need to rush, might be confused here because enumerates you, you can't you can't like set the start value.

A

What you're doing is you just use just now? We learned that? That's it just use it.

B

Before that, I think you may want to uh explain why an array is not an iterator. You cannot. You cannot do a enumerate call directly on the array.

A

I think you can. I know why I can't think you should be able to do right.

B

No, not this one, the previous step.

A

Okay, didn't let me figure out: okay.

A

Okay, so it works.

A

Maybe I'll explain this first, so this one is this is the range range pattern in rust essentially means it starts from the number one to don't know what number in to add infinitum whatever so you'll see if we, so it will be one two three, so this like it just it's just like this will be similar to the e numerator, but usually we just we use in a way.

A

Okay. Back to that.

A

Why this doesn't work.

A

Let me look into.

A

A

Okay, this is because iterator is implement is not implemented, for the iterator is not implemented for the for the array, yet the original value, but iterator is implemented for this.

A

So let's look at this.

C

B

uh You probably want to explain what do you mean by those.

A

B

We'll go back to the compiler error yeah. What's the square bracket integer.

A

Oh okay, okay, so this thing means array.

A

This means this array, but it doesn't show the each let's show the length. This is a mutable reference, which means a reference and mutable yeah. This one can can be another top, but it just it's just like a mutable, it's a mutable pointer, but it's an immutable reference to an array, but this one is to a slide, no, this to a slice, but this is to an array of three items.

A

But iterator is, I don't think iterator is implemented for this, which is why you get this error.

A

Yeah because iterator is not implemented for this, so I think, maybe later they will edit the editors because the iterator this one is it's very recent that they just add a lot. They just add a lot of the iterator.

A

This is why they call the cons, generic things considering or something so this array, if the number is here yeah they added this quite recent. So maybe not. Every item is implemented, which is why you get this error yeah and then, if you do, it means that it will directly make it into a reference which will work.

A

I think this might work. I don't it's not. I see that it didn't work to the white.

A

Yeah, this also didn't work. I didn't know why. I don't know why.

A

But yeah these have the same issue, because this is here, so you want to hide this so that it converts it into the something that is iterable. Let's do an array. Yes, that is array.

A

A

Yeah, I don't see that so that's why you can't directly use this for enumerate, which is one of the reasons.

A

A

Yeah, I don't. I don't think I can explain that at this point of time.

A

Maybe I'll continue is there any other question.

A

A

Yeah, which is why you can't you need to look into the types and which right you need this ether to cover it into an iterator and then there's all this map filter filter map for keep well skip that we just fold each item. You do a function and then give some output. I just I show.

A

Which we just like multiply by two for each item, do something and return the output filter, adjust filter item like.

A

You want to filter positive, so you'll filter all the items like. If you get the next value, it will skip all the items that is not not true here, yes and then filter map fitment is like both filter and map.

A

This is used quite often. Let's say you need you need something that is returning uh option. You see this signature, so the function is like a filter, but inside a map. If it's none means it will skip like, for example, if e is none, you can pass one. So you get one your pass pass return a result.

A

Okay, convert the result into option result result is that option, but you have okay and error type, but I won't explain a lot of that here, except is that option, but it's an option and okay just convert up the result of ok to sum sum value and error to none. I basically just convert this. If can, if it can pass this into integer, it will return the integer if it cannot pass it return.

A

None! That's how you see one and five. All these are embedded so there's filter map. It's quite useful for four is like uh four is like reduced, but it's not reduced so far you can. You can give the first value then it's like, except for you have an accumulator and for each of the value you you pass through. You can use the accumulator and do something with the value in this case. It's doing a sum here, your initial value of 0, accumulator and occasionally x.

A

If you are familiar with the javascript or if you are from your javascript, reduce your videos, I think half right python also has like a lot python. Have the two have produced as well? You might see this pattern or, if you are from functional language, my seniors pattern.

A

So with this basically just fold fold it inside doing some here and there's also a try alternative, which can fail and for reduce yeah. Both reduce is new, reduce its similar to fall. Think it's similar to four except it takes a. It takes all value.

A

It's like it's at four. First, I think radius is like four first, the first value as the accumulator reduce that's the difference and then there's peakable. We show pickup, but just now is to pick to the next value. Saturn is like. Let's say: if you have a array array of array, you see two array, you flatten it into single array: flatten you have inspect and by ref inspect. Unless you let let you do something that you can print print the value for debugging, it's very useful for debugging, like inspect the radio.

A

And this is very useful. I think I think this feature is very useful for debugging and then you have by ref just like explain by ref. You can, let's say you use the iterator use the iterator here, but you want to use it a second time.

A

So it's like it's something that you duplicate the iterator or something, but you can use both.

A

And then there's partition windows. These are very both high level function. Partition is like it will partition that you split and iterator into two place. Two partition like for in this example is you split, a single iterator into an odd, and even I think this is very rare, even in other languages. I see this very rare, but sometimes I think it's quite useful. You split this into odd and even split into two groups, and then you have windows.

A

I think e. Isn't here? No, why isn't here? I think, maybe slice only have this. You have windows as in like first two value. If your windows 2 means second value, two value, this windows and you have chunks chunks is like if you two means two value and two value yeah. This is chunk and that's chunk exact.

A

So you see that there are a lot of methods on the iterator. uh I think technically just now. The two is not it's not iterated, but it's similar to the functions provided by iterator and collect we mentioned about call light just now. The mechanism is it uses from iterator and into iterator to collect into any type of one.

A

That's all on any. Oh, oh and it's quite common. It's just. If all of the items fulfill the requirement, then you'll return. True yeah, it's just like that and if n, if all means, if one of them is false right, then the iterator will stop and it will directly return force inside a short circuit and there are any return force same for any enemies. If any of the items return true, then you return true.

A

Otherwise it will go through all the items to find if it goes through all the items and cannot find anything in return false.

A

So that's all any questions I think it's quite common in even the other language is quite common for all in any, and you have some product cycle some some. You know you just your summit yeah. Basically just some literature add them together here. This is the sum trade also. So if you have anything that implements the sum trade, you can add them together, product is just multiply them and cycle. Is you can just like like look again loop again, this is like you see.

A

There are quite a lot of functions and you have skipped wow take a while yeah quite a lot of functions over here, so these are quite huge and if you use them, if you know them like sometimes max, you can even max buy something.

A

So there are quite a lot of functions here and if you know them, then you can keep the code very short and concise, but you might spend some time reading reading through them understanding like what do. What do what yeah any questions on this thing go through yeah go through quite a few of them, which I commonly use. I think I just got a lot. I use a lot of them frequently different ones, yeah, it's quite it's quite common, though I mentioned at least I use them quite frequently.

A

So usually I don't so like when there's a case user I use both. I use both uh imperative loop inside for loop and I use both. I use the iterator as well, so I think which one if something is small, which method is more readable, I'll use that method more readable and more understandable.

A

So in some places I can't use iterator, because I need to collect multiple items, sub-multiple aggregation in those cases, I can't use I'll fall back to use normal loops, but in user case I will just use an iterator which is quite quite useful, can simplify the code. A lot made, it quite short.

A

And yeah you see that it's very high level. You get all this certain very high level, but all this in run time, they don't cost stuff like so is related to the zero cost. Abstraction, as in everything, is types in during compile time, and it will be combined into the so the runtime. All these types are gone in run time. You, you don't get the types unless, unless you do that, unless you do reflection or something but uh yeah, you don't get the types in runtime.

A

That's why there's no cost you can use as much as possible, but if you use too much of these types, sometimes like uh the types might be hard to, it might be hard to debug the compiled time error. The types are like so long and you don't know which part have error, but if you are used to it, then you can easily pinpoint. Where is the error during compile time and yeah live code.

A

Okay, life code, an example.

A

What should I do live code.

A

A

Is okay: let's look into implementing.

A

Why is that called one two one one two.

A

Okay: let's do that.

A

Okay, so if you want to see sequence, take the last two number: we will build an iterator using the with fibonacci sequence,.

A

So fibonacci sequence have you have the first.

A

Is and the second also your size.

A

So I'll create a new value, yeah I'll just do default, but yeah I just created.

A

A

Should start from 1 1.

A

So now implement and we implement iterator for people upon rc sequence.

A

Okay: first, let's create a new fibonacci.

A

A

This method create a new structure x is 0, 1 and y is.

A

A

First rate is one one.

A

I think I think it'll start from zero.

A

uh Lazy to type this 50 right, so what this does is you generate a code for fibonacci, which is zero, zero start from zero in payment iterator for fibonacci the type item? You know you see iterator right, you need to implement the type and item and you need to implement the next method, which takes a mutable itself to return option of the cell item.

A

Total item will be view size and I'll I'll have the next method clear function, next method that takes in.

A

Self immutability and return option of self item.

A

Okay, so fibonacci is you add the two numbers together and then you move them. One number backwards.

A

So first I'll extract the x and y value out.

A

Can you do it and let's do it.

A

Create a new fibonacci type instance, and next.

A

Yeah, just let's get the next value a few times, save it compiles.

A

Oh, so it says that if you want to iterate the next value, because the next right, the next takes a mutable self, you can't just next without being mutable, so fibonacci needs to be mutable as in it needs to be unique and other other items can change it. So next can take the next. I can change the next value. It will iterate our fibonacci and change the next value.

A

Which is why you need this.

A

Okay, so works now we have x and y value. Next is your match uh just here's two mesh? Okay, let's say if you wanna see x y, let's say if two number is zero: if zero.

A

A

So I want to return two items and to first return the value v, and so this will be so. The value will be.

A

One and then I'll make it like zero one.

A

And if fibonacci, I will special case the first two option, because, like the first two options, we will just special case it and this again because the first value is also one I'll just return these two special case and for the others x given x y, because the y is the next value and x is the second o value.

A

I don't like this supposedly.

A

I think I got a reverse manual mine, so the fibonacci for this is x, plus y right, x, plus y, and then the fibonacci is, for example, x.

A

A

X is y and then y is x, plus y.

A

Like this right.

A

So what you do is you set the cell.

A

Cell value to click, and then you return v return value v.

A

Let's see, I don't think this.

A

A

Yeah, because this is uh because this.

A

This is the structure, so you need to specify the value you need to specify the name here unless, unless this is x, unless this x, you can ignite.

A

I don't think this works either.

C

A

I think I should just remove the xy. I don't know why I did this okay and.

A

Cannot add this.

A

Together, I think we need to do this.

A

Yeah, so we need to reference this to get the already to assign it back to the original value rather than the.

A

Reference, so the other error is because it needs to return option, so you need to return a sum, because it has some.

A

A

Okay, so you see the first value is one second. I just thought right here, one one, two: okay, so it works. But right now we just want to get like 10 values.

A

And we want to enumerate as.

A

A

So we will enumerate fibonacci.

A

Fibonacci and the value, so we want to print the value.

A

And- and we so print out for those that not not for me for us, this is a macro.

A

Yeah, so you get one one, two, three five, eight.

A

Wait there's too many yeah, we didn't limit it, so we want to limit it to 10 values. I think we did too too much. We just won the first 10 menus. We just you know. This is how we have take right.

A

Yeah, we just take the 10 values yeah. So just take one one: okay, okay, yeah see just just say this with the iterator but of course the more correct version. I don't think my motion is very correct here. Supposedly you can just do just the x and y doesn't make sense. Just just do the x y here you can just change it to x and.

A

Y, you can just replace this with hex.

A

Yeah, I think this is simple, because the order matters.

C

A

This way, you'll be.

A

C

A

A

Strong yeah see this is more, it's clearer.

A

I think we could make it clearer by doing.

C

C

A

A

Yeah, so is there any question on this? So this is how you write the fibonacci and yeah. Now you don't need multiple here, because you don't run the next value just to enumerate and in your way it takes self. It takes the original original structure, the whole ownership of the original structure and then enumerated. That's how you don't need the build and yeah this output.

A

This is like sort of implementing your own iterator for like custom type, you just take the value. Add them together, set the next type, see any question on this.

B

Also, it's worth mentioning that enumerate itself is also implementing uh iterator. That's why you can.

A

Yeah yeah yeah enumerate implement iterator.

A

Now you see the item is fuel, size and original item. Thank you.

A

Is the other question.

B

Please stop switching to your chat. Please.

A

Is there another question.

B

A

For now, or maybe someone wants to give an other.

A

Example, someone wants other how another example.

A

Like this is implementing your own iterator, what.

C

B

I guess we do another simple one, then we close the stream because it's kind of late already it's almost 9 30.

A

Okay, do a simple one: let's say you want to sum.

A

A

You want to sum even number up to 100: do you think? That's okay, sure or or sum the first or you want to sum the first 100 even number? Okay, let's do that.

A

Sum the first 100 even number.

A

So you will start from one, so you have a range and you want to filter out the even number.

A

So we should be able to modulus by 2 equals 0. It should be e1. This means easy ones, there's no remainder after being divided by 2., and you want to sum the first two hundred. So no what what hundred two hundred! I don't know why just this one 100 this summer.

A

A

Method, you can just put a type here for the.

A

For integer, okay, I need a type here yeah, just I think the sum is a type and when I just give it.

A

Okay, one zero one: zero zero. Is that correct.

A

So it's one zero one: zero.

A

Zero, okay, so that is the algebraic algebraic progression just add just take. The first number is two and the last the 100 number one hundred two is uh two hundred right.

A

I guess two plus four hundred multiplied by fifty, so one zero zero, two yeah yeah. This is correct. Is it too simple or why it's too simple? Is there any other example?

A

Any other question.

B

Please don't I can't change.

A

Let's see question any other question.

B

uh It's almost 9 30.

B

Does anyone have any questions to ask.

A

B

Of the day, if not, we are closing the stream soon and if you uh want to talk to ivan, please join our discord. The link is in the description of the video.

B

uh I don't see any questions so.

A

Maybe a lot of this is due to the types you see all of these types.

B

Yeah, well, I guess that's it for today.

A

B

We will see you in september ivan for the presentation and thanks everyone for watching we'll see you again in september. Do join us on telegram and discord if you want to.

B

Get updates stay updated to our events. That's it thanks. Everyone bye.

A