What is Span in C# and why you should be using it
Вставка
- Опубліковано 12 лип 2021
- Become a Patreon and get source code access: / nickchapsas
Check out my courses: dometrain.com
Hello everybody I'm Nick and in this video I am going to talk about Span of T in C#. Span was introduced in C# and .NET alongside a series of optimisations all the way back in .NET 2.1 and it has come to my attention that it is really confusing to understand if you just read the documentation for it. In this video I will break it down for you and help you understand exactly what it is, how it works and how you can use it in your applications today!
Don't forget to comment, like and subscribe :)
Social Media:
Follow me on GitHub: bit.ly/ChapsasGitHub
Follow me on Twitter: bit.ly/ChapsasTwitter
Connect on LinkedIn: bit.ly/ChapsasLinkedIn
Keep coding merch: keepcoding.shop
#csharp #dotnet #span
Hello everybody. As some people have already pointed out, after 13:50, when I'm returning a ReadOnlySpan, the ToString() on line 21 should be removed. I didn't notice because of the implicit operator. If you leave the ToString() in then you still allocate the string you return.
- Keep coding
Great video. But where could I check source code?
@@lipatovsa7 The source code is available to my Patreons
Thought you were just testing if we paid attention!
@@nickchapsas What would be the benefit/differences in using method(in string text) method(ref string text) ?
@@roflex2 strings are immutable in C#. Even if you pass down explicitly by reference your can't change the value of the string. You just point to a new string reference type
I'm a systems developer and primarily work with low-level languages like ASM, C, C++, etc., so I don't have a lot of experience on the intricate details of optimizations for managed languages. Explanations like these are invaluable and I find them immeasurably useful so I thank you very much for this.
Yeah typically managed languages have to add complex features to get the same speeds we get using lower level languages. Its a big trade off in my opinion. In those languages you usually either have to swallow performance penalties or readability penalties. In languages like c++, using a pointer isnt going to confuse anyone. So really I prefer those languages, but it is cool to learn the intricacies of these languages. In particular because I use it for a my job
C++ also introduced similar concepts as standard, and before that there were 3rd party libraries providing some kind of span. See std::string_view and std::span
@@minciNashu arent those features kind of just bloat though. We have void* and thats all we really need
My first thought was, I mean, great, but why do I get the impression that C# has a lot of tricks for solving problems caused by C# in the first place? ;) Why not just have a span method on the string that returns a readonly reference to a section of the string?
I'm so impressed by these explanations. Honestly, the best thing to happen for a junior dev, is to find your channel. Great explanations, very helpful videos with in-depth knowledge and analysis. Thank you, please keep doing them!!
Span is a really under-rated feature, not just for performance, but also my favourite pun in C# - TimeSpan
Spantastic pun 😉
If you have a dog... it's probably a Spaniel ? 🐕
Lately the news has been so dull I've tuned into Cspan
@@jeffwilson8246 This whole thread makes as much sense as watching C while only wearing a .
I don't get the OP. I understand span and TimeSpan, but I don't see the pun. I am the dummy cause 40+ people saw it. Guess I need to get out of my static internal scope of thought.
When I first watched Microsoft themselves explain Span, I was lost and confused. They have a knack for making something sound way more convoluted and complex than needed when they explain new concepts. This video made it all click instantly. Thank you very much, I can't wait to start using Span in my own work.
The same is true for their documentation
ye microsoft love overcomplicating every single one of their examples.....they need to hire people to teach them to keep things simple. Especially for their documentation.
I love these types of videos that explain the standard classes in the framework that help you write more efficent code. These are the types of things I don't stumble upon when researching how to solve a problem. I would love to see more videos like these!
Great video. This is a topic I've been half-aware of for a while, but seeing it in context helps a lot.
Great explanation, Nick, yet again! Thank you for taking the time, in your very informative videos, to show us what's happening "behind the scenes". On the heap, stack, etc. 🙏🏻
hey Nick, I have been watching your videos for a while. Just want to thank you so much for the learnings I got from them. You cannot imagine how helpful you are to people like me. I am using these learnings in a software solution I am developing myself already for a year. Again, thanks!!!
Andre from Portugal
Didn't learn anything new in this, but I'm impressed by your presentation. I would have more quickly learned how Span works if this was the first video I saw about it.
Your video's are awesome. I very much appreciate the technical in-depth explanations of them. Thanks a ton!
As usual, you explain the nuts and bolts, the theory, and the benefits, all better than the documentation and anything else I’ve found.
Thank you for this video.
As Unity 2021 LTS now supports spans this gave a really good introduction and explanation on how to use them.
Also the ref struct is valuable information since I have some very short lived structs for triangles and other mesh-related objects that only live for the duration of the method.
To me, these are the best technical videos on the net, even though about 95 percent of them are over my head. This video, however, was worthy of getting a bowl of popcorn, sitting back and just watching. Thanks a bunch, Nick.
Had missed the memo on this one. Thanks, this was really informative!
this is the first explanation of Span that made sense to me.
thank you!!
Nothing new under the sun but very well explained. It's useful to have this kind of videos around, proper knowledge should be distributed like this.
I'm a senior developer with 10+ years of experience and I learn so much from your videos thanks in advance
I really appreciate your content Nick, helped me a lot improve the way i code :)
Very nice explanation. Will surely be helpful in my work. Thanks!
Great Video, thanks for the board explanation, it was awesome!
Amazing delivery. Thank you!
Brilliantly explained, as per usual.
Quality content as always !
Great explanation, Nick 👍
Great presentation, thank you!
Another great video Nick!
A small question - is there a way to split Span into array\list of Spans?
Or actually, what is the best way to do it, without iterating by myself over the Span?
Nice explanation. Thank you!
Curious to learn how it deals with byte arrays, compared to working with strings like you demo'd.
Excellent video and presentation.
Excellent explanation, Nick! Thank you very much. :)
You are amazing! Thanks for your explanations
Span is a powerful structure but has some limitation, what about Memory and what is the difference between them
Nicely explained!
Wow, I never see before but I will us it in the next projects -- thank you.
Very cool explanation! Thank you!
Your short brought me here. Good stuff!
Hi Nick, great video as usual :)
Great vid easy to follow thankyou
Marvelous ! To the point. Respect !
Thank you Nick. Your videos are easy to understand, neat and to the point. I looked on your web site at the courses and wanted to know if the Dependency injection was based on a third party app. I was unable to locate a way to contact you there.
you should make a follow up video on ref structs if you haven't already.
Great video as always!
in 2:26 how do you get to that debugger window with the memory tab ?
Good video, very insightful !
good stuff! thanks for sharing.
Brilliant as usally, Nick :-)
That memory view in rider is awesome
Isn't the ReadonlySpan allocating at least a copy of the string on the heap?
For example, consider:
string s = "123";
ReadOnlySpan span = s;
Console.WriteLine(int.Parse(span.Slice(0,2)));
s = "321";
Console.WriteLine(int.Parse(span.Slice(0,2)));
produces:
12
12
Oh, since strings are immutable in this language, the second time we assign to »s« we actually perform a second heap allocation, and »span« can happily use the address to which »s« pointed to when »span« was defined. And some smart-pointer-like stuff.
if I understand this well, when you use the span, like:
Readonly dateAsSpan = _dateAsText;
you make a new variable called "dateAsSpan" but it's not allocating new memory for the "value" itself (in the heap), but instead just basically stores a reference, similarly when in c (normal C, not ++ or sharp) if I had a function like this
void double_it(int *j) { j *=2 }
using "int *j" instead of "int j" (* means that you pass by memory address reference instead of value)
Hi. Thank you for the tutorials. I have learned a lot. I wanted to know how do you get the results inline. Thank you.
That is a great video, thanks, Nick. May I ask what IDE you are using?
thank you so much
great example m8 ty for the presentation
Nice, thanks for that!!!
Should we refer to the _Span_ Slice method's start _index_ as... *"Spandex"* ? 😜
When a (male) programmer is very good at using Span... is he "SpanKing" ? 👑
(Or even Span heh). 🤴🏻
wtf
Thank you very much
When you talk about allocation, it's important to stress that span doesn't really copy the source memory on the stack. The span object itself - containing probably a starting pointer and a max size - is created on the stack.
Very true👆He should explaine it, otherwise it's misleading
The Span.ToString() method would allocate heap memory because strings are immutable, right?
Great video. I believe application only stops (completely) for GC when using workstation GC as opposed to server GC - you should compare the differences as they are quite striking and too long to go into here.
This is great.
Hi, Thanks for your great video,
Please record a view about diagnostics and tracing in .NET 5.
Its nice to see that D is benefiting C#
Nice! I learned a new C# feature. Span look like pointers, so I'd have liked to know what would have happened if dateAsText had changed, and make sure if span really works like pointers. (:
BTW i appreciate your video.
Phenomenal!!!
wohoaaaa!!! now i understand a bit more the use span and garbage collectors....thanks nick!!
Very nice and helpfully 🎉
This is excellent for string manipulations
Nice content!
This is... I can't believe how many times this could've helped me...
In comparison to C- span sound like a pair of pointers, and slice moves one of them.
I hope I got the idea of this feature thanks!
You should have maybe ran the test with a changing date every time. Strings in c# are instantiated once per instance and re-used. So "foo" in variable a and "foo" in variable b are both the same "foo" in memory. This likely will show a more realistic real world use case.
So this is how you use memory tab in Rider!
Is Span.Slice.ToArray() slightly faster than Buffer.BlockCopy or byte array copy using unsafe methods? (seems to be yes)
At 10:40, wouldnt it be an offset of 2 instead of 3? Im guessing that its a 0 based offset, so the first value in the string would be index 0 and the third would be 2, so if we want to start reading the month which starts at index 2 the offset would then have to be 2 right? (+2 offset and 2 length).
Great vid Nick, and there's me still working in c#7, .NET 4.8 & WinForms... I'm so far behind these days, but good to see new features in c#. I just wish the company I worked for didn't work on 15-20 year old projects...
don't worry, it's the same for me. Working with .Net framework 4.6.1 ! But i never stop learning new technologies. It is the only way
@@aminejadid2702 Same here, but I'm now losing interest doing it in my spare time, so unless I do it at work I lose the new skill.
Add the System.Memory Nuget package. It's not everything that you get with core, but you get some benefit.
@@harag9 You can always find a better job.
.net 4.0 for me at work 🤣
The best👌👌
this is brillant
Hello, great video as usual :)
Do the .ToString() really needed in the YearAsText() method return ? (end part of the video)
It did because at that point it is a ReadOnlySpan not a string and the Console.WriteLine method doesn't have an overload for it.
@@nickchapsas I was wondering about YearAsText, line 21. Would the `ToString()` there still allocate on the heap and then implicitly converts the string to the ReadOnlySpan? I would’ve expected `return yearAsText` without the `ToString()`
@@TylerEich No you don’t need to do ToString(). What you would be returning is a readonly ref struct to the caller and in there you can do the ToString()
@@TylerEich Sorry, Yeah not that you pointed out the line I understand what you mean. No that's a mistake. I added a pinned comment to explain that. I missed it because of ReadOnlySpan's implicit operator.
Does int.Parse accept a Span? or theres a implicit conversion from Span back to string?
There is an overload with Span yeah
Awesome!
Span? I only know Div
Amazing
perfect
Very nice
I guess that Parse method have span override. But what is happening if we need to use method which will except string.
I guess it will have implicit conversion there.
If the method needs a string then the return string will be allocated but you can prevent any potential allocation during the mid-way processing in the method, depending on the workload. Also yeah, int.Parse has a span overload, including many other things that used to accept string.
Great content!
how does span compare to string builder in such scenarios?
string builder is used to create a string without having to deal with the immutability concerns. Span, even tho it could technically be used for something similar, primarily does the opposite.
So it's like a StringView, providing a view into the string pretty much.
True
For those who are not aware - It's std::string_view (after you #include ) in C++ 17
Very good video - has helped to clear up my understanding of Span :)
6:15 - You could access String as an array too!
But if you would try to change even just one char like that:
string myString = "Hello!";
myString [0] = 'h'; // This will not be compiled!
It still would allocate a whole new string. StringBuilder do not behave like that tho so you could do that:
var sb = new StringBuilder("Hello!");
sb[0] = 'h';
The indexer of a string in C# is get only.
What you wrote is invalid code.
@@Dennis19901 Thanks! I pointed out that this line can't be compiled.
Good video
Great video as always. Is this only ever useful for strings? Thanks
A span resembles an array so it can work as a byte array, int array and so on
In order to make use of the Span 'value' you still need to convert it ToString(), which as you say loses the value in Span, unless the resulting string is a concatination of a bunch of Span.Slice functions. So you could take the fact that Span is basically an Array and then join them together to produce the final result.
As you can see in the pinned comment the ToString() in the method that returns a ReadOnlySpan was a mistake. You don't need it and if you don't use it you don't allocate until the final ToString()
good stuff ;)
Does anyone know the implications this has with C# and Unity and/or if we can use this to our advantage with game optimizations?
In your last example you changed the return type but forgot to remove the ToString call from the method, negating the performance benefit. But thanks for explaining Span (and ref struct) in an understandable way; neither of those ever made sense to me until now!
**Everyone** Nice video!
**Me** Wait... A method can return multiple values?!
Tuples!
Even before tuples you could use *out* or *ref* parameters to return multiple valuess without the need of making a new type every time. I miss both of those greatly when I have to work in java (among lots of other things).
Sure. You also can return Monad like F# Descriminated Unions. In nutshell this is the simple value containers, but technically it incapsulated multiple values indeed. Tuples is the famous example of monads.
Yeah but don't. Unless they are related. Like co-ordinates or dates (you would use DateTime instead imo)
@@Qrzychu92 If want to return two unrelated things, they belong in seperate methods entirely. If it was co-ordinates or something like that, then Tuple would be fine.
at 00:03:14 ,if it slows things down, can you ask garbage collection not to free up the garbage until the entire program is done running? i mean, thats what memory is for right? let the entire program finish before doing any sort of expensive garbage collection?
So I wonder if there is a similar use of SPAN for a more common task of parsing CSV comma delimited strings?
Yeah that's a great usecase
4:18 benchy was such a cute name 😍😂 didn't see that coming
Does the StringBuilder class (which I have been told is more efficient than simply using String) use Span behind the scenes?
Some of the StringBuilder implementation has been updated to use that behind the scenes where possible