I think the performance improvement is one of the most key benefits of Python version 3.11! I hope that this trend will continue in subsequent versions as people continue to compare python to more performant languages.
Microsoft has tasked -- and paid -- 2 Python Code Devs to speed up CPython. Mike Shannon and Guido van Rossum. And Mike outright stated that the target is 5x performance!! At such speed, the performance penalty of Python will be less and less important, and people will emphasize on Python's ease of use, or even the "compile-less" toolchain will be preferable!! And with a parallel effort to run Python in browsers (via Wasm), we might just be on the cusp of seeing Python as a first-class Front End language!
This new “Self” type is amazing! I had to explain the whole core of “__future__” and “annotations” for an intern a couple of days ago and I can’t even imagine how simple it would be with this type, even though what you said about classmethods are true
I watched your video all the way through and sent it to my colleague who is learning python too, and wihtin ten minutes I also got an email reminding me you had a new video out! Thanks a bunch Arjan! The absolute ONLY youtuber I look froward to getting emails from
I haven't written any python since 2.4, but I watch this channel religiously to keep myself in the loop. One of these days I'll want to use python again and what I learn here is going to make the process of catching up much less painful :)
The first thing that sprang to mind re performance improvements is how that will eventually (hopefully) translate to lower bills for AWS lambda functions and equivalents, for those that make heavy use of python in that use case. Nice 👍
This speed improvement will definitely help in existing projects. (For example Django). Since classes variable initialisation all these being used extensively used in the existing frameworks and with this performance I believe definitely the frameworks performance is gonna improve too.. so for this reason I'm looking forward to use Python 3.11 Thanks Arjan for this informative video. 🙏🏽
I love the way you explain things, very calmly and professionally, not as the nowadays kids that only know how to explain hello world with weird music in the background and make some shitty TikTok video. Thanks for being alive! Keep it going! You are doing a really good job.
I so much wished that the error positions in tracebacks somehow would have used the more common line:column formatting that most compilers use. On scale the verbose 2 extra lines just adds unnecessary cost to logging unless you disable it.
taking enjoyment in writing complicated one-liners just for fun I didn't had any way of debugging or figuring out how can I fix them when they didn't work. Now with new trackbacks it would be easier.
I've been waiting for this version after Guido's presentation ( you can Google the presentation, I believe it's under his GH). I need to benchmark my scripts' performance and see the improvement. Thanks for sharing
Hey Arjan, I think that in your first example of performance boost in python3.11 you are misinterpreting actual speed boost to 50% as it usually happens with this counter intuitive thing. TL;DR u have 0.095 sec/code and 0.053 sec/code. Then in 1 sec you achieve about ~10 and ~18 codes done which actually is an 80-90% speed boost.
this is subjective - the new performance can be seen as 50% faster in comparison to the old performance, but the old performance can be seen as 90% slower in comparison to the new performance.
> StrEnum saves typing > proceeds to type about the same as the old names = auto() is ugly boilerplate, but it does reduce the chance you’ll misspell the particular enums. They should have done it as a context, EnumDefiner or something.
i like your thumbnail of actually excited for something whereas others people's thumbnail goes something like "you should not be developer" or "you maybe not a data science to be at all" so its good to see fresh new face
Awesome video, thank you. The example of faster 3.11 that demonstrates it using class get-set-delete in 8:25 is about the same speed as what you'd gain with __slots__ on 3.10, and it looks like slots are no longer faster in 3.11. Do you know what happened there?
Thanks David! When I run a test with Python 3.11 and compare slots vs no slots, I still get a 20% speed improvement when using slots (comparing the median execution time). So it seems that even in Python 3.11, using slots is still quite a bit faster.
@@ArjanCodes Thank you for the reply. Interesting, I checked again and you are probably right, there's still a difference, although it seems much smaller than what I had in python3.10.
I find your videos extremely helpful and well executed. You should do a full coursera / udemy course on "coding well". By that I mean content focusing on the design of code and the aspect of problem solving before you even write a line of code. You do this in the videos here on youtube as well, but a full-blown structured course would be awesome. I have seen so many Intro to Programming / python courses which do not focus on the importance of structuring your code.
3.10 was really cool feature wise but 3.11 doesn't add anything I depend on. The speed improvements are awesome though, they're going in the right direction.
I see a lot of influences from rust in this. I like it a lot. Soooo... algebraic data types when? It's honestly the feature I miss by far the most in python.
the reason I hate hate hate computer programming is that absolutely nothing any of millions of online tutorials or videos can help me do the specific thing *I* need the computer to do: computer differential algebra.
Interesting take. I've always found exceptions in Java to be far more cryptic. Granted most of my work in Java has been heavily dependent on libraries, so maybe the library authors are more to blame than the exception handling.
Funny for me, leaving C++ for python many years ago due to the ridiculous cycle of adding functionality just so the C++ committees can show off every year and now these same people are following me. :(
bruh haha 3.11 is so nice I remember using hacks to implement all these things like StrEnum and so on myself. Also i think with 3.11 there should be shape hinting for np arrays or tensors which is a game changer in ds / ml.
Self is an instance, Type[Self] is the type of the class you're in. The example is on a class method, whose first parameter is different from the normal method with a self parameter
Lol. Nice feature of python. I am still trying to learn and just think f strings are awesome. Just think I am wrapping my head around stuff and then the f bomb ohh SQL injection so we do different now. I will not french "f string" but yeah f bombs is like a pun intended. Nice video. But yeah, programming languages like Linux is just a rabbit hole without end. Poor me being newb. But will get there. But as always. Nice good video with not all info but it does allow good topics for research. That I think is a very good point. Not giving do like this but making clear here is how you can. And that is worth a lot. That is key to education I think. Not giving answers but give enough in way to make it useful but also there is work to be done and learn
Exactly! Though the problem is that once you're here, you'll be surrounded by great computer scientists who are going to continuously point out any minor thing you're doing wrong. It's like real-life Reddit :).
I find your content quite engaging, but it is way too advanced for me. I am new to Python and don't come from a CS background. What book or resources would you consider to be to your standards, to get a newbie like me in the correct path for solid python coding. (I don't think the Python documentation is really for beginners)
Dijkstra, Guido, now Arjan. The Dutch gave many great things, and so I'm saying "hello" from (formerly) Pays Kolonie 😄 Anyways, 3.11 performance improvement is really amazing in asyncio! I had script I originally wrote in 3.10, it runs a bit faster in PyPy, but MUCH faster in 3.11! So much so that I broke my rule of not using anything ".0" for production 😜 Edit: Maybe not asyncio itself per se, but every asyncio task I queued will end up instantiating a class, so I believe that's where 3.11's speedup benefitted me greatly.
👷 Join the FREE Code Diagnosis Workshop to help you review code more effectively using my 3-Factor Diagnosis Framework: www.arjancodes.com/diagnosis
I think the performance improvement is one of the most key benefits of Python version 3.11! I hope that this trend will continue in subsequent versions as people continue to compare python to more performant languages.
Microsoft has tasked -- and paid -- 2 Python Code Devs to speed up CPython. Mike Shannon and Guido van Rossum. And Mike outright stated that the target is 5x performance!!
At such speed, the performance penalty of Python will be less and less important, and people will emphasize on Python's ease of use, or even the "compile-less" toolchain will be preferable!!
And with a parallel effort to run Python in browsers (via Wasm), we might just be on the cusp of seeing Python as a first-class Front End language!
@@PanduPoluan CPython is not Python (anymore at least)
@@JP-re3bc The heck are you talking about?
CPython is the reference implementation of Python, written in C.
I think you're confusing it with Cython.
This new “Self” type is amazing! I had to explain the whole core of “__future__” and “annotations” for an intern a couple of days ago and I can’t even imagine how simple it would be with this type, even though what you said about classmethods are true
Just found your channel earlier this week and I am impressed by the quality of your videos! Keep doing what you do sir.
I watched your video all the way through and sent it to my colleague who is learning python too, and wihtin ten minutes I also got an email reminding me you had a new video out! Thanks a bunch Arjan! The absolute ONLY youtuber I look froward to getting emails from
Thanks bro finally someone who isn't posting malware or fake stuff, you deserve my subscribe!
I haven't written any python since 2.4, but I watch this channel religiously to keep myself in the loop. One of these days I'll want to use python again and what I learn here is going to make the process of catching up much less painful :)
Concise and understandable information. Aesthetically pleasing video. Good audio. Man, where has this channel been? Subbed.
Thanks for subbing! :)
The first thing that sprang to mind re performance improvements is how that will eventually (hopefully) translate to lower bills for AWS lambda functions and equivalents, for those that make heavy use of python in that use case. Nice 👍
Right now only if you use docker containers. Python runtime is stuck at 3.9 for AWS Lambdas and 3.10 is one year old.
This speed improvement will definitely help in existing projects. (For example Django). Since classes variable initialisation all these being used extensively used in the existing frameworks and with this performance I believe definitely the frameworks performance is gonna improve too.. so for this reason I'm looking forward to use Python 3.11
Thanks Arjan for this informative video. 🙏🏽
great video Arjan.
you might want to add the link to the chess roast video 5:55
I love the way you explain things, very calmly and professionally, not as the nowadays kids that only know how to explain hello world with weird music in the background and make some shitty TikTok video.
Thanks for being alive! Keep it going! You are doing a really good job.
Thanks
Thank you so much, glad you enjoyed the video!
9:18 - The home of both Arjan and off course Guido, the creator of Python.
Next, we will see Arjan start showing us "frikadel" and "kroketten".
Great video! Could you please also cover the new asyncio TaskGroups and the GroupErrors with the except* syntax? Thanks!
Great suggestion, I would also like to see a video about the new TaskGroups !
This is just amazing I just updated my version and tested these what you just showed, Thanks
Thanks Aman, Glad it was helpful.
You coulve showed how the add exception notes looks in the trace when running it
Guessing they're not showing at all. You can probably print then yourself though.
Oh, I also use the StrEnum pattern myself as well, but I'm never quite satisfied that I'm not screwing it up somehow. Great to see that built in
I always use this
In his piece class, what does the colon do in x: int or color: Color = Color.None? Ive never seen that before.
I so much wished that the error positions in tracebacks somehow would have used the more common line:column formatting that most compilers use. On scale the verbose 2 extra lines just adds unnecessary cost to logging unless you disable it.
that was exactly what I needed , thank you so much
taking enjoyment in writing complicated one-liners just for fun I didn't had any way of debugging or figuring out how can I fix them when they didn't work. Now with new trackbacks it would be easier.
At 5:25 how did you add those unicode symbols easily?
I've been waiting for this version after Guido's presentation ( you can Google the presentation, I believe it's under his GH). I need to benchmark my scripts' performance and see the improvement. Thanks for sharing
Thank you Arjan! Always promote your videos between colleges (win-win)
Worked smoothly, tysm
Self type isn't introduced for renaming, it's introduced to type annotate methods of inherited classes.
Totally not related to Python 3.11, but I love that sweatshirt-where’d you pick it up?
Hey Arjan, I think that in your first example of performance boost in python3.11 you are misinterpreting actual speed boost to 50% as it usually happens with this counter intuitive thing.
TL;DR u have 0.095 sec/code and 0.053 sec/code. Then in 1 sec you achieve about ~10 and ~18 codes done which actually is an 80-90% speed boost.
You’re right - that’s a good point. And it demonstrates the performance improvements even more clearly.
this is subjective - the new performance can be seen as 50% faster in comparison to the old performance, but the old performance can be seen as 90% slower in comparison to the new performance.
@@nivimani But that’s not subjective.
@@russianbotfarm3036 you're right. i meant something like, it depends on someone's preferred point of view.
Hey! Thanks so much for this video!
> StrEnum saves typing
> proceeds to type about the same as the old names
= auto() is ugly boilerplate, but it does reduce the chance you’ll misspell the particular enums. They should have done it as a context, EnumDefiner or something.
Love your stuff dude. Thanks.
Thanks, happy you’re enjoying the content!
Wow! it's Amazing I did it well ! Perfect work !
i like your thumbnail of actually excited for something whereas others people's thumbnail goes something like "you should not be developer" or "you maybe not a data science to be at all" so its good to see fresh new face
Great video - I enjoy the content, you do a great job of it.
When the execution time halves, it means the speed has doubled. That's a 100% performance improvement, not a 50% performance improvement.
I love the feature in 3.11 where its give you precisely where the the actually error.
That was great. Thank you.
Thanks so much, glad the content is helpful!
So good, thanks Arjan.
Clearly explained, I understood, and I'm not a Python user but will be soon
This is the top channel for anything python, great video as always
THANK YOU SO MUCH FOR THIS VIDEO
So we now know the next versions of python will be 95, 98 and se.
Haha, good one!
Awesome video, thank you.
The example of faster 3.11 that demonstrates it using class get-set-delete in 8:25 is about the same speed as what you'd gain with __slots__ on 3.10, and it looks like slots are no longer faster in 3.11. Do you know what happened there?
Thanks David! When I run a test with Python 3.11 and compare slots vs no slots, I still get a 20% speed improvement when using slots (comparing the median execution time). So it seems that even in Python 3.11, using slots is still quite a bit faster.
@@ArjanCodes Thank you for the reply. Interesting, I checked again and you are probably right, there's still a difference, although it seems much smaller than what I had in python3.10.
Hey Arjan, can you please make a tutorial about logging? thanks!
I like your content and the jumper!
Somehow "version 3.11" has a very nostalgic ring to it.
Only if you are old
Which color theme is this in ?
9:00 actually that would be nearly a 100% improvement (half the time is twice as fast).
Thanks it helped me install it
Works good, tnx
Can you do a video on Type annotations for Pandas and the new 3.11 Generics changes?
Looks great!
LiteralString didn't work for me. I tried the example from documentation - and it didn't work.
I find your videos extremely helpful and well executed. You should do a full coursera / udemy course on "coding well". By that I mean content focusing on the design of code and the aspect of problem solving before you even write a line of code. You do this in the videos here on youtube as well, but a full-blown structured course would be awesome. I have seen so many Intro to Programming / python courses which do not focus on the importance of structuring your code.
Glad you like the content! There is a course actually if you’re interested: www.arjancodes.com/mindset.
WAITING FOR 3.12 VERSION, TKS!
With toml support do your recommend using this instead of hydra with yaml files for configuration?
Do Self type also dynamicaly change if I use a subclass ?
Good vid!
Thanks so much
What is the benefit of toml ove ini files? The structure looks the same. Just import configparser....
How big are the pulls on your hoodie?!
Galaxy size pulls for galaxy size hoodie for galaxy size brain 🤯
I’m compensating.
@@ArjanCodes 🤣 for being so brilliant?
@@ArjanCodes 😂😂😂
those pulls are like karate belts
Those are some nice performance improvements.
Thanks Michael, happy you’re enjoying the content!
3.10 was really cool feature wise but 3.11 doesn't add anything I depend on. The speed improvements are awesome though, they're going in the right direction.
I see a lot of influences from rust in this. I like it a lot.
Soooo... algebraic data types when? It's honestly the feature I miss by far the most in python.
Nice video so I clicked subscribe 🎉
what do u think of carbon?
Almost here, but as per tradition you'll have to wait 6 months until all the packages you need support it.
Thank you!
Thanks so much, glad the content is helpful!
Those performance improvements are crisp
Is it possible to read an xml file with tomlib ?
Does it remain slooooow? Most of this new bloatware will only make it even slower.
Almost here!
You are awesome 👌
It's cool to get a definitively correct pronunciation for "Edsger Dijkstra" after all these years.
in StrEnum it automatically lowercase?
yes
What would performance it be like compared to 3.7
Hi, Can you please share the source of the hoodie? My husband really wants one :)
the reason I hate hate hate computer programming is that absolutely nothing any of millions of online tutorials or videos can help me do the specific thing *I* need the computer to do: computer differential algebra.
Is it possible to have a portable Python, i.e. no need to install, just download 1 single file and put it in the PATH?
Ive always wondered why exceptions in python don't have more info. Coming from the java world it felt a bit weak.
Interesting take. I've always found exceptions in Java to be far more cryptic. Granted most of my work in Java has been heavily dependent on libraries, so maybe the library authors are more to blame than the exception handling.
Funny for me, leaving C++ for python many years ago due to the ridiculous cycle of adding functionality just so the C++ committees can show off every year and now these same people are following me. :(
I'm so glad they finally put Tail Call Elimination in CPython!... Oh....wait...
Very helpful. Please prepare the video on Kubernetes. Thank you and lots of love from India.
Happy you’re enjoying the content!
This is starstruck moment for me. I love your content so much.
bruh haha 3.11 is so nice
I remember using hacks to implement all these things like StrEnum and so on myself.
Also i think with 3.11 there should be shape hinting for np arrays or tensors which is a game changer in ds / ml.
I saw a proposal for this the other day, though I'm not sure if it made it into 3.11
Almost here 🎉
What’s the difference between Type[Self] and Self?
Self is an instance, Type[Self] is the type of the class you're in. The example is on a class method, whose first parameter is different from the normal method with a self parameter
So the fact that that’s a class method necessitates the use of Type[Self], am I right? When would I use Type[] otherwise?
@@__sassan__ Type[blah] is useful also for class decorators, factories, and reflection
This video clip has some parts where video and audio are unsynchronized.
Thanks for noticing. We will work on it.
Lol. Nice feature of python. I am still trying to learn and just think f strings are awesome. Just think I am wrapping my head around stuff and then the f bomb ohh SQL injection so we do different now. I will not french "f string" but yeah f bombs is like a pun intended. Nice video. But yeah, programming languages like Linux is just a rabbit hole without end. Poor me being newb. But will get there. But as always. Nice good video with not all info but it does allow good topics for research. That I think is a very good point. Not giving do like this but making clear here is how you can. And that is worth a lot. That is key to education I think. Not giving answers but give enough in way to make it useful but also there is work to be done and learn
packages like psycopg2 and greenlet are failing in python 3.11
So all I have to do to becoming a great computer scientist is moving to the Netherlands! Got it!
Exactly! Though the problem is that once you're here, you'll be surrounded by great computer scientists who are going to continuously point out any minor thing you're doing wrong. It's like real-life Reddit :).
@@ArjanCodes 🤣🤣🤣. I love your sense of humor 😂😂
I find your content quite engaging, but it is way too advanced for me. I am new to Python and don't come from a CS background. What book or resources would you consider to be to your standards, to get a newbie like me in the correct path for solid python coding. (I don't think the Python documentation is really for beginners)
Python 3.11 almost here and AWS Lambda still didn't implement v3.10 runtime :)
They might be stuck on the current stable version of Debian, which has Python 3.9.
Best tutorials ever
I love how your outfit spells “11” :-)
I use the "@classmethod from_foo() -> Self" pattern all the time. Since you can't overload the constructor, this is the best alternative.
Dijkstra, Guido, now Arjan. The Dutch gave many great things, and so I'm saying "hello" from (formerly) Pays Kolonie 😄
Anyways, 3.11 performance improvement is really amazing in asyncio! I had script I originally wrote in 3.10, it runs a bit faster in PyPy, but MUCH faster in 3.11! So much so that I broke my rule of not using anything ".0" for production 😜
Edit: Maybe not asyncio itself per se, but every asyncio task I queued will end up instantiating a class, so I believe that's where 3.11's speedup benefitted me greatly.
toml config files... so like rust? That's pretty sweet
I implement Python in Azure Functions, which maxes out currently at 3.9 :(
Honestly, I cannot wait to get this new version.
Only a few days now.
honestly? release candidate 2 is available for download on the official website