Python Bytes - #317 Most loved and most dreaded dev tools of 2022

Episode Date: January 3, 2023

Topics covered in this episode: StackOverflow 2022 Developer Survey PePy.tech - PyPI download stats with package version breakdown Codon Python Compiler 8 Levels of Using Type Hints in Python Extras Joke See the full show notes for this episode on the website at pythonbytes.fm/317

Hello and welcome to Python Bytes, where we deliver Python news and headlines directly to your earbuds. This is episode 317, recorded January 3rd, 2023, and I am Brian Ocken. Hello, I'm Michael Kennedy. Welcome to 2023. How exciting. It is very exciting. It's a lot like 2022, but I think maybe a little more optimism. We'll see. Yeah, but I've had one extra shower since 2022. Do you remember 2022? It just feels like a few days ago. It honestly, it really flew by. Yeah. I mean, we stayed inside for like two years and they're like, hey, look, we get to go outside. Oh, time flies when you're not just stuck inside. I even actually rode a bicycle the other day.

Nice. And then Oregon said we're going to rain for two weeks. Of course. Of course it did. Yeah. Well, let's kick it off off what do you got for us first all right let's talk about the stack overflow 2022 developer survey now this is one of the better surveys that gives you a pulse on the developer community it's you know we do the psf survey every year and that one i think is really well done this one's interesting because it's not just Python, obviously. It's all of Stack Overflow people.

Yeah. And it's really, really long. Notice the scroll bar. So I'm going to try to go through these and just hit on some of the high points for us. All right. So they did that as well for us at the top here. They kind of summarized some little things

you can call out in the jump to other sections saying learning to code online increased by 10% year over year. However, respondents older than 45 years are most likely to have learned from books, whereas the youngins are more likely to use some kind of online course or YouTube video or something along those lines. Does that surprise you? No. Yeah, me either. So last year they said we saw Git rise as a fundamental tool of software development. So like if you are a software developer, you basically just need to know Git.

Surprisingly this year they said that's Docker this year, increasing from 55% usage to 69% usage. And I kind of feel like the hype and the heyday of Docker was actually a couple years ago. And I hear a little bit less, maybe it's just less hype, but it's just more like, yeah, we're just using it. But I don't know, that was surprising and interesting to me. I'm using it more in places where I didn't used to. Like what? Build chains for just to keep them consistent. We're just, and it isn't even a cloud thing.

It's just having, being able to have a consistent build system no matter what machine you're on. Yeah, that makes a lot of sense. CI systems are utilizing Docker more. Sure. I've heard of even people using Docker to manage and employ. Is that how you would phrase it?

Little apps to Raspberry Pis for like home automation and stuff. Really? I'm sure there would be more of that if you could actually get a Raspberry Pi, but it's more of a theoretical. Those who already have the Raspberry Pi. One of my favorite sections of this is the most loved, most dreaded, and most wanted. Yeah.

Because it really tells you, you know, how do people feel about how things are now and how do they feel about where they're going, right? What would they do if they were unencumbered by, you know, their boss saying, nope, we're using this particular database or nope, it's all Angular and you're on the team that does Angular, so I don't care what you want, you're doing Angular.

And so Rust is on its seventh year as the most loved language. Seven years? Is it even seven years old? I know. I know. It's my thought as well. But 87% of developers say they want to continue using it. Wow.

I mean, yeah, anyway, that's pretty interesting. But it also, Python and Rust are basically exactly tied as the most wanted programming language. That's cool. I can see that. Yeah, absolutely. Phoenix. Phoenix is some web framework I've never heard of

that is apparently the most popular. AngularJS, it continues into its third year of being the most dreaded. All right, what else? Go ahead. Oh, just React is most wanted, but who knows? Yeah, exactly. So some of these questions and some of these phrasings and analyses are super, good and others are really bad so i'll try to highlight some of these so let me jump down here we'll i don't really care about the education stuff um i mean it tells you education profile it tells you things like yeah there's an insane number of guys versus women no that's fine i just thought it was funny that you said, I don't care about education. Teacher, you know, leave the kids alone. All right. So no, like it's, it's fine, but it's,

you know, people can check it out if they care about it. Right. Like there's, there's stuff about the demographics, like age and gender. That is pretty interesting. I wish some of these numbers were better. Like, yeah, it totally is. Right. But you know, that's also were better. Yeah, that's embarrassing. Yeah, it totally is, right? But, you know, that's also not news. So, yeah, some of the most popular technologies here. Now, I know, Brian, I might just be getting older and grumpy. But when I look at this, there's like, it's almost all these responses are broken into two segments. Professional developers and learning to code people. And the default is to take all those

together and i guess that's the biggest picture but i don't know i just when i overlook this i'm just like professional developers tell me about this like yeah you know if some if you're on your third week of a coding tutorial and that tutorial says use this like how how much does that really speak to the state of the community i don't't think it, you know. Agreed. So I'm checking, like when I think about these, I'm just going to check professional developer. And if you want the same numbers, you can do so as well. So where are we?

We are on programming languages, scripting and markup. JavaScript is definitely the most popular here. This is one of those areas where I don't, I think you need to phrase it a little bit different. Let's see what the question exactly is. Which programming languages have you done extensive development work in over the year, over the last year? Okay.

I don't know how, quite how to solve this problem, but there's a ton of Python people who did JavaScript work. Yeah. There is a ton of C++ people who did SQL, right? There's also Python people who did SQL. So SQL ranks above Python, but there's very few people whose job it is as a developer to write SQL and not use something else, right?

So that's- It's weird to have SQL even in there. Yeah, I know. Well, HTML and CSS, right, is like a little bit weirder as well. And so I feel like it should say, like, what is your primary programming language or something like, you know, something like that, right? I feel like it's... It's kind of like favorite movie. Like, what's your favorite movie, do Node.js and JavaScript, and that is what I think this top bar where JavaScript is should represent. But it's that plus everyone else, right? Aggregated. So I don't know. It's, these are a little bit weird, but I'll just go to the

time. Yeah, backend. So like, unless you're primarily just a front-end developer, what backend languages or whatever? Right, right. Or maybe you're using Node on the backend, but then other than that. So JavaScript, number one one html top programming language okay uh sql third most and then python and then typescript so you know like a bunch of interesting ones and then you've got your your um your vm um garbage collected javascript java and c-sharp back there and it starts to fall fall off pretty quick after that. Down at the very bottom, let's see the very bottom one is Crystal.

Okay. And APL, oh my gosh. APL, APL is something. All right. Another one that's worth talking about is databases. So we've got Postgres, MySQL, SQLite, SQL Server, MongoDB as the top five. So it's pretty clear that Postgres has the momentum.

Like MySQL was higher and now Postgres is passing it. But, you know, that's kind of going up here. So this is which environments have you done extensive work in. We'll also see that in like the loved versus dreaded section. And by the way, on the audience, Marco says, I've already aged myself with that Pink Floyd reference.

Indeed. All right. So cloud platforms, you know, AWS, I'm going to surprise you that AWS has over half of the entire market. No.

No. But only, it's only a smidge over half. It's 55, so. Yeah, exactly. Web technologies. Here's another one um no not this this one's kind of kind of okay but like for example they've got node.js um but they don't have python right so i don't know it's it's a little bit funky like they've node and express

and express runs on node whereas they have flask and django but they don't have the thing that Django and Flask run on, which is, I don't know. Anyway, it's a little bit of a challenging one, but Node and React are quite high up there. Look at this artifact, this living fossil right here, Brian. What's number three? jQuery. jQuery, you know it.

Dollar document ready. Let's go. All right. I learned jQuery a long time ago. Yeah, I still use a little bit of jQuery, you know it. Dollar document ready. Let's go. All right. I learned jQuery a long time ago. Yeah, I still use a little bit of jQuery if there's some site that's already using it. I'm like, yeah, whatever. I'm just going to put one handler in here

and not worry about it. Fast APIs on the list. Yeah, fast APIs on the list. Django and Flask are 14.6%. Both fast APIs, 6% and going up. Fast APIs just actually had a Ruby on Rails, which is pretty interesting. Wait, so Phoenix is down at 2%, 2.13%,

but I thought it was like... It's the most loved. Okay. It's the most wanted. I'm not, I looked at it, I'm not sure I agree with that, but it's a web framework built on the Elixir language. So there you have it. And we'll see.

So yeah, another one here that's worth pointing out, I'm not going to go over too many more. Other frameworks and libraries. This one is pretty weird in terms of the partition. This is what I had in mind when I'm like, there's something really weird about this. Yeah,.NET and NumPy in the same bucket. Exactly. So.NET is the equivalent of Python. It's the equivalent of C++. It's the equivalent

of Ruby, right? It's like a huge and broad. And it's the thing that all the other stuff in that ecosystem runs on, right? ASP.NET or whatever. And so they say, what other language do you work with? Do you work or other libraries? Do they say, what other language do you work with? Do you work or other libraries? Do you work with.NET or do you work with NumPy or do you work with Panda? There's no Python here, right? And there's no Go and there's no, no, I don't know. Like it's just, it's, I feel like this just fully is like out of place here. So if you just go past that, the top two are NumPy and Pandas, which by quite a good margin. And then Spring for Java, React, Native for mobile with JavaScript, Flutter.

But then back to a bunch of Python libraries. Yeah, like Learn, PyTorch. Yeah, so Python actually ranks pretty high in the other frameworks and libraries you might be using. Let's just, let's jump to the loved, dreaded, wanted section. And then that's probably enough. People can check out the rest. All right. So they've got this sort of, you know, like what is driving the market greed versus fear type of thing. And so they put each technology up and it's got a,

how much love versus how much hate does this technology get in a single bar, which is the, the, the graphics here are pretty nice um so rust 86 percent love so people use rust really like rust and they don't strongly dislike it it's probably a self-selecting group right if you made if you took i don't know javascript people and told them they had to do rust they might not like it but rust people seem to really like Rust. Elixir, I've never done anything with Elixir, but it's also really high. Closure TypeScript in Python is there at about two thirds loved,

one third dreaded, I suppose, or not loved. And you can see it goes down from there. Yeah, I kind of want to talk to those 32% of the people and say, what is it that you don't like? Well, I agree. I feel like there's a lot of people who end up, so for example, if you look at just the third that says they don't like it, that's 11 times as many people using Elixir. And that's two times all the people using Rust. So I think it's a little bit, you know,

a scale thing, right? Like there's a bunch of people who were working in one technology and then they were told you're going to do Python. And some of them wanted that change. And some of them probably like, but I like C++, or I like whatever I'm doing. And you know, they're unhappy to be in that situation. So I imagine there's some of that going on here, where there's less people like being moved into Elixir projects, unless they really, you know, want to pretty small group. Anyway, there's that. And then if you switch to the wanted, though, Python and Rust tie for number one in most wanted, along with TypeScript is showing very notable there.

Yeah, nice. Yeah. Last one, we'll say databases, I suppose. Postgres, most loved, least dreaded. Redis, again, one of the things like, does this really belong on this list? Redis is cool. It's mostly for caching and message queues,

not really as a standalone independent database. So can it really be the most wanted database? You technically, I know, can make it work that way, but it's not really built for it. So I don't know. Anyway, Redis is number two, whether it belongs there or not. And then MongoDB and SQLite and so on,

in terms of wanted, Postgres and Mongo. Nice. Not too much of a change. All right. I think that's pretty much it. I'll give just a quick shout out to some graphics that are down here that people can play with. If you really want to go explore,

Brian, check this out. Which one is this? This database one's probably fair. So there's this giant, I don't know what section is this. It is work with versus want to work with. Okay. So it has this wheel.

And as you move around the wheel, it'll show you, for example, 5,000 people who worked with Postgres want to work with SQLite. 8,000 people who work with MongoDB want to stay with it. Whereas 7,000 people who work with MongoDB want to stay with it, whereas 7,000 people who work with MySQL want to move to MongoDB. And similarly, there's like, it shows you all these interesting trends and flows from this technology to that technology.

Like if you go up here to the language one, you go to pros and you click on C Sharp, it's pretty interesting because like the C Sharp people, they want to go to TypeScript or Python or JavaScript or HTML, but they don't want to go to Java. Like there's no arc over to Java, right? Because these are basically

the same technology with slight different trade-offs. So if you wanted to be in Java, you would just be in Java, I feel like. Yeah. You can sort of check out these flows and stuff.

So anyway, I think that's more than enough on this survey, but people should go check it out. There's a lot of data here and a lot of things to learn. So does that say HypeScript? It doesn't fit on the screen, that's all. It's HypeScript.

Yeah, there's just YPE script, TypeScript. It's just a scaling. Nice. Okay. Before we move on real quick, John out in the audience says it's interesting that MySQL and MariaDB are perceived so perceived so differently yeah and that is true yeah i don't know much i think that my sequel like has a problem that people used it a long time ago and maybe it's different now but i mean that's what i'm basing it my my experience with my sequels like all

over 10 years ago so yeah same all right what you got for us next well i um i wanted to talk about peppy or p i don't know it's a pepy.tech so uh i found out about this um who did i find from from the uh the person that wrote it petru rare is um so it's it's we've talked about something similar what What we've talked about before is, is PyPI download stats or pypistats.org. And so I was like, well, what's different? So Pepe does download stats for, for packages. So what I'm showing, we're showing here is the,

the stats for PyTest. And what the PyPI stats does is it shows you a breakdown of like you've got the overall packages but then you can also look at downloads per Python 2 versus 3 and then major and minor versions of Python and so that's kind of interesting but but I actually now after we're like so much far past the two three versus three split like i'm not even supporting two anymore um so uh what i like is peppy doesn't do that so peppy's same stats are same for pi test what it's looking at is it grabs like the top handful so let's do the default i have it all customized right now. So let's drop those off. By default, it shows us

the last few versions of PyTest plus all of the downloads for the last major one. So all the 7x downloads. And so the numbers are a little different because we're not showing everything. So right off the bat, I kind of like that. I'm just caring about the last few versions. That's kind of nice. And it also has a daily, weekly, monthly chart so that you can kind of, uh, see trends going on, um, which is nice. Um, the, uh, I didn't realize at first is, uh, that you can, you can select, uh, different

down different statistics. So if I wanted to see everything on all versions, I can just go ahead and type a star there. I think that works. Yeah. There you go. And then it does all of the downloads. And then also when I just dropped that, I don't know if you noticed, but it shows all the different download stats already so you can kind of get a glimpse of what you're going to find um right off the bat now going down through the chart we've got both a chart and a table for for when they when they grab the data um and this this split up of i'm not looking at python version i'm looking at version for the package that I care about, or like, you know, possibly my package. So I did this recently for PyTest check. And I was thinking about it for PyTest

of like, which version should I support? And clearly, like most, like tons of people have switched up to most people have switched to 7x already. So yeah, I think it's perfectly fair game to not support 6x pytest anymore. So why not? Anyway, so just a different take on download stats. Plus, it includes some cool badges that if you want to grab your, like, add this to your repo to say how many downloads per week or per month. Kind of neat. Yeah, that's really cool. If you're maintaining some open source package,

and you're like, and it depends on something like this, like PyTest, you can quickly look at that and make a call like, well, how important is it for me to support this old version? Or how ready am I to move to the new thing? Adopt this feature that won't allow older versions of Python or whatever it is you're looking at, right? Like, should we add types?

Well, that might not work on, like, where is that? 3.3 and before? Yeah, well, like on PyTest Check, I've got a feature that I'm going to deprecate and it's only been in since like the last version. I just didn't think through the API enough, I think. But so I wanted to take a look how many people are using the most recent version. And actually, so most of the people, or quite a few people are using the most recent version. But it isn't most. If I do a total of every version,

it's definitely not most people using the most recent one. But I don't know. And of those, I have no idea if they're using the feature or not. but I don't know. And of those, I have no idea if they're using the feature or not. Anyway, interesting. Take a look on it. And I appreciate it. And it's an open source project as well. And some of these graphs have been updated by a contributor.

So it's kind of nice to have it live and well on open source. Peppy.tech. Very nice. Good find. I like it. All right. Off to our source. Peppy.tech. Yeah. Very nice. Good find. I like it. All right. Off to our sponsor for the week. Yes.

Thank you, Microsoft, for sponsoring this episode. This episode is brought to you by Microsoft for Startups, Founders Hub. Starting a business is hard, but by some estimates, over 90% of startups will go out of business in the first year. Ouch. With this in mind, Microsoft for Startups set out to understand what startups need to be successful and create a digital platform to help overcome those challenges. And Startups Founders Hub was born. Microsoft for Startups Founders Hub provides all founders at any stage with free resources to help solve startup challenges. The platform provides technology benefits, access to expert guidance and skilling resources,

mentorship and network connections, and much more. Unlike others in the industry, Microsoft for Startups Founders Hub does not require startups to be investor-backed or third-party validated to participate. It is truly open to all. So what do you get? You can speed up development with free access to GitHub and the Microsoft Cloud with the ability to unlock credits over time.

To help with your startup and to help innovate, Founders Hub is partnering with innovative companies like OpenAI, a global leader in AI research and development, to provide exclusive benefits and discounts. Through Founders Hub, becoming a founder is no longer about who you know. You'll have access to their mentorship network, giving you access to a pool of hundreds of mentors across a range of disciplines, across areas like idea validation, fundraising, management, and coaching, sales and marketing, as well as specific technical stress points. You'll be able to book a one-on-one meeting with the mentors, many of whom are former founders themselves. Make your ideas a reality today with the critical support you'll get from Microsoft for Startups Founders Hub. To join the program, visit pythonbytes.fm slash foundershub 2022 or click the link in your show notes. Indeed. microsoft for startups let's move over here so this next one

comes to us from several folks recommending it and to be honest i'm not sure if i would cover it if there wasn't a bunch of people said hey you should cover this is really interesting so i'll sort of take their lead and say this does look pretty interesting so jeff Jeff Hutchins and Abdulaziz Al-Skwasim sent this in. Thank you both for sending it. And this is the Codon Python compiler. Have you heard about this, Brian? Just from people submitting it. Yeah. And people talk about Python performance all the time. Sometimes, I would say most of the time, it doesn't really matter to you. Most of the time, you're spending way more time waiting on a network, waiting on an API call, waiting on a database. But sometimes it really does matter and you need your code to go faster, right? So, you know,

traditional options have included things like Numba or Cython or even PyPy, P-Y-P-Y, or the JIT compiled version of Python. So here's one more thing to put into that category of options called Codon from Exaloop. And it's a high-performance, zero-overhead, extensible Python compiler using LLVM. And LLVM is one of the compilers that gets used all the time, of course, for a bunch of different things.

And they claim that it is on the order of 10 to 100 times or more faster than regular Python for single core performance. And because it really compiles to some native code, let's say also, by the way, sometimes better than C++ in terms of performance. But it also, because it compiles to some native representation, it doesn't involve the GIL. So you also get better scaling. And it comes out of this, I think, genetics project called SEC, bioinformatics rather, which is this language for bioinformatics. And I think what they've decided is like, you know, let's just go and just do straight

Python, right? And if you go to exaloop.io, they've got a couple of sort of selling points here. You know, performance SE is a Python. It adopts Python syntax and is fully interoperable with Python. And they've got this little flow graphic type thing, shows your Python code. And if you look, it's not even type annotated,

which kind of surprised me, Brian. I figured they would at least require you add, you know, Python type ends. Yeah, but no, it says you take Python syntax and semantics, and you feed it in, it builds up an abstract syntax tree, it does some kind of type inference to figure out what the types are like ints comes up with an intermediate representation and then applies a bunch of JIT style optimizations and domain specific optimizations and outcomes pretty interesting native code as I said but also where does it show if it doesn't have a better picture because it's cut off here outcomes native code but also web assembly and GPU targeted code think about that you want to write some code? This Python code does matrix stuff.

I want it to run on my NVIDIA GPU. You know, dash dash GPU or whatever the thing you do to make it come out with that. It's pretty wild. Or put it on the web with WebAssembly. It sounds great. It does sound great. They've got some examples. It says we're going to

have a list of fruits, just three strings, banana, apple, and lime, and some prices. In a dictionary, banana has a price, apple has a price. And it says we're going to write standard for fruit in fruits. F string, print out the price of the fruit. And then it's going to put a bunch of stuff in the cart by doing list multiplication and addition like you would expect. Call the sum with a comprehension and print it out. And then, boom, no problem.

Compile that to native code if you want. It says if you know Python, there's nothing to learn. Yeah, there's a bunch of, yeah, go ahead. Question is which Python? It says Python syntax, but which Python syntax? Is it 3.10, 3.9, 3.11? It's not clear other than 3.6 or above because of the F strings.

The only giveaway I see here. And possibly lots of people are fine with not exactly knowing which version of the syntax. Yeah, it didn't say. I mean, maybe if you go into the docs, it'll exactly say, right? But not on the landing page. It was initially dissolved to solve

computing challenges in genomics, which it on the landing page. It was initially dissolved to solve computing challenges in genomics, which it's still focused on. Also, it's got a bunch of stuff focused on finance and GPUs for CUDA programming, as well as WebAssembly, which is pretty dope. Also, if you scroll down a little bit, there's a little video. Let me go ahead and tell it to go. There's a video.

I'll tell it to stop. I don't know. Where they show it running as a kernel in jupiter notebooks just doing standard standard python stuff with like kind of mass intensive things and then pulling in matplotlib and you know showing the various things you get with matplotlib so yeah it looks pretty good so uh what do you think uh i think even if it i'd like to know what version of python it's compatible with but other than, it looks pretty good. So what do you think? I think even if it, I'd like to know what version of Python it's compatible with. But other than that, it looks fun.

It looks really good. It's on GitHub. And so it is technically at least source open. What is the license here? The license is a business source license. What the heck is a business source license? Okay, from MariaDB Corp. Yeah, I'm gonna have to figure out what the business source license? Okay. From the MariaDB Corp. Yeah.

I'm going to have to figure out what the business source license is. I don't know. You can read the source, but it's hard. I don't know. That's what I said. Is it source open or is it not? Yeah. And so all this stuff is good. And the reason I'm hesitant to recommend it, but I think these benefits are really big

and people can use it, then great. You go a little further. It says Exaloop and you, this is free for non-production use. If you want to use it for things that are commercial, then you know what it costs? Don't know. Contact us. I don't know either. If you got to ask, it's too much.

It's too much for you. Let's just, let's just put it there. That's what my dad always said. If you got to ask, you can't afford it. Exactly. Which I don't know. I don you got to ask, you can't afford it. Exactly. Which I don't know. I don't, I'm not against them charging for it. I understand that this is probably an insane amount of work to put together.

And if that's their model, that's, that's their program. It's fine. I don't think everything has to be given away free. There's tons of work that goes into a lot of stuff. And sometimes charging for it is the right answer. But just tell people what it costs. You know, don't make them email you. That is really not good. So Robert mentioned, Robert Robinson, maybe it's a sliding scale.

And it's like, it depends on who's asking. You know, Microsoft calls up and asks. It's a different price. I'm sure it is a different price. And I'm sure that's, oh, do you want to run it on all of Azure? Well, here's the price. But put some common cases in there. Are you a data scientist and you want to use it for an API?

Here's what it costs. Are you trying to host it as a server? Well, here, then maybe you need to... There should be at least a little bit of a, here's the realm of where you start, right? A lot of things have, here's the base price, here's the realm of where you start right a lot of things have you know here's the base price here's the pro price here's the enterprise price and here's the contact us if that doesn't fit you right but to say it costs money we won't tell you what it costs i don't know it just sounds well and i hope people so that the free non-production non-commercial versus

commercial that's still a fuzzy line. And I wish there was more detail whenever anybody does that. Because I've been in situations a lot where I'm just some dude writing a script in a company. And the company's big, but I just want to get this thing to work. And it's just a small piece of my build chain or something. I'm not going to get okayed to do a corporate license for something just to save me five minutes a day. Yeah. And maybe you're not selling it, right? Yeah. Yeah. That's another thing is, is it part of the thing I'm selling or is it just part of my

work to get something else done? Yep. Indeed. So anyway, pluses and minuses, but it's an interesting contribution and thanks folks for sending that in. Definitely. Indeed. Well,, pluses and minuses, but it's an interesting contribution. And thanks, folks, for sending that in. Definitely. Indeed. Well, so I want to talk about type hints a little bit. I've been thinking about type hints a lot and ran across this article called Eight Levels of Using Type Hints in Python. And it's from Yang Cho.

And first off, I love the graphic. This is an unsplash graphic, but I'm kind of a fan of it. It looks like posters, like band posters or something like that on something. And then half of them are peeled off. So cool image. Anyway, complete side note. But so I was attracted to this because this eight stage, I kind of thought of it as like eight stages of using type hints

because that's kind of how I started using type hints. And it almost got the order the same. So let's go through these orders. So the first one is, did I skip over it? The first was type hints for basic data types. So basically, like you kind of do this if you're using data classes at all. You just say, well, I've got my age as an int

and it defaults to 29. Just to kind of tell people what you're thinking about. And while my hosts are a tuple and they default to an empty tuple or something. So this is, this is pretty much how I started as well. And I started mostly because I was started using type data classes. And this is what they look like when you're declaring the contents of a data class. The, and then, so where do you go from there the next uh the next stage is uh defining a constant using final type actually i've never used this um have you ever used final i've wanted to use it but it's so it's so not exactly enforced type of thing it's like

right i would love to have a constant in python but doing it through typing means the only time it's going to be validated or checked is if i run my pi or something and like most of the time i'm not you know so it's like well okay whatever i'm not yeah i'm actually increasing the amount of times i'm using uh my pi or other or other type checking things but um yeah i wouldn't have put this number two but minor gripe it's kind of a cool feature so it is a cool feature though yeah i don't i don't knock it's existence i just wish it was more supported like in the language yeah uh the third is uh adding multiple type hints to one variable um like it shows here a couple different ways, the union of an int and a float or the bar of int and float. And I, my, the most, the number one way I use this, I mean, I use things like this

also, like it could be an int or a float, but I mostly use it for like int or none or something or none, because I've got a default value of none and i kind of wish there was another way to do it because i when you when you have a i don't know it's kind of convenient anyway but like let's say i've got a parameter to a function and it defaults when i internally i assign it it defaults to none if nobody passes anything in but i also don't really want somebody to pass none in you know um uh so i wish there was a like the default is none but users can't pass in or something like that um yeah but anyway it works um and and in the end it helps with testing anyway to be able to pass in all all the types but so

yeah the the and i've totally switched to the thing. I think it was 310 that added that. Do you remember? I think it was 310. I think it was 310. It says right here. It even says there, yeah. Yeah. So anything I'm writing, I support 310 and above because I like the bar.

The next thing is using general type hints, which also this is, yeah, I think now we're starting to increase the stage of understanding of how type hints can help you. And the notion, like in the example, they use iterable as, you know, from typing import iterable. So you say that, you know, I'm accepting something like numbers and it takes an iterable. I love this use. And I, you know, it takes a little bit more of advanced use so you get a little more comfortable with type hints and uh to be able to say i'm gonna do

iteration like for in and nums doesn't really matter what it is if it's a list or a tuple or or something else as long as i can iterate over it it's it's good so um and the other thing is uh probably and i'm not sure why they didn't put, is optional, but along the same as iterable, I'm going to often say it's iterable, but the type inside is an int or something like that.

But anyway. Right. The fifth one is like... Interable bracket int or something. Yeah. Or bracket strings, you know what you get out, yeah. The fifth is type ints for functions.

And I probably would have put this as like number two because I started using this right away. Yeah, just to say, because that's one of the things that's hard when I came to Python from C++ is knowing like, where do I put the return value? Like the return type is, you know, maybe it's multiple return types,

which I don't really want it to be, but it's nice to be able to see what the return type is and type hints add that with type hints for functions. Right, or you read a function, you know, like I want to call that, but what do I do? Yeah. Do I have to read the function and try to understand it

or can I just read the definition? Right. I think this is one of the main uses, right? And it's also nice to be able to see right there with a union type is none a possibility. Do I have to deal with nothing coming back as something is possible? Number six is alias for type hints.

And I kind of love this. I haven't been using this much, but I'm going to start using it more. Especially it says in Python 3.10. So before Python 3.10, you had to do like from typing import type alias and have this weird syntax thing around it. But it's not really that weird. But after 3.10, you just sort of declare a thing. It says like post type.

Post type is a dict that goes from int to string. That's great. And then you can use that as a type hint at other places. That's interesting. That's news to me. Yeah, that's really interesting. So definitely, I'll definitely start using that. So I'm glad I read the article for that.

Type hints for a class itself, which I'm not sure why it was titled this, but basically it's the self type. And this is very helpful if you have a class that has objects that refer to other objects of the same type. I don't know how you did that before then. I guess I wasn't caring about it. But you had to, there was a hack where you put it into quotes. So the example from this article is a class called list node,

and it's going to be past the previous node, right? So what is the type of that? Well, it should be a list node. Yeah. Python is in the process of building up the list node class, so it doesn't yet know what it is. But you're trying to define a function that has that. So there's this weird patch 22 thing.

Yeah. It can anything. So what you put is you would put list node in quotes. That's the old version. And of course it was just a string, but then it's like, well, but it's a string that says that. So we're going to just guess that that's what you want. It was weird.

Okay. Well, the eighth and final one is to provide literals for variables. And I haven't done this, but it's kind of nice also. And I didn't know you could do this. Just to say, in the example, it's from typing import literal. And then you define like a weekend that's a literal that can only be the strings Saturday or Sunday. And so this isn't an enumeration. It's just saying that it can only have these particular values. So I think I got that right. I don't think it has to be an enumeration it's a it's just saying that it can only have these particular values

so i think i got that right i don't think it has to be an enumeration but no no so this this is in the end so like the example says well if you assign that variable to monday it's gonna your your type checkers are going to complain because you didn't assign the one of the known ones and i probably would use this in conjunction with enums, or maybe I don't have to with enums. I don't know. But there's a lot of stuff that is passed around that's kind of like an enum,

but it's passed around as just a handful of strings that it can only be one of these strings. Yeah, it'd be nice for this to support it. Henry on the audience points out that x colon final equals true also maps to one of these literal checks so is literal of true rather than just a bool so the type checker can make sure it's always true not true or false so there's an interesting tie together of those as well

and robert points out that it sounds like a case of the mondays oh dear um and henry mentions that last time i checked self is not supported by MyPi. So we'll have to check on that. Anybody from MyPi listening? Let us know. Let us know. Well, that's it for this. I just thought it was a good... And I do think this is a good idea

is to not try to jump in the deep end of type hints. Try some of the simple stuff first and then you get to caring about it. Yeah, that was a good little discussion. end of type hints, try some of the simple stuff first, and then you get to caring about it. Yeah. That was a good little discussion around it. That was fun. Yeah. All right. Got any extras for us you want to throw out there?

The only extra I wanted to bring up is the testing code hasn't had an episode since October or something, and it's not dead. I just wanted to take the holiday off without telling anybody and it's coming back. So sometime in January, there'll be a Test and Code episode. Oh, excellent. Looking forward to it.

I have a few extras and one leads to another, which leads to another as is the nature of all these things. So I ran across a couple of articles, let me start that way, That sort of talked about RSS and how RSS is actually kind of important as a foundation of the independent web in ways maybe bigger than like I really realized. So there's this one called Back to RSS.

I'll just read a little excerpt from each of them. It says, it's 2023 or 1999 or whatever. Personal sites are back, blogs are back, RSS is back and owning your data is becoming real, which kind of a cool thing. And then also there's this one the same day on The Verge called Bring Back Personal Blogging. And yeah, anyway, so I ended up writing

a Welcome Back RSS article that kind of riffed on those ideas and talked about some more, talked about Reader, which I talked about before and stuff. So yeah, anyway, people can go check out that article. And it's interesting. And as part of the conversation, I learned about Reader, as in reader.readthedocs.io, which is a Python feed reader library. So it does things like let you subscribe with Python to RSS and Atom and JSON feeds,

mark articles as read and important, filter them, get stats, do all of it. Like basically if you wanted the logical internals of an RSS reader, here's the Python library to do that, which is cool. Now what we need to do is have a listener, like take reader and put a textual front end on it. So we have like a- Yes, actually, that's a pretty interesting idea.

That'd be cool. This guy over here built this thing called kustos which is uh basically like instapaper uh self-hosted python web app built on reader so people can check that out and then readwise looks really interesting are you familiar with readwise no so it's like an rss reader but it also lets you take notes mark stuff as read for later even if it doesn't have an rss feed and highlight like your research as you're going through it and save it so think how good that would be for people that say do podcaster you gotta remember a bunch of stuff about an article pull that thing back up and it's got your highlights and your notes on it so i'm liking this i this. I'm going to have to check it out.

Yeah, that's looking pretty good. So anyway, all that stuff came from this little article that I wrote. People can check that out. Another very important thing. There's this Mastodon account called Year Progress, and it just goes along and every day it just posts a progress bar for how far we are in the daily progress. So it's like a little ASCII, you know, TQDM style progress bar of the year. And it's just notable that we just had 100% progress for the year. Now that we're done, it's reset to zero, but it was 100%. Actually, it looks like it doesn't actually do it every day,

but it does it when it goes to a new percentage. So it's every few days or two or three days, right? Yeah, it looks like it, which I guess is probably good. I mean, thank you for not doing it. It's like saving 3.6, 2.6 posts, something like that. Yeah. All right. And don't just watch it.

Wait for it. It's not that exciting. It doesn't actually that often. It has no decimal point. So it's got to work in big ASCII blocks. So it does what it does. All right. This is something that people might want to keep an eye out

who had automated builds running over the winter break and also use PyTorch. So one of the thoughts thoughts I had was like, I was doing, I was just kind of kicking back. We're hanging around. I'm like, you know, I'm going to work on some of these little weird projects. Like I'd like the login page on TalkPython training

to just be a little smoother and cleaner. Maybe I'll just go play with that. Like I'm not really going to do real work, but it'd be kind of fun to do some web design and just sort of like no pressure, redesign that a little bit. And I thought, but I'm kind of be careful about adding new libraries, updating existing ones

like pip tools, style, update them. Because is anybody watching? And if something bad got in there, would anybody know or are they all on break, right? Well, it turns out my paranoid side was onto something. Luckily, not anything I was using. But PyTorch discloses malicious dependency chain compromise over the holidays this is bad uh if you pip installed somehow the pytorch stuff i don't

know the exact how it links together but it says pytorch has identified a malicious dependency with the same name as the frameworks torch triton library so that was like somehow shipped or or directly linked in pi torch in a way that i don't fully understand because i don't use it but what happened is on merry merry christmas the um on december 25th users who installed pi torch nightly uh should ensure their systems are not compromised because this trititon Torch. What they did is somebody put a malicious version on PyPI and when PyTorch was getting set up, instead of picking its own version, it preferred the PyPI version, which was a virus.

So anyway, people can check that out. It's sub-ideal. Since the PyPI index takes precedence, this malicious package was being installed instead of the version from the official repository. This enables somebody to register a package to do bad things. Anyway, people can read all about that, but

if you have PyTorch and you pip installed it over the point of break, read that. Yeah. Plus, kudos for them for coming up with a great domain name, bleepingcomputer.com. Yes, I know, bleepingcomputer. There's a lot of good articles on bleepingcomputer. It. Yes, I know, bleepingcomputer. There's a lot of good articles on bleepingcomputer. It's in my RSS feed. All right. Well, that's all I got for extras. How about a joke?

Yeah, sounds great. So first of all, let me, I don't want to kill the punchline before we set the stage. Okay. So I pulled up just the picture. Can you tell people what this is on the screen, Brian? You remember these things? these things well remember they're still there uh it's not my laptop not oh right okay it's on the back of like a computer or something a a desktop computer next to the power cord there's a toggle switch with a on and off it's on and off toggle switch right it's like a physical power button directly on the power supply on a desktop computer, right? Yep. Yeah, and Robert got it exactly right. So here is the joke. They finally made a switch to exit Vim.

And they've highlighted that physical power button on the back of the power supply. Nice. You're a Vim lover, so I won't rub it in too much, but that's a pretty effective way to exit Vim. Vim runs in a window, so just close the window. You know, whatever. The funny thing about power switches, so I used to work with a lot of, so I work with embedded systems,

and some of them run Windows Embedded. And embedded systems are supposed to be able to toggle the power off, just like from the power supply. And now normal Windows, you're not supposed to do that because you want to let it shut down gracefully. And if the disk is moving at the time, and even with Windows Embedded, there's some issues, or any embedded operating system,

there are a possibility you could do some damage to the disk if you power off at exactly the right time. But there's supposed to be safeguards in place to take care of it. So I'm just in the habit of using these things all the time, of like toggle off and on. And if we get like a new batch of people in that are mostly from DevOps or from non-embedded worlds, they're like, what are you doing? You're going to break it?

Yeah, you're going to break it. Like, no, these are good. Yeah. That's awesome. Yeah, my little circuit Python thing, it doesn't have a power switch. You just disconnect it from power if you want it off. That's how it works. well and one of the reasons uh now i'm going in the weeds but one of the reasons for the the the need for the ability to just power off is because a lot of times instruments are in a huge

rack and you don't go through and power off each individual instrument you just turn the rack off um so wow interesting cool well it's good to be back in 2023 yeah it is good to be back. So welcome back. And thanks, everybody, for tuning in. Yeah, you bet. Bye. Bye.