#160: Your JSON shall be streamed

00:00 Hello and welcome to Python Bytes, where we deliver Python news and headlines directly to

00:04 your earbuds. This is episode 160, recorded December 4th, 2019. I'm Michael Kennedy.

00:10 And I'm Brian Okken.

00:11 And this episode is brought to you by DigitalOcean. Check them out at pythonbytes.fm

00:15 slash DigitalOcean. $50 credit for new users. Let's start with a topic that we've covered

00:21 quite a bit, Brian. Type hints. Or do you call them type annotations?

00:24 I usually call them type hints.

00:27 Yeah, I go back and forth. And the article you're going to mention actually says you can call

00:30 them both. But yeah, it's kind of funky that they don't have like a definitive name. Maybe they do.

00:34 And I just don't have certainty around it.

00:37 Yeah, one of them's wrong. And whatever the one I'm using is probably the wrong one.

00:41 So we have covered it a lot. And there's a lot of tools that we've talked about, like mypy and

00:45 others. So hopefully we've convinced everybody that they should at least do a little bit of type

00:50 hinting. If you're in that boat, then this article is perfect for you. Al Swigert wrote an article

00:56 called Type Hints for Busy Python Programmers. And that's everybody, right? I mean...

01:01 We're all busy.

01:02 We're all busy. I really like the article. It's a kind of a... When I first went to it,

01:07 I was a little shocked because there's a lot of words there. But it reads like a conversation.

01:11 It's a conversational thing as if it's a programmer talking to an expert. So that sort of tone of,

01:17 I think I should do this, but what do I do? And so he just sort of walks through a whole bunch

01:22 of questions with it. He covers using mypy and typing as the two modules you need and how to use

01:29 those. And there are other tools, but really most people are using mypy. And that's what... That's

01:35 Al's saying that. I don't know if most people are using mypy or some of the other tools.

01:40 We don't... I don't think there's any way to gain statistics on that. But I think mypy is a good way

01:45 to start. And a reminder that using types doesn't affect runtime. You have to run a tool like mypy

01:52 to get it to work. And so I don't think he covers this, but I think that people should hook running

02:00 mypy into their talks system or into their continuous integration pipeline to make sure that these are

02:08 met.

02:08 Yeah. Everything's hanging together. Yeah.

02:10 Yeah. There isn't any compiler to catch you. So one of the things I liked about it is it gets you

02:16 started and it covers some of the basics. But there's... As I've started running it, started using

02:21 type hints, the exact things that I've run into, Al covers. For instance, you've got a function that

02:28 takes an integer, but it also could be none or the default is none or it's optional or something.

02:34 How do you deal with that? There is stuff built in the type int system for unions. So you can say

02:39 it's either an int or none and there's optional parameters. And then a shout out to stuff that

02:46 if you kind of get into it and you might... Like if you're passing callables or sequences around,

02:50 yes, there are ways to type that. And so I think this is actually just a really great quick read and

02:57 it'll get you started pretty fast.

02:59 Yeah. I really like it. I love the conversational style. It seems super approachable and it's kind

03:03 of from the skeptics perspective. It seems a little bit like, oh, I want to do this, but what about...

03:08 I don't know if I'm sure about this. Why can I do this? And yeah, it seems like a lot of fun. I'm

03:13 definitely a fan of type hints these days, not for the mypy reasons. I understand that makes sense for

03:19 very large projects. But for me, like the editors are just so much better if you can just give them

03:25 these little hints here and there. I'll give you an example. Like for the database for Python Bytes,

03:32 we are talking to MongoDB with Mongo Engine. And the way you specify it is much like SQLAlchemy

03:37 or even maybe more so like Django ORM where you say this column is a Mongo Engine column of type such

03:45 and such. Well, the type editor, the editors think the type is like a column, but it's not a column.

03:50 It's probably a date time or it's a string or it's a list or it's anything, but what do they think

03:54 it is? And so you can use type hints to say this column is an integer and it's equal to one of these

04:00 column things. Oh, okay. That'll tell all the editors when I hit dot on this, don't tell me it's

04:04 one of these column things. I don't want to know that. I want to know what its value is because

04:07 that's what I program against when I run it. So there's just all these different angles that make

04:11 it super valuable. I'll have to take a look at how you're using that. That's pretty cool.

04:14 Yeah, for sure. Speaking of cool, another thing that we have beaten to death, but is very tough

04:19 and leathery and has not died is the ability to create an application out of Python code,

04:25 right?

04:25 Yes.

04:26 Well, I stumbled across this thing. I don't believe it is new, but it actually is interesting

04:32 on layers. There are layers of interest here called auto-py-2-exe. Auto-py-exe.

04:40 Okay.

04:40 Okay. So this is a GUI and it's built with eel. First of all, it's a GUI Python thing built

04:47 with eel. Remember eel?

04:48 I know. Remind me.

04:49 Yeah. It lives in the ocean. Some of them are electrical. You shouldn't touch them. No

04:53 wait, hold on. It's a library for making electron-like HTML and JavaScript GUI apps with Python.

04:58 Oh, okay. It says electric snake. Got it.

05:02 Yeah. Electron. Yeah, exactly. So it's very much like that. I think it's for kind of simplistic

05:07 ones, but it certainly works for this app, which is pretty cool. So here's a Python app built in eel.

05:12 That's cool. And it's kind of meta in the sense that it is, its job is to make other Python

05:18 applications out of native Python code.

05:20 Okay. Interesting.

05:21 Okay. Yeah. So what it does is basically is this is automation over top py installer. And you might

05:26 think, okay, well, I could just, you know, pass the command line arguments and just do that.

05:30 But there's something really nice about having a GUI say, here's the icon. I just put the icon

05:36 into the app. And here's the name I want for the title and such and such. And, you know,

05:41 here are the additional binary files or additional resource files I want bundled and shipped with my

05:47 application. We're using this JSON file and we're using this, I don't know, this image that we're

05:51 going to embed somewhere. Just include those. And like, so there's all sorts of settings in this

05:56 auto py.dxe where you can just pick all that stuff out and you don't really have to be an expert at

06:00 all with py installer, but you can build pretty nice stuff. Like there's a checkbox. Do you want just a loose

06:05 set of files or do you want it bundled into a single .exe that I can just give and it'll contain

06:10 Python and the dependencies and all that? Okay. This does sound neat. Yeah. It's pretty neat, right?

06:14 Yeah. And so really people want to, if this sounds interesting to you and you want to find out,

06:19 Hey, is this cool? Is this interesting? There's a three minute YouTube video. You can just check that

06:23 out. Okay, cool. So I put the link to the three minute video at the end. People can click on that,

06:27 watch it. And then within three minutes, you'll decide whether or not this is useful for you. But

06:31 some of the conversation I saw people saying, yeah, I know it just automates PyInstaller and

06:36 that's cool. But there were things I couldn't get to work with PyInstaller, but doing it here,

06:39 it worked fine. So I feel like there's a lot of value in this. Yeah. Nice. Yeah. Speaking of value,

06:44 DigitalOcean, they are adding a lot of value to all the hosted things of the world, including our stuff

06:49 that's hosted there across all the different sites and whatever. I believe we're doing like 15 terabytes

06:54 of traffic a month through DigitalOcean. That is not messing around. That's a lot. So been rock solid.

07:00 The other thing that they have that they're just launching is the Kubernetes, the container, their own

07:05 container registry for things like their Kubernetes service. So we've got like Docker hub and places

07:11 that you can put your containers, but that's on the public internet and things like that. It'd be nice

07:16 to have a private little one for your private code that you want to make sure doesn't somehow leak with,

07:21 you know, making it public or private on accident or whatever. So you can use their DigitalOcean

07:26 container registry and have nice private container definitions, which is pretty cool, among other things.

07:31 So check them out at pythonbytes.fm/DigitalOcean. Like I said, $50 credit for new users to get started

07:37 and play around with this. Speaking of playing around, I've never played with Sphinx. Oh, you never have?

07:41 I haven't. I've had no reason to do it, but it seems something like something that's useful. Maybe you should learn about it.

07:46 Yeah. So one of the things like, so I have, what I've wanted to use Sphinx for,

07:51 so Sphinx can be used for building up documentation and makes like a documentation website based on

07:57 either REST or Markdown or not the REST in that. Restructured text. Restructured text.

08:01 Yeah. Thanks a lot, programming world, for having two things that could be abbreviated REST.

08:06 In the C++ land, there was some auto-documentation tools to where you can take, and I know other

08:13 programming languages have this too, where you can take information about the structure of a program

08:18 and automatically generate some web documentation about how to use it and what the parameters are

08:25 for different functions and stuff like that. So you can do that with Sphinx, and that's the thing

08:30 I've been wanting to do, and I've tried several times and failed. And it's usually because,

08:35 probably on my part, but the tutorials get me part of the way, and then, you know, they reference a

08:40 bunch of documentation and stuff, and I don't want to become a Sphinx expert. I want, that's kind of

08:45 the point of this. I want to be able to just point it at my code and go, go do it. And like, for instance,

08:51 pulling out, so what does a function do? What do different tests do? I'd like that to be embedded

08:57 in the doc string so that when I'm coding, I can make sure that that's correct, and then that's just

09:02 automatically generated. So what I'm highlighting today is an article called How to Document Python

09:08 Code with Sphinx. It's a short tutorial by Mosh Zadka. That's a cool name. And so, walks through,

09:16 it has a little example of a little module with the doc string, and then the small example of how to set

09:23 up your index.rst. And I like Markdown better, but this is so short, I'd probably just use the

09:29 restructured text version. Points out that you need a few extensions. You need Autodoc, Napoleon,

09:34 and ViewCode. And then it gives you, just gives you the conf.py file you need to make this work.

09:40 And then how do you generate this stuff? Well, there's this long, weird command. It's not that

09:45 weird, but I don't want to memorize it. So this tutorial also shows you how to set up a tox.any

09:50 file with this Sphinx built in. So you can just say tox and your docs get rebuilt. So I like this

09:57 tutorial, and I'm going to give it a shot. Yeah, it's a really cool thing, and automatically

10:01 generating your documentation. Not terrible at all. Of course, it doesn't cover, like,

10:05 okay, well, now that you've generated it, it's a bunch of web pages and stuff. Where do you put

10:09 that? But for internal documentation within a team, you got your own way to do it, whether you're

10:15 pushing it into a, like a small web server, or just leaving it with your code, or whatever. So

10:21 leave that to the user. Yeah, or put it on DigitalOcean, or not DigitalOcean,

10:24 GitHub Pages, right? Yeah. That's one thing you do. You put it on Netlify. A lot of options for just

10:30 hosting HTML. Yeah, or throw it into a GitHub cluster, or a Kubernetes cluster. Maybe not. That's right. So, you know, something that's always

10:39 amused me is the association of Python ecosystem and language with snakes. Because that's not the

10:48 origin, right? That's not what Python is named after. It's named after Monty Python. All these

10:52 apps and stuff should have, like, medieval knight type of names, but they don't. So I'm going to tell

10:57 you about one that is called Snake. Snake? Snake? S-N-E-K. Snake? Don't quite know the right

11:05 accent to put on it to pull it off, but Snake? It's pretty awesome. So it's a PowerShell module

11:10 for integrating standard CPython, which is Python code, and various packages with PowerShell. So if

11:18 you're doing anything on Windows, any sort of automation or advanced shell stuff, you're probably

11:23 using PowerShell. Although the regular terminal is getting better as well. And, you know, the story of

11:29 Python on Windows has always had a little caveat, or maybe a little star, like, yeah, you always do this,

11:35 except for on Windows, you type this other thing. And it's slowly getting better, right? Steve Dower's

11:39 done a lot of work to improve that. Yeah. With, say, the Python installer that comes off the Windows 10

11:44 store has Python 3 as a command, not just Python and so on. But here's one more thing that you can do to

11:51 make working on Windows with Python better. If you do anything with PowerShell, you can write scripts

11:57 now that have sections of those scripts or libraries for PowerShell written in Python. And it uses this

12:04 library called Python for .NET, which is a new modern way to share the entire last library of .NET and all

12:14 the objects you might create there with Python code. And similarly, say, this Python thing returns some

12:20 funky .NET type that can be converted so, like, it looks normal to the rest of the PowerShell land,

12:25 which is generally .NET stuff. So if you're doing stuff with PowerShell, check this out. It's pretty

12:31 cool. The syntax is funky, but, you know, it's PowerShell syntax. So there you have it.

12:35 Yeah. Well, okay. So I think that people can't be blamed for doing the snake thing because our logo

12:41 is two snakes. I know. I think that that is absolutely true. I think it should be like maybe

12:47 some cheese. Yeah, exactly. Or maybe a shield. Yeah, I don't know. Whatever. I don't mind having it

12:52 being snake. The snake is cool. And there's the, you know, the Python snake. It's just funny.

12:55 That's not actually the name. No. Killer Rabbit would be nice. Yes. The Killer Rabbit would be

13:00 fantastic. If you just got an exception and that appeared as an emoji, that would be beautiful.

13:05 Yeah. Or maybe a failed test in pytest, right? You could write a plugin. It unleashes the killer,

13:10 the killer rabbit and all that. All right, we're going down a bit of a hole. So maybe get us out and

13:16 talk, get us back to the databases. Okay. So databases. Since in Python world, one of the things we like to do

13:22 is use Pandas because Pandas makes dealing with data like so much easier with data frames and stuff.

13:29 And also one of the things with databases that we can use is SQLAlchemy if it's a SQL based database.

13:35 And most of them are. So you can just take Pandas and put it together with SQLAlchemy and put it

13:42 together with your database and just slurp your entire data tables into Pandas data frames and then

13:48 manipulate them there. So you're good to go, right? Right. Unless you have like five gigs of data,

13:52 then you're good to swap and be very slow. Yeah. Or if terabytes of data, that'd be bad. Yeah.

13:58 Very bad. So all hope is not lost because I'm going to guess Arena Trong wrote this article called

14:05 How to Use Pandas to Access Databases. And it's really great. One of the recommendations she uses

14:11 is don't be afraid of the command line for SQL because you can play around with queries and look

14:16 to see how big the shape of your data is to start with, which is a good thing. But then goes on to

14:23 tips on how to put this together correctly. And I'm just going to read out the highlights, but

14:30 there's examples of all these. So I think it's great to have a look. Don't grab a whole table limit

14:35 to just the fields that you care about that you're going to actually use. Tables often have a lot of

14:40 columns that you don't need. So don't grab everything. Limit the number of records. You can either use the

14:45 limit field to say, I only want to grab like a thousand elements or something. You can also combine

14:50 that with just the number of fields, but also you can also pick out, say, I know I'm going to already

14:56 throw away all the rows that don't have a particular value in this one field. So you can

15:02 limit that in your query. One of the things you'll often run into with SQL databases, of course, is to

15:07 need the need to use joins. And the quick recommendation there is let the database do that.

15:13 Put it in your query. Don't do the join within Pandas because you can do it with Pandas and SQL

15:19 coming together, but don't. It's going to be slower.

15:23 Right. Load it on a memory and have the data table merge it. That seems probably not like the

15:28 best way to do it.

15:29 Right.

15:29 Databases are pretty good at joining stuff.

15:32 Yeah. And I mean, if you find, of course, it didn't talk about this, but if you find you're doing

15:36 that a lot, you can kind of restructure some of your queries within. I think you can build that in,

15:41 like make a join be faster. I don't know. I'm not a database person, so.

15:46 You could put indexes. You could write a store procedure and call that. There's a couple options.

15:50 Okay. Ask Michael how to do it. Anyway, maybe not. So a lot of this is we're caring about memory. So

15:57 how do you figure it out? Like how much memory you're going to use? And I didn't know you do this,

16:01 but there's, she has a tip of using query called memory, memory underscore usage, and then summing it.

16:07 This is, I don't know what the field comes back, but you can do a query and grab the memory, but you do

16:13 like a small query, like just a hundred, a hundred rows or something, and then take that. And then you

16:18 can use that to estimate the total size. If when you're going to read out like a thousand or

16:24 2000 or something. And then the last one, I'm actually didn't quite follow, but I think I kind

16:30 of followed. The tip is to talk that the types get converted, like integer types get converted to like

16:37 64 bit types. I think, and you don't really need that for a lot of stuff. So there's a,

16:43 I'm calling this out because it's definitely a, for the user read this if you need it, but

16:48 there's a little bit on using chunking to, and then combining that with specifying what types,

16:55 pandas types to use so that you get smaller data footprints.

17:01 Yeah. Yeah. So for example, maybe one of the columns in the data frame that just the way it

17:05 came back from the database is a 64 bit type integer and you know, everything can actually fit into an

17:11 untyped and unsigned 16 bit integer. Well, that'll save you. It'll be a quarter of the size of the data

17:18 if you'd make that conversion. Right. So got a million of them, a quarter of the size of a million,

17:22 it's possibly meaningful.

17:24 Yeah. I mean, a lot of stuff that like, for instance, a lot of stuff I'm, one of the tables I deal with has,

17:28 has a numbers for the different specifying, which kind of a result we're looking at, for

17:37 instance. And it's, they each have like a number code. There's only like six. So the number can be

17:43 one through six. I don't need 64 bits to store that. So yeah, for sure. So it's good. Do you use a

17:48 pandas with SQLAlchemy and any of our doing it?

17:51 Not really. I, there's a few times I've been thinking I should probably try to do some BI,

17:57 like graphing type stuff, maybe about popularity of episodes or things like that, you know, figure

18:03 out what day I should release our podcast on, but I haven't done anything. So no.

18:06 Okay. But you do use a SQLAlchemy though, right?

18:08 Some, but not for the main site because that's MongoDB.

18:12 Oh, right.

18:13 I'm using something equivalent, like we're using Mongo Engine, which is super equivalent.

18:16 Okay.

18:16 Yeah.

18:17 Nice.

18:17 So yeah, I suspect there's probably some interesting MongoDB integration with pandas

18:22 that I just don't know about, but I don't know any equivalent of there. Well, this next

18:26 one I'm going to cover is very similar to what you talked about. It talks about how do I basically

18:31 efficiently work with data from this one source databases? And how do I not just select star

18:37 from all the tables into pandas and work on it? Another problem you might have is you might

18:43 have JSON data. JSON is sweet, but one thing JSON is not very obviously good at is streaming

18:50 data, right? Because how do you know where the closing curly braces? Maybe you've got some

18:56 key and then it's got an array of stuff. And the closing bit of that array is not until, you

19:01 know, maybe 500 megabytes down the file. That's a problem potentially, right? So if you load

19:07 it with, if you just load up a, you know, import JSON and do a JSON load, give it a file pointer,

19:13 it's just going to hoover that all up and jam into it in one giant dictionary and spit it out.

19:18 But if it's huge, obviously what you want to do is iterate over it and process it item by

19:23 item. Yeah. I mean, that's what you would do for a regular file if you could take it line

19:27 by line, but because it's this formatted thing, it's much harder. And something I ran

19:31 across last time, time before we were talking about some other project and it said, oh,

19:37 this project uses iJSON. I'm like, wait a minute, what is iJSON? That's interesting.

19:42 Yeah. Yeah. So iJSON is an iterative JSON parser with the standard Python iterator interface. So

19:49 what you can do is basically point it at a file stream, some like search expression kind of in

19:57 the JSON document, and then it will stream those back to you generator style one at a time.

20:02 That's so cool.

20:03 Yeah. So there's a little example I put in the show notes. It says, imagine we've got some object

20:09 called earth in our JSON document. It has a bunch of continents and within each continent, it lists

20:15 out say cities and states and so on. If that was huge, probably it's not, but if it were, you would

20:20 want to iterate over it. So you just go to the iJSON and go, give me the items of this from this file

20:25 for earth.europe.item. And item is just a way to say, stop there. And then what you do is just say

20:32 for thing in, you know, whatever came back from that load that you did and you just iterate over

20:38 it in a for in loop.

20:38 That's really neat.

20:39 Yeah.

20:40 And it's doing, like I was looking at the website, there's an example with grabbing it from like a

20:46 URL because that's, you might be getting it from an API endpoint.

20:50 Yeah.

20:51 So I'm guessing that it's not grabbing everything from the endpoint at first.

20:55 It probably depends on how the endpoint is set up or even maybe how the object that's

21:01 querying it back. So I think they're using URL lib or something like that there.

21:05 Okay.

21:05 If you would set that up to do streaming, then you pass the stream over to iJSON, then it would

21:10 probably stream it from the server and never keep it all in memory at once anyway. That'd be cool.

21:14 Yeah, that'd be cool.

21:15 Yeah. But there's probably several layers you got to set that up as.

21:17 I'm using generators and iterators more and more because this lazy, doing lazy work is

21:23 awesome. It speeds things up quite a bit sometimes.

21:26 It's so awesome. And you basically no longer have a memory issue, no matter how much data

21:30 you have. You might still have a time issue, but you're not going to run out of memory, right?

21:34 You're just going to blast through it one at a time instead of blasting through them all

21:37 and then working on them.

21:37 Oh, nice. I got to check this out. Cool.

21:39 Yeah. So people, if you have large JSON, bring out the iJSON.

21:43 Bring out the iJSON.

21:44 Yep.

21:45 For sure. All right. Well, that's it for all of our main items, Brian. I have, for some reason,

21:50 it's like a crazy news week and I actually decided to not throw a bunch of stuff in here,

21:55 but to save it till next time because I just don't want to overwhelm everyone. But yeah,

21:59 I've got a bunch of stuff to talk about. How about you?

22:01 I am just kind of in a lull after Thanksgiving.

22:06 Still in the turkey coma?

22:08 Turkey coma. Yeah. My head's spinning. There's a lot of stuff coming up in the next couple of

22:12 months and I'm pretty excited about it.

22:15 That's awesome. Yeah. I'm trying to juggle between getting over that holiday break,

22:19 doing all the Black Friday stuff at Talk Python training, and then getting ready for vacation.

22:23 It's a lot. But in that little gap where I'm really still paying attention, there's a bunch

22:27 of stuff that I ran across this week. So first of all, let's start with this. Django 3 was released.

22:34 Oh, sweet. Yeah. So that's pretty cool. And Django 3, there's a couple of things that are

22:38 important about it. One is it has official support for MariaDB, which is pretty cool. The other one is

22:45 that it begins the journey to the async world.

22:48 Dun, dun, dun.

22:49 Dun, dun. We'll cross over to the ASGI world.

22:52 Yeah. So things like async and await are now making their way to Django.

22:58 I saw some messages from some of the developers there saying that this is not like the full story.

23:02 This is the first steps. So we've got a long ways to go, but it's first steps, right? It's great to

23:07 see them moving that way so that you get super high scalability if you need it. Most people actually

23:13 don't need it. But if you do, it's cool to say the answer is not, well, now you move to go or node.

23:19 It's you put the async keyword in front of your view method and you're good again.

23:22 Yeah. I'm excited to watch this. And actually the Django 3 release, I had dipped my toes in Django

23:28 before and I wanted to sort of wait until I knew the 3 was coming. So I was sort of waiting for this

23:34 to be official before I started up again. So now I got no excuse. I can get started again.

23:39 Yeah. That's awesome. Speaking of excuses running out, the death clock does toll, doesn't it?

23:44 Yeah.

23:45 Pretty soon the death clock is going to toll for Python 2.7 or 2 in general and in particular

23:51 2.7. And as part of that, Python 2 will be removed from Homebrew.

23:56 Wow. That's big.

23:57 Yeah. That's pretty. It's just another sign that, you know, you probably shouldn't be using

24:00 this. It used to be installable. Homebrew install Python at 2 or 2 or whatever the keyword was.

24:05 Yeah. That doesn't work. Not yet. Very soon that's going to stop working. So that's cool.

24:12 Big news, although I'm not really sure what spin to put on it, so I'll just give you the

24:16 news, is Guido is stepping down from the steering council. And this came to us from Vincent Polio.

24:21 Great. It's a thing. I don't really know how to spin it other than that.

24:25 Well, yeah. I mean, basically, from what I got out of doing a little bit of reading is

24:29 there were new elections to re-elect people to the steering council. And Guido didn't put his

24:35 name in the hat. And people said, he didn't put his name in the hat. Here, let me put it in a

24:39 forum. And then there was a conversation that followed. Like, you know, a lot of people are

24:42 enjoying being part of the steering council. I've been on that part of Python for a long

24:47 time. And it's not, it's just not fun for me right now. And I've earned some fun in the

24:52 programming world. So I want to work on projects and not politics and all that kind of stuff.

24:56 So I'm going to let someone else take it. And he also said, the reason I'm comfortable doing

25:00 this is it's in good hands. The people already, the other council members there are great.

25:06 So y'all don't need me. I'm still going to be around and working on stuff, but I just

25:09 don't want to be on the steering council. So anyway, thanks for sending that over. That's

25:12 big news, but that's about all I got to say. There wasn't a whole lot written about it really.

25:16 GitHub is freezing open source for hundreds of thousands of years to preserve it for humanity.

25:25 That's pretty awesome.

25:25 Yeah. So the GitHub archive program, I watched this video.

25:29 It's well done, right? It's like a polished little video.

25:31 Yeah. So I don't know what to make of it other than it's a thing. They're going to put a whole

25:37 bunch, all the, as a, like they've already put a bunch of stuff in there, but I guess next

25:43 spring they're going to, or when next summer or something, they're going to go up and take

25:47 all open source code on GitHub and digitize it into these film, these weird.

25:53 Microfish code.

25:55 Yeah.

25:55 Digital microfish or something that'll last for very long.

25:58 Yeah. It's supposed to last a thousand years or something like that. And they're going to put

26:01 it in like some cave in, on an Island somewhere.

26:05 Greenland, Iceland, some, one of those two, I can't remember.

26:07 Right. So why?

26:09 To preserve it. I don't know. I'm pretty excited. I have over a hundred public repos that are going

26:13 to be up there though. That's going to be awesome. I'm excited for them to get cold and.

26:16 Yeah. If, all of, so if civilization collapses, we're counting on future civilizations

26:23 to be able to find this cave and figure out some technology to read these microfiche things.

26:29 Wow. Those ancient creatures wrote in a very weird symbol form.

26:33 Yeah. Okay. Well, good luck future selves.

26:37 Yeah. Yeah. Good luck future selves. Hopefully they got, some, some hints, but anyway,

26:40 that's pretty cool. And the last thing I had talked about this Python for decision makers course,

26:46 that I'm working on. And basically it's a high level of why Python, when keys it, what it's good

26:53 for, what it's not good for. I'm also going to do a live webcast that people can check out after the

26:57 winter break. So mid January, I'm doing just a free webcast. People can sign up and come ask questions

27:03 and we can just share our thoughts on, you know, whether Python makes sense for whatever they're

27:07 doing. So people can check that out, click the link and register for free. Cool. That's a lot of stuff,

27:11 man. It is. I think we should wrap it up with a joke though. Yeah, let's do that.

27:15 This is our tradition now. All right, you go first.

27:18 Okay. So, okay. Web developer walks into a restaurant. He immediately leaves in disgust

27:24 as the restaurant is laid out in tables.

27:27 Oh, must've been a no SQL web developer. Does it want that relational stuff in there?

27:33 Yeah, that's pretty funny. That comes from a joke API that Sinjitsu sent over. And so appreciate

27:38 that. And this next one as well is from there, but I kind of like it. So I'm going to cover it.

27:42 So last time when we were talking about pint, you talked about using prefixes and suffixes

27:46 for variable names. Like if I'm going to have something in meters, I'd use an underscore M

27:51 potentially or centimeters underscore CM, whatever convention you use. So what is the best prefix

27:57 for global variables?

27:59 I'm not sure. What? Maybe G?

28:00 That would be that. G. G underscore something like that. If it's global. Or hash.

28:06 Like you shouldn't have global variables. So let's just comment those babies.

28:11 Comment those out.

28:12 Pretty sweet. Yeah. Thanks, Sinjitsu, for sending those in. Those are funny.

28:15 Yep.

28:15 All right, Brian. Well, thanks as always.

28:17 Thank you. Bye.

28:18 Yep. Bye.

28:18 Thank you for listening to Python Bytes. Follow the show on Twitter via at Python Bytes. That's

28:23 Python Bytes as in B-Y-T-E-S. And get the full show notes at pythonbytes.fm. If you have a news

28:30 item you want featured, just visit pythonbytes.fm and send it our way. We're always on the lookout

28:34 for sharing something cool. On behalf of myself and Brian Okken, this is Michael Kennedy.

28:39 Thank you for listening and sharing this podcast with your friends and colleagues.