Building Flask-based Web Apps

00:00 When you think of Python web micro-frameworks, Flask is definitely near the top of the list.

00:04 With almost 19,000 stars on GitHub, it's a powerful and extensible web framework.

00:10 It even powers the bandwidth-intensive audio delivery of the Talk Python to Me podcast.

00:15 In this episode number 48, we'll talk with Miguel Grinberg, who has written some amazing

00:20 Flask tutorials, books, and open-source projects. This is episode number 48, recorded February 2nd, 2016.

00:28 Developers, developers, developers, developers.

00:32 I'm a developer in many senses of the word, because I make these applications, but I also

00:37 use these verbs to make this music. I construct it line by line, just like when I'm coding another

00:42 software design. In both cases, it's about design patterns. Anyone can get the job done. It's

00:48 the execution that matters. I have many interests.

00:51 Welcome to Talk Python to Me, a weekly podcast on Python, the language, the libraries, the

00:57 ecosystem, and the personalities. This is your host, Michael Kennedy. Follow me on Twitter,

01:01 where I'm @mkennedy. Keep up with the show and listen to past episodes at talkpython.fm,

01:06 and follow the show on Twitter via at Talk Python. This episode is brought to you by Hired and SnapCI.

01:12 Thank them for supporting the show on Twitter via at Hired underscore HQ and at Snap underscore CI.

01:20 Hey everyone, I have a couple of news items before we get to Miguel. First, Miguel has generously

01:26 agreed to give away an e-book copy of his Flask web development book published by O'Reilly. To be

01:32 eligible, just become a friend of the show at talkpython.fm, and I'll randomly select a winner

01:37 during the week. It's been two weeks since the Kickstarter to launch my Python online training

01:42 adventure kicked off. I have a few updates you might be interested in. The project is funded and well

01:48 underway. The Kickstarter is over $14,000 and going strong. I launched a new website,

01:53 training.talkpython.fm. Check it out. You can see the next set of courses I'm working on after the

01:59 Kickstarter closes. Just click on courses in the nav bar. Once again, thanks to everyone who backed

02:05 the project so far. If you want to get access to my course, Python Jumpstart, by building 10 apps

02:09 for over half off at $29, hurry over to talkpython.fm/course and back the Kickstarter before it

02:17 closes in the next two weeks. Now, let's get on to an excellent conversation with Miguel.

02:22 Miguel, welcome to the show.

02:25 Thank you. Glad to be here.

02:27 You have a lot of stuff going on. This is going to be a fun conversation.

02:30 I hope so. I'll try to make it fun.

02:32 Yeah, yeah. We're going to focus this around Flask and look at all the different things you have going

02:37 on kind of in that space. Yeah?

02:39 Mm-hmm. Sounds good.

02:40 Okay, cool. Now, before we get into the details, let's talk about how you got started in programming

02:46 Python and all that stuff. What's your story?

02:48 Yeah. So, I first got into programming in the probably mid-80s, 1980s, with a TI-99 computer.

02:57 You can Google it, if you don't know what that is. But it used to be a very cool computer in the age

03:02 where, you know, computers were really new and most people didn't have them in their homes. And I

03:07 started writing games in BASIC on a TI-99 computer. Eventually, that seemed slow, so I decided to learn

03:15 the second language you can program on that machine. There were only two, and that was assembly language.

03:20 And I started writing games in assembly.

03:22 My gosh. So, you took, like, a real easy sort of gentle introduction to game programming, yeah?

03:27 Yeah. Those were the two steps you had on, you know, on the TI machine.

03:32 That is a big step.

03:33 Right? There was no other thing. And BASIC was slow, so, you know, what was the other choice,

03:38 right? If you needed speed, assembly was the answer. So, yeah, I wrote a Tron lifecycle game

03:45 on assembly language on a TI-99 computer.

03:48 That's cool. Did you release it?

03:50 I released it to friends only back in that day. So, yeah, not commercially.

03:56 Yeah, I mean, that basically predates the web, right? So, what do you release it to in some

04:00 sense? BBSs maybe, right?

04:01 Right. There were BBSs in those days, right?

04:04 Yeah.

04:04 Yeah.

04:05 That may be aging us.

04:06 Yeah.

04:06 I think I can still tell the actual connection speed based on the sound the modem made when

04:12 it connected.

04:12 Right.

04:15 Yeah, that was...

04:16 Yeah. So, my first board speed, the BBS speed was 1200.

04:20 Oh, yeah.

04:21 You can put an age to me, you know, by knowing that, right?

04:25 Yeah, I'm just a little younger. I came in at the 9,600 days.

04:28 Yeah, I'm old.

04:29 Okay. So, you got into BASIC and then assembly, and then where'd it go from there?

04:36 So, and then, so this was, I was a teenager, so before college. So, then in college, I learned,

04:43 you know, the languages that you normally learn in college, at least in those days, Pascal,

04:48 C, C++, Smalltalk, basically get a job. So, I'm from Argentina, and one of the things that's

04:54 great about college in Argentina is that you can do it at night and work full-time while you

04:59 study. So, I was lucky to have found a job, a programming job, you know, from the early days

05:05 in my college time. I had basically both sides, the academic side and the, you know, having to do

05:12 something production-wise, having to get something that works and that makes money all at the same

05:17 time. So, it was pretty good.

05:18 Yeah, that's a really cool opportunity to have a chance to write, like, professional code or

05:23 whatever you want to think of it as a student, because a lot of people come out of college

05:26 and they're like, now let me have, try to go find some experience so I can get a job,

05:31 right? Which is, it's a kind of tough place to be.

05:33 Right. So, thanks to the TI-99, I had some experience. I could get a job without having a

05:39 college degree. So, you know, I basically, those two complemented and, you know, allowed me to

05:45 have a, you know, a full experience from early on in my life. So, then Python came much, much later.

05:52 Maybe eight, nine years ago, I was working at a company where we had a really large C++ code base

05:59 that was really painful to test. This code base was exposed as a DLL, a C++ API on Windows. So,

06:07 I decided to create a Python wrapper so that the testers could write tests more efficiently.

06:13 Right. They didn't have to write it in C++ and all that stuff. Yeah.

06:17 Exactly. Right. And compile it and test it and crash and then do it all over again. Right. So,

06:22 I decided to create a, the Python wrapper and basically I got hooked. I really liked it. So,

06:28 it started from there.

06:29 Interesting. So, you're working in C++ and you made these wrappers and I guess the thought process

06:34 went something like, why don't I just keep working over here? Why do I have to go back to those

06:37 compilers and headers and libs and all that stuff, right?

06:40 Right. So, exactly. And in the same way, I studied and worked at the same time when I'm working,

06:47 even now, I always have lots of side projects that I do just for fun. So, at those times,

06:53 you know, eight, nine years ago, doing stuff on the web seemed to be the right thing.

06:57 Yeah. Excellent.

06:58 Okay. So, what can I do with Python for the web? You know, that led to, well, you guessed it, Flask.

07:05 Yeah. So, you have a lot going on with Flask. Why did you pick Flask over the other options? And,

07:09 you know, given the timing, it might've been, well, there were not many other choices,

07:12 but Django was probably around at that point, yeah?

07:14 Django was there. Bottle was there, believe it or not. I did look at Bottle at the time. So,

07:20 basically, this was a, this is a funny story. I started, it was sort of a friendly competition

07:25 with my wife. My wife, who is not a technical person, she was blogging and she was having,

07:31 you know, pretty good success with her blog. And she was always teasing me that she was doing

07:35 something technical and I was behind. I had to basically start blogging. Basically, that was

07:40 the dare that she put in front of me. So, I said, if I'm going to blog, I'm going to blog on my terms.

07:45 So, I had to write my blog engine.

07:47 Of course.

07:48 Right? I'm not going to use WordPress or, you know, one of those.

07:52 No, no, no. That has PHP in it.

07:54 Exactly. Right. So, I went and looked at Django and Django seemed overkill for a little blog

07:59 that I wanted to write. I looked at Bottle and Bottle seemed too little. Right? I think that

08:04 there were some things like writing more than one file. It wasn't clear how you would do it. You

08:08 will have to start inventing things. So, I settled with Flask. You know, Flask was right in the middle

08:13 and it seemed like the right balance.

08:15 Yeah. I think, you know, I think finding that balance sort of depends on your experience,

08:20 how you like to code, how much hand-holding you like it in some sense. Right? Absolutely.

08:26 I personally love to write my HTML just so and my CSS and have lots of control over the

08:32 URLs and everything. And so, these micro frameworks like Flask and Pyramid, I'm a huge fan of.

08:38 Yes. I completely agree with that. With that, I micromanage my code. I really like to write

08:44 myself and be aware of everything that's going on, make it very clear and make it very me. Yeah.

08:49 Flask seemed to be the right balance of not having to deal with all the pain of HTTP requests

08:55 and responses. And, you know, Flask abstracts you from dealing with that. But at the same time,

09:01 it doesn't hand-hold me and I can do what I want, which is, you know, it was a requirement for me.

09:06 You know, we had Armin Ronecker on show, I think it was 13, one of the earlier shows and sort of dug

09:13 down into a lot of the pieces of Flask. But maybe, you know, that's a long time ago. Maybe you could give

09:18 people like a high-level look at just what the moving parts of Flask are and kind of like how

09:22 you, how much work is it to get a website up and running? Right. There isn't really much. So,

09:28 Flask is very small. You can go read the code and you will probably understand most of it,

09:32 you know, in your first read. It's, you know, super clear code. It's code that I could write,

09:37 the kind of code that I wish I could write, actually, because it's so clear. And basically,

09:42 the main thing that Flask does for you is to abstract you from the interface with a web server.

09:48 You don't have to think about parsing HTTP requests or formatting the responses. Basically,

09:55 you get a decorator that allows you to associate a function that you have to write with an HTTP request.

10:02 That's basically pretty much, you know, the big thing that Flask has for, you know, for web development.

10:09 And then you write your application. You supposedly need to know how to write it.

10:14 Right. Sure. So, I'm trying to remember, like, the Hello World app on Flask is probably five,

10:19 eight lines?

10:20 I think it's, yeah, seven lines, if I remember correctly.

10:23 Seven lines, yeah. At least I got the interval right.

10:25 Yes. So, you have seven lines, four of which are boilerplate. So, your Flask, your typical Flask

10:32 boilerplate is four lines, four lines of code. And then you write your functions. You write typically

10:37 one per HTTP endpoint that you want to support. And in the function, you do any processing you need to

10:44 do and then return your response, which usually is, I mean, if you're doing a web application,

10:50 it's a string containing HTML that you can generate yourself or through a template engine.

10:55 Right. And the default is Jinja 2, right?

10:58 Jinja 2 comes with Flask. It doesn't need to, really. You can use any template you want. And

11:04 this is a theme with Flask. For any X feature, you can use any package that does X. Flask really

11:12 doesn't care much about the stack that you use. So, yes, Jinja comes integrated. There's a very

11:19 light integration. And that helps you with generating the HTML. So, you write your HTML in files and put

11:26 placeholders for the dynamic components. And then you generate those, you know, dynamic parts in your

11:31 function logic.

11:32 Right. And do you pass just like a dictionary off to the...

11:35 Correct. Yes. And then Jinja 2, that's the replacement.

11:38 Yeah. And then the view engine will put it together, right?

11:41 Yeah.

11:41 Exactly.

11:42 So, really, for a beginner, I mean, being able to get, you know, the Hello World thing with seven

11:47 lines and then start hacking it and make it do more things, most beginners can get it right away,

11:52 which is fantastic, right?

11:54 Yeah, it is quite easy. I'm also a fan of Pyramid. I like Pyramid a lot, but that step from zero to

12:01 kind of, I've got something I'm ready to start working with is much bigger than it is in Flask.

12:07 I think to a large degree because of the packaging, you've got to kind of deploy or set up your website

12:12 as a package.

12:13 Right. Honestly, I see Pyramid more closer to Django than to Flask.

12:20 Yeah. Yeah. I agree. I think Flask is much smaller, but I think one of the sort of superpowers of Flask is

12:27 it's super small, but there's so many things you can plug into it, including a number of your open

12:33 source projects, right?

12:34 Yes. I have written a number of things that you can add to Flask, which are all optional. And that's a

12:41 big thing, right? In Flask, as I said before, you pick your components. So, yes, I've written stuff to

12:47 do WebSockets in Flask. That's actually the one that my most popular extension so that you can do

12:52 WebSocket and does it in a way that is very Flask friendly, not the typical WebSocket thing that you

12:59 see. But basically, the WebSocket connection is represented by a function that runs forever,

13:04 or at least through the life of the connection. And in this extension that I built, you handle events.

13:11 And these are short events that are built on top of the Socket.io protocol.

13:16 Yeah. Okay. Very cool.

13:18 So, it has a good side and a bad side. The good side is that, you know, if you're familiar with

13:22 HTTP routes, it's very easy for you to start doing WebSocket because with this extension,

13:28 you do it more or less in the same way. The bad side is that, you know, they're not the same. So,

13:32 a lot of people get confused about, I mean, that they think that they can do whatever they can do

13:36 with HTTP requests. And the fact that the extension makes it simple doesn't mean that,

13:42 you know, you do have a request context, which you don't when you're doing WebSockets, right?

13:47 Yeah. Right. Interesting. So, that is something I wanted to ask you about while we're still down

13:51 in the details of Flask and so on. And frameworks like Pyramid, they will pass a request object that

13:58 has the request, the response, the headers, all that kind of stuff into your view methods. Or if

14:02 you're using Pyramid handlers, it'll pass it to your class and you store it for later. Those kinds

14:06 of things. Flask doesn't do that. It uses these sort of contexts that are ambient to your code. And

14:13 those are really cool. Can you talk about that a bit? Maybe how that works?

14:16 I can. So, this is possibly the only aspect of Flask that can be thought of somewhat magical.

14:23 It's only magical because a lot of people don't understand it. Once you understand it, of course,

14:27 there's no magic. But yes, the idea is that if Flask didn't do what it does with the request objects

14:33 and, you know, other variables that are request specific, you will have, for each of your handler

14:40 functions, you will have to receive a bunch of arguments. The request, and then you need

14:46 to return the response, but you may need the session, and sometimes you don't need them. So,

14:51 the way Flask handles that is that it makes those available as context variables. They're present

14:57 in every request. You don't get them as argument. You see them as global variables. And I said they're

15:03 magical because even though they look like global variables, they're not. They're really thread-specific

15:09 global variables. So, if you have a production web server running multiple threads, each thread will see,

15:15 look at this, for example, the request global variable, and it will see a different one. It

15:19 will see the one that applies to that thread. That's a little bit magic that a lot of people

15:24 get confused about. But other than that, basically, you know that when you get into a request handler,

15:29 you have access to the request as a variable called request. You have the session as a variable

15:35 called session. There's a couple more. And then extensions can add more things using the same

15:41 idea. So, for example, if you use Flask...

15:43 Do they maybe just pile on top of the same pieces?

15:46 Yes. They use the same...

15:48 Like the shared global context?

15:50 Correct. So, for example, Flask Login, which is an extension to do authentication,

15:54 has a current user. So, you know, that's always there if you are in a request. I think it's a very

16:00 clean way to have those things out of the way if you don't need them and then handy when you do.

16:05 I really like that.

16:06 Yeah.

16:07 I enjoy, you know, not having those as arguments in my functions.

16:10 Yeah. It's definitely nice and clean. Yeah. That's cool. So, let's take a moment and talk

16:15 about your blog. So, you said it starts by you writing your own blog engine because you're not

16:20 going to be ruled by one of these pre-built ones, right?

16:24 Correct. Yeah. That's not my way. So, I wrote the engine and, you know, okay, so I'm done. I have a

16:31 blog now and my wife was still saying, well, you're not blogging. So, I had to start writing.

16:37 There's no... It doesn't matter if it runs. There has to be a post on it, right?

16:40 Right. So, right there with all the Flask experience of building the blog in my mind,

16:46 what made more sense was to blog about Flask. So, I started writing articles about writing,

16:51 you know, a complete application, you know, using Flask and that was how this mega tutorial was born.

16:59 This is a 17-part, very detailed Flask tutorial. It's starting to age a little bit. It's been written...

17:07 I started maybe four years ago and I wrote it during a period of a year and a half.

17:12 So, yeah, that basically covers a bunch of areas. For example, I'm really proud that I have very good

17:19 coverage for internationalization. Not very many people talk about, so you can learn how to do that

17:25 with Flask. Ajax, which is more common now with single-page apps. That was basically a source of

17:32 not only a way to teach what I learned, but also to, I mean, for myself to learn, because it pushed me to

17:38 learn more things, right? I wanted to do something more... I didn't want it to look as a Flask beginner,

17:44 which I was at the time.

17:45 Yeah, that's so true. At the time, but then you go through teaching this stuff and like, well, if I'm going to talk

17:51 about this, I got to know all the little details. And maybe if you were going to write an app,

17:55 you might only learn just enough to get the app working. But if you're going to try to teach it,

18:00 you got to dig deeper.

18:01 Exactly. I always say that teaching is the best way to learn.

18:04 Yeah, it absolutely is. It absolutely is.

18:06 I learn by teaching.

18:07 Yeah, I think one of the ways that I think of it, sort of conceptualize it, is if you were,

18:12 say, a consultant and you went to work at a place and they say, you need to solve this problem.

18:16 If you know one way to solve that problem, that's probably good enough, right? You're ready for

18:21 your job. But if you're going to go teach other people and there happens to be three ways to solve

18:26 it, not only do you need to know the three ways, you need to know when to use which, what are the

18:30 trade-offs, all that. Those are the kind of details you would just not bother with unless you have to

18:34 teach it. So that's great.

18:35 Yeah, correct.

18:36 This episode is brought to you by Hired. Hired is a two-sided curated marketplace that connects

18:52 the world's knowledge workers to the best opportunities. Each offer you receive has

18:56 salary and equity presented right up front and you can view the offers to accept or reject them before

19:01 you even talk to the company. Typically, candidates receive five or more offers within the first week

19:06 and there are no obligations ever. Sounds awesome, doesn't it? Well, did I mention the signing bonus?

19:11 Everyone who accepts a job from Hired gets a $1,000 signing bonus. And as Talk Python listeners,

19:16 it gets way sweeter. Use the link Hired.com slash Talk Python to me and Hired will double the signing

19:22 bonus to $2,000. Opportunity is knocking. Visit Hired.com slash Talk Python to me and answer the call.

19:29 At the time at work, still in the C++ domain, we were starting to do some things with REST APIs.

19:41 I also got an interest sort of in parallel when I got tired of writing mega tutorial articles.

19:46 I will take a vacation from that and write about REST.

19:49 Right. So you basically have two really large areas on your blog. And one is this Flask mega

19:55 tutorial? And that'd be a good place for if my listeners are out there and like, hey, I'm getting

19:59 into Python. Maybe web apps are cool. Maybe Flask is cool. How should I get started? Your tutorial is a

20:04 good place?

20:04 I think it's a good place. There are warnings in the proper places in the tutorial where things

20:09 got a little bit out of date. I'm going to give you just one example. Back then, I thought using

20:14 OpenID for authentication was a cool idea. So OAuth wasn't that...

20:19 We were told OpenID was cool. We were told.

20:22 Yes, we were told. And it looked...

20:24 It was cool.

20:25 I went with that. And of course, today, it's kind of an old thing.

20:29 Yeah, that's right.

20:30 One day, I'm probably going to go refresh that portion of the article, that chapter. But yeah,

20:35 right now, you can still implement it. But of course, if you're doing anything for real,

20:39 you will not do what I did back then.

20:41 Sure. You could say... You can log with OpenID and your users would say, what's OpenID?

20:45 Exactly. Right. I mean, probably there's one or two providers that still support it. And

20:50 it's decreasing very quickly. So yeah.

20:53 Okay, cool. So the Flask tutorial sounds great. And then the other thing you focused on a lot

20:58 was REST APIs. Yeah?

21:01 Yeah. This was mind-blowing when I learned about REST. It seemed great. And I said, well, why

21:09 didn't I think of this? It seemed like a very clear way to organize and do the separation

21:14 of concerns between the client and the server and write more richer client applications. I

21:19 really made an effort to learn. If you go Google REST, you're going to find there's a lot of

21:24 discussions about what's RESTful, what's not RESTful, people who are very angry at others because

21:31 they're calling something REST and they think it's not.

21:35 So let me see if I got it right. If I have a service, like a website and it returns something,

21:40 so if it returns like JSON over HTTP, it's a REST service, right?

21:44 Wrong.

21:46 Just kidding.

21:47 Oh, you're teasing.

21:49 I'm going to start some kind of Rastafarian war.

21:52 Yeah, don't worry.

21:54 So the REST ideas are great, but even if you go read the paper by Roy Fielding, it seems like he's

22:02 saying simple things, but hidden behind those very sparse words, there are very complex ideas

22:09 that you really have to think and make sure you understand. I did try to get to the bottom of it.

22:15 I'm really not one of the angry people who think that you should call REST things that are 100% pure REST.

22:23 I think REST has ideas that are very good. Even if you implement some of them, you're still winning.

22:28 And I'm not going to complain if you call it REST, right? Because everybody does.

22:32 Yeah, absolutely. I think that REST as a black and white concept was much stronger several years ago.

22:39 And now people see it as a spectrum and they're like, you know, let's adopt HTTP, JSON sort of APIs

22:47 so we can talk to both iPhones, web apps, and other... It's almost a practical drive to adopt something like REST.

22:56 Yeah, I think that it's a collection of ideas and some ideas are very easy to grasp and, you know,

23:02 almost everybody say, yeah, this is great. The idea of having resources, right? Everybody gets that.

23:07 And how to use get post, put, and delete, that makes perfect sense. So almost everybody get that.

23:13 But if you start thinking about hypermedia, for example, which I think it has only five or six

23:19 words in the original paper by fielding, it's like it's mentioned in passing, right? I mean,

23:26 not everybody or almost nobody get what you gain by implementing hypermedia. And for many projects,

23:33 it doesn't make sense. Right. But for some, it does. So I wanted to understand it, even if I don't always

23:39 agree that it is a good idea, I wanted to understand what it meant and what you could do. And I even did

23:45 a one of my PyCon talks, the one I did last year, I showed, you know, a fully implemented hypermedia enabled

23:52 REST API, which was a pretty fun project to do. Another way that forced me to learn something,

23:59 you know, to keep learning. I had to implement that in a way that the few hundred people that

24:03 were there and then the people who watch it on YouTube won't put my fingers at me saying,

24:08 yeah, that's wrong.

24:08 Yeah, that's cool. And I have seen that talk. That was, that was a good one. And the title is,

24:13 "Is your REST API RESTful?" which I think is a clever title, right?

24:18 Right. Right. Yeah. I mean, it plays on this continuously ongoing discussion between the

24:23 people who want your REST to be called REST and, you know, the people who say, "Well, okay,

24:28 it's a REST API," whatever, right? I didn't implement all the items in the list, but still REST.

24:34 Yeah. So one of the principles of REST is to sort of embrace the HTTP-ness of the transport, right?

24:42 Like you said, use the verbs and the URIs and all those kinds of things, right?

24:46 This is very interesting because if you go read the paper that, you know, the fielding paper,

24:52 he almost briefly mentions HTTP and the message there is that you could implement REST over

25:00 other transports. So if you go by that document, there's nothing that says that you should use

25:06 GET requests for this or that or POST requests because that's HTTP specific. When you go and implement

25:13 REST, you're supposed to map the ideas of REST into the transport that you're using. So, you know,

25:19 all these rules that say that you should use... I think part of the reason that HTTP gets so tied

25:25 up with REST is it's the only practical choice for a broad reach on the internet. That's correct. So,

25:32 to my knowledge, nobody is doing REST over something that is not HTTP. I'm sort of a little bit disappointed

25:40 that fielding or one of his colleagues didn't write something more specific to HTTP because I think

25:48 that's the source of all the confusion. It's that mapping, how you map what fielding said in that

25:55 original paper to the HTTP protocol, which allows you to do the same thing in many different ways

26:00 ways in a lot of cases. Right. So... Yeah, it's unfortunate. I think you even mentioned this in

26:06 your talk that there's not like a reference implementation client server that you could go

26:10 look at, right? Right. And there is one because really it's very difficult to make something that

26:16 complies with every little thing that is mentioned in the original REST documentation. And some things,

26:22 you have to do the mapping, right? So they're not clearly specified so that you can find out how to

26:27 implement that with HTTP. So yeah, it's very interesting how the technology developed around

26:32 it. And I think REST took off because at least some parts of it, as I said before, are very easy to grasp,

26:39 and they're very powerful. Right. And if you look at the alternatives, they were like XML soap

26:44 web services with WS star specifications, making it even more complicated and all that kind of stuff, right?

26:50 Right. It's like, yeah, I mean, you have boilerplate and you need to learn a lot of things. I mean,

26:55 it's the same thing that goes on with FLASC. FLASC, you can pick up really quick and REST,

27:00 at least a portion of it, you can pick up really quick too.

27:02 Yeah, I think REST is one of those things you can pick up quickly, but it takes a long time to master.

27:07 Yes. If you wanted to get all the advantages of REST, definitely it takes a long time. I hope I did a

27:13 good job on that talk because I really tried to explain in a very simple language what each of the six

27:19 up characteristics of REST mean and how they benefit you.

27:23 I was still waiting for a picture on stateless.

27:24 Yes.

27:25 Because you had pictures with it, you're like, this one I can't do in a picture, it's stateless. What

27:29 does that look like?

27:30 I started with the idea of making pictures for everything, but yeah, no. It would have been pretty

27:37 hard to do stateless. Yeah.

27:38 Yeah, well, I didn't have an idea. But yeah, the ones that I did pictures, I think the pictures in

27:43 some of them, they do help drive the point home. I mean, especially caching and those things you

27:48 normally don't think about. Yeah.

27:50 And they are kind of important. I think they're important, especially if you're doing something

27:53 big, right? If you're doing a REST API for controlling your garage door on a Raspberry Pi,

28:00 you probably don't care about caching. Yeah.

28:04 But if you're getting paid to do a REST API, you should probably care about it.

28:08 Yeah, you definitely should. So all of these posts and articles sort of led you to your next

28:14 endeavor, which was your book, right? So you wrote all these articles and O'Reilly reached out to you

28:19 and said, "Hey, that stuff's cool. You want to write a book?" That's actually pretty much how it went. Yeah. This editor from O'Reilly, she was basically looking

28:29 for Python content and she ended up on my blog at the time. At some point, she came to Portland for

28:37 a conference and we met. And basically after that meeting, yeah, she said, "Well, send a proposal.

28:43 We really like to have a book on Flask." And it-

28:46 Nice. Yeah. And I looked through it so much. I haven't made it all the way through, but I really

28:52 like it. I think it has a really nice practical style. It's not too verbose. It doesn't go into too

28:58 many details, but it doesn't feel like you're skipping over so much. I think it's a really

29:02 nice book. Do you maybe want to kind of talk about the theme? And you have this one kind of

29:06 major app that you build up throughout this book, right? Yeah. So it's probably a typical

29:11 book for learning a framework. I decided to write it in a kind of a tutorial way. Basically,

29:20 each chapter builds on the previous one. And by the end of the book, you have a complete application,

29:25 which is sort of a Twitter kind of clone. I decided to do it that way because people learn,

29:32 in my experience, it's easier to learn, you know, gradually. And if each chapter starts with a new

29:40 project, you have, you know, you lose your context. So you have to sort of start over.

29:46 Doing it that way also allowed me to do something that I think it's also very original, which is how I

29:51 have the code for the book on GitHub. And people tell me I'm crazy because it's a lot of work. But

29:58 basically, what I did is I engineered the history. I mean, using get rebases, I engineered the history of

30:05 the repository to be, you know, one commit per hit point, basically, there are a few in each chapter. But

30:13 each time you reach a point where you can go test something, there's a commit there. If I need to make

30:18 a fix, I need to insert it in the right place in the history. So it's I developed a few tools to help me

30:24 with that so that I don't make mistakes. Right? Because it's crazy. I agree. I'm kind of crazy.

30:29 Well, it's really cool from a reader perspective. So basically, you start out the book saying,

30:35 look, you need to know get welcome to the 21st century. And you're a programmer. So you got to

30:39 know get. And here's how you get the code, right? You check this out, the you check out the repository,

30:45 but then you go to tags, right, that are basically save points. Correct. Yeah. So you have spread

30:51 through the book. At any point where you are ready to try something, it tells you, okay, now you will get

30:55 commit tag seven a, you know, they're named with the chapter number and then a letter. And so you get

31:02 that. And if you want to see what changed from the previous tag, you can do a diff, which is awesome.

31:08 Even on the GitHub website, you can diff one commit and see exactly that these are pretty small incremental

31:15 changes. So if you want to find out how I did authentication, for example, you find the commit

31:20 that does it and diff it against the previous one. And you see all the changes, not just, you know,

31:25 the big stuff, every little thing, including template changes, every little detail is there. So if you

31:31 want to learn with a lot of detail, what you need to do to add that feature, you can find it in the diff.

31:36 Yeah, I think it's a really, really nice way to do it instead of having a zip file with,

31:40 you know, chapter one, chapter two, chapter three in it, right?

31:43 Right. I started that way. And it seemed wrong. Right. And it's also difficult to maintain. So,

31:48 yeah. So, you know, overall, I mean, people criticize the book. Some people did so far in

31:53 the almost two years it's been out. Nobody criticized the the Git organization, right? The repository with

32:00 the code. Interesting. That, you know, it's almost unanimous. You know, people like it. They seem to like

32:05 it a lot. Yeah, yeah, that's really cool. And so, you know, as far as I made it, I really like the book.

32:09 It's tough to do stuff in public, out on the internet and other places like it. You can't

32:13 please everybody. So you just do the best you can. Right. Yeah, exactly. I mean, it's obviously

32:17 it's going to resonate with some people and it's not going to with some others. And that's fine. Right.

32:23 And these days, I mean, back in those days, this was the only book. But now it's not the only book.

32:28 There are like five or six or seven. So there's plenty of choice. It's a great ecosystem.

32:34 Yeah, it is a great ecosystem. Yeah. And like I was saying, I think that's some of the power of

32:38 a flask itself. It's kind of mirrored in the books. So your work on your blog led to your book. Your

32:43 work on your book led to some videos that you did. Right. So these as these things go, you do some

32:49 minor thing and it just snowballs over time if you keep at it in ways you don't expect, maybe. Right.

32:54 Yeah. In ways I did not expect. That's actually true. The book went out, did really well. So O'Reilly

32:58 was super happy with it. And they asked me to do webcasts, which I did a couple. But then I said,

33:05 well, maybe, you know, we should do more flask content. So they asked me to think about if I

33:10 wanted to do a video tutorial, basically a video version of the book. And that didn't seem right to

33:16 me. Right. Repeat the thing. I mean, I don't know. I was kind of like, I wrote it down. Why do I need

33:22 to do this? Yeah. I mean, it's kind of actually boring for me. Right. To do it all over again,

33:27 you know, in a video. Yeah. Sure. Instead, I decided to think about what else could I do in

33:32 video format. So one thing that some people found the book was too easy and some people found that the

33:38 book was too hard. Right. Both sets of people were right. Right. Because each is the same person.

33:43 It's a relative statement. That's right. Right. I decided to write a introductory flask course on video,

33:49 which started at a much lower. So the book is kind of for intermediate. You need to know your way in

33:55 Python to really take advantage of it. So this is much more basic. And the being on video, my screen,

34:04 you can see my screen as I do the exercises. For a lot of people that visual beginners find the visual

34:10 component very inviting. Right. That's how they avoid messing up. If they're reading it from a book,

34:15 Yeah. They don't see my terminal. They don't know what I type. So seeing, you know, the very basic steps

34:21 to build a web application being played in front of them. I thought it was a good idea. Yeah. I think

34:26 these video tutorials where you're doing screen sharing and especially when people are new, they can

34:31 watch. I think it's really, I feel almost like, you know, the blood pressure goes down, the stress goes

34:37 down. It's just like this thing that seemed hard. I saw, I saw that guy build it up and I saw all the

34:44 steps. None of them were hard. None of them were confusing. Exactly. Started from here, went to there,

34:49 and I can do that. Why did I think this was hard? But of course, if you're like running through the

34:53 docs and you don't know where to start, it's a whole different type of thing. So that sounds really cool,

34:57 really helpful.

34:59 SnapCI is a continuous delivery tool from ThoughtWorks that lets you reliably test and deploy your code

35:17 through multi-stage pipelines in the cloud without the hassle of managing hardware. Automate and visualize

35:23 your deployments with ease and make pushing to production an effortless item on your to-do list.

35:28 Snap also supports Docker and M browser debugging, and they integrate with AWS and Heroku. Thanks

35:35 SnapCI for sponsoring this episode by trying them with no obligation for 30 days by going to snap.ci/talkpython.

35:51 So that was video one? That was video one. Complete beginner with only a little experience in Python.

35:56 You will do that. That will give you the basis. You will know how to write simple applications with

36:01 Flask, and then you can go to the book. Hopefully at that time the book will make sense if it didn't before.

36:06 And then the second video that I did was specifically on REST APIs. And this is because when I was writing

36:12 the book, I was in a constant fight with my O'Reilly friendly fight, I should say. She's awesome. But a constant

36:19 fight regarding page count. They wanted a relatively small-ish book. And I tend to put a lot of detail. I

36:27 want to make sure things are clear, so I write a lot. So one of the things that suffered in the book was the

36:34 chapter on REST APIs. It's much smaller than I thought it should be. So I decided to expand on that on the video.

36:40 And because I wanted it also to be fun for me, that video has also APIs for Raspberry Pi. So I have my

36:48 Pi on screen showing you cool stuff you can do with the camera. It was a blast to do it. I enjoyed it very much.

36:55 Oh, that's really cool. Yeah. It's tough when you're starting out, or you're teaching people who are

37:00 starting out, to make something that's interesting and engaging. And so if you can do things like that,

37:05 like grab the Raspberry Pi and show how you're controlling the camera. And even though the code

37:09 might not be super complex, it feels like you're learning more than just a loop or a print statement or something.

37:14 Right. You bring up a good point. I don't think that video is for advanced users. It's probably

37:20 still in the intermediate level, but it is dedicated to REST APIs. I think that there's a lot more to say

37:27 about APIs than what you get from the book. So it was a good thing. And being able to show how to

37:32 implement APIs on devices like the Raspberry Pi, I think that was great. And that will be something

37:38 more difficult to do in book format, I think. I mean, especially playing with the camera. I was in part of the

37:43 video. I show how to do a time lapse. There's an API to trigger a time lapse. So I have the camera

37:48 pointed at myself doing the class, doing the course. And then I play the time lapse. It's pretty cool.

37:57 It was really fun. Yeah. Really nice. Really nice. So we have time for a few more questions. One thing I

38:04 did want to talk to you about while we were together is your open source projects. But I kind of feel like

38:11 maybe the way to do it is to give people a sense of what you've done, what's out there,

38:14 and maybe some other time we'll come talk in depth about them. Yeah. So I'll tell, I'll say,

38:20 I'll say one of your projects and you give me like the elevator pitch one to two sentences. Like,

38:24 what is this? How's that work? Sounds good. All right. Flask Socket IO. I briefly mentioned

38:29 that before. It allows you to do WebSocket applications with Flask. It makes it super easy. Right. That's the,

38:36 the continuously running function with events. Correct. Great. And then Flask Migrate.

38:39 Flask Migrate allows you to do database migrations using Alembic, which is the migration engine for

38:46 SQLAlchemy. So this extension makes it very, very easy to implement migrations in Flask. Okay.

38:52 Interesting. I'll have to check that out. Flask HTTP auth. For REST APIs to do authentication.

38:58 Again, decorator based. Like OAuth, that kind of stuff? This is not OAuth. This is a basic and digest

39:03 authentication for an API. So if you're doing a straight API, say you don't have a web application,

39:09 so you cannot do OAuth, which requires the web so that the user logs in. So you implement token-based

39:15 authentication, for example. This will give you decorators that basically prevent you from having

39:21 to deal with the headers 100%. So you can do everything in Python land. Right. You could deal with

39:26 with them yourself, but why would you want to, right? Exactly. Yeah, exactly. Yeah. Python Socket.io.

39:32 Python Socket.io is one of my two non-Flask specific projects. This is the engine to Flask Socket.io,

39:39 the one that I mentioned before. So Python Socket.io is a general purpose Socket.io server. So you can do

39:45 socket applications for any framework or just a plain, you know, no framework, just the Socket.io server.

39:53 Since we're talking about it, I would like to invite anyone who wishes to write a wrapper for

39:58 other frameworks. I maintain the one for Flask. But if there are any, anyone interesting in supporting

40:05 Python Socket.io in Django, Pyramid, Bottle, whatever else, I would be more than happy to help. I see. So

40:11 so that would be the core. That's the core, correct. That you would use to build that and not have to

40:16 basically maintain it. Yeah. Yeah. It has no relation with Flask. Yeah. Okay. Awesome. It's

40:20 Python pure framework. Sure. Okay. Flask Moment. Oh, Flask Moment. Flask Moment allows you to use Moment.js

40:28 on the client to render. For example, you can do this very cool three seconds ago, three minutes ago,

40:36 timestamps that are very friendly. With this extension, you can implement those in Jinja templates. So the

40:44 template generates the JavaScript code that then runs on the client to use Moment.js. Yeah, that's

40:49 really nice. And Moment.js is live, right? So like, if it said five seconds ago, then it would say six,

40:53 and then seven and so on. Is that right? Yeah, it's awesome. Yes. The way I implement it is, it's

40:59 optional. You can request not to do that. For some things, it doesn't make sense. But this extension will

41:04 allow you to do that. If you want to say, keep updating, you send refresh equals true in the template.

41:09 And that's all it takes. Yeah. And then it updates on its own. Yeah, that's really cool. So you render

41:14 the page, you look at it says 30 seconds ago, go grab a coffee. And then it says two and a half minutes

41:19 ago, right? It's beautiful. Yeah, it changes. It's really nice. Yeah, flask page down page down. So

41:25 if you use the stack overflow, the little editor where you write your, your answers, that automatically

41:33 updates below, you see the render representation, you type mark down, and you see it update immediately. So

41:39 this is an open source project. It's a JavaScript library. So flask page down allows you to use that

41:45 in your applications. And again, implement it, sort of, you think you're doing it in the server,

41:51 but you do it in the Jinja template. And all the extension does is generates the JavaScript that

41:56 then runs on the client. But at least you can have control of it from the server side. So that's what

42:02 it does. Yeah, okay. Yeah, that's really cool. It's coming to a climax with the open source

42:06 project. The last one is actually called climax.

42:08 climax is, this was a really fun one. I implemented it like in a weekend, two weeks ago, to be honest.

42:18 So I really like click, I mean, Runikers command line, decorator based command line parser. But

42:25 there's one thing that I don't like. Yeah, it's cool. And we talked about that on his show when he was on,

42:29 and we talked about his work with click. So, right. So it's another one of those things that I think,

42:34 damn, I should have thought of that, right? I hate the guy because he thought of it.

42:38 But anyway, one thing that I don't like is Armin's distaste of our parts, right? He doesn't like our

42:45 parts. And for some things that I'm doing, it will be useful for me to be able to integrate with existing

42:52 applications that use our parts. So I created climax as a clone. It's a blatant copy of Armin's click,

43:01 but as engine, it uses our parts. So that allows you, for example, you could take a tool like say,

43:09 Alembic, which uses an art parts, a parser for its command line, and import it into a bigger command

43:16 that you own, and offer, you know, Alembic's set of options as a subcommand, because it nicely

43:22 integrates and you can compose bigger parsers by taking, you know, our parts, parsers from different

43:27 sources. Okay, that's really cool. Yeah, yeah, very nice. Okay. So I'll try to have links to all

43:32 that stuff on and everything else we talked to on the episode page. Cool. Okay, so let me ask you two

43:39 more questions before we wrap it up. So last week or two weeks ago, you wrote climax. What editor did you

43:46 use to write it? I use Vim with the Python mode plugin. That's my always go to editor. Okay,

43:54 I'm gonna be honest, I know probably, I would say, five to six keys in Vim. And it makes me super

44:03 proactive, even though I know very little about it. I don't use any of the advanced things. But Python

44:08 mode makes it really cool. I also like PyCharm. When I need to debug something really complex, I tend to go

44:14 go to PyCharm, but more for the debugger than the editor. Yeah, okay, cool. Yeah, the debugger in

44:18 there is pretty sweet. So the other question is, there's 60,000 or however many packages on PyPI.

44:25 What ones would you recommend people check out that maybe they haven't heard of? So, you know,

44:30 they probably heard of requests, maybe Flask. But you know, what else? What else should people go check

44:35 out that is maybe not well known yet that you want to tell them about? One package that I really,

44:41 really like, speaking of debuggers, I started my career working on Windows long ago. So I'm used to

44:47 Visual Studio and the Visual Studio debugger, which is really, really good. So I'm used to interactive

44:52 debugging more than GDB or, you know, console type debugging. So there's a package called PUDB.

44:59 I'm not sure how popular it is. This is an interactive debugger, but it works in text mode.

45:05 Say you are debugging something on a remote server or maybe on an AWS instance, you know, something that

45:11 you can only get through a shell. You can pop this thing. So pip install PUDB. And then it works like

45:17 PDB, the command line Python debugger, but it brings up a GUI, text mode GUI. And it's really, really decent

45:24 to work. And it runs anywhere, right? Because it's text. I go to that debugger a lot when I'm

45:30 debugging, you know, remote sites. It's extremely useful. Yeah, that's cool. You're SSHed somewhere.

45:35 There you go, right? You don't want to install. There's no way really to reasonably install PyCharm out

45:39 there anyway, right? So... Right, exactly. You can't. So it doesn't work for that. I mean,

45:43 you may be able to set up remote debugging, but it's a hassle. If you need to debug something... Yeah,

45:48 exactly. Just put the PUDB package and, you know, you're good to go. It's, you know, super easy to use.

45:54 Yeah, okay. Oh, good recommendation. All right. So I guess, you know, people should go check out your book,

46:01 check out your blog. What else should they be up to now that they've heard all about what you're up to?

46:05 They're definitely invited to contact me, to interact, tweet me, questions. You know, I spend a lot of time

46:13 answering questions, whether it is Stack Overflow or Twitter or comments on my blog. Probably spend,

46:21 you know, on average, I'd say an hour a day responding to questions. And I enjoy that. That's another source,

46:27 another motivation for me to learn and stay, you know, current on things. I'm really a person that I'd

46:33 like to be engaged. And so I invite everyone, you know, to send me questions, problems, ideas.

46:39 Awesome.

46:39 Yeah. And yeah, we'll all learn together.

46:42 Yeah. Very cool. PyCon this year is coming to Portland.

46:46 Yay.

46:46 Your hometown, my hometown.

46:48 Yeah, I can still be my bad.

46:49 Yay.

46:50 I love that.

46:50 Are you going to make it?

46:51 Yes, definitely. I have my ticket.

46:53 That's awesome. Are you going to make it?

46:54 Yes. I'm going to be there. I think I sent four proposals. So we'll see what happens. I hope I get to talk.

46:59 Excellent. Yeah. Good luck with that.

47:01 I've been doing tutorials, the last two PyCons. And this year, I'm doing, I guess,

47:05 an advanced tutorial. I hope I get it. It's to address the question that there's a common

47:09 misconception that Flask does not scale.

47:12 Interesting. Okay.

47:13 I created a tutorial dedicated specifically to address all those concerns and show how to scale

47:18 Flask in three hours.

47:20 All right. That sounds really cool. Like a really good follow on to your other ones. So

47:23 that'll be great. I'll see you in Portland.

47:25 Mm-hmm. Yeah.

47:25 Thanks for being on the show.

47:26 Yeah. It was a pleasure. Thank you for inviting me.

47:29 This has been another episode of Talk Python to Me. Today's guest was Miguel Grinberg. And this

47:34 episode has been sponsored by Hired and SnapCI. Thank you guys for supporting the show. Hired

47:40 wants to help you find your next big thing. Visit Hired.com/Talk Python To Me to get five or more offers

47:46 with salary and equity presented right up front and a special listener signing bonus of $2,000.

47:51 SnapCI is modern continuous integration and delivery. Build, test, and deploy your code directly from

47:58 GitHub, all in your browser with debugging, Docker, and parallelism included. Try them for free at

48:03 snap.ci/talkpython. Do check out the video course I'm building. The Kickstarter is open until March 18th,

48:11 and you'll find all the details at talkpython.fm/course. You can find the links from this show at

48:17 talkpython.fm/episodes/show/48. Be sure to subscribe to the show. Open your favorite podcatcher and search

48:25 for Python. It should be right at the top. You can also find the iTunes and direct RSS feeds in the

48:30 footer of the website. Our theme music is Developers, Developers, Developers by Corey Smith, who goes by

48:36 Smix. You can hear the entire song on talkpython.fm. Once again, this is your host, Michael Kennedy.

48:42 Thank you so much for listening. Smix, take us out of here.