Thank you. Okay. Um, I'm going to talk about Streamlit and Python and Pythony stuff.

How many Python programmers do we have in the room? Okay, good number. Um, how many people are happy with pandas?

Smaller number. I might spend a bit of time on pandas. Okay, cool. This is my one slide. This is all I need because this is going to be all live coding.

I'm going to write some software. Um, 2025. This may be the last time you see someone write software by hand as a human.

So, let's enjoy it while we can. You were here that day before it all fell down and everything got written by the robots, right?

Um, there's one thing I have to tell you before we get going.

So, I'm going to talk about Streamlit. Streamlit's an open-source project, and it's one of those open-source projects that had like a company saying, "We'll do the open source project, but if you want to host it, we'll make that easy." And that's the commercial thing that supports the open-source thing.

They got bought by Snowflake, and slightly to my surprise, I started working for Snowflake last week.

So there is a connection, a financial connection between me and Snowflake, which I disclose, but let's talk about data.

So, Streamlit, why is Streamlit interesting? If you've got some data, and you all have some data, and you want to stick it in front of my mum or like users, end users, the public, you kind of know your options.

there. You're probably going to end up building a JSON API and like sticking it out to a React front end. You spin up a React team or maybe Swift and you're doing an iOS or something.

There's a well-trodden path to taking data to end users and the public.

I think there's a much less well-explored design space for I'm a programmer or a data scientist.

I've got some data. I just want to see it and maybe show it to some colleagues and maybe put it on my company intranet and that space from can I just see this right now really quickly up to one concurrent user a second, that is an underserved design space and it's the design space that Streamlit is trying to get at and I think it does it very well, makes it really easy to say just show me the data.

That's my pitch for Streamlit, but I'm not going to talk about it more. I'm going to show you and hopefully by the end of this you'll say the next time you've got some data you want to look at, you'll say, I know a fast way to do this and it's not hard.

So, I always like to begin live coding demos. Can everyone see that well enough? If you're at the back, you might want to move forwards.

If you're at the back and you're shortsighted, you definitely want to move forwards.

Let's do a bit larger. I like to begin my live coding demos by making an empty directory and going into it.

because that way, you know, there's no trickery. There's nothing up my sleeve or pre-prepared.

Um, let's do, actually, let's do this. Let's do that. That'll be a bit easier, right?

Okay. So, um, Python, we need a new Python package. Who's started using UV? Not many people. Okay, so you've used pip and you've probably used virtual env if you're a Python person.

Package manager plus a package isolator. I recently started using UV, which does both of those.

It's quite nice. The thing that it's brilliant for, you can do UV init and create a new project and you get a few things for free.

The thing that it's really terrific for and why I like it is if I say UV add streamlit, it does all the project things you want, but it also does that thing where whenever you download a package, it puts it in a central um, directory on your laptop and then the next time you want that package, it just goes straight to the cache, which means as a speaker starting new projects and showing people, you don't need to rely on the conference Wi-Fi.

It just downloads the package instantly. This is good. So I now have um, a Streamlit project running.

If I do source uh, it also creates a virtual environment for me automatically. So if I do source VM bin activate, I can then call Streamlit run my main function and that opens a browser for me.

Not very interesting at the moment, but can everyone see that?

Got my terminal. I've got my browser. I've got live reloading all ready to go.

So, let's use some Streamlit. Give me another window. Uh oh, I've got some old stuff going there.

And I'm going to edit that main file. Yeah, there we go. Right. So, that's a default UVI project.

Let's do some Streamlit with it. So I will import streamlit as st is the convention.

At the moment all this uh UVI script does is print to the command line. I need it to print to a web page. So I'm going to use our first typography part of Streamlit, st.title.

Right. So there's some typography options. Uh when I save that, the web page notices that I've made a file change.

I'm going to select always rerun. So for the rest of this demo, whenever I save the file, that will reload.

So we got titles. Let me check that's working by saying hello to.

Right. Cool. I'm a web developer now. Brilliant. Um you've got a few typography options.

So you can say header. Um how is it going? Yeah. Um there's also subheader, I think.

Yeah, you've got a few options. Um, and you can do st.text for the regular stuff.

Okay. Um, what else can I show you that's interesting typography-wise? There's this thing in Streamlit which I find a bit of magic, and I'm not sure how you feel about magic in code, but you're allowed to say for some expressions if it realizes the expression is something it could display to the screen easily, it will just display it.

So you can have kind of a naked string there and when you save it, it will show it to the screen for just a string.

The idea there is that you can start to do a literate coding style where the source code closely matches what the final output will look like.

You might like it, you might not, up to you, but it's there.

Um, it doesn't just work with strings, it works with a few other things.

So I could say 1 + 2 + 3 + 4 + 5 + 6, and when I save that, let's make it a bit larger, I get the result 21.

So it works with a few things like strings, numbers.

If you want to do that literate coding style, there's an `st`, there's a really nice little utility called `st.echo`, and if you say with `st.echo`, this will let you show what you're about to evaluate and show the result.

So you can actually do that kind of literate coding.

Here's the expression. Here's what it means. That, um, if you're anything like me, you may be wondering if it crashes when you do something like that, right?

Can you do the inception thing? And yeah, it works, right?

So, that's good. Is there anything else typography-wise I'd like to show you? Um, there are a few widgets like you can do a badge and you get like a nice little thing.

There are a handful of them. Go and check the docs, but there's lots of typography stuff you might want to use.

Ah, and you get a horizontal rule. That will do for just basic layout. Set me up a web page. So, let's draw a line under that.

Um, uh, call that typography. And I'm going to check that's working. And I haven't broken anything yet. And I'm going to comment that out.

Okay. So we can build a basic page with some words on it. Brilliant base camp. That will do.

What we really want to do is start visualizing data and manipulating it. So you could connect to a database for this or you know, um, a million and one ways to get data.

You go to another API, but I'm assuming probably in Python you're going to use, at least for this demo, I'm going to use Pandas.

I'm going to pick Pandas as my way of reading and manipulating data.

So if I import pandas as pd, as is the convention, Pandas has a bunch of utilities um for reading in data. I'm going to use pandas read CSV, right?

And what data should we look at? I needed a dataset. I know I'm also in a couple of weeks I've got to give a talk to some public sector workers in the UK, UK government, right?

So I need something serious and UK government-y, and I need something Copenhagen appropriate.

I thought I'd choose the dataset from the single most important transnational diplomatic project of the 20th century, the Eurovision Song Contest.

Right. Yes. Yes. So, let me copy my, um, sorry, my votes, uh, to here.

And now I can read the CSV file votes.csv. Right? Pandas returns a data frame. What it calls a data frame. If you're not used, if you're not used to pandas, think of it as like a virtual table.

And, um, that returns a data frame. By convention, that's often called DF. You'll see that in the docs if you start using it.

I'm going to call it votes because I like proper names. Okay. So, if I save that. Good. Remember I said that, um, Streamlit will see some expressions as things it can display.

If I just have naked the data frame votes.

Now we have actual some data visualization. So, so far the actual Streamlit part of this to get a table on the screen where everyone who isn't a programmer can see it is basically two lines and you get you get the kind of widgets you'd expect from a table.

Uh, you can download it as CSV, which people who like Excel will like. Um, I won't, I won't labor that.

So, we have a big old table. One interesting thing that Streamlit will also display is a data frame's shape property.

So we can see that we've got 51,000 odd rows and nine columns. Let's pretend I'm a data scientist and manipulate some of this data.

What would I like to do with this data? The first thing is, um, this is a round column which mixes uh we hang on there we go the round can be the final or the semifinals or whatever.

I'm not that fussed about Eurovision semi-finals.

I'm not that big a fan, so I'm just going to look at the final data. How do we filter a pandas data frame for just the final?

And I say this for the people that don't use pandas. I personally find it a bit weird. I kind of expected it to look a bit more like SQL and it doesn't and I can never 100% get past that.

You may love pandas and think SQL looks weird. But if you want to filter, you need to say, okay, take the votes and look at the round column and tell me, have a look when that is equal to final.

If you do that, you basically get a bit mask.

You get 51,000 true or false values, right? You can use that bit mask to filter out the votes table and you will just end up with the votes where round is equal to final.

And then that returns a new data frame, but I don't want to.

So I'm going to reassign it to votes. And that is how you do a WHERE clause in pandas.

That which I still feel is a little bit odd, but I've got to get past that.

Okay, so we can filter for just the finals. Um, I'm going to do a bit more data manipulation, then we'll go back to real Streamlit.

But as so often with real world data, you find it's not quite in the shape you want it.

And when you can see that thing like staring you in the face, you know what you need to do with it.

So a problem I have with this data set is sometimes it has total points set and these ones not set, and sometimes it's different because Eurovision Song Contest for years, they used to just go around a panel of music experts around Europe and ask them to vote on their favorite songs, and they would give it a score from 1 to 12.

And then in about 2015, they said, let's do a telephone vote where the public can vote for their favorite songs too.

This may be the democratization of the Eurovision Song Contest, or it may just be an enormous money spinner because those premium rate phone lines bring in a lot of cash.

But we end up with this data set where kind of annoyingly, the jury points is sometimes set after 2015, but before then it's just in the total points.

And if you try and compare pre-2015 with post-2015, post-2015, the number of possible votes doubles, and it gets really awkward.

So, we need to fix that. And it's actually not that hard. What I'm going to do is say, give me the votes jury points, right?

That will sometimes be none. So, fill it in where it's none with the value of the votes total points, right?

And reassign that to votes. And if I do that, I end up with just a table of the jury points for each year.

Not quite what I want because I've clobbered the rest of the um table. So I need to assign that to I need to reassign that to jury points.

And now we have the shape of data I want. Feelings about pandas, anyone? I like it, but I do find it a bit strange.

Okay, now I've got some data I actually want to do something with. What shall I do with it?

It would be I think it'd be quite interesting to different countries vote for each other, right? I think it'd be interesting to look at the country voting patterns and say who's voting for who when it matters.

Building Data Visualisations in Python in Minutes • Kris Jenkins • GOTO 2025

GOTO gathers the brightest minds in the software community to help developers tackle projects today, plan for tomorrow and create a better future.

Learn from thought leaders and innovators with top-rated videos released daily, and at our year-round conferences, masterclasses and meetups.

Get started: https://gotopia.tech

GOTO is produced by Trifork, a leading full-service supplier of high-quality, custom-built applications and end-to-end solutions.

GOTO YouTube Channel Code of Conduct: https://blog.gotocon.com/yt-coc


Building Data Visualisations in Python in Minutes • Kris Jenkins • GOTO 2025

Scripts