Max Kuhn -SHINYLIVE IS SO EASY

Transcript#

This transcript was generated automatically and may contain errors.

All right, our next speaker is Representing the year 2019 and he's also spoke at 15, 16, 18, 20, 21, 22, 23 And I always put him toward the very end to make sure people stick around because he's awesome All right, you thought I was gonna insult him a little bit. No, that was awesome because he's awesome guy And he's been held at gunpoint twice But he's never been as scared as when he was teaching his son to drive Please everyone a really big welcome for Max. Hello. You okay?

Great so maybe for the first time here. I'm not talking about modeling. It's a new experience for me Hopefully for you too. I'm gonna talk about shinylive Has anybody ever heard of that yet? Okay, good. All right

It does almost all the heavy lifting for you and we'll see the almost in just a second

So just to clarify we kind of seem obvious at this point shiny server all the things happen on the server You're just sending messages back and forth and it's serving you images and and whatever it is that you're looking for in your output Shinylive downloads R in packages runs everything locally

Setting up shinylive with Quarto

Now if you want to get set up with Quarto the first thing you have to do is go to a terminal inside of your Quarto project and run this command and Essentially Quarto has a bunch of extensions you can use for various things like if you want to use like font awesome there's an extension for that and so on but the shinylive extension does is gives you all the infrastructure to run Shiny in the way that we're going to do and then one more thing is the way Quarto runs is Everything is done in markdown. So when you like When you process your files, you get a markdown file at the end of that and then we use Pandoc to basically convert it to HTML or TeX or whatever it is that you're gonna like render it in like your target Output format. And so what you have to do is you have to sort of shim that process and so a Quarto filter Is something that will basically not interrupt but sort of get in the inside of that process of making the markdown Converting to Pandoc and then compiling it into the final format And so you just have somewhere in your Quarto yaml file You have to have a filters argument and just tell it to treat shinylive a little bit more differently

And after that you're done you can start writing code chunks So in code chunks and R at least if you're used to like knitter and R markdown You would just use like bracket R bracket and that's telling it that you know I have an R chunk instead of doing that you shiny live dash R and I should also say That everything I'm telling you about R is also true about Python. So we have Shiny for Python We have shinylive Python was the original sort of like first version of that and so on Of course Quarto works with Python So if you're more of a Python person all this will also be true for you

One other thing you have to do is you have to declare this one option called standalone and set it to true which makes it Work inside of Quarto and then just put your Shiny app in so it's it's pretty it's pretty simple Like if you're used to writing Shiny and you just do the setup then now you have Shiny locally, which is kind of amazing

Declaring packages and loading data

How do you declare packages so you're gonna use various packages like in that that you map example I have like ggplot like one of the color palette packages as well as dplyr loading in there. And so Shinylive in webR need to know what you need so it can go out and get them And so the way you would do that is it uses renv if you've ever used that so You just say like library dplyr and it knows to go out and get it now There might be we do this in tidymodels a lot You might have packages you went there, but you're not going to directly load attach them You want to call them by namespace or you know, it's an imported package for you and so what you can do in that case is you can just declare it as a library like you'd normally would just comment at That out and then renv will pick that up

So right now what it does is when you build when you render your your Shiny app It goes out it downloads R and then it goes to the the wasm code repository and gets The current version of that package and that's what you use, but we think a lot about reproducibility So we're continually George we here's George We're continually working on making that happen more. So the there's I don't think it's been committed yet But he was telling me what they're gonna do is at the time that you build the app not the render time for the build Time that they'll go out to the wasm repository get the versions that are like the most recent versions then when you built it and then sort of like either cache them or Or he'll just keep keep he'll retain all the old versions all the binary versions these packages So you can get probably get the package version that you started with Originally or something that's pretty close to it and I imagine this is gonna get more sophisticated as time goes on because there's a lot we could probably do with renv to get Around this like a like a lock file or snapshot But we're not quite there yet

All right, so how do you get your code in like if you're not just gonna if you have like a long Shiny app You just don't want to paste that into some Quarto doc Probably and how do you get data in like you might want to upload a CSV file or let the users do that and analyze? It and this is where it gets like the almost this is where it gets interesting There's really one simple simple principle that you have to remember, which is this

This you're in JavaScript actually your own WebAssembly world now But you know that you're sort of constrained a little bit So getting data in is a little bit more difficult because if you've used Quarto at least when I started using this I was like well if I have a code chunk that makes some data frame or some model object and I go to do the Next code chunk. I have that just laying around just go ahead and use that right doesn't work like that It's basically spawning another R process inside the browser that is basically clean So that's why we list the packages we just don't Inherit the ones we've already loaded or attached. So you have to do a little bit extra a little bit differently

Techniques to load data or source files and things like that. So WebAssembly doesn't really let you just like open a network connection So you can see if you want to learn more because I don't understand this part You can go to the github issue around this but curl is definitely a no-no. So if you've been using curl That's not gonna be allowed and what George has done is he's patched download dot file and base R so what we'll use Basically a separate protocol to get the files and that works pretty well For me, it's a little like more simple than maybe what you're gonna deal with because I keep all my stuff on github in a public Repo, so you'll see how I do in a minute But but this is the only thing that might trip you up a little bit in terms of like well How do I get my data in in a way that shinylive and WebAssembly will permit and again? If you want to read more there's an issue here that you could upvote or get feedback on

Security considerations

Now Gordon Shotwell who's in the Shiny group like succinctly put he's like I'm terrified of this in a way Because we're gonna have people that are going to you know Not realize that anything data or code that's in their Shiny app is available locally to whoever has the URL, right? Yeah, so he's like And so any data you have goes to the client any code you have goes to the client if you Accidentally maybe have an object that has some API key that you need for the app. It's available to the client So you're happy you like super careful about what you put in the app even if it's like you're giving somebody access to upload CSV files to analyze or something like that or Let them connect to a database. You just have to be cognizant that anything that gets in there Theoretically could be used by somebody else or accessed by somebody else

Not realize that anything data or code that's in their Shiny app is available locally to whoever has the URL, right? And so any data you have goes to the client any code you have goes to the client if you Accidentally maybe have an object that has some API key that you need for the app. It's available to the client