IDE-ntity Crisis: Choosing the Right Tool for Me (Isabel Zimmerman, Posit)

Transcript#

This transcript was generated automatically and may contain errors.

So hello everyone in person and online. Hello as well and to all your cats and dogs on the couch with you. I am Isabel Zimmerman . I am a software engineer on the Positron team. But before that, I was a data scientist. And I kind of had this identity crisis going from data practitioner, you know, was I still a data practitioner? Like I make data science tools now, can I be both? Was I both?

Something I did know about myself, though, was that I strongly believe pretty much everything can be solved by a good checklist. So I made something like this. I wrote down the things I did as a data scientist and as a software engineer. And I kind of realized, you know, after all my check marks and everything, it's not the label that mattered. You know, the label doesn't make the job. It's the tools you use and the tasks that you're doing.

And there's something even more high level about this, you know, checklist. You know, it's using some sort of tool to gather a bunch of information and make a decision about it, either for me or for someone else. And that's what I did as a data scientist and that's kind of what I helped build as a tool maker. But my love of using some sort of tool to organize information to make a decision did not come from computers. It came from magazine quizzes.

You can tell this is younger me because there's no pink hair. So if you're not a Teen Vogue subscriber, these are quizzes. They're tools that you can use to solve the identity crises of your youth. Of my youth, I wanted to learn, you know, what band would I be a good lead singer for? Or maybe was I a risk taker? Or what pet should I get? So today, we're going to help determine the best fit for potentially the largest tool that you use every day as a data science practitioner.

And that's your IDE. So in honor of Teen Vogue subscribers everywhere, I'm going to help you find your IDE soul mate today. And the first question that we have to think about is maybe how interactive is your coding experience?

Interactivity and the console experience

So I do have like a true little quiz going on. If you want, you can go to the link at the bottom. You can take the QR code, whatever. You can remember how many A, B, or C you have. There's no right answers. This is just trying to figure out where you fit into this lifestyle. Or you can just come along for the ride.

So I went to school for data science. And in my research there, and as you know, my first roles as a data scientist, I had to think about, you know, was I going to be using Python or R scripts? Was I going to be using kind of a mix of scripts and a notebook and a console? And I started out by using a lot of notebooks. So I naturally gravitated towards JupyterLab.

Now, this is a fantastic, like, huge community of folks who are using these notebooks. It's really the standard for notebooks everywhere. And that means all of my data science packages, including the really complex ones that were like widgets and maybe like interactive plots, they all worked really well right out of the box in this JupyterLab experience. You know, with a primarily notebook workflow, this was exactly what I needed.

Now, I also sometimes wanted a place to do quick experiments that maybe wasn't a whole JupyterNotebook. And JupyterLab also has a console. And it's kind of interesting. There's two really big ways that you can have some sort of, like, console or repl, which is read, execute, print loop in Python. And one is having just a native Python experience. So this is just using the Python language. And there's also this other package called IPython or interactive Python. And you can kind of think about it as, like, JupyterNotebook minus the markdown. Really quick experiments. And the nice thing about IPython is that, once again, a lot of your data science packages will work right out of the box. So you don't have to do anything weird for setup or, like, have something else pop up in a different window if you've ever ran matplotlib in a native Python repl. It just kind of works in this repl.

And so this console is an IPython console. The thing about this console is it's kind of hard to get your code out of this experience into some sort of reproducible script. You can do maybe a little bit of copy pasting or some sort of export options. But it's just not a robust experience.

And I was starting to use less notebooks and more Python scripts as I was transitioning into being a package builder. So JupyterLab didn't quite feel right. The building blocks of a Python package are Python scripts, not notebooks. And so, you know, the notebook experience just wasn't quite for me anymore. My next step was to move to a different IDE called VS Code. And VS Code is a general purpose IDE. So you can do pretty much anything in it.

As long as you're willing to install one too many extensions. So I installed a Python extension and a Jupyter extension and, like, a Ruff extension and a Quarto extension. And all of these together made a good experience for me. I also still got to hold on to a little bit of this console REPL experience by using these hash percent percents. I could add these into my Python scripts. And then this interactive window spawns. And you can see I can, like, run my plot line code, see my output, all of that. And if the interactive window is sparking some connections, this is also an iPython console.

This is really interesting because I can use my Python script as, like, sort of a notebooky experience. But the thing is, I spent a little bit of time deleting these hash percents out of my Python scripts. So there was still a little bit of uncomfortability here. Even though it felt really good for most of my Python experience.

Now, VS Code does actually have the second type of console we talked about where you can use this native REPL. But because I build a lot of data science tools, I really need that interactive experience to have something that feels good for me.

And nowadays, I primarily build Positron. But I still spend a lot of time, I explore data, I get to teach others, like, today. I still maintain Python packages. And I have fallen in love with this Positron console. I can run everything line by line with command enter. This is an iPython console. So once again, all of my widgets and my plots, like Altair and plot 9, work right out of the box. And of course, there is Jupyter Notebook support in Positron. And there is Jupyter Notebook support in VS Code. But, you know, for my daily practice, those aren't the tools that are ergonomic for me. So I just love to hang out in this console. It has all the features I need. It's got, you know, really good syntax highlighting. It's got autocomplete. And there's even some, like, AI integrations to help me fix errors and write some code.

And nowadays, I primarily build Positron. But I still spend a lot of time, I explore data, I get to teach others, like, today. And I have fallen in love with this Positron console.

But when it comes to IDEs, people feel like they can only have one. IDEs are tools just like packages. All of these IDEs are free even for commercial use, same as RStudio, so there's no harm in trying things out and seeing where they might fit into your lifestyle.

For me, I use Positron whenever I'm writing documentation or Quarto, I use it for all my package management now when I'm writing Python code, but for Positron work, you know, sometimes I'm writing TypeScript or other languages, and for that, I might pick up VS code, and it feels like a right fit for me. I also would not be a Posit employee if I did not tell you these are all available on Posit Workbench, so you can even not have to convince any of your team members to make a swap. So I'll say thank you all for joining me today. I hope you've learned a little bit about all of these different IDEs, and I hope that you all have found your IDE soul mate. Thank you.