Skip to main content

Pretty Good Hat

Tag: code

I learned today that one of the magical things pandoc can do under the hood for Quarto is use a Powerpoint template to render slides. I can work in RStudio to make slides in my company’s style; this is amazing!

(Footnote: This assumes a template that uses at least some standard powerpoint slide notation, which my company’s official templates do … not quite do. But a little bit of revision of the slide master got me there!)

I’m really taken by this line in Paul Ford’s newest essay at Wired:

What I’m going to work on, for the rest of my career in the tech industry, hand to God (OK, I’m an atheist and easily distracted, so caveat lector), is making nice little tutorials and tools—better sticks for kinder monkeys.

“better sticks for kinder monkeys” is such an admirable, and needed, call to center empathy and humanity in what we make.

Two windows from the RStudio profvis tool, showing the times of several procesess. The second window shows the same process being completed dramatically more quickly than the first.

I used the RStudio tool profvis this weekend to find speed improvement opportunities in Armorer. I suspected that I could rewrite a big operation that calculates the maximum of many columns across several thousand rows. Holy smokes: Using matrixStats::rowMaxs cuts processing time by an amazing amount!

Screenshot of a web application showing two items selected from a form field. They are circled in bright purple marker with an arrow pointing to them to draw the eye.

I finished a big update to Armorer this week, to enable inclusion of subclass fragments in stat calculations. I learned a ton with this release and laid good groundwork for additional mod management. I’m pretty pleased!

I’m happy to have improved some R code today using group_split, replacing my prior method of cycling through a data frame to build several new subsets. I’m not sure it’s more efficient given the size of the data set, but it sure reads more nicely.

Super-pleased to see that a couple of small Shiny app improvements I made this weekend correctly picked up on some source data changes today and automatically handled them. Pretty cool!

Having seen how easy it was to convert a couple of hobby projects to use {pins}, I’m now daydreaming about putting it at the center of an S3 data lake operation.

Sharing a quick tip that I’ve found useful while building with Shiny, recently: It’s handy to be able to save off the current state of a data set for bringing over to a scratch file. I made the download link appear only when running in my local environment. This way I can easily snapshot my in process data set for experimenting with visualization in my scratch file.

In ui.R:

    if(interactive()) {
      downloadLink("downloadData", "💾️ Download data set") 
    }

And in server.R:

    output$downloadData <- downloadHandler(
      filename = function() {
        paste("shiny-export_", Sys.Date(), ".csv", sep="")
      },
      content = function(file) {
        write.csv(globaldata$armor, file)
      }
    )