First, thank you to Bob Rudis and his webr code, especially https://github.com/hrbrmstr/slightly-more-than-minimal. Second, thanks to George Stagg for WebR. Thanks to Hadley Wickham and all the other authors of the tidyverse for dplyr. And thanks to the maintainers of blockly. I have done basically nothing but glue these tools together (poorly).
This is a proof of concept experiment. It's in no way well coded or even fully functional. Everything starts somewhere and this is where this starts.
Years ago it seemed clear that blockly's ability to connect blocks paired perfectly with dplyr's ability to pipe functions. I was working on the Verizon Data Breach Investigations Report and was looking for a way to make simple DBIR analysis available to everyone. Almost all DBIR analysis looks like:
dbir |>
dplyr::filter(plus.dbir_year==2022, subset.2022dbir) |> # yearly analysis subset
dplyr::filter(attribute.confidentiality.data_disclosure.Yes) |> # breaches
verisr::getenumCI("action", ci.method="bootstrap") # analysis queryThis pattern fits incredibly well with blockly, however at the time, it would have had to be tied to a back end R kernel and anyone able to get an R kernel up and running and connected to the browser, really didn't need blockly for the analysis.
Fast forward to 2023 and we now have WebR, an ASM implementation of R. With this we can bring functional data analysis in R to flks w/o needing to write a line of code.
The basic concept is to focus on the ability to create or open data, apply a chain of dplyr functions to it, and view the output. Maybe we'll add ggplot or vega visualization later.
Currently, we have a very small number of functions created. The current project is a combination of Bob Rudis' https://github.com/hrbrmstr/slightly-more-than-minimal WebR implementation and the codelabs blockly example app. It draws heavily on the example of adding a JSON generator to start adding an R generator.
At this initial starting point there are a very small number of R functions implemented and a very small number of the existing blockly blocks implemented. Just enough to provide the concept. Additionally nothing is particularly tested. Also, no attempts have been made at efficiency. Plus, the UI is still copy/paste from the parent projects with basically no clean-up. And I'm far more proficient with R than with JS. This is truly a starting point. You have been warned.
Hopefully I'll get the opportunity to continue to add blocks (both core and R) until it's far more robust, as well as improving the quality of the code and the user experience. Only time will tell.