experiment for JSX

[timelyportfolio]

This pull is solely intended as a point of discussion. It would be really nice (but probably an edge case) to be able to produce JSX. In JSX attributes are often unquoted JavaScript. To achieve this in a naive way with ugly code, I made the changes in this pull. I am willing to work much harder on this, but I want to make sure it is acceptable before I expend the effort.

Example

I am working on the reactR package that would ease working with React/JSX in R. As I was experimenting, since tags does not allow unquoted attributes, I currently have to do something like this.

library(htmltools)
library(reactR)

attachDependencies(
  browsable(
    tagList(
      tags$head(tags$script(src="//d3js.org/d3.v3.min.js")),
      tags$body(tags$script(babel_transform(
'
ReactDOM.render(
  <MicroBarChart data={[1,5,2,4]} hoverColor="rgb(161,130,214)"
  fillColor="rgb(210,193,237)" />,
  document.body
)
'
      )))
    )
  ),
  list(
    html_dependency_react(),
    htmlDependency(
      name="microbarchart",
      version="0.1",
      src=file.path(getwd(),"src"),
      script="react-micro-bar-chart.js"
    )
  )
)

It would be really nice since I am so spoiled by htmltools to do this instead.

htmltools::tag("MicroBarChart", list(data=noquote('{[1,5,2,4]}'), hoverColor="rgb(161,130,214)", fillColor="rgb(210,193,237)"))

which after this pull gives us this valid JSX.

<MicroBarChart data={[1,5,2,4]} hoverColor="rgb(161,130,214)" fillColor="rgb(210,193,237)"></MicroBarChart>

[timelyportfolio]

ping to see if there are any thoughts regarding this idea

[jcheng5]

Sorry this slipped through the net the first time. I like it. I'd maybe change the function/class name from noquote to something more unique, like rawattrib or something? noquote sounds like something out of the rlang package.

I'd also refactor the code slightly so we test for this case before testing for !is.na(attribValue), so you don't have to repeat the test twice.

@wch, any thoughts?

[timelyportfolio]

@jcheng5 no problem at all, I had considered it a long shot anyways but it sounds like maybe there is more chance than I thought. I chose noquote since it is built-in to R core, it implies what a user wants (in a different context though), and means we don't have to manage a new class or object. If you think another way of handling is better or more intuitive, I am happy to change. Also, makes sense on refactoring and will change after @wch comments.

Thanks!!!!!!!

[timelyportfolio]

any thoughts no this especially now that reactR is more viable for htmlwidgets and Shiny inputs?

[timelyportfolio]

one last effort - any hope?

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[cpsievert]

I think I agree with @jcheng5 on this one. It seems safer to introduce a new rawattrib() (or similar) class to avoid the chance of breaking any existing code. I think we'd be willing to take it on with that in mind (we should also do the check once and include unit test(s))

[jcheng5]

How about HTML(), as suggested in #17?

[timelyportfolio]

@cpsievert @jcheng5 thanks so much for the consideration. Given the lack of any other user expression of interest or desire, I think better to just close this out and not try to squeeze JSX into HTML.

Really appreciate all the effort on htmltools.