Plotly: LaTeX Typesetting in R doesn't work

[statist-bhfz]

I can't reproduce an example from https://plot.ly/r/LaTeX/ with latest versions of R, RStudio and all packages on Windows (including development version of plotly). It looks like something wrong with MathJax support in the current version of package.
I have provided some screenshots in my question on Russian StackOverflow: http://ru.stackoverflow.com/questions/478133/latex-%D0%B2-plotly-r/

[artemklevtsov]

Should be fixed if add mathjax library to the package and htmlwidgets config.

[cpsievert]

Mathjax is too big to bundle with the R package, so we're looking at a situation similar to #302, where we want to conditionally download/include relevant stuff under https://github.com/plotly/plotly.js/tree/master/dist before rendering

The thing is, with the current interface, there is no way to detect whether MathJax is needed. Maybe we provide a simple function, say

tex <- function(x) {
  structure(x, class = "LaTeX")
}

which could then be used essentially as a flag to include MathJax:

plot_ly(x = c(1, 2, 3, 4), y = c(1, 4, 9, 16),
        name = tex("\\alpha_{1c}"))

Any thoughts @chriddyp @cldougl @etpinard?

[cpsievert]

@timelyportfolio a while ago I tried to resolve this in #411. I'm sure you can come up with a better approach :)

[timelyportfolio]

Would it be reasonable to expect RStudio and rmarkdown availability? If so, perhaps we could leverage the rmarkdown:::pandoc_mathjax_local_path() function or just educate users on mathjax: local option in html_document. Then we wouldn't have to worry about doing all that RStudio does (see lines).

So, this works for me

``````

title: "MathJax Test"
runtime: shiny
output:
html_document:

mathjax: local

$$
\frac{d}{dx}\left( \int_{0}^{x} f(u)\,du\right)=f(x).
$$

```{r, echo=FALSE} 
sliderInput("x", label = "Value:",  min = 1, max = 10, value = 5)

renderText({
   input$x
})
```

``````

To confirm local, here is a screenshot of the network request.

Then the trick becomes how do we know to add mathjax if not in rmarkdown. We could employ a method like crosstalk to add dependencies with a add_mathjax() function.

[cpsievert]

That would certainly be nice if we could let rmarkdown do its thing in Rmd docs, but unfortunately, plotly.js doesn't seem play nicely.

``````

title: "MathJax Test"
output:
html_document:
self_contained: false

mathjax: local

$$
\frac{d}{dx}\left( \int_{0}^{x} f(u)\,du\right)=f(x).
$$

```{r, message = FALSE}
library(plotly)
plot_ly(x = 1:2, y = 1:2) %>%
layout(xaxis = list(title = "$x$"))

``````

> Would it be reasonable to expect RStudio and rmarkdown availability?

I wouldn't mind adding an rmarkdown dependency if need be, but I know there will be Jupyter/IPython/non-RStudio users that will want to generate static HTML that can be viewed locally.

We also need some way of including geoassets (#356), for local preview, [if a geo or choropleth trace type is used](https://github.com/ropensci/plotly/pull/411/files#diff-cbb7235afee10c3646161358a7655b0aR286). Unfortunately, due to the size of the external files, I think the best way to go about this is to require a clone of plotly.js, or at least a path to [its distributed files](https://github.com/plotly/plotly.js/tree/master/dist), and add them add dependencies when need be.

I was also thinking that, similar to rmarkdown, users should be able to specify whether or not they want to use local MathJax or the cdn, so something like:

``` r
plot_ly(x = 1:2, y = 1:2) %>%
  layout(xaxis = list(title = "$x$")) %>%
  config(mathjax = "local")

versus

plot_ly(x = 1:2, y = 1:2) %>%
  layout(xaxis = list(title = "$x$")) %>%
  config(mathjax = "cdn")

[baptiste]

FWIW KaTeX may be a good alternative to Mathjax,

https://khan.github.io/KaTeX/

(one of its strengths is that it's light, self-contained and easy to bundle)

[dswatson]

Any updates on this? I'm a big fan of ggplot2 and plotly, but it's a real shame to lose figure titles, axis labels, and/or legends that use plotmath expressions when attempting to add interactive features with ggplotly. Workaround hacks very welcome.

[jackparmer]

Similar discussion and solution proposal for the offline Python library:
https://github.com/plotly/plotly.py/issues/515

[cpsievert]

I'm likely going to create another R package, which will make easy to control plotly.js dependencies tied to R plotly objects, which will hopefully solve this and #356. The idea is to include just what you need from dist/ (and I might even do minimized bundles for each trace type to help minimize the file size)

[jackparmer]

Any companies out there interested in sponsoring this one?
https://plot.ly/products/consulting-and-oem/

@baptiste we looked at KaTeX, but there's still a lot of MathJax functionality that would need to be filled-in. Unfortunately it looks like the project is largely inactive right now.