Is there a way to compile interactive tables into an HTML book in Julia?

Tooling
1

My lab uses HTML notebooks to communicate information to stakeholders.

I am curious about migrating my workflow from R to Julia but am having trouble replicating an interactive table that I can include in the book like DT + bookdown in R or itables + jupyter-book in Python.

I’ve tried using ipynbs and making an R or Python call via RCall or PyCall just for the corresponding table function I need but the table isn’t displayed (I am assuming there a is a type issue here). What I tried:

PyCall


py"""

# example from the docs
import itables 
import pandas as pd
import numpy as np

df = pd.DataFrame({
    'cups_of_coffee': [0, 1, 2, 3, 4, 5, 6, 7, 8, 9],
    'productivity': [2, 5, 6, 8, 9, 8, 0, 1, 0, -1]
})

itables.show(df)

"""py

This was probably a naive experiment and doesn’t return or display anything in the book.

RCall

using RCall

R"

library(DT)
x <- datatable(mtcars)
"
# this gives me an ordered dict of the dataset (fair enough!)
@rget x

TableView.jl

I’ve also tried TableView.jl which works great when I’m using the live Jupyter Lab interface…

using RDatasets
using TableView

# great!
mtcars = dataset("datasets", "mtcars");

…but doesn’t appear in the finished book when I compile to jupyter-book. It gives an error about not detecting the required extension in the environment (again, only when building the book):

WebIO not detected.

Please read the troubleshooting guide for more information on how to resolve this issue.

https://juliagizmos.github.io/WebIO.jl/latest/troubleshooting/not-detected/

This suggests perhaps I could resolve the issue by including the extension in the compilation environment somehow, but before I spent a bunch more time on it I wanted to make sure I hadn’t overlooked an existing solution.

Are there solutions I’m missing, and, if not, what are some steps I might try to get an interactive table compiled into an HTML book?

EDIT: typo and…
NOTE: the issue with TableView.jl happens when I use jupyter-book and weave.jl, so I am guessing it’s just not designed with this in mind.

2

Maybe Books.jl might be what you are looking for? https://github.com/JuliaBooks/Books.jl

@rikh would certainly know more about it than I.

2 Likes
3

Thank you! This looks like the right idea, I’ll give it a test!

4

Interactive tables are not in Books.jl and I also suspect that you’re more into single page conversions which is not a focus of Books.jl

PlutoSliderServer has some more dynamic capabilities. Or what does an “interactive table” mean? Is it an HTML table with a bunch of Javascript on top (static) or does it also require a server?

1 Like
5

Thanks!

In terms of what I need to be able to replicate to leave R, it would be a table my end users can filter and search like those made by DT or itables.

And they’re an HTML table with a bunch of JS on top. They are compiled into the HTML book and and do not require a server (though we host the books centrally to avoid having more than one version of the truth if the reports are updated).

1 Like
6

I’m 99% sure that the news unfortunately is that such things packages aren’t available in Julia at this moment.

I don’t have much experience with your particular problem, but I do know that sometimes when language wrappers do not work then there is luckily CSV, that is, store CSV in Julia and read it with R or vice versa.

1 Like
7

Does TableView do what you are asking for?

9

Ok, I’ll see if there is workaround with the intermediaries that works for now. It might also be worth asking some of the other authors how I might go about it if I was going to attempt to write something that would be compatible with existing tools.

Thanks for your time! :slight_smile:

10

Unfortunately, no - that’s the example that doesn’t end up in the book. It does seem like it’s the closest so I’ll see if I can fix the issue with the compilation environment.

11

I’ve been playing with it and…frustration. When I load the notebook file and execute it interactively, it’s fine, but when you export or nbconvert to html, it no longer can see the webio notebook extension that TableView needs.

1 Like
12

Maybe have a look at quarto, it’s a brilliant system for converting Markdown and notebooks (including IJulia notebooks) to different formats. It’s done by the RStudio people based on pandoc (like Bookdown I think). You can execute Julia code using notebooks, or with code blocks in Markdown files.

I don’t know if it does what you need, but it’s worth having a look at the project anyway :slight_smile:

3 Likes
13

That’s awesome! Long time user of RMarkdown here and what a happy journey it’s been. This looks like the next level to me and i look forward to creating new reports based on julia code as well.

1 Like
14

Same! I’ve documented my efforts here.

15

Thanks for bringing this to my attention I will def take a peek!

16

I am infuriatingly close. I am using RCall to generate a beautiful table in a standalone HTML file:

using RDatasets
using RCall

x = dataset("datasets", "mtcars");

@rput x;

R"

library(DT)

htmlwidgets::saveWidget(x %>% DT::datatable(.), 'result.html')

"

Thus, a more refined version of my question is “In ipynb, does anyone know how I can get a markdown cell or a Julia cell to display this HTML file?”

WebIO is out, I get the same thing from before in the books, I just recreate it myself:

using WebIO

htmlString = read("result.html", String)

Scope(dom=node(:p, htmlString))

Looks like I should ask the WebIO folks about this.

At least now I can take the partial victory of knowing I can likely get this to work by switching back to Rmd where I’d have cells-by-language or switch to Python where I’d have magics instead of calls from a library.

I might not be able to escape to Julia but I do now know I can at least escape to Python lol.

17

Kind of like escaping from Pepsi to Coke instead of Prosecco. :laughing:

You could avoid the R part by writing a translator to client-side HTML/JS, like Tabulator, which does not look difficult. I still don’t see how to inject that into a jupyter-book or weave page, but you could write a script to post-process the initial HTML outputs.

1 Like
18

LMAO that’s good. I’d much prefer to be working in Python over R so to me it’s more like escaping from Diet Pepsi to a PBR when I was looking for a Boston Lager lol.

I will look into this other angle and open a question issue on WebIO. Thanks all!