Invalid HTML in prerendered shiny documents #1912

dakep · 2020-10-07T01:20:09Z

rmarkdown::run() on a Rmd document with runtime: shiny_prerendered serves invalid HTML.

Consider the follwing Rmd file invalidhtml/invalidhtml.Rmd with contents

---
title: "Invalid HTML"
output: html_document
runtime: shiny_prerendered
---

This creates invalid HTML.

Running this document with

rmarkdown::run("invalidhtml/invalidhtml.Rmd")

Will serve a document with the follwing invalid HTML (I removed most content for brevity)

<!DOCTYPE html>
<html>
<head>
  <meta http-equiv="Content-Type" content="text/html; charset=utf-8"/>
  <script type="application/shiny-singletons"></script>
  <!-- text snippets removed -->
</head>
<body><!DOCTYPE html>

<html>

<head>
  <!-- text snippets removed -->
</head>
  <!-- text snippets removed -->
</body>
</html></body>
</html>

The problem is that rmarkdown is rendering the Rmd file to a standalone HTML document, and returning the whole HTML document in the shiny app's UI function. The UI function, however, expects only the body of the document.

I am not sure if this issue has not been discovered until now, or if it's accepted for simplicity (I think most major browsers accept this invalid HTML).

R session info:

R version 4.0.2 Patched (2020-07-13 r78838)
Platform: x86_64-apple-darwin17.0 (64-bit)
Running under: macOS Catalina 10.15.6

Locale: en_US.UTF-8 / en_US.UTF-8 / en_US.UTF-8 / C / en_US.UTF-8 / en_US.UTF-8

Package version:
  base64enc_0.1.3 digest_0.6.25   evaluate_0.14   glue_1.4.2     
  graphics_4.0.2  grDevices_4.0.2 highr_0.8       htmltools_0.5.0
  jsonlite_1.7.1  knitr_1.30      magrittr_1.5    markdown_1.1   
  methods_4.0.2   mime_0.9        rlang_0.4.7     rmarkdown_2.4.1
  stats_4.0.2     stringi_1.5.3   stringr_1.4.0   tinytex_0.26   
  tools_4.0.2     utils_4.0.2     xfun_0.18       yaml_2.2.1     

Pandoc version: 2.7.3

By filing an issue to this repo, I promise that

I have fully read the issue guide at https://yihui.org/issue/.
I have provided the necessary information about my issue.
- If I'm asking a question, I have already asked it on Stack Overflow or RStudio Community, waited for at least 24 hours, and included a link to my question there.
- If I'm filing a bug report, I have included a minimal, self-contained, and reproducible example, and have also included xfun::session_info('rmarkdown'). I have upgraded all my packages to their latest versions (e.g., R, RStudio, and R packages), and also tried the development version: remotes::install_github('rstudio/rmarkdown').
- If I have posted the same issue elsewhere, I have also mentioned it in this issue.
I have learned the Github Markdown syntax, and formatted my issue correctly.

I understand that my issue may be closed if I don't fulfill my promises.

The text was updated successfully, but these errors were encountered:

cderv · 2020-10-07T09:23:31Z

Hi, thank you for the feedback.

The problem is that rmarkdown is rendering the Rmd file to a standalone HTML document, and returning the whole HTML document in the shiny app's UI function. The UI function, however, expects only the body of the document.

I am not sure to follow here.
AFAIK, you can pass a full html document to ui component in shiny::runApp as a tagList object.
See https://shiny.rstudio.com/articles/html-ui.html
And this is what shiny_prerendered_app() is doing with the rendered html file.

Will serve a document with the follwing invalid HTML

Where do you see this invalid html ? I can't reproduce this on my end, but not sure I look in the right place. Can you point me in the correct direction ?

dakep · 2020-10-07T16:24:36Z

Where do you see this invalid html ? I can't reproduce this on my end, but not sure I look in the right place. Can you point me in the correct direction ?

The steps that work for me to see the invalid HTML are

Run the interactive document with rmarkdown::run("invalidhtml/invalidhtml.Rmd") as above
Either (a) open the URL in any browser and view the page source (not in the development tools, but the actual source) or (b) use cURL, e.g., curl {SHINY_URL}. Firefox is helpful here as it highlights invalid HTML in bold red.

The difference to the examples you pointed to is that rmarkdown:::shiny_prerendered_app() serves the HTML as character vector with attribute attr(*, "html")= logi TRUE, not as tag list. The article uses shiny::htmlTemplate() to let shiny know the UI is a standalone HTML document (it appends the class html_document). I guess shiny then takes care of merging the contents in <head> and <body> with the content from shiny.

gadenbuie · 2021-02-23T21:31:34Z

I can reproduce this, and the issue was separately reported in rstudio/learnr#498. This issue doesn't visibly impact the page (browsers do everything they can not to fail), but the HTML syntax is invalid: there should be only one <head></head> or <body></body> element.

ghost · 2021-02-24T07:43:40Z

It's impacting all the prerendered shiny documents. It also has negative impact in terms of SEO

cderv · 2021-02-24T09:25:38Z

Thanks for the precisions. I understand now better what is happening. We can't reproduce anymore with the first code above because having a document with runtime: shiny_prerendered without server context will now throw an error.

With the learn example I could reproduce and look more into it. The fix in #1942 help a lot to see where the issue is. I just want to get to the bottom of it and see if it impacts also runtime: shiny. I'll check with the shiny team to confirm what should be the full fix for this, as I don't yet the full codebase for shiny support in rmardown.

Thanks for your patience.

gadenbuie · 2021-02-24T11:49:20Z

We can't reproduce anymore with the first code above because having a document with runtime: shiny_prerendered without server context will now throw an error.

Yesterday I added one context = "server" chunk to reproduce without learnr. It seems like the entire pre-rendered html document is being sent, rather than processing the head and body content, which I believe is wheat #1942 addresses.

I did not see similar behavior with runtime: shiny.

cderv · 2021-02-24T12:06:03Z

Yes if I add some chunk for server context I can reproduce. This is what I have done this morning to investigate further.

This issue is highlighting the fact that using Rmd, a full HTML document is rendered and it should be dealt correctly after. I do not see the same with runtime: shiny but the full rendered document HTML code will be passed to shiny::renderUI() - I do not know if this is problematic or not. It seems that it's handled correctly. (The source HTML will have a shiny::uiOutput() with runtime: shiny that is why it is not exactly the same)

ghost · 2021-02-24T13:20:48Z

I tried using #1942 but it's a few commits behind master. When I use it with learnr, it returns object 'kbd' not found. I guess it's because 18n_translations has been added recently

jcheng5 · 2021-02-25T02:21:59Z

I think runtime: shiny is technically doing the wrong thing too, in the sense that it's starting with a full HTML document, and then shoving another full HTML document inside of the uiOutput. It's just not as obvious as in the shiny_prerendered case, because for shiny, "View Source" is too early (document has not yet been injected) and DOM Inspector is too late (the browser or maybe jQuery has already ignored the spurious doctype, <html> and <head>).

For the same reasons though, I think fixing the shiny_prerendered case is the higher priority.

cderv · 2021-02-25T13:24:48Z

Thanks @jcheng5 for this comment. It confirms what I thought about runtime: shiny and that is good to know! I'll look at it more closely after fixing the shiny_prerendered thanks to @dakep PR!

github-actions · 2021-08-31T05:13:10Z

This old thread has been automatically locked. If you think you have found something related to this, please open a new issue by following the issue guide (https://yihui.org/issue/), and link to this old issue if necessary.

Special comment `` is now added in `$header-includes$` and there is not need anymore to add it in template. This will make it work with any custom template. As a reminder, this comments is used by htmltools to insert head content - which shiny does for html dependencies found during knitting. They will now be inserted at the same place as with non shiny document. This offers a better fix than #2064 that aimed to be an improvement over #1942 to fix #1912. #2064 workaround is still kept though as safety measure but should not be needed Co-authored-by: Christophe Dervieux <christophe.dervieux@gmail.com>

dakep mentioned this issue Nov 6, 2020

fix invalid HTML code in shiny_prerendered documents #1942

Merged

gadenbuie mentioned this issue Feb 23, 2021

Generate multiple html, head and body tags rstudio/learnr#498

Closed

cderv closed this as completed in #1942 Feb 25, 2021

cderv mentioned this issue Feb 25, 2021

runtime: shiny passed a full HTML document to shiny::renderUI() #2057

Open

cderv added bug an unexpected problem or unintended behavior theme: shiny regarding shiny support in R Markdown document labels Feb 25, 2021

This was referenced Mar 1, 2021

Wrong order of dependencies in header for shiny prerendered document ? #2060

Closed

Allow to add specific shiny-prerendered parts in HTML template #2064

Merged

github-actions bot locked as resolved and limited conversation to collaborators Aug 31, 2021

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Invalid HTML in prerendered shiny documents #1912

Invalid HTML in prerendered shiny documents #1912

dakep commented Oct 7, 2020

cderv commented Oct 7, 2020

dakep commented Oct 7, 2020

gadenbuie commented Feb 23, 2021

ghost commented Feb 24, 2021

cderv commented Feb 24, 2021

gadenbuie commented Feb 24, 2021

cderv commented Feb 24, 2021

ghost commented Feb 24, 2021

jcheng5 commented Feb 25, 2021 •

edited

Loading

cderv commented Feb 25, 2021

github-actions bot commented Aug 31, 2021

Invalid HTML in prerendered shiny documents #1912

Invalid HTML in prerendered shiny documents #1912

Comments

dakep commented Oct 7, 2020

cderv commented Oct 7, 2020

dakep commented Oct 7, 2020

gadenbuie commented Feb 23, 2021

ghost commented Feb 24, 2021

cderv commented Feb 24, 2021

gadenbuie commented Feb 24, 2021

cderv commented Feb 24, 2021

ghost commented Feb 24, 2021

jcheng5 commented Feb 25, 2021 • edited Loading

cderv commented Feb 25, 2021

github-actions bot commented Aug 31, 2021

jcheng5 commented Feb 25, 2021 •

edited

Loading