How do I prevent output: asis chunks from modifying htmltools outputs?

[AlbertRapp]

Description

This is related to #7600. I'm pretty confused why output: asis modifies HTML code. In this reprex, the rendered HTML-doc has additional p-tags even though they are not in the output that is generated from the R code chunk. Is there a way to circumvent that to get the generated HTML code from the R chunk "as is"?

```{r}
#| echo: false
#| output: asis

htmltools::div(
  htmltools::a('Some link')
) |> as.character() |> cat(sep = "\n")
```

The output from this chunk (if executed in R in the console) is

<div>
  <a>Some link</a>
</div>

But once I render in Quarto I get

<div>
  <p>
    <a href>Some link</a>
  </p>
</div>

[cscheid]

Can you provide a full example of your input to quarto? Thank you.

[AlbertRapp]

The R code chunk that I provided is the only thing in my Quarto doc. And the two chunks that I posted after the R code chunk just describe what I see in the R console vs the HTML output.

[mcanouil]

You don't have frontmatter specifying at least the format?
Below is a complete small reproducible example that anyone can simply copy, paste, and run without having to "guess" anything.

````qmd
---
title: "Reproducible Quarto Document"
format: html
---

This is a reproducible Quarto document using `format: html`.
It is written in Markdown and contains embedded R code.
When you run the code, it will produce a plot.

```{r}
plot(cars)
```

The end.
````

[AlbertRapp]

Nope. No YAML header. Trying to keep the example as minimal as possible by taking a blank doc and inserting that one code chunk. But I can add a header next time and share it.

[mcanouil]

You still render to one format 😅
Adding the format in the frontmatter ensure we are on the same page.

[cscheid]

What's happening here is that Pandoc doesn't like <a> elements to be directly inside divs, and so it inserts a paragraph. This is happening because your output (use keep-md: true to inspect it) doesn't bracket the content in a raw block:

---
keep-md: true
---


<div>
  <a>Some link</a>
</div>

If you need your HTML output to be unprocessed by Pandoc, you need to emit a raw block. I don't know the htmltools syntax to do it, but knitr::raw_html will take a string as input and produce a raw block.

---
keep-md: true
---

```{r}
#| echo: false
#| output: asis

knitr::raw_html(
htmltools::div(
  htmltools::a('Some link')
) |> as.character() |> paste())
```

produces

---
keep-md: true
---


::: {.cell-output-display}


```{=html}
<div>
  <a>Some link</a>
</div>
```


:::

which then is unprocessed by Pandoc.

[cscheid]

(nevermind the cell-output-display business, that's for the other issue you've raised)

[AlbertRapp]

Ah perfect. That's the answer I was looking for. FYI it also works to replace paste() by raw_html(), i.e.

```{r}
#| echo: false
#| output: asis


htmltools::div(
  htmltools::a('Some link'),
) |> as.character() |> knitr::raw_html()
```

[cderv]

See below how I would recommend to do it (#7614 (comment))

[cderv]

This should be as simple as that to use knitr and htmltools to produce HTML content

---
title: "test"
format: html
keep-md: true
---

```{r}
#| echo: false
#| output: asis
htmltools::div(
  htmltools::a('Some link')
)
```

htmltools + knitr takes care for you about all the thing you need to do to print the right markdown.

---
title: "test"
format: html
keep-md: true
---


::: {.cell-output-display}


```{=html}
<div>
<a>Some link</a>
</div>
```


:::

(and as said, the .cell-output-display will be removed when fixing #7600).

So no need of as.character() or raw_block() because knitr and htmltools does it for you (through the clever custom print logic from knitr.

That is to me the right way to do that.

Does it work for you correctly ? Did you try it ? If not working, which version of the tools are you using ?

[AlbertRapp]

Yes, that's how I feel too. And it works just like you said. But it doesn't work with the |> as.character() |> cat(sep = "\n") approach. And I used that approach because I had to use that way in #7600 and so I thought this is the safer version. Turns out that was wrong 😆 Good thing that the .cell-output-display thing will be taken care of. Thank you for all the help 🥳

[cderv]

But it doesn't work with the |> as.character() |> cat(sep = "\n") approach.

it does not do the same thing. Some information are lost in translation when you do that, like the fact a raw HTML block is required - hence why you need to add it yourself using raw_block() if you go this road.