Peculiar styling of messages in example output

[jennybc]

I can't figure out how downlit is deciding to show some messages in bold and not others.

I'm seeing this in the googlesheets4 examples. Here are few examples as of tidyverse/googlesheets4@9309c2a

Probably relevant: googlesheets4 is using cli (therefore rlang::inform()) for messages now.

The basic pattern is: the first googlesheets4 message is bold, but no others.

Messages originating elsewhere, presumably coming via message() are ?always? bold. ("Files trashed..." is always coming from googledrive, for example).

Below is a sampler.

---

https://googlesheets4.tidyverse.org/reference/range_read.html#examples

tidyverse/googlesheets4@9309c2a#diff-c089033b92eaa2e50242f2188feedb0d449c5fd93c1765301867ee1c9bb1ae39

---

https://googlesheets4.tidyverse.org/reference/range_speedread.html#examples

tidyverse/googlesheets4@9309c2a#diff-c755ed40978311fdb955d656db19f8477496e343ec19095b0cfcf64716ca26ec

---

https://googlesheets4.tidyverse.org/reference/sheet_write.html

tidyverse/googlesheets4@9309c2a#diff-19e7740a2d283dc1311c374e0bb4b937e9a904f3dbef7acb84432b9053abff3f

[cderv]

Just sharing what I saw yesterday when the question was ask.

pkgdown is setting a CSS for not of class message: https://github.com/r-lib/pkgdown/blob/dd94073575aaee2fd480892d24dfe0be85f81a03/inst/assets/pkgdown.css#L308

.message { color: black;   font-weight: bolder;}

This seem to lead to the bold text.

About not everything is bold : downlit does not seem to produce valid HTML in this case: the closing </span> for class message is not at the right place.

See below,

• With message we correctly get </span></div>
• Using gs4_success, we get </div></span> which is not correct because it opens with <div class='output co'><span class='message'>.
  This could lead the browser to not auto correct the same on the first line and other lines, hence the difference seen.

> extract_output_part <- function(html) {
+     s <- stringr::str_locate(html, "<div class='output co'>")[, "start"]
+     stringr::str_sub(html, s)
+ }
> 
> html1 <- downlit::evaluate_and_highlight('message("a")')
> extract_output_part(html1)
[1] "<div class='output co'>#&gt; <span class='message'>a</span></div>"
> 
> html2 <- downlit::evaluate_and_highlight('googlesheets4:::gs4_success("a")')
> extract_output_part(html2)
[1] "<div class='output co'>#&gt; <span class='message'><span style='color: #00BB00;'>v</span><span> a</span></div></span>"

I can't do a reprex above because this does not happens inside reprex::reprex() call - not interactive session ?

extract_output_part <- function(html) {
  s <- stringr::str_locate(html, "<div class='output co'>")[, "start"]
  stringr::str_sub(html, s)
}

html1 <- downlit::evaluate_and_highlight('message("a")')
extract_output_part(html1)
#> [1] "<div class='output co'>#&gt; <span class='message'>a</span></div>"

html2 <- downlit::evaluate_and_highlight('googlesheets4:::gs4_success("a")')
extract_output_part(html2)
#> [1] "<div class='output co'>#&gt; <span class='message'>v a</span></div>"

^{Created on 2021-03-09 by the reprex package (v1.0.0.9002)}

Hope it helps.

[hadley]

Somewhat more minimal reprex:

library(downlit)

f <- function() cli::cli_alert_danger("x")
cat(evaluate_and_highlight("f()"))
#> <div class='input'><span class='fu'>f</span><span class='op'>(</span><span class='op'>)</span></div><div class='output co'>#&gt; <span class='message'>x x</span></div>

^{Created on 2021-04-27 by the reprex package (v2.0.0)}

[hadley]

Even simpler reprex:

x <- "<div>\033[31mx\033[39m y</div>"
fansi::sgr_to_html(x)
#> [1] "<div><span style='color: #BB0000;'>x</span><span> y</div></span>"

^{Created on 2021-04-27 by the reprex package (v2.0.0)}

[hadley]

@jennybc this makes the output consistent, but consistently bold. Removing the bold in r-lib/pkgdown#1644 .