How to harden the formatting of URLs? (issue with # encoding)

[mjsteinbaugh]

I came across an edge case where some URLs are not rendering as expected.

Here's a minimal reprex:

## From googlesheets4 vignette:
## https://googlesheets4.tidyverse.org/articles/googlesheets4.html

url <- "https://docs.google.com/spreadsheets/d/1U6Cf_qEOhiR9AZqTqS3mbMF3zt2db48ZP5v3rkrAEJY/edit#gid=780868077"

cli::cli_alert(sprintf("Downloading '%s'.", url))
## → Downloading 'https://docs.google.com/spreadsheets/d/1U6Cf_qEOhiR9AZqTqS3mbMF3zt2db48ZP5v3rkrAEJY/edit#gid=780868077'.

cli::cli_alert(sprintf("Downloading {.file %s}.", url))
## → Downloading

cli::cli_alert(sprintf("Downloading {.url %s}.", url))
## → Downloading

cli::cli_alert(sprintf("Downloading {.val %s}.", url))
## → Downloading

cli::cli_alert(sprintf("Downloading {.var %s}.", url))
## → Downloading

Is there a way to harden the encoding of URLs to work better with the cli package?

[mjsteinbaugh]

It's the # character that's the problem, of course. Removal of this works, but I can't figure out the correct way to escape/encode:

cli::cli_alert(sprintf(
    fmt = "Downloading {.url %s}.",
    gsub(
        pattern = "#",
        replacement = "",
        x = url,
        fixed = TRUE
    )
))
## → Downloading <https://docs.google.com/spreadsheets/d/1U6Cf_qEOhiR9AZqTqS3mbMF3zt2db48ZP5v3rkrAEJY/editgid=780868077>.

cli::cli_alert(sprintf(
    fmt = "Downloading {.url %s}.",
    gsub(
        pattern = "#",
        replacement = "\\#",
        x = url,
        fixed = TRUE
    )
))
## → Downloading

[mjsteinbaugh]

Here's a simpler reprex that's the main issue:

cli::cli_alert("#")
## → #

cli::cli_alert("{.val #}")
## →

[gaborcsardi]

This is a bug in glue: tidyverse/glue#224

But you don't need sprintf() in cli, it supports substitution natively:

url <- "https://docs.google.com/spreadsheets/d/1U6Cf_qEOhiR9AZqTqS3mbMF3zt2db48ZP5v3rkrAEJY/edit#gid=780868077"
cli::cli_alert("Downloading {.url {url}}.")

#> → Downloading <https://docs.google.com/spreadsheets/d/1U6Cf_qEOhiR9AZqTqS3mbMF3zt2db48ZP5v3rkrAEJY/edit#gid=780868077>.

[gaborcsardi]

This is now "fixed" in dev glue, but unterminated # characters within {} will be errors now. (tidyverse/glue#227)

I am closing this now.