Prettyprint.char does not count newlines

[DavZim]

This is most likely a minor issue as it only affects prints.

Setting data.table.prettyprint.char to some value will not result in a character vector of that length as it ignores newline characters.
I've found that it ignores \n, but also \t.

MWE

options(datatable.prettyprint.char = 3)

data.table::data.table(x = c("1234", "12\n\n\n\n\n34", "12\t\t\t\t\t\t34"))
#>                     x
#>                <char>
#> 1:             123...
#> 2:   12\n\n\n\n\n3...
#> 3: 12\t\t\t\t\t\t3...

I would expect to see

data.table::data.table(x = c("1234", "12\n\n\n\n\n34", "12\t\t\t\t\t\t34"))
#>           x
#>      <char>
#> 1:   123...
#> 2:  12\n...
#> 3:  12\t...

Session info

Note I have also tested and reproduced the bug with version 1.16.99 (at the time of writing this corresponds to 3a7ec2d)

sessioninfo::session_info()
#> ─ Session info ───────────────────────────────────────────────────────────────
#>  setting  value
#>  version  R version 4.4.0 (2024-04-24)
#>  os       Ubuntu 22.04.4 LTS
#>  system   x86_64, linux-gnu
#>  ui       X11
#>  language (EN)
#>  collate  en_US.UTF-8
#>  ctype    en_US.UTF-8
#>  tz       Etc/UTC
#>  date     2025-01-20
#>  pandoc   3.1.1 @ /usr/lib/rstudio-server/bin/quarto/bin/tools/ (via rmarkdown)
#> 
#> ─ Packages ───────────────────────────────────────────────────────────────────
#>  package     * version date (UTC) lib source
#>  cli           3.6.3   2024-06-21 [1] RSPM (R 4.4.0)
#>  data.table    1.16.4  2024-12-06 [1] RSPM (R 4.4.0)
#>  rstudioapi    0.16.0  2024-03-24 [1] RSPM (R 4.4.0)
#>  sessioninfo   1.2.2   2021-12-06 [1] RSPM (R 4.4.0)
#> 
#>  [1] /home/dazimme/R/x86_64-pc-linux-gnu-library/4.4
#>  [2] /opt/R/4.4.0/lib/R/library
#> 
#> ──────────────────────────────────────────────────────────────────────────────

[ben-schwen]

Thanks for the report. I just checked and this previously worked (as you expected) and was introduced in #6048.

[joshhwuu]

I believe this is due to the change to using strtrim() instead of substring(), which has different behaviour for non-printable characters (such as tabs) depending on locale:

from ?strtrim()

What happens with non-printable characters (such as backspace, tab) is implementation-dependent and may depend on the locale (e.g., they may be included in the count or they may be omitted).

[MichaelChirico]

Thanks for investigating @joshhwuu! As a reminder, we switched to strtrim() over substring() in #6048:

This PR changes trunc.char to recognize when the char is combining or full-width and indexes accordingly. Additionally, strtrim() is used now in place of substr() because as mentioned in ?substr:

These functions are often used with nchar to truncate a display. That does not really work (you want to limit the width, not the number of characters, so it would be better to use strtrim), but at least make sure you use the default nchar(type = "c").

My first reaction is this is "WAI" due to the limitations of the underlying strtrim() routine. I definitely think it's preferable to substr(), whose documentation explicitly recommends against using it for truncation. We could potentially devise some clobbering code to force strtrim() to account for invisible characters, but I'm not sure it's a good idea -- better to just document the issue and users can clobber in their own way as needed.

[MichaelChirico]

Closed by #6743