fansi - ANSI Control Sequence Aware String Functions

Counterparts to R string manipulation functions that account for the effects of ANSI text formatting control sequences.

Formatting Strings with Control Sequences

Many terminals will recognize special sequences of characters in strings and change display behavior as a result. For example, on my terminal the sequences "\033[3?m" and "\033[4?m", where "?" is a digit in 1-7, change the foreground and background colors of text respectively:

fansi <- "\033[30m\033[41mF\033[42mA\033[43mN\033[44mS\033[45mI\033[m"

This type of sequence is called an ANSI CSI SGR control sequence. Most *nix terminals support them, and newer versions of Windows and Rstudio consoles do too. You can check whether your display supports them by running term_cap_test().

Whether the fansi functions behave as expected depends on many factors, including how your particular display handles Control Sequences. See ?fansi for details, particularly if you are getting unexpected results.

Manipulation of Formatted Strings

ANSI control characters and sequences (Control Sequences hereafter) break the relationship between byte/character position in a string and display position. For example, to extract the “ANS” part of our colored “FANSI”, we would need to carefully compute the character positions:

With fansi we can select directly based on display position:

If you look closely you’ll notice that the text color for the substr version is wrong as the naïve string extraction loses the initial"\033[37m" that sets the foreground color. Additionally, the color from the last letter bleeds out into the next line.

fansi Functions

fansi provides counterparts to the following string functions:

• substr (and substr<-)
• strsplit
• strtrim
• strwrap
• nchar / nzchar
• trimws

These are drop-in replacements that behave (almost) identically to the base counterparts, except for the Control Sequence awareness. There are also utility functions such as strip_ctl to remove Control Sequences and has_ctl to detect whether strings contain them.

Much of fansi is written in C so you should find performance of the fansi functions to be slightly slower than the corresponding base functions, with the exception that strwrap_ctl is much faster. Operations involving type = "width" will be slower still. We have prioritized convenience and safety over raw speed in the C code, but unless your code is primarily engaged in string manipulation fansi should be fast enough to avoid attention in benchmarking traces.

Width Based Substrings

fansi also includes improved versions of some of those functions, such as substr2_ctl which allows for width based substrings. To illustrate, let’s create an emoji string made up of two wide characters:

pizza.grin <- sprintf("\033[46m%s\033[m", strrep("\U1F355\U1F600", 10))

And a colorful background made up of one wide characters:

raw <- paste0("\033[45m", strrep("FANSI", 40))
wrapped <- strwrap2_ctl(raw, 41, wrap.always=TRUE)

When we inject the 2-wide emoji into the 1-wide background their widths are accounted for as shown by the result remaining rectangular:

starts <- c(18, 13, 8, 13, 18)
ends <-   c(23, 28, 33, 28, 23)
substr2_ctl(wrapped, type='width', starts, ends) <- pizza.grin

fansi width calculations use heuristics to account for graphemes, including combining emoji:

emo <- c(
  "\U1F468",
  "\U1F468\U1F3FD",
  "\U1F468\U1F3FD\u200D\U1F9B3",
  "\U1F468\u200D\U1F469\u200D\U1F467\u200D\U1F466"
)
writeLines(
  paste(
    emo,
    paste("base:", nchar(emo, type='width')),
    paste("fansi:", nchar_ctl(emo, type='width'))
) )
## 👨 base: 2 fansi: 2
## 👨🏽 base: 4 fansi: 2
## 👨🏽‍🦳 base: 6 fansi: 2
## 👨‍👩‍👧‍👦 base: 8 fansi: 2

HTML Translation

You can translate ANSI CSI SGR formatted strings into their HTML counterparts with to_html:

Translate to HTML

Rmarkdown

It is possible to set knitr hooks such that R output that contains ANSI CSI SGR is automatically converted to the HTML formatted equivalent and displayed as intended. See the vignette for details.

Installation

This package is available on CRAN:

install.packages('fansi')

It has no runtime dependencies.

For the development version use remotes::install_github('brodieg/fansi@development') or:

f.dl <- tempfile()
f.uz <- tempfile()
github.url <- 'https://github.com/brodieG/fansi/archive/development.zip'
download.file(github.url, f.dl)
unzip(f.dl, exdir=f.uz)
install.packages(file.path(f.uz, 'fansi-development'), repos=NULL, type='source')
unlink(c(f.dl, f.uz))

There is no guarantee that development versions are stable or even working. The master branch typically mirrors CRAN and should be stable.

Related Packages and References

• crayon, the library that started it all.
• ansistrings, which implements similar functionality.
• ECMA-48 - Control Functions For Coded Character Sets, in particular pages 10-12, and 61.
• CCITT Recommendation T.416
• ANSI Escape Code - Wikipedia for a gentler introduction.

Acknowledgments

• R Core for developing and maintaining such a wonderful language.
• CRAN maintainers, for patiently shepherding packages onto CRAN and maintaining the repository, and Uwe Ligges in particular for maintaining Winbuilder.
• Gábor Csárdi for getting me started on the journey ANSI control sequences, and for many of the ideas on how to process them.
• Jim Hester for covr, and with Rstudio for r-lib/actions.
• Dirk Eddelbuettel and Carl Boettiger for the rocker project, and Gábor Csárdi and the R-consortium for Rhub, without which testing bugs on R-devel and other platforms would be a nightmare.
• Tomas Kalibera for rchk and the accompanying vagrant image, and rcnst to help detect errors in compiled code.
• Winston Chang for the r-debug docker container, in particular because of the valgrind level 2 instrumented version of R.
• George Nachman etal. for Iterm2, a Free terminal emulator that supports truecolor CSI SGR.
• Hadley Wickham and Peter Danenberg for roxygen2.
• Yihui Xie for knitr and J.J. Allaire et al. for rmarkdown, and by extension John MacFarlane for pandoc.
• Gábor Csárdi, the R-consortium, et al. for revdepcheck to simplify reverse dependency checks.
• Olaf Mersmann for microbenchmark, because microsecond matter, and Joshua Ulrich for making it lightweight.
• All open source developers out there that make their work freely available for others to use.
• Github, Codecov, Vagrant, Docker, Ubuntu, Brew for providing infrastructure that greatly simplifies open source development.
• Free Software Foundation for developing the GPL license and promotion of the free software movement.