Originally written 2015-10-28; Updated 2017-12-10, 2018-05-24, 2019-01-06, 2019-06-09 and 2020-06-16
We show and discuss a few of the files included in the inst/examples/
source directory of littler (which becomes the examples/
directory once installed). In a few cases we remove comment lines to keep things more concise on this page. We use $
to denote a shell (i.e. terminal) prompt.
Note that some systems (such as macOS) cannot install littler as r
(as lower-case and upper-case are by default the same; not a great idea). And for example the zsh
has r
as a builtin so you would have to use /usr/bin/r
. See the littler FAQ for more.
littler can be used directly on the command-line just like, say, bc
, easily consuming standard input from a pipe:
$ echo 'cat(pi^2,"\n")' | r
9.869604
Equivalently, commands that are to be evaluated can be given on the command-line
$ r -e 'cat(pi^2, "\n")'
9.869604
But unlike bc(1), GNU R has a vast number of statistical functions. For example, we can quickly compute a summary()
and show a stem-and-leaf plot for file sizes in a given directory via
$ ls -l /boot | awk 'BEGIN {print "size"} !/^total/ {print $5}' | \
r -de "print(summary(X[,1])); stem(X[,1])"
which produces something like
Min. 1st Qu. Median Mean 3rd Qu. Max.
13 512 110100 486900 768400 4735000
The decimal point is 6 digit(s) to the right of the |
0 | 0000001122222279222
2 | 79444
4 | 71888
6 |
8 |
10 |
12 |
14 | 8
16 | 4
18 |
20 | 333
(Note that some systems may not have /boot
in which case you can try /sbin
or another directory.)
As we saw in the preceding example, the program can also be shortened like using the new -d
option which reads from stdin and assigns to a data.frame
named X
.
And, last but not least, this (somewhat unwieldy) expression can be stored in a helper script (where we now switch to using an explicit readLines()
on stdin
):
#!/usr/bin/env r
<- as.integer(readLines(file("stdin")))
fsizes print(summary(fsizes))
stem(fsizes)
(where calling #!/usr/bin/env
is a trick from Python which allows one to forget whether r is installed in /usr/bin/r
, /usr/local/bin/r
, ~/bin/r
, …).
This is one of my favourite littler scripts which I use frequently to install packages off CRAN.
#!/usr/bin/env r
if (is.null(argv) | length(argv)<1) {
cat("Usage: installr.r pkg1 [pkg2 pkg3 ...]\n")
q()
}
## adjust as necessary, see help('download.packages')
<- "https://cran.rstudio.com"
repos
## this makes sense on Debian where no packages touch /usr/local
<- "/usr/local/lib/R/site-library"
lib.loc
install.packages(argv, lib.loc, repos)
I invoke it all the time with one, two or more packages to install (or reinstall).
$ install.r digest RcppCNPy
It conveniently installs all dependencies, and uses the chosen target directory, all while keeping my R prompt (or prompts with multiple sessions) free to do other things. Also, if used with options("Ncpu")
set, then (remote CRAN) packages will be installed in parallel.
Thanks to the fabulous docopt package, we also have a variant install2.r
with optional settings of repo and location. It was first installed with littler 0.2.1. The current (i.e., 0.3.3 as of this writing) version is more featureful and longer and included to keep this brief. Some usage examples are
$ install2.r -l /tmp/lib Rcpp BH # install into given library
$ install2.r -- --with-keep.source drat # keep the source
$ install2.r -- --data-compress=bzip2 stringdist # prefer bz2 compression
Another very useful option is -n N
or --ncpus N
which will parallelize the installation across N
processes. This can be set automagically by setting options("Ncpus")
.
Starting with version 0.2.2, install.r
and install2.r
now recognise installable source files. So one can also do this:
$ install.r digest_0.6.8.tar.gz
and the local source file will the installed via a call to R CMD INSTALL
.
A related use case is to check packages via check.r
. This script runs R CMD check
, but also installs package dependencies first as tests may have dependencies not yet satisfied on the test machine. It offers a number of options:
$ check.r -h
Usage: check.r [-h] [-x] [--as-cran] [--repo REPO] [--install-deps] [--install-kitchen] [--deb-pkgs PKGS...] [--use-sudo] [--library LIB] [--setwd DIR] [TARGZ ...]
-a --as-cran customization similar to CRAN's incoming [default: FALSE]
-r --repo REPO repository to use, or NULL for file [default: https://cran.rstudio.com]
-i --install-deps also install packages along with their dependencies [default: FALSE]
-k --install-kitchen even install packages 'kitchen sink'-style up to suggests [default: FALSE]
-l --library LIB when installing use this library [default: /usr/local/lib/R/site-library]
-s --setwd DIR change to this directoru before undertaking the test [default: ]
-d --deb-pkgs PKGS also install binary .deb packages with their dependencies [default: FALSE]
-u --use-sudo use sudo when installing .deb packages [default: TRUE]
-h --help show this help text
-x --usage show help and short example usage
$
The script rcc.r
will also check a source tarball, and offers another set of options:
$ rcc.r -h
Usage: rcc.r [-h] [-x] [-c] [-f] [-q] [--args ARGS] [--libpath LIBP] [--repos REPO] [PATH...]
-c --as-cran should '--as-cran' be added to ARGS [default: FALSE]
-a --args ARGS additional arguments to be passed to 'R CMD CHECK' [default: ]
-l --libpath LIBP additional library path to be used by 'R CMD CHECK' [default: ]
-r --repos REPO additional repositories to be used by 'R CMD CHECK' [default: ]
-f --fast should vignettes and manuals be skipped [default: FALSE]
-q --quiet should 'rcmdcheck' be called qietly [default: FALSE]
-h --help show this help text
-x --usage show help and short example usage
$
A related script is build.r
which I often use inside a source repository to quickly build a source tarball. Like rcc.r
, it has a switch -f
or --fast
to omit building of vignettes.
$ build.r # without argument works on current directory
$ build.r digest/ # with directory argument builds tar.gz from repo in directory
Installation directly from GitHub is also popular. Here is an example:
$ installGithub.r RcppCore/RcppEigen
Installing from github is supported via the following helper script:
#!/usr/bin/env r
#
# A simple example to install one or more packages from GitHub
#
# Copyright (C) 2014 - 2015 Carl Boettiger and Dirk Eddelbuettel
#
# Released under GPL (>= 2)
## load docopt and remotes (or devtools) from CRAN
suppressMessages(library(docopt)) # we need docopt (>= 0.3) as on CRAN
suppressMessages(library(remotes)) # can use devtools as a fallback
## configuration for docopt
<- "Usage: installGithub.r [-h] [-d DEPS] REPOS...
doc
-d --deps DEPS Install suggested dependencies as well? [default: NA]
-h --help show this help text
where REPOS... is one or more GitHub repositories.
Examples:
installGithub.r RcppCore/RcppEigen
installGithub.r is part of littler which brings 'r' to the command-line.
See http://dirk.eddelbuettel.com/code/littler.html for more information.
"
## docopt parsing
<- docopt(doc)
opt if (opt$deps == "TRUE" || opt$deps == "FALSE") {
$deps <- as.logical(opt$deps)
optelse if (opt$deps == "NA") {
} $deps <- NA
opt
}
invisible(sapply(opt$REPOS, function(r) install_github(r, dependencies = opt$deps)))
One of the scripts I use the most (interactively) is update.r
which updates installed packages. An earlier version looked like the following example, the current version is a little longer and has more features:
#!/usr/bin/env r
#
# a simple example to update packages in /usr/local/lib/R/site-library
# parameters are easily adjustable
## adjust as necessary, see help('download.packages')
<- "https://cloud.r-project.org"
repos
## this makes sense on Debian where no package touch /usr/local
<- "/usr/local/lib/R/site-library"
lib.loc
## r use requires non-interactive use
update.packages(repos=repos, ask=FALSE, lib.loc=lib.loc)
As above, it has my preferred mirror and library location hard-wired.
Another very useful option is -n N
or --ncpus N
which will parallelize the upgrade across N
processes. This can be set automagically by setting options(Ncpus)
.
Here is another convenience script, knit.r
, which knits a given file after testing the file actually exists.
#!/usr/bin/r
#
# Simple helper script for knitr
#
# Dirk Eddelbuettel, May 2013
#
# GPL-2 or later
if (is.null(argv)) {
cat("Need an argument FILE.Rnw\n")
q(status=-1)
}
<- argv[1]
file if (!file.exists(file)) {
cat("File not found: ", file, "\n")
q(status=-1)
}
require(knitr)
knit2pdf(file)
Similar to the previous example, the script roxy.r
one uses roxygen to extract documentation from R files – either in the current directory, or in the given directory or directories.
#!/usr/bin/r
#
# Simple helper script for roxygen2::roxygenize()
#
# Dirk Eddelbuettel, August 2013
#
# GPL-2 or later
## load roxygen
library(roxygen2)
## check all command-line arguments (if any are given) for directory status
<- Filter(function(x) file.info(x)$is.dir, argv)
argv
## loop over all argument, with fallback of the current directory, and
## call compileAttributes() on the given directory
sapply(ifelse(length(argv) > 0, argv, "."), FUN=roxygenize, roclets="rd")
The next script, compAttr.
, can be used with Rcpp, and particularly is powerful Attributes feature, in order to auto-generate helper code. It is similar to the preceding script, but invokes compileAttributes()
instead.
#!/usr/bin/r
#
# Simple helper script for compileAttributes()
#
# Dirk Eddelbuettel, July 2014
#
# GPL-2 or later
## load Rcpp
suppressMessages(library(Rcpp))
## check all command-line arguments (if any are given) for directory status
<- Filter(function(x) file.info(x)$is.dir, argv)
argv
## loop over all argument, with fallback of the current directory, and
## call compileAttributes() on the given directory
sapply(ifelse(length(argv) > 0, argv, "."), compileAttributes)
The render.r
script generalizes the earlier knit.r
to convert Markdown into its designated output.
#!/usr/bin/env r
#
# Another example to run a shiny app
#
# Copyright (C) 2016 Dirk Eddelbuettel
#
# Released under GPL (>= 2)
suppressMessages(library(docopt)) # we need docopt (>= 0.3) as on CRAN
## configuration for docopt
<- "Usage: render.r [-h] [-x] [FILES...]
doc
-h --help show this help text
-x --usage show help and short example usage"
<- docopt(doc) # docopt parsing
opt
if (opt$usage) {
cat(doc, "\n\n")
cat("Examples:
render.r foo.Rmd bar.Rmd # convert two given files
render.r is part of littler which brings 'r' to the command-line.
See http://dirk.eddelbuettel.com/code/littler.html for more information.\n")
q("no")
}
library(rmarkdown)
## helper function
<- function(p) {
renderArg if (!file.exists(p)) stop("No file '", p, "' found. Aborting.", call.=FALSE)
render(p)
}
## render files using helper function
sapply(opt$FILES, renderArg)