# Littler FAQ

2019-03-23, updated 2020-06-16

### Well, why?

Glad you asked. When the initial work started, Rscript did not exist so there was no scripting tool. We take a small amount of pride into beating Rscript narrowly to a first release. Of course, by now, Rscript is more widely used as it ships with every R installation but we still have some aspects we like about r: it is simpler and shorter, deals (in our biased) view more sanely with command-line arguments via the argv vector, always loaded package methods (which Rscript finally came around to a good decade later) and still starts faster. And r is just to cool a command (in relation to the trusted R) so someone had to!

### No Printing

In the very beginning of littler, we made an executive decision to not echo each command output when we evaluate commands in the read-evaluate-print loop. So 2+2 will be evaluated silently: four will be computed, but not shown.

That made sense for scripts we wanted to be silent. The idea is that user will add a print() or cat() as needed. If you desire each step to printed, add a -p or --verbose argument.

### No .Renviron

Another early decision was to make r start faster than the alternatives. Our r is “just” a small little binary. Whereas R and Rscript are front-end shell scripts sourcing a number of things and setting them dynamically, we determine values at compilation time and freeze them into the binary. That may seem risqué but worked out just fine. However, while we later added the ability to source Rprofile files (which contain R code we can evaluate) there is no way for us to source the Renviron files and to modify the already running process. So if you want to source Renviron you could add an explicit readEnviron() to ~/.littler.r or /etc/littler/r.

### Lower-case / upper-case

This is mostly an issue on macOS where the brain surgeons behind the OS decided that r is the same as R. What can we say? On that OS you may need to rename the build to lr instead. Please send your complaints to Cupertino, California.

When the package is installed from source, it displays

*
* new binary $(TGT) installed in bin/ subdirectory" * consider adding a symbolic link from, e.g., /usr/local/bin" * on OS X, you may have to name this 'lr' instead" * see the littler-faq vignette for more details" * where the variable $(TGT) normally expands to r. As the text suggests you may want to create a symbolic link from a directory in the to make the newly built binary available to users. On Linux you may do

cd /usr/local/bin
sudo ln -s /usr/local/lib/R/site-library/littler/bin/r .

to create a link for r in /usr/local/bin. On macOS, as suggested, you may want

cd /usr/local/bin
sudo ln -s /usr/local/lib/R/site-library/littler/bin/r lr

to create a command lr instead; see the previous question as to why. In either case adjust the source part of the ln command to where your binary is – which R can tell you via system.file("bin", "r", package="littler").

Also note that e.g. the zsh has a builtin command r which may conflict, so using the explicit path of, say, /usr/bin/r or /usr/local/bin/r or … may provide an alternative.