← SASnRLoad an R package (library)

Introduction

Installing a package puts its files on disk. It does not make them usable yet.
To actually call the functions in a package, we have to attach it to the current R session. This is called loading the package.
In SAS, most procedures (PROC MEANS, PROC FREQ, PROC SORT and so on) are already built in. We almost never think about loading anything. R is the opposite by default — the base install is small, and anything beyond it has to be loaded every session.
The closest SAS analogue is attaching a format catalog or an autocall macro library: the code exists on disk, but SAS cannot see it until we point a LIBNAME or FMTSEARCH option at it.
In R, the command that does this is library().
Key things to remember before running the code below:
- Loading is per session. Every time we start a fresh R session we have to load again.
- If the package is not installed, library() throws an error. In that case, install first (see lesson 0110) and then load.
- Once loaded, we can call a function directly by name (for example filter()) without the package:: prefix.

Load an R package (library)

SAS (Base SAS)+

*==============================================================================;
* SAS has no direct equivalent of library() because most procs are built in. ;
* The closest analogues are pointing SAS at external code that it cannot see ;
* until we attach it.                                                        ;
*==============================================================================;

*------------------------------------------------------------------------------;
* 1. Autocall macro library (similar to loading user-written code).           ;
*------------------------------------------------------------------------------;
options mautosource sasautos = ("C:\my_macros", sasautos);

*------------------------------------------------------------------------------;
* 2. Format catalog (similar to attaching saved formats).                     ;
*------------------------------------------------------------------------------;
libname fmtlib "C:\my_formats";
options fmtsearch = (fmtlib work library);

* After these statements, macros in C:\my_macros and formats stored in fmtlib ;
* are available for the rest of the session.                                  ;

SAS does not have a single one-line equivalent of R’s library() call.
That is because most of what we use in clinical programming (DATA step, PROC MEANS, PROC SORT, PROC TRANSPOSE and so on) is already compiled into SAS and available by default.
For user-written autocall macros, we point SAS at the folder with options mautosource sasautos = (...).
This is the closest match to library() — it extends the session so new function-like code becomes callable by plain name.
For user-defined formats, we use libname and options fmtsearch = (...) together to tell SAS where to look.
Nothing new gets copied into the program. SAS just becomes aware of the extra location, the same way R becomes aware of a package after library().
We put these statements near the top of the program, just like library() calls sit at the top of an R script.

R (tidyverse)+

#==============================================================================
# Load the tidyverse meta-package
#==============================================================================
library(tidyverse)

#------------------------------------------------------------------------------
# Optional: hide startup messages (useful in reports and Rmd chunks)
#------------------------------------------------------------------------------
suppressPackageStartupMessages(library(tidyverse))

#------------------------------------------------------------------------------
# Optional: install once if not already installed, then load
#------------------------------------------------------------------------------
if (!requireNamespace("tidyverse", quietly = TRUE)) {
  install.packages("tidyverse")
}
library(tidyverse)

To load the whole tidyverse in one call, we use library(tidyverse). It is a meta-package that pulls in dplyr, tidyr, readr, purrr, ggplot2, tibble, stringr and forcats together.
After that single call, we can reach for filter(), mutate(), ggplot() and the rest directly by name for the remainder of the session.
If we want to hide the “Attaching packages…” banner that tidyverse prints on load, we wrap the call in suppressPackageStartupMessages().
That is most useful when knitting reports or running non-interactive scripts, where we do not want the banner in the output.
To make a script resilient on another machine, we wrap the load in an if (!requireNamespace(...)) guard. On first run it installs the package; on every run afterwards it is a no-op.
Handy when sharing with a teammate who may not have the package yet.
One gotcha when combining packages: if two loaded packages both define a function with the same name, only the most recently loaded version answers to the plain name.
A common example is filter. Base R already has stats::filter(), and when we load dplyr it brings its own dplyr::filter().
R calls this “masking” and prints a Conflicts note at load time listing every function that got shadowed.
We read those messages carefully.
If the note says dplyr::filter() masks stats::filter(), from then on a bare filter(...) call is the dplyr one.
If we actually want the base R version, we have to write stats::filter(...) explicitly.

Happy learning!

Search lessons

Introduction

Load an R package (library)