Elegant way to check for missing packages and install them?

Question

I seem to be sharing a lot of code with coauthors these days. Many of them are novice/intermediate R users and don't realize that they have to install packages they don't already have.

Is there an elegant way to call installed.packages(), compare that to the ones I am loading and install if missing?

Answer 1

Yes. If you have your list of packages, compare it to the output from installed.packages()[,"Package"] and install the missing packages. Something like this:

list.of.packages <- c("ggplot2", "Rcpp")
new.packages <- list.of.packages[!(list.of.packages %in% installed.packages()[,"Package"])]
if(length(new.packages)) install.packages(new.packages)

Otherwise:

If you put your code in a package and make them dependencies, then they will automatically be installed when you install your package.

Answer 2

Dason K. and I have the pacman package that can do this nicely. The function p_load in the package does this. The first line is just to ensure that pacman is installed.

if (!require("pacman")) install.packages("pacman")
pacman::p_load(package1, package2, package_n)

Answer 3

You can just use the return value of require:

if(!require(somepackage)){
    install.packages("somepackage")
    library(somepackage)
}

I use library after the install because it will throw an exception if the install wasn't successful or the package can't be loaded for some other reason. You make this more robust and reuseable:

dynamic_require <- function(package){
  if(eval(parse(text=paste("require(",package,")")))) return(TRUE)
  
  install.packages(package)
  return(eval(parse(text=paste("require(",package,")"))))
}

The downside to this method is that you have to pass the package name in quotes, which you don't do for the real require.

Answer 4

A lot of the answers above (and on duplicates of this question) rely on installed.packages which is bad form. From the documentation:

This can be slow when thousands of packages are installed, so do not use this to find out if a named package is installed (use system.file or find.package) nor to find out if a package is usable (call require and check the return value) nor to find details of a small number of packages (use packageDescription). It needs to read several files per installed package, which will be slow on Windows and on some network-mounted file systems.

So, a better approach is to attempt to load the package using require and and install if loading fails (require will return FALSE if it isn't found). I prefer this implementation:

using<-function(...) {
    libs<-unlist(list(...))
    req<-unlist(lapply(libs,require,character.only=TRUE))
    need<-libs[req==FALSE]
    if(length(need)>0){ 
        install.packages(need)
        lapply(need,require,character.only=TRUE)
    }
}

which can be used like this:

using("RCurl","ggplot2","jsonlite","magrittr")

This way it loads all the packages, then goes back and installs all the missing packages (which if you want, is a handy place to insert a prompt to ask if the user wants to install packages). Instead of calling install.packages separately for each package it passes the whole vector of uninstalled packages just once.

Here's the same function but with a windows dialog that asks if the user wants to install the missing packages

using<-function(...) {
    libs<-unlist(list(...))
    req<-unlist(lapply(libs,require,character.only=TRUE))
    need<-libs[req==FALSE]
    n<-length(need)
    if(n>0){
        libsmsg<-if(n>2) paste(paste(need[1:(n-1)],collapse=", "),",",sep="") else need[1]
        print(libsmsg)
        if(n>1){
            libsmsg<-paste(libsmsg," and ", need[n],sep="")
        }
        libsmsg<-paste("The following packages could not be found: ",libsmsg,"\n\r\n\rInstall missing packages?",collapse="")
        if(winDialog(type = c("yesno"), libsmsg)=="YES"){       
            install.packages(need)
            lapply(need,require,character.only=TRUE)
        }
    }
}

Answer 5

if (!require('ggplot2')) install.packages('ggplot2'); library('ggplot2')

"ggplot2" is the package. It checks to see if the package is installed, if it is not it installs it. It then loads the package regardless of which branch it took.

Answer 6

TL;DR you can use find.package() for this.

Almost all the answers here rely on either (1) require() or (2) installed.packages() to check if a given package is already installed or not.

I'm adding an answer because these are unsatisfactory for a lightweight approach to answering this question.

• require has the side effect of loading the package's namespace, which may not always be desirable
• installed.packages is a bazooka to light a candle -- it will check the universe of installed packages first, then we check if our one (or few) package(s) are "in stock" at this library. No need to build a haystack just to find a needle.

This answer was also inspired by @ArtemKlevtsov's great answer in a similar spirit on a duplicated version of this question. He noted that system.file(package=x) can have the desired affect of returning '' if the package isn't installed, and something with nchar > 1 otherwise.

If we look under the hood of how system.file accomplishes this, we can see it uses a different base function, find.package, which we could use directly:

# a package that exists
find.package('data.table', quiet=TRUE)
# [1] "/Library/Frameworks/R.framework/Versions/4.0/Resources/library/data.table"

# a package that does not
find.package('InstantaneousWorldPeace', quiet=TRUE)
# character(0)

We can also look under the hood at find.package to see how it works, but this is mainly an instructive exercise -- the only ways to slim down the function that I see would be to skip some robustness checks. But the basic idea is: look in .libPaths() -- any installed package pkg will have a DESCRIPTION file at file.path(.libPaths(), pkg), so a quick-and-dirty check is file.exists(file.path(.libPaths(), pkg, 'DESCRIPTION').

Answer 7

This solution will take a character vector of package names and attempt to load them, or install them if loading fails. It relies on the return behaviour of require to do this because...

require returns (invisibly) a logical indicating whether the required package is available

Therefore we can simply see if we were able to load the required package and if not, install it with dependencies. So given a character vector of packages you wish to load...

foo <- function(x){
  for( i in x ){
    #  require returns TRUE invisibly if it was able to load package
    if( ! require( i , character.only = TRUE ) ){
      #  If package was not able to be loaded then re-install
      install.packages( i , dependencies = TRUE )
      #  Load package after installing
      require( i , character.only = TRUE )
    }
  }
}

#  Then try/install packages...
foo( c("ggplot2" , "reshape2" , "data.table" ) )

Answer 8

Although the answer of Shane is really good, for one of my project I needed to remove the ouput messages, warnings and install packages automagically. I have finally managed to get this script:

InstalledPackage <- function(package) 
{
    available <- suppressMessages(suppressWarnings(sapply(package, require, quietly = TRUE, character.only = TRUE, warn.conflicts = FALSE)))
    missing <- package[!available]
    if (length(missing) > 0) return(FALSE)
    return(TRUE)
}

CRANChoosen <- function()
{
    return(getOption("repos")["CRAN"] != "@CRAN@")
}

UsePackage <- function(package, defaultCRANmirror = "http://cran.at.r-project.org") 
{
    if(!InstalledPackage(package))
    {
        if(!CRANChoosen())
        {       
            chooseCRANmirror()
            if(!CRANChoosen())
            {
                options(repos = c(CRAN = defaultCRANmirror))
            }
        }

        suppressMessages(suppressWarnings(install.packages(package)))
        if(!InstalledPackage(package)) return(FALSE)
    }
    return(TRUE)
}

Use:

libraries <- c("ReadImages", "ggplot2")
for(library in libraries) 
{ 
    if(!UsePackage(library))
    {
        stop("Error!", library)
    }
}

Answer 9

# List of packages for session
.packages = c("ggplot2", "plyr", "rms")

# Install CRAN packages (if not already installed)
.inst <- .packages %in% installed.packages()
if(length(.packages[!.inst]) > 0) install.packages(.packages[!.inst])

# Load packages into session 
lapply(.packages, require, character.only=TRUE)

Answer 10

'Packrat' has been replaced in RStudio by 'renv' now. See comment bellow. So my answer is no longer relavent.

------------------------------------  Use packrat so that the shared libraries are exactly the same and not changing other's environment.

In terms of elegance and best practice I think you're fundamentally going about it the wrong way. The package packrat was designed for these issues. It is developed by RStudio by Hadley Wickham. Instead of them having to install dependencies and possibly mess up someone's environment system, packrat uses its own directory and installs all the dependencies for your programs in there and doesn't touch someone's environment.

Packrat is a dependency management system for R.

R package dependencies can be frustrating. Have you ever had to use trial-and-error to figure out what R packages you need to install to make someone else’s code work–and then been left with those packages globally installed forever, because now you’re not sure whether you need them? Have you ever updated a package to get code in one of your projects to work, only to find that the updated package makes code in another project stop working?

We built packrat to solve these problems. Use packrat to make your R projects more:

• Isolated: Installing a new or updated package for one project won’t break your other projects, and vice versa. That’s because packrat gives each project its own private package library.
• Portable: Easily transport your projects from one computer to another, even across different platforms. Packrat makes it easy to install the packages your project depends on.
• Reproducible: Packrat records the exact package versions you depend on, and ensures those exact versions are the ones that get installed wherever you go.

https://rstudio.github.io/packrat/

Answer 11

This is the purpose of the rbundler package: to provide a way to control the packages that are installed for a specific project. Right now the package works with the devtools functionality to install packages to your project's directory. The functionality is similar to Ruby's bundler.

If your project is a package (recommended) then all you have to do is load rbundler and bundle the packages. The bundle function will look at your package's DESCRIPTION file to determine which packages to bundle.

library(rbundler)
bundle('.', repos="http://cran.us.r-project.org")

Now the packages will be installed in the .Rbundle directory.

If your project isn't a package, then you can fake it by creating a DESCRIPTION file in your project's root directory with a Depends field that lists the packages that you want installed (with optional version information):

Depends: ggplot2 (>= 0.9.2), arm, glmnet

Here's the github repo for the project if you're interested in contributing: rbundler.

Answer 12

You can simply use the setdiff function to get the packages that aren't installed and then install them. In the sample below, we check if the ggplot2 and Rcpp packages are installed before installing them.

unavailable <- setdiff(c("ggplot2", "Rcpp"), rownames(installed.packages()))
install.packages(unavailable)

In one line, the above can be written as:

install.packages(setdiff(c("ggplot2", "Rcpp"), rownames(installed.packages())))

Answer 13

The current version of RStudio (>=1.2) includes a feature to detect missing packages in library() and require() calls, and prompts the user to install them:

Detect missing R packages

Many R scripts open with calls to library() and require() to load the packages they need in order to execute. If you open an R script that references packages that you don’t have installed, RStudio will now offer to install all the needed packages in a single click. No more typing install.packages() repeatedly until the errors go away!
https://blog.rstudio.com/2018/11/19/rstudio-1-2-preview-the-little-things/

This seems to address the original concern of OP particularly well:

Many of them are novice/intermediate R users and don't realize that they have to install packages they don't already have.

Answer 14

Sure.

You need to compare 'installed packages' with 'desired packages'. That's very close to what I do with CRANberries as I need to compare 'stored known packages' with 'currently known packages' to determine new and/or updated packages.

So do something like

AP <- available.packages(contrib.url(repos[i,"url"]))   # available t repos[i]

to get all known packages, simular call for currently installed packages and compare that to a given set of target packages.

Answer 15

Today, I stumbled on two handy function provided by the rlang package, namely, is_installed() and check_installed().

From the help page (emphasis added):

These functions check that packages are installed with minimal side effects. If installed, the packages will be loaded but not attached.

is_installed() doesn't interact with the user. It simply returns TRUE or FALSE depending on whether the packages are installed.

In interactive sessions, check_installed() asks the user whether to install missing packages. If the user accepts, the packages are installed [...]. If the session is non interactive or if the user chooses not to install the packages, the current evaluation is aborted.

interactive()
#> [1] FALSE
rlang::is_installed(c("dplyr"))
#> [1] TRUE
rlang::is_installed(c("foobarbaz"))
#> [1] FALSE
rlang::check_installed(c("dplyr"))
rlang::check_installed(c("foobarbaz"))
#> Error:
#> ! The package `foobarbaz` is required.

^{Created on 2022-03-25 by the reprex package (v2.0.1)}

Answer 16

The following simple function works like a charm:

  usePackage<-function(p){
      # load a package if installed, else load after installation.
      # Args:
      #   p: package name in quotes

      if (!is.element(p, installed.packages()[,1])){
        print(paste('Package:',p,'Not found, Installing Now...'))
        install.packages(p, dep = TRUE)}
      print(paste('Loading Package :',p))
      require(p, character.only = TRUE)  
    }

(not mine, found this on the web some time back and had been using it since then. not sure of the original source)

Answer 17

I use following function to install package if require("<package>") exits with package not found error. It will query both - CRAN and Bioconductor repositories for missing package.

Adapted from the original work by Joshua Wiley, http://r.789695.n4.nabble.com/Install-package-automatically-if-not-there-td2267532.html

install.packages.auto <- function(x) { 
  x <- as.character(substitute(x)) 
  if(isTRUE(x %in% .packages(all.available=TRUE))) { 
    eval(parse(text = sprintf("require(\"%s\")", x)))
  } else { 
    #update.packages(ask= FALSE) #update installed packages.
    eval(parse(text = sprintf("install.packages(\"%s\", dependencies = TRUE)", x)))
  }
  if(isTRUE(x %in% .packages(all.available=TRUE))) { 
    eval(parse(text = sprintf("require(\"%s\")", x)))
  } else {
    source("http://bioconductor.org/biocLite.R")
    #biocLite(character(), ask=FALSE) #update installed packages.
    eval(parse(text = sprintf("biocLite(\"%s\")", x)))
    eval(parse(text = sprintf("require(\"%s\")", x)))
  }
}

Example:

install.packages.auto(qvalue) # from bioconductor
install.packages.auto(rNMF) # from CRAN

PS: update.packages(ask = FALSE) & biocLite(character(), ask=FALSE) will update all installed packages on the system. This can take a long time and consider it as a full R upgrade which may not be warranted all the time!

Answer 18

Thought I'd contribute the one I use:

testin <- function(package){if (!package %in% installed.packages())    
install.packages(package)}
testin("packagename")

Answer 19

I have implemented the function to install and load required R packages silently. Hope might help. Here is the code:

# Function to Install and Load R Packages
Install_And_Load <- function(Required_Packages)
{
    Remaining_Packages <- Required_Packages[!(Required_Packages %in% installed.packages()[,"Package"])];

    if(length(Remaining_Packages)) 
    {
        install.packages(Remaining_Packages);
    }
    for(package_name in Required_Packages)
    {
        library(package_name,character.only=TRUE,quietly=TRUE);
    }
}

# Specify the list of required packages to be installed and load    
Required_Packages=c("ggplot2", "Rcpp");

# Call the Function
Install_And_Load(Required_Packages);

Answer 20

Quite basic one.

pkgs = c("pacman","data.table")
if(length(new.pkgs <- setdiff(pkgs, rownames(installed.packages())))) install.packages(new.pkgs)

Answer 21

Regarding your main objective " to install libraries they don't already have. " and regardless of using " instllaed.packages() ". The following function mask the original function of require. It tries to load and check the named package "x" , if it's not installed, install it directly including dependencies; and lastly load it normaly. you rename the function name from 'require' to 'library' to maintain integrity . The only limitation is packages names should be quoted.

require <- function(x) { 
  if (!base::require(x, character.only = TRUE)) {
  install.packages(x, dep = TRUE) ; 
  base::require(x, character.only = TRUE)
  } 
}

So you can load and installed package the old fashion way of R. require ("ggplot2") require ("Rcpp")

Answer 22

 48 lapply_install_and_load <- function (package1, ...)
 49 {
 50     #
 51     # convert arguments to vector
 52     #
 53     packages <- c(package1, ...)
 54     #
 55     # check if loaded and installed
 56     #
 57     loaded        <- packages %in% (.packages())
 58     names(loaded) <- packages
 59     #
 60     installed        <- packages %in% rownames(installed.packages())
 61     names(installed) <- packages
 62     #
 63     # start loop to determine if each package is installed
 64     #
 65     load_it <- function (p, loaded, installed)
 66     {
 67         if (loaded[p])
 68         {
 69             print(paste(p, "loaded"))
 70         }
 71         else
 72         {
 73             print(paste(p, "not loaded"))
 74             if (installed[p])
 75             {
 76                 print(paste(p, "installed"))
 77                 do.call("library", list(p))
 78             }
 79             else
 80             {
 81                 print(paste(p, "not installed"))
 82                 install.packages(p)
 83                 do.call("library", list(p))
 84             }
 85         }
 86     }
 87     #
 88     lapply(packages, load_it, loaded, installed)
 89 }

Answer 23

source("https://bioconductor.org/biocLite.R")
if (!require("ggsci")) biocLite("ggsci")

Answer 24

Using lapply family and anonymous function approach you may:

1. Try to attach all listed packages.
2. Install missing only (using || lazy evaluation).
3. Attempt to attach again those were missing in step 1 and installed in step 2.

4. Print each package final load status (TRUE / FALSE).

   req <- substitute(require(x, character.only = TRUE))
   lbs <- c("plyr", "psych", "tm")
   sapply(lbs, function(x) eval(req) || {install.packages(x); eval(req)})
   
   plyr psych    tm 
   TRUE  TRUE  TRUE 
   

Answer 25

I realize this is an older question, but much of the technology in R has been greatly improved in recent months and years to address this exact issue. The generally accepted and modern approach for handling the sharing of code with package dependencies that may or may not be installed on a collaborator's computer or the target computer is to use the renv package ("renv" is short for "Reproducible Environment") which was first made available as renv 0.8.0 in late October of 2019. Version 1.0.0 was officially released on July 7, 2023, and, as of this post, the current CRAN release is Version 1.0.2, which was made available last month on August 15, 2023.

This package greatly aids R collaborators in sharing code, packages (their exact version and their dependencies and their versions), and even the tracking of the exact version of R that was used to build and execute code in an R project.

To use it, typically the initial programmer creates an R project in R Studio/Posit and then initializes renv within the project using the command renv::init(). When this is executed, R does the following:

1. Creates an project specific package library within the project, rather than using the general/global library or libraries (e.g., .libPaths()). As part of this step, renv also creates an renv directory which contains some settings and files that most users don't generally bother with as renv takes care of this for you.
2. Creates an .Rprofile file, which is a project-specific R project profile that configures R to use the project specific library created in step 1 when you open the project.
3. Creates a lockfile, called renv.lock in the project home directory, which logs details about all the packages and versions used by the project -- this is what makes using renv the most attractive option for sharing code with package and package/version dependencies.

After initializing the project environment with the init() command, the initial programmer develops and writes code as typical, installing packages in the usual manner (e.g., by simply entering install.packages("PackageName") into the console) along they way as they are needed. These packages are stored into the project library, however, rather than any global library. When the initial programmer feels the code is in a satisfactory state to share the entire project, he simply issues the renv::snapshot() command which simply updates the lockfile created in step 3 above, with the names, version numbers, etc. of all the packages used in the project.

At this point the initial programmer can share the project with other collaborators (ideally this is done through GitHub, Bitbucket, Azure DevOps, etc.). When any new collaborators open the R project, they simply enter the command renv::restore() at the console. This will install the exact same version of every package used in the project (and it will be restored into the project library rather than the user's global package library).

This is an excellent workflow because it will ensure all users are working with the same version of the packages the initial developer used. Without renv, it's possible that a collaborator has a totally different version of an R package installed in their global library, and it may function differently from the initial developer's package. This could result in the code breaking, or worse yet, running, but resulting in different output, possibly unbeknownst to the end users.

As perfectly described on the Introduction to renv vignette, renv benefits users by ensuring package isolation, reproducibility across users and environments, and portability. renv() also enjoys some other benefits like caching of R packages so that they don't have to be downloaded multiple times when used across projects. It also works with Python and Python packages. If you also want to ensure that every detail of an enviroment—including the operating system and other external (to R/R Studio) dependencies—are in place, you can easily do so too by combining renv with a Docker image to fully containerize a solution!

One final note: if renv seems similar to packrat, that's because it is, somewhat. In fact the deverlopers of rev first attempted to build a reproducible environment package using packrat but gave up and "rolled their own" after struggling with its limitations. But it's vastly better too, I'd argue. The good news is, if you're already using packrat and want to upgrade to this modern tool, you can easily do so with the renv::migrate() command!

Good luck and happy collaborating!

Answer 26

I use the following which will check if package is installed and if dependencies are updated, then loads the package.

p<-c('ggplot2','Rcpp')
install_package<-function(pack)
{if(!(pack %in% row.names(installed.packages())))
{
  update.packages(ask=F)
  install.packages(pack,dependencies=T)
}
 require(pack,character.only=TRUE)
}
for(pack in p) {install_package(pack)}

completeFun <- function(data, desiredCols) {
  completeVec <- complete.cases(data[, desiredCols])
  return(data[completeVec, ])
}

Answer 27

Here's my code for it:

packages <- c("dplyr", "gridBase", "gridExtra")
package_loader <- function(x){
    for (i in 1:length(x)){
        if (!identical((x[i], installed.packages()[x[i],1])){
            install.packages(x[i], dep = TRUE)
        } else {
            require(x[i], character.only = TRUE)
        }
    }
}
package_loader(packages)

Answer 28

library <- function(x){
  x = toString(substitute(x))
if(!require(x,character.only=TRUE)){
  install.packages(x)
  base::library(x,character.only=TRUE)
}}

This works with unquoted package names and is fairly elegant (cf. GeoObserver's answer)

Answer 29

In my case, I wanted a one liner that I could run from the commandline (actually via a Makefile). Here is an example installing "VGAM" and "feather" if they are not already installed:

R -e 'for (p in c("VGAM", "feather")) if (!require(p, character.only=TRUE)) install.packages(p, repos="http://cran.us.r-project.org")'

From within R it would just be:

for (p in c("VGAM", "feather")) if (!require(p, character.only=TRUE)) install.packages(p, repos="http://cran.us.r-project.org")

There is nothing here beyond the previous solutions except that:

• I keep it to a single line
• I hard code the repos parameter (to avoid any popups asking about the mirror to use)
• I don't bother to define a function to be used elsewhere

Also note the important character.only=TRUE (without it, the require would try to load the package p).

Answer 30

Let me share a bit of madness:

c("ggplot2","ggsci", "hrbrthemes", "gghighlight", "dplyr") %>%  # What will you need to load for this script?
  (function (x) ifelse(t =!(x %in% installed.packages()), 
    install.packages(x[t]),
    lapply(x, require)))