Install xml2 R package on MacOS

Installing xml2 R package often fails due to missing or incompatible library issues. In this post, I describe why this problem occurs and provide two solutions to solve this problem.

What goes wrong?

xml2 R package depends on libxml2. When you install xml2 using install.packages and default options, xml2 package queries your system to determine where libxml2 headers and library files are located (more on this later). The error occurs when libxml2 is not found or a wrong version of libxml2. This is how the error looks.

> install.packages("xml2")
...
* installing *source* package ‘xml2’ ...
** package ‘xml2’ successfully unpacked and MD5 sums checked
Found pkg-config cflags and libs!
Using PKG_CFLAGS=-I/Users/<username>/anaconda/include/libxml2
Using PKG_LIBS=-L/Users/<username>/anaconda/lib -lxml2 -lz -liconv -lm
** libs
clang++ -I/usr/local/Cellar/r/3.2.2_1/R.framework/Resources/include -DNDEBUG -I/usr/local/include -I/Users/<username>/anaconda/include/libxml2 -I/usr/local/opt/gettext/include -I/usr/local/opt/readline/include -I/usr/local/opt/openssl/include -I/usr/local/include -I"/usr/local/lib/R/3.2/site-library/Rcpp/include" -I"/usr/local/lib/R/3.2/site-library/BH/include" -I/usr/local/include   -fPIC  -g -O2  -c RcppExports.cpp -o RcppExports.o
...
installing to /usr/local/lib/R/3.2/site-library/xml2/libs
...
Error in dyn.load(file, DLLpath = DLLpath, ...) :
  unable to load shared object '/usr/local/lib/R/3.2/site-library/xml2/libs/xml2.so':
  dlopen(/usr/local/lib/R/3.2/site-library/xml2/libs/xml2.so, 6): Library not loaded: libxml2.2.dylib
  Referenced from: /usr/local/lib/R/3.2/site-library/xml2/libs/xml2.so
  Reason: image not found
Error: loading failed
Execution halted

What happens behind the scenes?

xml2 package contains a bash script called configure which determines the location of libxml2. You can take a look at this configure on GitHub. This script first attempts to use xml2-config to determine the location. On my system, xml2-config is installed by anaconda.

$ which xml2-config
/Users/<username>/anaconda/bin/xml2-config

On my system, xml2-config tool points to the libxml2 installed by anaconda, which happens to be incompatible with R xml2 package.

$ xml2-config --libs
-L/Users/<username>/anaconda/lib -lxml2 -lz -liconv -lm
$ xml2-config --cflags
-I/Users/<username>/anaconda/include/libxml2

If xml2-config is not available, then configure script checks for libxml2 using the pkg-config tool.

$ which pkg-config
/usr/local/bin/pkg-config

Again, in my case, this pkg-config tool points to the MacOS system libxml2 headers and libraries. These may or may not be compatible with R xml2 package. In my case, these happened to be compatible.

$ pkg-config --cflags libxml-2.0
-I/usr/include/libxml2
$ pkg-config --libs libxml-2.0
-lxml2

We need to make sure that the /usr/include/libxml2 location actually exists. If this location does not exist, then first solve this problem as described in the footnotes¹.

Overall, the solution is based on the simple principle

Use the correct location of libxml2 headers and libraries.

There are various ways to solve this problem. I describe two ways in this post.

Possible solutions

For both solutions, we will need to download the source code of xml2 package from CRAN and install the package from source. Extract the original source code file and navigate to the extracted xml2 folder.

Solution 1

We will modify the configure script in the source code. The modification will cause configure script to use the libxml2 location provided by pkg-config. This can be done by commenting out the following relevant lines in the configure file

# Use xml2-config if available
if [ $(command -v xml2-config) ]; then
  PKGCONFIG_CFLAGS=$(xml2-config --cflags)
  PKGCONFIG_LIBS=$(xml2-config --libs)
elif [ $(command -v pkg-config) ]; then
  PKGCONFIG_CFLAGS=$(pkg-config --cflags $PKG_CONFIG_NAME)
  PKGCONFIG_LIBS=$(pkg-config --libs $PKG_CONFIG_NAME)
fi

and replacing them by these lines

if [ $(command -v pkg-config) ]; then
  PKGCONFIG_CFLAGS=$(pkg-config --cflags $PKG_CONFIG_NAME)
  PKGCONFIG_LIBS=$(pkg-config --libs $PKG_CONFIG_NAME)
fi

I recommend reading the code and then made the changes if you understand them.

Now that we have modified the source code of the package, we need to re-build a new .tar.gz file.

# Navigate to inside the xml2 folder
~/Downloads/xml2 $ R CMD build .
# This will create a xmlx_x.x.x.tar.gz file in the current folder

and then install this package from from source

# Start R in the folder that contains the xmlx_x.x.x.tar.gz file
> install.packages("xml2_0.1.2.tar.gz", repos=NULL, type="source")

If the location of libxml2 as specified by pkg-config is correct, then this solution should correctly install xml2 package.

Solution 2

If you do not want to modify the configure script and you know the correct location of libxml2 header and libraries, then you can simply perform a custom install of xml2 package by using the following command (as specified by the configure script)

# Navigate to inside the xml2 folder
R CMD INSTALL --configure-vars='INCLUDE_DIR=/usr/include LIB_DIR=/usr/lib' .

The above locations for INCLUDE_DIR and LIB_DIR correspond to the default MacOS locations but you may need to modify these. You can find out these locations using the tips provided in footnotes². This should successfully install xml2 package. This form of solution works correctly for other packages as well, albeit with some modifications³.

Footnotes

The exact options required by configure may be seen by running ./configure --help within the source directory of the package. Note that if you have run configure within the source folder using the incorect options, then you might see R CMD INSTALL fail even with the correct options. This problem can be solved by re-running configure using the correct options or getting a fresh copy of the source.