In the following replace <os-models> with the installation directory an <user> with your home directory.
This is a maven project:
To build: mvn install will download all dependencies and compile the project
To build without running tests mvn -Dmaven.test.skip=true -DskipTests install
To force update mvn clean install -U or mvn -Dmaven.test.skip=true -DskipTests clean install -U
This will build a uber jar file in the target directory, target/uber-os-models-1.0-SNAPSHOT.jar that can be loaded by the Spark notebook, or which will run in the spark shell. The jar does not include spark libraries. A second assembly jar target/os-models-1.0-SNAPSHOT-all.jar includes all spark libararies and is suitable for running standalone applications
After building the project install the spark notebook as described below. You should then be able to view notebooks
Source available from git clone https://github.com/spark-notebook/spark-notebook.git
Get binary distribution from http://spark-notebook.io/. Project spark version is 2.1.1, Scala version is 2.11.8
To start use bin/spark-notebook (*NIX) or bin\spark-notebook.bat (Windows). In order that the notebook find the project jar file set the environment variable EXTRA_CLASSPATH to the full path of the uber jar, or pass it to the notebook command:
EXTRA_CLASSPATH=<os-models>/target/uber-os-models-1.0-SNAPSHOT.jar ./bin/spark-notebook
Open your browser to localhost:9001
To use the notebooks in project directory you will need to copy them into the notebooks folder under the spark notebook install, or alternatively create a symbolic link from this project to the notebook install. E.g, run a command like this the the project directory.
ln -s `pwd`/notebooks /home/packages/spark-notebook-0.7.0-scala-2.11.8-spark-2.1.1-hadoop-2.7.2/notebooks/os-models
Spark comes with pure Java linear algebra libraries, but will perform better with native libraries
Linux
sudo apt-get install libatlas3-base libopenblas-base gfortran
To build Java RDKit bindings. In the shell commands below you will need to adjust paths to reflect your installation.
OS X
Does not work with a newer version of cmake (the brew make). Download and install cmake 2.8 from older version repository on cmake website (https://cmake.org/files/v2.8/) Use archived version of old brew boost.
cd /Users/gareth/packages
git clone https://github.com/rdkit/rdkit.git
brew install boost@1.55
export RD_BASE=/Users/gareth/packages/rdkit
export LD_LIBRARY_PATH=/home/packages/rdkit/lib:$LD_LIBRARY_PATH
export PATH=/usr/local/bin:$PATH
cd rdkit
mkdir build
cd build
cmake -D BOOST_ROOT=/usr/local/opt/boost@1.55 -D RDK_BUILD_SWIG_WRAPPERS=ON \
-D RDK_BUILD_INCHI_SUPPORT=ON -D RDK_BUILD_AVALON_SUPPORT=ON \
-D RDK_BUILD_PYTHON_WRAPPERS=OFF ..
make
Linux
As for OS X but (for Ubuntu 16.04) repository versions of boost and cmake are fine
Before building resolve junit issue discussed here https://sourceforge.net/p/rdkit/mailman/message/35306409/
It is easiest just to fix in /usr/share/java:
rm junit.jar
ln -s junit4-4.12.jar junit.jar
cd /home/packages
git clone https://github.com/rdkit/rdkit.git
cd rdkit
sudo apt-get install build-essential python-numpy cmake python-dev \
sqlite3 libsqlite3-dev libboost-dev libboost-system-dev libboost-thread-dev \
libboost-serialization-dev libboost-python-dev libboost-regex-dev
sudo apt-get install swig
export RD_BASE=/home/packages/rdkit
export LD_LIBRARY_PATH=/home/packages/rdkit/lib:$LD_LIBRARY_PATH
mkdir build
cd build
cmake -D RDK_BUILD_SWIG_WRAPPERS=ON -D RDK_BUILD_INCHI_SUPPORT=ON \
-D RDK_BUILD_AVALON_SUPPORT=ON -D RDK_BUILD_PYTHON_WRAPPERS=OFF ..
Copy jar file to local project repository (adjusting project path as required):
cp ../Code/JavaWrappers/gmwrapper/org.RDKit.jar \
<user>/src/os-models/repo/org/rdkit/rdkit/0.1.1/rdkit-0.1.1.jar
Move ../Code/JavaWrappers/gmwrapper/libGraphMolWrap.so to src/main/resources/lib/Linux_amd64
for native loading.
May need to revisit the native loading from the java resource path as boost libraries need to be present