setting_up_an_rdkit_build_env_with_conda.md

These instructions have been tested on linux and the Mac, both using the system C++ compilers. Note that you need a compiler which supports C++17.

Part 0: checkout RDKit github repo

git clone https://github.com/rdkit/rdkit.git RDKit_git

this will create a directory called RDKit_git with the RDKit source.

Part 1: setup conda environment

mamba create -n py310_rdkit_build python=3.10 boost-cpp boost cairo pandas pillow freetype cmake numpy eigen matplotlib 

Part 2: run cmake and build

from inside the RDKit source directory you checked out above:

conda activate py310_rdkit_build
mkdir build_demo
cd build_demo
cmake -DRDK_BUILD_INCHI_SUPPORT=ON -DRDK_BUILD_YAEHMOP_SUPPORT=ON -DRDK_BUILD_XYZ2MOL_SUPPORT=ON ..

That should run without errors. It’ll produce a bunch of warnings, but should end with something like:

-- Generating done
-- Build files have been written to: /localhome/glandrum/RDKit_git/build_demo

Now build it:

make -j8 install

That will take a while and generate a few warnings, but shouldn’t have any errors. The last line of the output should be something like this:

-- Installing: /localhome/glandrum/RDKit_git/rdkit/RDPaths.py

Part 3: setup your environment and then run the tests

Do this from inside the RDKit source directory (not the build directory). On linux:

export RDBASE=`pwd`
export PYTHONPATH=$RDBASE
export LD_LIBRARY_PATH=$RDBASE/lib:$CONDA_PREFIX/lib

on the Mac, that is:

export RDBASE=`pwd`
export PYTHONPATH=$RDBASE
export DYLD_LIBRARY_PATH=$RDBASE/lib:$CONDA_PREFIX/lib

Now you can run the tests:

cd build_demo
ctest -j6 –output-on-failure

This will take a bit and generate a lot of output, but the last bit should look something like this:

100% tests passed, 0 tests failed out of 219

Total Test time (real) =  63.92 sec

That’s it!

Part 4: Using the environment

In order to use the code you just need to repeat the export commands shown above. I find the following aliases really useful on linux:

alias remote_rdkit='unset RDBASE;export PYTHONPATH="";export LD_LIBRARY_PATH=""'
alias local_rdkit='conda activate py310_rdkit_build;export RDBASE="/localhome/glandrum/RDKit_git";export PYTHONPATH="$RDBASE";export LD_LIBRARY_PATH="$RDBASE/lib:$CONDA_PREFIX/lib"'
alias rdkit_vers='python -c "import rdkit;print(rdkit.__version__,rdkit.__file__)"'

local_rdkit sets up the shell to use my local RDKit build. remote_rdkit clears that stuff out so that you can use a conda rdkit install rdkit_vers shows you the current rdkit version and where the files are coming from