Building and installing ISCE3
Installing from conda-forge
If you just want to use ISCE3, you can avoid building it yourself by installing from our conda-forge package.
conda install -c conda-forge isce3
Install from source with pip
If you need to develop on ISCE3, you will need to build ISCE3 from source using pip.
git clone https://github.com/isce-framework/isce3
cd isce3
Make sure you have all the requirements for ISCE3 installed, and all the necessary packages available in your Python environment. If you don't have such an environment already, we recommend using miniforge to install all of these packages to a self-contained conda environment.
Using our list of dependencies from environment.yml, create an environment
for ISCE3 and activate it:
conda env create -f environment.yml
conda activate isce3
Now that you have all the prerequisites, it is time to run the build. Pip should handle detecting your python installation and the prerequisites you just installed.
pip install .
If that command completes successfully, ISCE3 will now be available in your environment.
python3 -c 'import isce3; print(isce3.__version__)'
Building with CMake (Advanced)
Developers who are familiar with CMake can run the standalone build with that instead of pip. This is a more advanced build procedure that allows more control over the build process, but requires more configuration.
mkdir build && cd build
export CUDAHOSTCXX=$CXX
export CUDACXX=/usr/local/cuda/bin/nvcc # or wherever your CUDA compiler is located
cmake .. -GNinja -DWITH_CUDA=ON -DCMAKE_INSTALL_PREFIX=./install
ninja install
| Environment variable | Description |
|---|---|
| CUDACXX | Path to CUDA compiler |
| CUDAHOSTCXX | Host compiler used for CUDA code, should match $CXX |
| CMake option | Description |
|---|---|
| -DWITH_CUDA=ON/OFF | Enable/disable CUDA support |
After installing, make sure to run the unit tests to check that ISCE3 is behaving as expected:
ctest --output-on-failure
Tip
If many of these unit tests are failing, it is a good indicator that some Python package is missing from your environment, or something has been configured incorrectly.
Once ISCE3 is installed and passing unit tests, making it available for import is more difficult than when using scikit-build-core. You will need to add the built Python packages / extensions to your PYTHONPATH, and also add the ISCE3 C++ library to your library loader path.
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$(realpath install/lib*)
export PYTHONPATH=$PYTHONPATH:$(realpath install/packages)
export DYLD_FALLBACK_LIBRARY_PATH=$DYLD_FALLBACK_LIBRARY_PATH:$(realpath install/lib*)
export PYTHONPATH=$PYTHONPATH:$(realpath install/packages)
ISCE3 should now be available in your environment.
python3 -c 'import isce3; print(isce3.__version__)'