diff options
Diffstat (limited to '.github')
-rw-r--r-- | .github/for_maintainers/tests_strategy.md | 28 | ||||
-rw-r--r-- | .github/how_to_compile_gudhi_in_a_conda_env.md | 93 |
2 files changed, 107 insertions, 14 deletions
diff --git a/.github/for_maintainers/tests_strategy.md b/.github/for_maintainers/tests_strategy.md index c25acf9b..338d4282 100644 --- a/.github/for_maintainers/tests_strategy.md +++ b/.github/for_maintainers/tests_strategy.md @@ -8,13 +8,13 @@ The aim is to help maintainers to anticipate third parties modifications, update ### Linux -As all the third parties are already installed (thanks to docker), the compilations has been seperated by categories to be parallelized: +As all the third parties are already installed (thanks to docker), the compilations have been separated in categories to be parallelized: * examples (C++) * tests (C++) * utils (C++) * doxygen (C++ documentation that is available in the artefacts) -* python (including documentation and code coverage that are available in the artefacts) +* python (including documentation and code coverage that are available in the artefacts; here the WITH_GUDHI_REMOTE_TEST option is enabled which adds datasets fetching test) (cf. `.circleci/config.yml`) @@ -25,9 +25,9 @@ Without CGAL, and, with or without Eigen builds are performed inside the docker #### Update docker images -C++ third parties installation are done thanks to apt on Ubuntu latest LTS. +C++ third parties installation is done thanks to apt on Ubuntu latest LTS. -Docker images need to be rebuild and push each time `.github/build-requirements`, `.github/test-requirements`, when a new third party is added, when a new CGAL version improves gudhi performances, ... +Docker images need to be rebuilt and pushed each time `.github/build-requirements`, `.github/test-requirements`, when a new third party is added, when a new CGAL version improves gudhi performances, ... ```bash docker build -f Dockerfile_for_circleci_image -t gudhi/ci_for_gudhi:latest . @@ -39,35 +39,35 @@ docker push gudhi/ci_for_gudhi_wo_cgal:latest ### Windows -The compilations are not parallelized, as installation time (about 30 minutes) is too much compare to +The compilations are not parallelized, as installation time (about 30 minutes) is too much compared to build and tests timings (about 30 minutes). Builds and tests include: * examples (C++) * tests (C++) * utils (C++) -* python +* python (here the WITH_GUDHI_REMOTE_TEST option is enabled which adds datasets fetching test) Doxygen (C++) is not generated. (cf. `azure-pipelines.yml`) -C++ third parties installation are done thanks to [vcpkg](https://github.com/microsoft/vcpkg/). -In case of installation issue, check in [vcpkg issues](https://github.com/microsoft/vcpkg/issues). +C++ third parties installation is done thanks to [vcpkg](https://github.com/microsoft/vcpkg/). +In case of an installation issue, check in [vcpkg issues](https://github.com/microsoft/vcpkg/issues). ### OSx The compilations are not parallelized, but they should, as installation time (about 4 minutes) is -negligeable compare to build and tests timings (about 30 minutes). Builds and tests include: +negligible compared to build and tests timings (about 30 minutes). Builds and tests include: * examples (C++) * tests (C++) * utils (C++) -* python +* python (here the WITH_GUDHI_REMOTE_TEST option is enabled which adds datasets fetching test) * Doxygen (C++) (cf. `azure-pipelines.yml`) -C++ third parties installation are done thanks to [brew](https://formulae.brew.sh/formula/). -In case of installation issue, check in formula issues. +C++ third parties installation is done thanks to [brew](https://formulae.brew.sh/formula/). +In case of an installation issue, check in formula issues. ## Pip packaging @@ -80,9 +80,9 @@ Only the Linux pip package is based on a docker image (`gudhi/pip_for_gudhi` bas ### Update docker image -C++ third parties installation are done thanks to yum on an image based on `quay.io/pypa/manylinux2014_x86_64`. +C++ third parties installation is done thanks to yum on an image based on `quay.io/pypa/manylinux2014_x86_64`. -Docker image need to be rebuild and push each time `.github/build-requirements`, when a new third party is added, when a new CGAL version improves gudhi performances, ... +Docker image needs to be rebuilt and pushed each time `.github/build-requirements`, when a new third party is added, when a new CGAL version improves gudhi performances, ... As `.github/test-requirements` is not installed, no need to rebuild image when this file is modified. ```bash diff --git a/.github/how_to_compile_gudhi_in_a_conda_env.md b/.github/how_to_compile_gudhi_in_a_conda_env.md new file mode 100644 index 00000000..4acfca2e --- /dev/null +++ b/.github/how_to_compile_gudhi_in_a_conda_env.md @@ -0,0 +1,93 @@ +# Install a conda development environment to compile GUDHI + +## Install miniconda + +Download the [installer](https://docs.conda.io/en/latest/miniconda.html) required by your system and follow the [instructions](https://conda.io/projects/conda/en/latest/user-guide/install/index.html). + +## Create a dedicated environment + +```bash +conda install -c conda-forge mamba # installation with mamba is faster +conda create --name gudhi +conda activate gudhi +mamba install -c conda-forge python cmake doxygen eigen cgal-cpp +``` + +Some of the requirements are in the gudhi-devel repository (please refer to +[how to use github to contribute to gudhi](how_to_use_github_to_contribute_to_gudhi.md)). +Once the gudhi-devel repository is cloned on your machine (`git clone...`) - let's call it `/workdir/gudhi-devel` i.e. - +and once the submodules are initialised (`git submodule update --init`): + +```bash +pip install -r ext/gudhi-deploy/build-requirements.txt +pip install -r ext/gudhi-deploy/test-requirements.txt # pytorch can be painful to install - not mandatory +``` + +## Compilation + +In order to compile all c++ utilities, examples, benchmarks, unitary tests, and python module: +```bash +cd /workdir/gudhi-devel +rm -rf build; mkdir build # /!\ any existing build folder will be removed +cd build +# To build all even examples and benchmarks +cmake -DCMAKE_BUILD_TYPE=Release -DCMAKE_PREFIX_PATH=$CONDA_PREFIX -DWITH_GUDHI_EXAMPLE=ON -DWITH_GUDHI_BENCHMARK=ON .. +``` + +### Specific python compilation + +In order to compile only python module +```bash +cd /workdir/gudhi-devel +rm -rf build; mkdir build # /!\ any existing build folder will be removed +cd build +cmake -DCMAKE_BUILD_TYPE=Release -DCMAKE_PREFIX_PATH=$CONDA_PREFIX .. +cd src/python +# To build python module in parallel +python setup.py build_ext -j 16 --inplace # 16 is the number of CPU that are used to compile the python module. Can be any other value. +# to clean the build +# python setup.py clean --all +``` + +In order to use freshly compiled gudhi python module: +```bash +PYTHONPATH=/workdir/gudhi-devel/build/src/python python # or ipython, jupyter, ... +``` + +### Specific C++ documentation generation + +```bash +cd /workdir/gudhi-devel +rm -rf build; mkdir build # /!\ any existing build folder will be removed +cd build +# python OFF to prevent python modules search makes cmake faster +cmake -DCMAKE_BUILD_TYPE=Release -DCMAKE_PREFIX_PATH=$CONDA_PREFIX -DWITH_GUDHI_PYTHON=OFF -DUSER_VERSION_DIR=version .. +make user_version; +cd version +mkdir build +cd build +# python OFF to prevent python modules search makes cmake faster +cmake -DCMAKE_BUILD_TYPE=Release -DCMAKE_PREFIX_PATH=$CONDA_PREFIX -DWITH_GUDHI_PYTHON=OFF .. +make doxygen 2>&1 | tee dox.log +grep warning dox.log # Warnings can be lost with parallel doxygen +firefox html/index.html # [optional] To display the c++ documentation. Anything else than firefox can be used. +``` + +### Specific python documentation generation + +```bash +cd /workdir/gudhi-devel +rm -rf build; mkdir build # /!\ any existing build folder will be removed +cd build +# python OFF to prevent python modules search makes cmake faster - it is the next cmake call in user version that matters +cmake -DCMAKE_BUILD_TYPE=Release -DCMAKE_PREFIX_PATH=$CONDA_PREFIX -DWITH_GUDHI_PYTHON=OFF -DUSER_VERSION_DIR=version .. +make user_version; +cd version +mkdir build +cd build +cmake -DCMAKE_BUILD_TYPE=Release -DCMAKE_PREFIX_PATH=$CONDA_PREFIX .. +cd python +# To build python module in parallel +python setup.py build_ext -j 16 --inplace # 16 is the number of CPU that are used to compile the python module. Can be any other value. +firefox sphinx/index.html # [optional] To display the python documentation. Anything else than firefox can be used. +```
\ No newline at end of file |