Build with Visual Studio

Build with Visual Studio#

Install “Visual Studio 14 2015”, “Visual Studio 15 2017” or “Visual Studio 16 2019” with the C/C++ components and cmake.

In case you want Python-bindings, make sure to use a “Visual Studio” version that generates dlite-libraries which are compatible to the Python version you are using. For Python>=3.5, any of the above versions will work: See mscompilers for a discussion on binary compatibilty between Visual Studio compiler versions. For older versions of Python see the correct versions here: pythonwindows.

In the examples below, we use “Visual Studio 15 2017” which is the compiler version with which Python>=3.7 is build on Windows.
(Optional) Install SWIG in case you want to build Python-bindings as described here swig. Add the swig-executable to your windows PATH
(Optional) Check versions

$ swig -version SWIG Version 4.0.2 Compiled with i686-w64-mingw32-g++ [i686-w64-mingw32] Configured options: +pcre Please see http://www.swig.org for reporting bugs and further information

$ cmake -version cmake version 3.20.21032501-MSVC_2 CMake suite maintained and supported by Kitware (kitware.com/cmake).
git clone dlite to a directory of your choice

Build and test

5.1. Build from the Windows commandline

When building with Python support, this will build against your default Python. Make sure you have the following Python packages installed: numpy, pyyaml, pandas, psycopg2 Use py -0p to check the Python default.

$ mkdir .\out\build\x64-Release-vs15
$ cd .\out\build\x64-Release-vs15
$ cmake -G "Visual Studio 15 2017" ^
        -A x64 ^
        -DWITH_DOC=OFF ^
        -DWITH_HDF5=OFF ^
        -DCMAKE_CONFIGURATION_TYPES:STRING="Release" ^
        -DCMAKE_INSTALL_PREFIX:PATH="%CD%\..\..\install\x64-Release-vs15" ^
        ..\..\..

 $ cmake --build . --config Release
 $ cmake --install .
 $ ctest -C Release

5.2 Build from Visual Studio

When building with Python support, this will again build against your default Python.

Open the dlite directory with Visual Studio, it will be recognized as a CMake project. Add a CMakeSettings.json file with the following minimum contents:

{
  "configurations": [
    {
      "name": "x64-Release-vs15",
      "generator": "Visual Studio 15 2017 Win64",
      "configurationType": "Release",
      "buildRoot": "${projectDir}\\out\\build\\${name}",
      "installRoot": "${projectDir}\\out\\install\\${name}",
      "cmakeCommandArgs": "-DWITH_DOC=OFF -DWITH_HDF5=OFF",
      "buildCommandArgs": "",
      "inheritEnvironments": [ "msvc_x64_x64" ]
    }
  ]
}

Select the “x64-Release-vs15” configuration
Right click on the top-level “CMakeLists.txt” -> “configure dlite”
Right click on the top-level “CMakeLists.txt” -> “Build”
Right click on the top-level “CMakeLists.txt” -> “Run Tests”

5.3 Build against Python in a virtual environment (Recommended)

This example uses Anaconda Python and conda environemnts.

$ conda create --name=py38dlite python=3.8 numpy pyyaml pandas psycopg2
$ conda activate py38dlite

$ mkdir .\out\build\x64-Release-vs15
$ cd .\out\build\x64-Release-vs15
$ cmake -G "Visual Studio 15 2017" ^
        -A x64 ^
        -DWITH_DOC=OFF ^
        -DWITH_HDF5=OFF ^
        -DCMAKE_CONFIGURATION_TYPES:STRING="Release" ^
        -DCMAKE_INSTALL_PREFIX:PATH="%CD%\..\..\install\x64-Release-vs15" ^
        ..\..\..

 $ cmake --build . --config Release
 $ cmake --install .
 $ ctest -C Release

Install the Python package via using ./python/setup.py from sources (legacy)

$ conda create --name=py38dlite python=3.8 numpy
$ conda activate py38dlite
$ cd python
$ python setup.py install

Use pip to install the Python package from sources (Recommended)

Make sure you have pip >= 21.3 in which in-tree-build is the default. For pip > 21.0.1 and < 21.3, you need an additional flag pip --use-feature=in-tree-build install .
```
$ conda create --name=py38dlite python=3.8
$ conda activate py38dlite
$ pip install --upgrade pip
$ pip -V # >= 21.3?
$ cd python
$ pip install .
```

Install dlite-python from GitHub sources

Make sure you have pip >= 21.3 again

$ conda create --name=py37dlite python=3.7
$ conda activate py37dlite

Install a specific commit:

$ pip install git+https://github.com/SINTEF/dlite.git@d19e93b7d69f486485e8d6c3468e6bc672d7590d#subdirectory=python

Install a specific tag:

$ pip install git+https://github.com/SINTEF/dlite.git@v0.3.2dev1#subdirectory=python

Install dlite-python via a pre-packaged wheel

Make sure you have pip >= 21.3 again

$ conda create --name=py37dlite python=3.7
$ conda activate py37dlite
$ pip install dlite_python-0.3.3-cp37-cp37m-win_amd64.whl