Installing#

Using pip#

For Linux and Mac, you can simply use pip that will install a wheel bundled with the required libraries:

$ pip3 install pyorc

Note

To install on Linux, you need pip 19.0 or newer. Earlier versions are unable to handle the manylinux2010 wheels, thus they try to install the package from source.

There could be some drawbacks of the bundled libraries in the package, when using together with other Python modules. If another module is loaded into the Python runtime besides PyORC that also pre-bundles one of the required C/C++ libraries but a slightly different version, then the two libraries will collide, and the interpreter will crash with segmentation fault at some point during the execution.

It’s easy to run into this situation. For example, libprotobuf is one of required library for ORC, and it’s quite popular for other projects as well. To avoid this, you have to make sure that the very same version of the common library is used by both of the modules, and therefore you might need to build PyORC from source.

Install from source#

To install from source, the module requires the Apache ORC C++ Core library. During the extension build step, the module will build the ORC core library before building the extension module itself. It requires cmake – in addition of a suitable C++ compiler. The following steps take place during the build_ext command:

  1. Downloading the Apache ORC release package.

  2. Extracting the package to a directory named deps into the project’s root directory.

  3. Running cmake to configure the ORC C++ library.

  4. Running the make package command.

  5. Finally, moving the C++ headers, ORC example files and ORC tools to the top level of the deps directory for the setup.py and tests to find.

  6. Building the C++ extension part of PyORC.

Note

The build_ext command has a --orc-version and a --source-url parameter for changing the default ORC library version or the URL of the source zip to download respectively. It also has a --skip-orc-build flag to skip ORC library build steps.

You also need the pybind11 Python package to be installed before running the installation:

$ pip3 install pybind11
$ python3 setup.py install

After the installation completes without errors, you have the module ready to use.