Nuitka Package Configuration 

Introduction 

For packaging, and compatibility, or some Python packages need to have special considerations in Nuitka. Some will not work without certain data files, sometimes modules depend on other modules in a hidden way, and for standalone DLLs might have to be included, that are loaded dynamically and therefore also invisible.

Another are is compatibility hacks, and removing bloat from code or just making sure, you are not using an unsupported version or wrong options for a package.

To make it easier to deal with missing DLLs, implicit imports, data files, bloat etc. Nuitka has a system with Yaml files. These ship inside of it and are located under plugins/standard and are designed to be easily be extended.

The structure of the filename is always *nuitka-package.config.yml . The standard file includes all things that are not in the standard library ( stdlib ) of Python. In stdlib2 and stdlib3 there are entries for the standard library. In stdlib2 there are only those for modules that are no longer available in Python3.

If you want to use your own configuration, you can do so by passing the filename of your Yaml file to --user-package-configuration-file .

Note

If this could be interesting for the whole user base of Nuitka, please do a PR that adds it to the general files. In this way, not every user has to repeat what you just did, and we can collectively maintain it.

The YAML Configuration File 

At the beginning of the file you will find the following lines, which you can ignore, they are basically only there to silence checkers about problems that are too hard to avoid.

# yamllint disable rule:line-length
# yamllint disable rule:indentation
# yamllint disable rule:comments-indentation
# too many spelling things, spell-checker: disable
---
										

An entry in the file look like this:

- module-name: 'pandas._libs'
  implicit-imports:
    - depends:
      - 'pandas._libs.tslibs.np_datetime'
      - 'pandas._libs.tslibs.nattype'
      - 'pandas._libs.tslibs.base'
										

The module-name value is the name of the affected module. We will show and explain to you everything the other things in detail later. But the key principle is that a declaration always references a module by name.

It is also important to know that you do not have to worry about formatting. We have programmed our own tool for this, which formats everything automatically. This is executed via bin\autoformat-nuitka-source and automatically when pushing with git if you install the git hook (see Developer Manual for that).

There is also a Yaml schema file to check your files against and that in Visual Code is automatically applied to the Yaml files and that then supports you with auto-completion in Visual Code. So actually doing the change in PR form can be easier than not.

Documentation 

Data Files 

data-files:
  dest_path: '.'
  dirs:
    - 'dir1'
  patterns:
    - 'file1'
    - '*.dat'
  empty_dirs:
    - 'empty_dir'
  empty_dir_structures:
    - 'empty_dir_structure'
  when: 'win32'
										

If a module needs data files, you can get Nuitka to copy them into the output with the following features.

Features 

dest_path : target directory

dirs : all directories that should be copied

patterns : all files that should be copied (filename can be a glob pattern )