Miscellaneous Details

Project Overview

The project is implemented in form of a small Python package. In the root directory, pyproject.toml contains the project dependencies. As a project management tool Poetry is used; poetry.lock is the corresponding dependency lockfile.

The bounded_domains directory contains the module implementation; the code is distributed into three submodules (see Reference Manual for a description of the modules and their members).

In the tests directory there are some test files that can be run with pytest. In fact, every time a commit is pushed to the repository on GitHub, the configuration in .github/workflows/ci.yml triggers the tests to be run on the instances tagged with ubuntu-latest, macos-latest, and windows-latest, each for Python version 3.7, 3.8, and 3.9.

The project also has a setup linter and code style check: for linting, flake8 configured via the .flake8 file is used, and the code style is checked and enforced via Black.

The docs directory contains the basic configuration for the documentation pages you are currently reading. The pages are deployed manually (because setting up automatic deployment did not quite pay off for this project).

And finally, demo.ipynb contains a Jupyter notebook illustrating basic usage of the package.

Note

notebook has not been specified as a dependency of bounded_domains. Be sure to pip install notebook in case you run into problems when trying to play the file.

Quadrilaterals instead of Triangles

The present implementation of PolygonalDomain relies on the passed Element objects describing triangles. Theoretically, more complex structures (like quadrilaterals) would be possible too. Depending on the concrete use case, the implementation has to be changed more or less:

• In case it is only important that users should be able to pass quadrilaterals (or general polygons), then I propose implementing Element such that internally a triangulation of the element is computed and stored. All other structures would then still use (and rely on) the fact that internally the domain is still triangulated. As an example: PolygonalDomain.closest_element() might then find that one of the internally created triangular elements is the closest one; it would be easy to return the index of the (more complex) original element instead of the determined sub-element.

• Otherwise, if for whatever reason computations have to happen with quadrilaterals, then several methods have to be changed substantially to handle the more complex situation – among them, for example, PolygonalDomain.build_mappings() (for the adjacent vertex mapping, etc.), and PolygonalDomain.distance_to_element().