Code vendored in PyDSM

Since version 0.9.0, PyDSM has vendored the original cvxpy package by Tomas Tinoco de Rubira.

This was likely the first Python package to provide a modeling framework for convex optimization problems. It has since been discontinued and superseded by another package —also called cvxpy— developed by Steven Diamond, Eric Chu, and Stephen Boyd, which follows a different API.

PyDSM was originally developed against Tinoco de Rubira’s cvxpy. When development of that package ceased and the new cvxpy emerged, there arose a need to maintain compatibility and ensure a smooth transition. The goals were:

• to preserve full reproducibility of published scientific results, avoiding significant variations caused by changes in the underlying optimization code;

• to allow meaningful comparison between the two codebases on the optimization problems relevant to PyDSM, both in terms of accuracy and computational efficiency.

For these reasons, PyDSM began supporting both versions of cvxpy by vendoring the original one. Tinoco de Rubira’s cvxpy was originally distributed under the GPLv3 license. Its original codebase has now disappeared from the web, but in PyDSM it is vendored as pydsm.cvxpy_tdr to distinguish it from the newer cvxpy. The licensing and copyright information is fully preserved. The vendored copy is based on the last release by the original author, with patches by Sergio Callegari and others. While some patches fix minor issues, most of them convert internal imports to relative imports, to enable renaming and vendoring of the package without conflicts.

Now that the new cvxpy has matured, offers much better performance, and alternatives such as PICOS have become available, the vendoring of the original cvxpy is expected to be phased out in the near future.