To make the khmer project easier to use and easier to build upon several fundamental changes need to happen. This document outlines our plan to do so while minimizing the impact of these changes on our existing users. The version numbers are approximate; there may be additional major number releases to support needed changes to the API along the way. This document has been updated for v2.0 onwards from the original roadmap.
The discussion that lead to this document can be read at https://github.com/dib-lab/khmer/issues/389
Continue transition to a single entrypoint named oxli . This
will be exempt from the project’s semantic versioning and will not be
advertised as it is experimental and unstable. (For the 2.0 version we removed
the oxli
script just prior to release and restored it afterwards to the
development tree.)
The Python script functionality will continue to migrate to a Python module named oxli. As the code moves over there will be no change to external script functionality or their command line interfaces (except for new features).
The oxli
command is now under semantic versioning. Scripts are still the
advertised and preferred entry point for users. Developers and workflow systems
can start to trial oxli
but need not switch until 4.0. New functionality is
added to both the scripts and the oxli
command.
Project renamed to ‘oxli’; all references to ‘khmer’ removed from the code and
documentation except for a single note in the docs. All scripts dropped as
their functionality has been moved to the oxli
command. Websites that we
maintain that have ‘khmer’ in the URL will have redirects installed.
Refinement of the Python API continues, however it is not part of the semantic versioning of the project.
The semantic versioning now extends to the Python API.
API for multiple container types and implementation of the same.
Cleanup of Python/C++ class hierarchy to cut down on boilerplate glue code.
Switch to new-style Python objects (see LabelHash & Hashbits)