Very nice!

Lynn Rees repo owner

Sure.

I've put up existing shove plugins in their own repositories, including those not currently ready for PyPi release. Most used to be part of the monolithic shove of old. Many have had their package specific infrastructure set up. Many have nothing more than testing stubs. Some have never been released (like the PostgreSQL HStore plugin).

Fork and hack whichever one scratches an itch. They should, at minimum, pass the existing test suite to be in a releasable state. Expanding the core test suite to handle corner cases for some plugins with peculiar idiosyncracies is a nice to have but not immediately required.

My target for existing and new shove plugins is that they must support CPython 2.7. They should support CPython 2.6, Python 3.4 and up, and PyPy if there's a Python library for the datastore that supports it. Swapping out whichever library isn't providing maximum flexibility in choice of backend for one that does is a supported way of fixing the limitations of an existing plugin.

The tox setup was decent 4 years ago. Whether it represents the cutting edge of Python I/O integration testing now is a question I haven't investigated. Recommendations are welcome.

2015-09-08T21:15:31+00:00

Comments (1)

Lynn Rees repo owner
Sure.

I've put up existing shove plugins in their own repositories, including those not currently ready for PyPi release. Most used to be part of the monolithic shove of old. Many have had their package specific infrastructure set up. Many have nothing more than testing stubs. Some have never been released (like the PostgreSQL HStore plugin).

Fork and hack whichever one scratches an itch. They should, at minimum, pass the existing test suite to be in a releasable state. Expanding the core test suite to handle corner cases for some plugins with peculiar idiosyncracies is a nice to have but not immediately required.

My target for existing and new shove plugins is that they must support CPython 2.7. They should support CPython 2.6, Python 3.4 and up, and PyPy if there's a Python library for the datastore that supports it. Swapping out whichever library isn't providing maximum flexibility in choice of backend for one that does is a supported way of fixing the limitations of an existing plugin.

The tox setup was decent 4 years ago. Whether it represents the cutting edge of Python I/O integration testing now is a question I haven't investigated. Recommendations are welcome.
- 2015-09-08T21:15:31+00:00
Log in to comment