notes from reorg meeting

• meeting on reorg (3/27/2017)
  • strategy
    • build makefile/automator
    • test limited attempt
    • finish submoduling
  • nightlies for pysal integration
    • check if something committed to a submodule affects something else in another module.
  • Readmes in each submodule allow us to redirect the potential first time developer
    • add to project template repo
  • importing from module vs. importing from pysal
  • legacy.py as the legacy api and then api.py be the new api, possibly py3 only
  • Documentation
    • sphinx for main project builds alone, refer users to sphinx for main redistributable project.
    • build outwards
      • submodules have their own sphinx stuff
      • make main project builds all subprojects
    • build across
      • each module is responsible for its own documentation structure and requirements.
      • project documentation links to each subproject
      • autodoc on sphinx would build each modules' docstrings anyway
    • We should build across, not out.
  • making
    • two commands, dist and dev
      • dist builds the source redistribution
        • git clone pysal
        • cd pysal/pysal && rm -r $SUBPROJECTS
        • git clone $SUBPROJECTS
        • rm $SUBPROJECTS/.git
      • dev builds the development tree
        • git clone ysal
        • cd pysal/pysal && rm -r $SUBPROJECTS
        • git clone $SUBPROJECTS
        • rm ../../.git
  • viz design
    • duck type & type dispatching,
    • plot functions should work with types and figure out how to plot them, not expose interfaces
    • like geoplot, they take structured data and plot it the way we want
    • leverage shim classes to convert objects into the format we need to plot
    • this way, things don't depend on viz and viz doesn't depend on things, it just consumes objects
    • you wouldn't try to pass a non-markov object to plot_markov, but all viz effort would be concentrated in a package with common dependences & a shared api.