-
-
Notifications
You must be signed in to change notification settings - Fork 8.5k
Dev meeting notes
tacaswell edited this page Nov 30, 2014
·
1 revision
Notes from meeting with mdboom, tacaswell and efiring
* figure serialization
- make containers work better
- visitor pattern
- json/object model needs a research project to see if/how it would work (long term)
- contingent on not placing any new constraint
* Release strategy
- more often, just tag
- set up upload from travis -> binstar on master
- nightly builds?
- sort out how to get windows nightly
* testing
- move to conda on travis
- make sure we cover most of old numpy versions
* Drafting pitch documents for fund raising + big picture of where we want library to go/priorities
- cilcky-clicky gui
- not a priority
- maintain leadership position
* external API redesign (follow on to the discussion on the mailing list)
- get auto-updating all api calls
- via decorators that keep a flag on canvas
- either put update checking logic in os hook or in decorator on each method/function
- make use of existing property call back system
- this needs a bit of TLC
* internal API redesign (insert bohek object model in the middle of our rendering stack? the 'controller' idea)
- see figure serialization
* Splitting some/all of the gui backends out of the main repo
- on hold for now
* macosx
- can we get it to save the figure to do automated tests?
* styles
- make default 'show off' function
- work with mailing list to develop this as part of new default discussion
* style cycle
- re-implement color cycle as list of dicts
* rcparams
- make sure that all rcparams happen at object creation time not at draw time.
- rationalize and minimize the rcparams/style complex
* fonts
- ship more fonts
- web-fonts?
* website/documentation overhaul
- ask Dora about SEO
- don't want google to be finding mpl.org/dev
- separate pages via autosummary for pyplot (and axes?)
- do something about the auto-generated kwarg table
- conversion to numpydoc is breaking these
- fix documentation documentation to not refer to ACCEPTS and refer to numpydoc
- expand glossary? Low priority.
- {tri}contour{f} documentation match each other
* issue tracker
- clean up issues
- turn on waffle.io
* default colors
- add diverging and sequential default color maps
- 'div' and 'seq' are special color map names,
- add category meta-data to colormap object
- make sure color map discussion gets folded into choosing colormap
docs
- adopt seaborn.show_palette into core
- add easy make-your-own color map tools (convince function)
- maybe adopt from seaborn
- add convenience functions for accessing color brewer/general color patterns
* Documentation repo
- don't care about history on documentation repo
- only keep tips of minor releases
- add robots.txt to dev folder
- auto-build docs go to mpl.org/dev
- maybe host in independent repo
* getp/setp
- deprecate MATLAB style input?
* clean up
- remove BrokenBarHCollection
- move logic to axes methed
* units framework
- this needs a major documentation / functionality overhaul
- why is so much heavy looking code hidden in testing.jpl_units and the examples?-