MEP22: Navigation by events

[fariza]

This PR is the implementation of MEP22, https://github.com/matplotlib/matplotlib/wiki/Mep22#implementation

This it supersedes the #2759

Toolbar has been relegated as listener/emiter of events that get dispatched to the Tools by Navigation.

The toolbar is easily reconfigurable at running time. new tools can be created/added/removed without modificaion of the backend code.

Example
examples/user_interfaces/navigation.py

[WeatherGod]

I am definitely very interested in looking through this this weekend. At a
high-level at least, this is along the lines of what I was thinking of
doing with the keymaps (i.e., map the keys to named events) which would
make for simpler code and configuration, and allows for user-extensibility.

On Thu, Oct 16, 2014 at 5:45 PM, Federico Ariza notifications@github.com
wrote:

This is an slightly different alternative to PR #2759
#2759 for MEP22 where
Navigation implements some events to communicate with the clients (Toolbar,
etc...), this allows to have multiple entry points for interacting with

Tools

You can merge this Pull Request by running

git pull https://github.com/fariza/matplotlib navigation-by-events

Or view, comment on, or merge it at:

#3652
Commit Summary

• navigation and toolbar coexistence
• mod keypress in figuremanager
• extra files
• helper methods in toolbar and navigation
• Adding doc to base methods
• property for active_toggle
• simulate click
• pep8 backend_tools
• activate renamed to trigger
• toggle tools using enable/disable from its trigger method
• simplifying _handle_toggle
• reducing number of locks
• pep8 correction
• changing toggle and persistent attributes for issubclass
• bug in combined key press
• untoggle zoom and pan from keypress while toggled
• classmethods for default tools modification
• six fixes
• adding zaxis and some pep8
• removing legacy method dynamic update
• tk backend
• pep8
• example working with Tk
• cleanup
• duplicate code in keymap tool initialization
• grammar corrections
• moving views and positions to tools
• The views positions mixin automatically adds the clear as axobserver
• bug when navigation was not defined
• Small refactor so that we first initiate the Navigation
  (ToolManager), before filling it with tools.
• Update for Sphinx documentation
• Moved default_tool initilisation to FigureManagerBase and cleaned.
• Fix navigation
• Temporary fix to backends
• Merge pull request Fix autofmt_xdate() when using in conjunction with twinx() #1 from
  OceanWolf/navigation-toolbar-coexistence-event-framework
• removing persistent tools
• removing unregister
• change cursor inmediately after toggle
• removing intoolbar
• events working

File Changes

• A doc/api/backend_tools_api.rst
  https://github.com/matplotlib/matplotlib/pull/3652/files#diff-0 (8)
• M doc/api/index_backend_api.rst
  https://github.com/matplotlib/matplotlib/pull/3652/files#diff-1 (1)
• A examples/user_interfaces/navigation.py
  https://github.com/matplotlib/matplotlib/pull/3652/files#diff-2 (59)
• M lib/matplotlib/backend_bases.py
  https://github.com/matplotlib/matplotlib/pull/3652/files#diff-3
  (510)
• A lib/matplotlib/backend_tools.py
  https://github.com/matplotlib/matplotlib/pull/3652/files#diff-4
  (751)
• M lib/matplotlib/backends/backend_gtk3.py
  https://github.com/matplotlib/matplotlib/pull/3652/files#diff-5
  (239)
• M lib/matplotlib/backends/backend_tkagg.py
  https://github.com/matplotlib/matplotlib/pull/3652/files#diff-6
  (216)
• M lib/matplotlib/rcsetup.py
  https://github.com/matplotlib/matplotlib/pull/3652/files#diff-7 (2)

Patch Links:

• https://github.com/matplotlib/matplotlib/pull/3652.patch
• https://github.com/matplotlib/matplotlib/pull/3652.diff

—
Reply to this email directly or view it on GitHub
#3652.

[fariza]

@tacaswell @WeatherGod @OceanWolf
After long traffic thinking time, I am pretty sure this is the way to go.
I applied the latest comments made by @tacaswell on #2759 except for

• dictionnary vs classes (tool definition)
• history singleton

If you agree, I will remove the other two PR's and we can keep going.

[tacaswell]

Sounds good to me. I am still not convinced that going with a zoo of classes is the right thing to do and am still not happy with adding more global state.

[fariza]

I removed the singleton, I added another tool, that keeps the "history". The other tools access this one to push things around in history

[fariza]

@tacaswell please tag this as "need revision" and remove it from #2759

[fariza]

@WeatherGod did you have time to check it out?

[WeatherGod]

Hopefully, I'll have time this weekend... (book-writing is hard!)

On Fri, Oct 31, 2014 at 9:41 AM, Federico Ariza notifications@github.com
wrote:

@WeatherGod https://github.com/WeatherGod did you have time to check it
out?

—
Reply to this email directly or view it on GitHub
#3652 (comment)
.

[tacaswell]

Patterns like this make me nervous, might as well just pass around a dict.

[fariza]

Removed

[fariza]

@tacaswell I gave up (too tempting).
I just added the radio_group attribute for ToolToggleBase derived tools, now it is possible to specify a group that transforms the tools in that group as "mutually exclusive" (untoggle one before toggling the other).
As per your suggestion, ToolGrid, ToolFullScreen, ToolYScale and ToolXScale are now toggle tools.

[OceanWolf]

Okay, merge time?

[WeatherGod]

Congratulations, @fariza and @OceanWolf ! You are now proud owners of some mpl real estate!

[fariza]

Thank you everybody.

Actually I am going to present a Lightning Talk at Pycon2015
Here is the notebook just in case somebody is interested https://github.com/fariza/pycon2015/blob/master/ToolDemo.ipynb

[OceanWolf]

@fariza nice, though the file looks nonsense on github ;). When do you give the talk? Perhaps I can get MEP27 ready by then ;).