Allow Artists to show pixel data in cursor display by blink1073 · Pull Request #3989 · matplotlib/matplotlib · GitHub

Also adds implementation for AxesImage. Closes #3984.
Inspired by mpldatacursor.

Should also look at ax.collections (that is where things like scatter live) and maybe ax.patches

and the order should be ax.collections + ax.patches + ax.lines + ax.artists + ax.images to match the order used for sorting out the rendering order in Axes.draw (around https://github.com/matplotlib/matplotlib/blob/master/lib/matplotlib/axes/_base.py#L1993 images are special cased around line 2030).

They then get ordered by zorder anyway: https://github.com/matplotlib/matplotlib/blob/master/lib/matplotlib/axes/_base.py#L2036

Right, but the ordering is a stable sort so ties are broken by which was in the list first. This is the source of at least one un-fixable (with out major overhaul of the internal data structure, see #3944) bug ( #409 ).

Ok then. So how to handle composite images? It does not seem feasible to recreate the composite every time.

The AxesImage could composite the specific pixel on demand.

I think to first order, just ignore the compositing and use the top image.

The in-place compositing is only for some of the vector backends which can't do it them selves (ex pdf) to try and save space (so you don't have multiple overlapping images) but it leads to other bugs as all images get smashed to the same zorder relative to the other lines etc. In Agg (which backs all but 3 (gtkcairo, macsox and windows) of the interactive backends) the whole if _do_composite block is skipped and the compsiting is taken care of at the c++ layer.

How's that? I noticed onRemove was doing something similar and found Artist.hitlist(), and standardized between the two.

However, it looks like Axes.get_children needs to return the same order as Axes.draw. New PR for that?

onRemove looks like another candidate (along with onHilite) for removal from FigureCanvasBase (now that you have simplified it)....

Probably worth having a top-level module for useful event handlers, which I think we greatly reduce the threshold to get people using the event system. @WeatherGod @joferkington thoughts?

Yes to separate PR for re-ordering Axes.get_children as that is a (subtle) api change.

I think we do need to handle compositing. For instance, if I run this example, I get the Zdata for the overlay layer. How about just checking for alpha=0 and then using the underlying in that case?

Well, the more I think about it the more I agree that the top image should get priority. The canny value really is the point of the example.

Added Artist.format_zdata to centralize formatting.

Do I still "need revision"?

Looks good to me. I am a tad wary of the format_z method which does not use self, but that seems like a reasonable place to put it.

I would like @WeatherGod to weigh in.

What do you mean "does not use self"?

Meaning it should be a static method?

Either way I think it belongs in the class so it can be overridden.

Yes, it is an instance method which does not refer to it's instance and I can't think of a use case where you would want to (but I may just have poor imagination). I thought about suggesting to move it to cbook, but it is much too specific for that.

I'd worry that adding a @staticmethod would give people the impression they could not subclass it as an instance method (I checked, and you can).

This also needs an entry in https://github.com/matplotlib/matplotlib/tree/master/doc/users/whats_new

Done.

@matplotlib/developers Can this get eyes from someone else but me?

This addresses a long standing and frequent feature request in a pleasingly general way.

Why do you want to use a string for this? Shouldn't "data" always be numeric?

This allows the Artist to control the display of numbers, similar to the existing ax.format_coord.

I agree with @efiring on this one. I can think of a lot of use cases (e.g. interactive sampling) where it would be very convenient to access a "raw" value. Returning a numerical value would make this method useful for a lot more than just display.

You already have a separate method for the formatting, so it's only slightly more verbose to flip the function calls on their head and make the displayed string using artist.format_zdata(artist.get_zdata()). In my opinion, this would be a better approach overall.

Ha! I originally meant to do that...

@joferkington again. Sorry to pester, but I see you are doing OSS stuff today and would like your feed back on this 😄

@tacaswell - Absolutely! Sorry I've been so out-of-touch with things lately, as well.

ENH: Allow Artists to show pixel data in cursor display

tacaswell added the status: needs review label Jan 9, 2015

tacaswell added this to the v1.5.x milestone Jan 9, 2015

blink1073 changed the title Allow Artists to Diplay Zdata in Cursor Allow Artists to Display Zdata in Cursor Jan 10, 2015

tacaswell reviewed Jan 19, 2015
View reviewed changes

tacaswell added status: needs revision and removed status: needs review labels Jan 19, 2015

tacaswell added status: needs review and removed status: needs revision labels Jan 20, 2015

blink1073 changed the title Allow Artists to Display Zdata in Cursor Allow Artists to display zdata in cursor Jan 26, 2015

tacaswell modified the milestones: next point release, proposed next point release Feb 17, 2015

efiring reviewed Feb 17, 2015
View reviewed changes

Steven Silvester added 20 commits March 17, 2015 15:56

Fix the artist sorting behaviour

8ea9504

Fix syntax error

02456b8

Specify hitlist on axes instead of figure

641304c

Put get_zdata on the correct class

3f20d3a

Ignore the axes patch

5850150

Create new function to format zdata

88b56be

Clean up string formatting

aece695

Add a whats new blurb.

07745c8

Add a test for image cursor data

7626394

Rename to get_pixel_data and return raw data

7b0b6b1

Update the whats new

f5ff73f

Fix name of method

ea565c6

Fix default pixel data and handle no pixel data

1429206

Update test for new method name

e62b0e3

Simplify transform

34b4df4

Update display

8f266a1

Simplify format_pixel_data

70983ff

Update name to cursor_data

cd0e52b

Move artist sort to within if block

8ce3a6f

Fix method name in cursor_test

bac8dff

blink1073 force-pushed the add-cursor-zdata branch from df4ee01 to bac8dff Compare March 17, 2015 20:56

This was referenced Mar 27, 2015

Twin axis message coordinates #4284

Closed

corrplot does not show cell coordinates with nbagg backend mwaskom/seaborn#495

Closed

michaelaye mentioned this pull request Apr 1, 2015

Heatmap mouseover mwaskom/seaborn#496

Closed

tacaswell added a commit that referenced this pull request May 24, 2015

Merge pull request #3989 from blink1073/add-cursor-zdata

c473c1b

ENH: Allow Artists to show pixel data in cursor display

tacaswell merged commit c473c1b into matplotlib:master May 24, 2015

tacaswell removed the status: needs review label May 24, 2015

tacaswell mentioned this pull request Jul 16, 2015

Display z value in matshow in addition of x, y. #3146

Closed

tacaswell mentioned this pull request Jul 26, 2015

Regression Qt5 Backend #4767

Closed

megies mentioned this pull request Jan 31, 2018

fix detecting which artist(s) the mouse is over #10356

Merged

5 of 6 tasks

Allow Artists to show pixel data in cursor display by blink1073 · Pull Request #3989 · matplotlib/matplotlib · GitHub

Repository navigation

Allow Artists to show pixel data in cursor display - #3989

Allow Artists to show pixel data in cursor display#3989tacaswell merged 26 commits intomatplotlib:mastermatplotlib/matplotlib:masterfrom blink1073:add-cursor-zdatablink1073/matplotlib:add-cursor-zdataCopy head branch name to clipboard

Conversation

blink1073 commented Jan 9, 2015

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

blink1073 commented Jan 19, 2015

Uh oh!

blink1073 commented Jan 19, 2015

Uh oh!

tacaswell commented Jan 19, 2015

Uh oh!

blink1073 commented Jan 19, 2015

Uh oh!

blink1073 commented Jan 19, 2015

Uh oh!

blink1073 commented Jan 19, 2015

Uh oh!

blink1073 commented Jan 20, 2015

Uh oh!

tacaswell commented Jan 20, 2015

Uh oh!

blink1073 commented Jan 20, 2015

Uh oh!

blink1073 commented Jan 20, 2015

Uh oh!

blink1073 commented Jan 20, 2015

Uh oh!

tacaswell commented Jan 20, 2015

Uh oh!

blink1073 commented Jan 20, 2015

Uh oh!

tacaswell commented Jan 20, 2015

Uh oh!

blink1073 commented Jan 20, 2015

Uh oh!

tacaswell commented Feb 17, 2015

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

tacaswell commented Feb 20, 2015

Uh oh!

joferkington commented Feb 20, 2015

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

7 participants

Allow Artists to show pixel data in cursor display#3989
tacaswell merged 26 commits into
matplotlib:masterfrom
blink1073:add-cursor-zdata