Specify custom tick space heuristic in MaxNLocator #11027

egpbos · 2018-04-11T14:35:29Z

PR Summary

Following discussion in #11025, I added an optional argument to MaxNLocator (and hence AutoLocator) to specify a custom heuristic for spacing ticks. It allows people to experiment with their own spacing rules without having to change Matplotlib source code. This way, one can benefit from the automatic spacing introduced in Matplotlib 2 (if I understand correctly), but still customize it if deemed necessary (which was the case for me as discussed in #11025).

PR Checklist

Has Pytest style unit tests
Code is PEP 8 compliant
New features are documented, with examples if plot related
Documentation is sphinx and numpydoc compliant
Added an entry to doc/users/next_whats_new/ if major new feature (follow instructions in README.rst there)
~~Documented in doc/api/api_changes.rst if API changed in a backward-incompatible way~~

jklymak

I'm 50/50 on the utility of this, and dubious about adding another kwarg given that the new kwarg is only used if the nbins = 'auto'.

I think this is a pretty obscure use case, but if you think its important, it will only get used if there is an example somewhere...

jklymak · 2018-04-11T17:18:43Z

lib/matplotlib/ticker.py

@@ -1821,6 +1822,12 @@ def __init__(self, *args, **kwargs):
            Relax `nbins` and `integer` constraints if necessary to
            obtain this minimum number of ticks.

+        *tick_space_heuristic*


I'd suggest that this just be another option for nbins. i.e. if nbins is callable it is used to determine the heuristic...

That's a great idea!

jklymak · 2018-04-11T17:19:31Z

lib/matplotlib/ticker.py

@@ -1882,11 +1889,17 @@ def set_params(self, **kwargs):
            self._extended_steps = self._staircase(self._steps)
        if 'integer' in kwargs:
            self._integer = kwargs['integer']
+        if 'tick_space_heuristic' in kwargs:
+            tick_space_heuristic = kwargs['tick_space_heuristic']


I think we just use kwarg.pop these days...

That's fine by me. I was just following the style used in the rest of the set_params function. Do you think I should convert the entire function then? Seems like something for a different PR, but I don't mind doing it here.

jklymak · 2018-04-11T17:22:29Z

lib/matplotlib/axis.py

@@ -1760,6 +1760,11 @@ def get_minpos(self):
        raise NotImplementedError()


+def tick_space_heuristic(axis_length, label_fontsize, label_fontsize_factor):


This should be private I think.

You mean prepend the function with an underscore?

tacaswell · 2018-04-11T18:34:40Z

I am 👎 on this going in as-is. My primary concern is expanding the public API of the Axis classes to serve the needs of one of the Locator classes. Even if I were convinced that adding an extra coupling between the locator and the ticker was the right thing to do, I am not sure that this is the right path. It might be better to ex add a method to the Axis that returns both the length of the axis, the font-size, and the orientation of the font relative to the axis (parallel, particularize, or rotated). If I were convinced that this is a good idea and this is the right API to add, it will need to be much better documented and motivated.

I think a better course path here would be to make it easier to set the default XAxis, YAxis, and tick-related classes used by Matplotlib (it looks like you can currently do it by over-riding a private method on Axes to use your prefered classes and then re-registering your new class Axes class with the projection framework, yikes!).

attn @timhoffm for more ticker-related issues.

tacaswell

See my comment for details, concerned about expanding the public API in this way.

jklymak · 2018-04-11T18:46:24Z

@tacaswell. The horse is already out of the barn on that one; axis.get_tick_spacing predates this PR and is explicitly used for MaxNLpcator as far as I can tell

egpbos · 2018-04-12T06:08:16Z

@jklymak

I think this is a pretty obscure use case, but if you think its important, it will only get used if there is an example somewhere...

Following #11025 (comment), I can make an example using AxesGrid that shows the problem I have with the ticks and compares to a plot using the custom tick heuristic to make it look nicer.

egpbos · 2018-04-12T06:44:14Z

@tacaswell: regarding extending public API, what if I follow @jklymak's suggestion of using the nbins parameter (check if it's callable) instead of adding a new parameter to MaxNLocator, would that be acceptable?

As @jklymak said, the coupling between X/YAxis and MaxNLocator was already there, I just added the heuristic parameter as a way to generalize/abstract this method. I can hide that into a private attribute of the Axis if you prefer.

The problem I see with setting different defaults (which I agree would be nicer) is that there is currently no option among the existing Locators to do something that acts like nbins='auto', which is desirable behavior, but with a different way of automatically estimating the tick label size in order to prevent overlapping labels, as the current hard-coded heuristic does, but, imho, does in a way that makes the axis look ugly when the labels are so small, i.e. have an aspect ratio more like 1:1 instead of the 3:1 that the XAxis heuristic assumes. So, since there is currently no such customization option, setting different defaults would still not be sufficient.

tacaswell · 2018-04-16T19:34:53Z

We discussed this on the weekly phone call and came to the following suggestion:

deprecated get_tick_space
replace it with _get_axis_length_and_font_size (name negotiable, maybe make this public?)
in the locator look at the Axis object and axis.axis_name to decide what ratio to use and then you can adjust that ratio by sub-classing the Locator and setting your locator onto the axis or possible some API on the Locator.

Notes are at https://paper.dropbox.com/doc/Matplotlib-meeting-agenda-aAmENlkgepgsMeDZtlsYu

jklymak · 2018-05-09T20:34:44Z

ping @egpbos are you still interested in keeping this alive given the feedback above?

egpbos · 2018-05-14T06:32:30Z

@jklymak I'm interested, but currently I have some other priorities, so won't get to it any time soon.

jklymak · 2019-02-07T17:39:14Z

OK, I'm going to close this as not having enough support. Maybe something like this can go in an example showing how to subclass MaxNLocator?

Specify custom tick space heuristic in MaxNLocator

d9d4406

jklymak reviewed Apr 11, 2018

View reviewed changes

tacaswell requested changes Apr 11, 2018

View reviewed changes

jklymak added the topic: ticks axis labels label Apr 11, 2018

jklymak added this to the v3.0 milestone Apr 11, 2018

egpbos mentioned this pull request Apr 12, 2018

AxesGrid ticks missing on x-axis #11025

Closed

jklymak added the status: needs revision label Apr 28, 2018

tacaswell modified the milestones: v3.0, v3.1 Jul 7, 2018

jklymak closed this Feb 7, 2019

anntzer mentioned this pull request Apr 4, 2020

Unnecessary drawing with NbAgg #13971

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Specify custom tick space heuristic in MaxNLocator #11027

Specify custom tick space heuristic in MaxNLocator #11027

egpbos commented Apr 11, 2018 •

edited

Loading

jklymak left a comment

jklymak Apr 11, 2018

egpbos Apr 12, 2018

jklymak Apr 11, 2018

egpbos Apr 12, 2018

jklymak Apr 11, 2018

egpbos Apr 12, 2018

tacaswell Apr 16, 2018

tacaswell commented Apr 11, 2018

tacaswell left a comment

jklymak commented Apr 11, 2018

egpbos commented Apr 12, 2018 •

edited

Loading

egpbos commented Apr 12, 2018

tacaswell commented Apr 16, 2018

jklymak commented May 9, 2018

egpbos commented May 14, 2018

jklymak commented Feb 7, 2019

		@@ -1760,6 +1760,11 @@ def get_minpos(self):
		raise NotImplementedError()


		def tick_space_heuristic(axis_length, label_fontsize, label_fontsize_factor):

Specify custom tick space heuristic in MaxNLocator #11027

Specify custom tick space heuristic in MaxNLocator #11027

Conversation

egpbos commented Apr 11, 2018 • edited Loading

PR Summary

PR Checklist

jklymak left a comment

Choose a reason for hiding this comment

jklymak Apr 11, 2018

Choose a reason for hiding this comment

egpbos Apr 12, 2018

Choose a reason for hiding this comment

jklymak Apr 11, 2018

Choose a reason for hiding this comment

egpbos Apr 12, 2018

Choose a reason for hiding this comment

jklymak Apr 11, 2018

Choose a reason for hiding this comment

egpbos Apr 12, 2018

Choose a reason for hiding this comment

tacaswell Apr 16, 2018

Choose a reason for hiding this comment

tacaswell commented Apr 11, 2018

tacaswell left a comment

Choose a reason for hiding this comment

jklymak commented Apr 11, 2018

egpbos commented Apr 12, 2018 • edited Loading

egpbos commented Apr 12, 2018

tacaswell commented Apr 16, 2018

jklymak commented May 9, 2018

egpbos commented May 14, 2018

jklymak commented Feb 7, 2019

egpbos commented Apr 11, 2018 •

edited

Loading

egpbos commented Apr 12, 2018 •

edited

Loading