Add a new `time_support` context manager for plotting times #8782

astrofrog · 2019-06-01T15:55:32Z

Matplotlib natively provides a mechanism for plotting dates and times on one or both of the axes, as described in Date tick labels. To make use of this, users need to explicitly access plot_date on time instances they want to plot. While this might be ok for a number of situations, I thought it would be nice to have a way of more seamlessly using times when plotting that gives more control over the time scale and format.

This PR implements a time_support context manager/function that is meant to be similar to quantity_support:

from matplotlib import pyplot as plt
from astropy.time import Time
from astropy.visualization import time_support

time_support()

plt.figure(figsize=(5,3))
plt.plot(Time([58000, 59000, 62000], format='mjd'), [1.2, 3.3, 2.3])

(by default the time is removed from the ISO string if it's always midnight for all tick labels, this can be controlled with the simplify argument).

This can be used as a context manager and can take the format and scale to use for the axis:

with time_support(format='mjd', scale='tai'):
    plt.figure(figsize=(5,3))
    plt.plot(Time([58000, 59000, 62000], format='mjd'))

Note that this also allows Time to be used in set_xlim/set_ylim for example. One of the benefits of this approach is that all times are automatically converted to a common scale/format.

A few notes:

Something which I'm not entirely happy with, which also affects quantity_support, is that even when using the context manager, once an axis has been used inside the context manager, it has a cache of the converter and it's impossible to change it again later. I think this is something that would need to be fixed in MPL though. As long as a new axis/figure is created, then the context manager works fine.
I'm also not really fond of the current names quantity_support and time_support, I feel there should be a plot or mpl in there somewhere so it's clear that it's related to plotting.
I've consolidated the docs to mention quantity_support and time_support in one place, in the visualization docs.
There is still some cleanup to do in various places and I think I can expand the docs and tests a little, but I don't want to invest too much time until I hear whether people are interested in this. I need this anyway for another project so I decided to implement it here, but it's ok if people would prefer not to include it!

mhvk

This looks very nice! Some in-line comments, the largest one of which is that it would be nice if the default scale and format were taken from the (first) input times.

On the names: I guess one could go for mpl_quantity_support, but that certainly doesn't seem worth the trouble of changing the name.

Should there be an matplotlib_astropy_support, though, where all these options get turned on?

But those questions are probably best addressed separately.

mhvk · 2019-06-02T15:02:46Z

astropy/visualization/time.py

+YMDHMS_FORMATS = ('fits', 'iso', 'isot')
+
+
+def time_support(scale='utc', format='isot', simplify=True):


Would it be possible to have the default set by the times first used? (say, scale=None here)

Agree, this will be consistent with quantity_support - fixed

mhvk · 2019-06-02T15:10:26Z

astropy/visualization/time.py

+        """
+        scaled = getattr(value, self.scale)
+        if self.format in YMDHMS_FORMATS:
+            return scaled.mjd


Is this done to make the values numerical? But, if so, shouldn't it happen for all string formats? For best precision for relevant times, one might consider using cxcsec or so.

I've now fixed compatibility for all string formats (and have tests for all of them).

mhvk · 2019-06-02T15:11:34Z

docs/visualization/matplotlib_integration.rst

@@ -0,0 +1,121 @@
+.. _quantity:


This link seems wrong.

astrofrog · 2019-06-02T20:36:29Z

@mhvk - thanks for the initial review! You mentioned that using MJD might not be ideal, so I was wondering whether, since you are intimately familiar with astropy.time, you would mind trying out this PR and finding cases where the precision doesn't work properly?

One use case I know of already that needs to be fixed is when looking at very small time intervals, the string-based formats need to have their precision updated (they are currently fixed to millisecond precision). But it would be great if you could let me know if you can find other cases that won't work properly/optimally!

astrofrog · 2019-06-02T21:54:41Z

This whole functionality doesn't work very well with Matplotlib 2.0 just because of the way that it dealt with non-numpy arrays back then (it split the Time array into a list of many times, and same for Quantity). Since Matplotlib 2.1, 2.2, 3.0 and 3.1 are out now and 2.0 is over two years old, maybe we can just drop it? (rather than have to add convoluted workarounds here).

larrybradley

@astrofrog LGTM, thanks! I'm fine with dropping support for matplotlib 2.0. I don't think it's worth the trouble.

larrybradley · 2019-06-03T16:25:16Z

The travis test failures look real.

mhvk · 2019-06-03T19:19:14Z

@astrofrog - using mjd means the best possible precision is

np.spacing(Time.now().mjd * u.day).to(u.us)
# <Quantity 0.62864274 us>

This is probably good enough for most purposes...

But it brings up a different question: do you also handle TimeDelta? I think those should just be treated as Quantity...

astrofrog · 2019-06-07T14:39:22Z

@mhvk - just to check, do you mean we should support it in quantity_support or in time_support? It would be much cleaner to support it in the former. I guess this wouldn't matter if we then introduce a context manager such as matplotlib_astropy_support which enables all options like you suggest.

mhvk · 2019-06-07T14:50:30Z

@astrofrog - yes, you're right, TimeDelta is more logically supported in quantity_support.

I do think an overall astropy_support makes sense; one really shouldn't want one without the other. Indeed, in implementing support for the new __array_function__ (which allows all numpy functions to work on quantities!), I'm finding that plain plot(q1, q2) starts breaking (because quantities are no longer interpreted as plain ndarray). So, the recommendation may well become to always have astropy_support on when plotting.

… with various formats/scales in Matplotlib

…manager is defined

…as a function

…ng-based formats

…ance

…module to import

astrofrog · 2019-06-14T15:49:09Z

@mhvk @larrybradley - this is now ready for a final review (I think TimeDelta and the joint context manager should be handled in a separate PR)

codecov · 2019-06-14T17:14:10Z

Codecov Report

Merging #8782 into master will increase coverage by 0.02%.
The diff coverage is 97.67%.

@@            Coverage Diff             @@
##           master    #8782      +/-   ##
==========================================
+ Coverage   86.99%   87.01%   +0.02%     
==========================================
  Files         400      401       +1     
  Lines       59469    59598     +129     
  Branches     1100     1100              
==========================================
+ Hits        51736    51861     +125     
- Misses       7092     7096       +4     
  Partials      641      641

Impacted Files	Coverage Δ
astropy/visualization/__init__.py	`100% <100%> (ø)`	⬆️
astropy/visualization/time.py	`97.65% <97.65%> (ø)`
astropy/samp/hub.py	`77.38% <0%> (-0.14%)`	⬇️

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 41930f7...899f0ee. Read the comment docs.

codecov · 2019-06-14T17:14:12Z

Codecov Report

Merging #8782 into master will increase coverage by 0.02%.
The diff coverage is 97.67%.

@@            Coverage Diff             @@
##           master    #8782      +/-   ##
==========================================
+ Coverage   86.99%   87.01%   +0.02%     
==========================================
  Files         400      401       +1     
  Lines       59469    59598     +129     
  Branches     1100     1100              
==========================================
+ Hits        51736    51861     +125     
- Misses       7092     7096       +4     
  Partials      641      641

Impacted Files	Coverage Δ
astropy/visualization/__init__.py	`100% <100%> (ø)`	⬆️
astropy/visualization/time.py	`97.65% <97.65%> (ø)`
astropy/samp/hub.py	`77.38% <0%> (-0.14%)`	⬇️

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 41930f7...5f7cdf7. Read the comment docs.

mhvk

This looks great! One in-line question about whether we should raise the minimum version of matplotlib to 2.2 for astropy 4.0. But obviously that's no reason to hold up this PR, so approving.

Would be good to raise separate issues for

Support for TimeDelta - in the next matplotlib release, these will be captured by this time_support (as a subclass of Time), so we need an explicit solution (this probably needs a 4.0 milestone).
Combined astropy_support. Possibly with the question of whether we should turn both on by default...

mhvk · 2019-06-14T17:24:14Z

astropy/visualization/time.py

+            Convert a Time value to a scalar or array.
+            """
+
+            # For Matplotlib < 2.2


Should we increase our minimum version?

I already just increased it to 2.1 from 2.0, but we could consider increasing further especially given the 2 year support for 4.0. I wish we had a written down policy for this :)

Probably OK to leave for later: #8822 (comment)

larrybradley

Thanks, @astrofrog! Just one little inline nit about the parameter order in the docstring.

astropy/visualization/time.py

astrofrog · 2019-06-16T11:55:55Z

@mhvk issues opened in #8859 and #8860

mhvk · 2019-06-16T12:40:45Z

Thanks for raising the issues. SInce @larrybradley's comment was addressed, and he was happy otherwise, I'll merge now - very nice to have this!!

Requested changes were made

mhvk · 2019-06-16T12:45:25Z

Hmm, having done it, not sure that the right way is to "dismiss a review"....

and remove unnecessary code added in astropy#8818 and astropy#8782

and remove unnecessary code added in astropy#8818 and astropy#8782 Add new JSON for figure tests.

astrofrog added time Enhancement visualization labels Jun 1, 2019

astrofrog requested review from taldcroft, mhvk and larrybradley June 1, 2019 15:55

astrofrog added this to the v4.0 milestone Jun 1, 2019

pllim added the whatsnew-needed label Jun 1, 2019

mhvk reviewed Jun 2, 2019

View reviewed changes

larrybradley reviewed Jun 3, 2019

View reviewed changes

astrofrog added 14 commits June 14, 2019 14:54

Started implementing time_support to allow Time objects to be plotted…

3a542c7

… with various formats/scales in Matplotlib

Implement support for intervals greater than a month

b9969ad

Don't show times for labels with all midnight times

0cd278f

Added a test for the simplify functionality and simplify how context …

0aa33d6

…manager is defined

Added more tests

f4bb4ec

Started adding documentation

497b116

Write more documentation and fix time_support so that it can be used …

5f6dd05

…as a function

Added changelog entry

c681a84

Remove unused code

9f60840

Fix context manager

0c22d71

Make doctest ignore output

b35df8a

Added test cases for all formats and fix issues with some of the stri…

595eac1

…ng-based formats

Use format/scale from first Time encountered for a specific Axes inst…

49e0aa3

…ance

Have a default label for the axis.

c18da5e

astrofrog added 4 commits June 14, 2019 14:54

Move another import to make sure we don't require Matplotlib for the …

311d5a0

…module to import

Skip all time_support tests if Matplotlib is not installed

265f477

Fix tests on Windows

edb456c

Fix case where time_support is nested

899f0ee

astrofrog force-pushed the time-support-mpl branch from ae97d26 to 899f0ee Compare June 14, 2019 15:10

astrofrog added the Ready-for-final-review label Jun 14, 2019

astrofrog requested review from mhvk and larrybradley June 14, 2019 15:49

mhvk approved these changes Jun 14, 2019

View reviewed changes

mhvk mentioned this pull request Jun 14, 2019

Upstream minimal versions to support for v4.0 LTS #8822

Closed

larrybradley previously requested changes Jun 14, 2019

View reviewed changes

astropy/visualization/time.py Outdated Show resolved Hide resolved

Fix ordering of parameters in docstring

5f7cdf7

astrofrog requested a review from larrybradley June 15, 2019 11:05

This was referenced Jun 16, 2019

Fix support for TimeDelta in time_support #8859

Open

Provide a way to enable both quantity_support and time_support in one go #8860

Open

mhvk merged commit 62c3857 into astropy:master Jun 16, 2019

mhvk mentioned this pull request Jun 22, 2019

Create equivalent of quantity_support() for plotting Time objects #7930

Closed

pllim added a commit to pllim/astropy that referenced this pull request Jan 25, 2023

MNT: Bump matplotlib minversion to 3.2

29c6274

and remove unnecessary code added in astropy#8818 and astropy#8782

pllim mentioned this pull request Jan 25, 2023

MNT: Bump matplotlib minversion to 3.2 #14321

Merged

10 tasks

pllim added a commit to pllim/astropy that referenced this pull request Jan 25, 2023

MNT: Bump matplotlib minversion to 3.2

cb9c319

and remove unnecessary code added in astropy#8818 and astropy#8782

pllim added a commit to pllim/astropy that referenced this pull request Jan 26, 2023

MNT: Bump matplotlib minversion to 3.2

074eadc

and remove unnecessary code added in astropy#8818 and astropy#8782

pllim added a commit to pllim/astropy that referenced this pull request Jan 30, 2023

MNT: Bump matplotlib minversion to 3.2

b58ca80

and remove unnecessary code added in astropy#8818 and astropy#8782 Add new JSON for figure tests.

dougbrn pushed a commit to dougbrn/astropy that referenced this pull request Mar 13, 2023

MNT: Bump matplotlib minversion to 3.2

01f8b92

and remove unnecessary code added in astropy#8818 and astropy#8782 Add new JSON for figure tests.

		YMDHMS_FORMATS = ('fits', 'iso', 'isot')


		def time_support(scale='utc', format='isot', simplify=True):

Uh oh!

Add a new time_support context manager for plotting times #8782

Add a new time_support context manager for plotting times #8782

Uh oh!

Conversation

astrofrog commented Jun 1, 2019 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

mhvk left a comment

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!

astrofrog commented Jun 2, 2019

Uh oh!

astrofrog commented Jun 2, 2019

Uh oh!

larrybradley left a comment

Choose a reason for hiding this comment

Uh oh!

larrybradley commented Jun 3, 2019 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

mhvk commented Jun 3, 2019

Uh oh!

astrofrog commented Jun 7, 2019

Uh oh!

mhvk commented Jun 7, 2019

Uh oh!

astrofrog commented Jun 14, 2019

Uh oh!

codecov bot commented Jun 14, 2019 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

codecov bot commented Jun 14, 2019 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

mhvk left a comment

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!

larrybradley left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

astrofrog commented Jun 16, 2019

Uh oh!

mhvk commented Jun 16, 2019

Uh oh!

mhvk commented Jun 16, 2019

Uh oh!

Uh oh!

Add a new `time_support` context manager for plotting times #8782

Add a new `time_support` context manager for plotting times #8782

astrofrog commented Jun 1, 2019 •

edited

Loading

larrybradley commented Jun 3, 2019 •

edited

Loading

codecov bot commented Jun 14, 2019 •

edited

Loading

codecov bot commented Jun 14, 2019 •

edited

Loading