Repository navigation
Unevenly spaced ticks in plot_date() #13183
Description
Bug report
When plotting with plot_date(), the x axis ticks can be unevenly spaced. It looks like when the tick spacing is less than a month, the start of the month is always present as a tick, which may mean the space between it and the previous tick is small. This can lead to overlapping tick labels and generally poor looking axes and grids.
Code for reproduction
import numpy as np
import matplotlib.pyplot as plt
start = np.datetime64('2018-01-01', 'D')
end = np.datetime64('2018-04-01', 'D')
dates = np.arange(start, end)
data = np.random.randn(len(dates))
plt.plot_date(dates, data)
plt.grid(True)
plt.gcf().autofmt_xdate()
plt.show()Actual outcome
Expected outcome
The plot should have evenly spaced gridlines and axis ticks, unless a custom locator has specified otherwise.
Matplotlib version
- Operating system: Arch Linux
- Matplotlib version: 3.0.2
- Matplotlib backend (print(matplotlib.get_backend())): Qt5Agg
- Python version: 3.7.2
- Jupyter version (if applicable): n/a
- Other libraries: matplotlib 1.15.4
Matplotlib, Python and numpy are installed with the OS package manager using the standard Arch Linux repositories.
Activity
jklymak commented on Jan 14, 2019
Yes, this was an unintended consequence of #9801. Its been "fixed" by #10841 which is on master...
Now "fixed" is still different to how it was before, which was (v2.2.3):
If you want the old behaviour back, you need to do a bit of a song and dance.
I agree that the way things are in 3.0.2 are undersirable because to intervals were not good (i.e. you could get 29, 1 ticks). OTOH, I think the new behaviour (as of master) is better than the 2.2.3 behaviour.
chrisjbillington commented on Jan 14, 2019
Oh, that sounds like a great improvement! Snapping the ticks to month edges makes a lot of sense (so long as they're still evenly spaced).
I have no complaints then, and will close this issue since it's fixed already.
jklymak commented on Jan 14, 2019
No problem - sorry for the inconvenience of the bad state right now...
jklymak commented on Jan 14, 2019
Oh, sorry, "evenly spaced" for months won't work, of course.... Note the ticks are not evenly spaced above...
chrisjbillington commented on Jan 14, 2019
Oh, right, since not all months have the same number of days. Well, it's close enough not to be jarring and is more understandable than exactly even spacing (I guess I didn't realise prior to now that this was a trade-off, I just though matplotlib had been choosing the ticks poorly for no reason).
jklymak commented on Jan 14, 2019
OK, thats what "we" decided, but its good to make sure the users are happy too. These things come down to esthetic decisions, and not everyone agrees on the best defaults.
BTW there will also be a new formatter available, so if you happen to be on master, your opinions on that would be welcome. https://matplotlib.org/devdocs/gallery/ticks_and_spines/date_concise_formatter.html#sphx-glr-gallery-ticks-and-spines-date-concise-formatter-py
chrisjbillington commented on Jan 18, 2019
The ConciseDateFormatter looks great! I tried it out and can't think of any complaints against it.