Matplotlib麻烦绘制x标签

Question

having issues using set_xlim . (possibly because of datetime objects??)

here's my code (doing it in ipython notebook):

%matplotlib inline
import matplotlib.pyplot as plt
import numpy as np
import datetime

date_list = [datetime.datetime(2015, 6, 20, 0, 0), datetime.datetime(2015, 6, 21, 0, 0), datetime.datetime(2015, 6, 22, 0, 0), datetime.datetime(2015, 6, 23, 0, 0), datetime.datetime(2015, 6, 24, 0, 0), datetime.datetime(2015, 6, 25, 0, 0), datetime.datetime(2015, 6, 26, 0, 0)]
count_list = [11590, 10743, 27369, 31023, 30569, 31937, 30205]

fig=plt.figure(figsize=(10,3.5))
ax=fig.add_subplot(111)

width = 0.8

tickLocations = np.arange(7)

ax.set_title("Turnstiles Totals for Lexington Station C/A A002 Unit R051 from 6/20/15-6/26-15")
ax.bar(date_list, count_list, width, color='wheat', edgecolor='#8B7E66', linewidth=4.0)
ax.set_xticklabels(date_list, rotation = 315, horizontalalignment = 'left')

This gives me:

But when I try to make some extra space on the leftmost and rightmost edges with this code:

ax.set_xlim(xmin=-0.6, xmax=0.6)

I get this huge error (this is just the bottom snippet):

    223         tz = _get_rc_timezone()
    224     ix = int(x)
--> 225     dt = datetime.datetime.fromordinal(ix)
    226     remainder = float(x) - ix
    227     hour, remainder = divmod(24 * remainder, 1)

ValueError: ordinal must be >= 1

Any idea what's going on guys? thanks!

Answer 1

For various historical reasons, matplotlib uses an internal numerical date format behind-the-scenes. The actual x-values are in this data format, where 0.0 is Jan 1st 1900, and a difference of 1.0 corresponds to 1 day. Negative values aren't allowed.

The error you're getting is because you're trying to set the x-limits to include a negative range. Even without the negative number, though, it would be a range on Jan. 1st, 1900.

Regardless, it sounds like what you're wanting isn't ax.set_xlim at all. Try ax.margins(x=0.05) to add 5% padding in the x-direction.

As an example:

import matplotlib.pyplot as plt
import numpy as np
import datetime

count_list = [11590, 10743, 27369, 31023, 30569, 31937, 30205]
date_list = [datetime.datetime(2015, 6, 20, 0, 0),
             datetime.datetime(2015, 6, 21, 0, 0),
             datetime.datetime(2015, 6, 22, 0, 0),
             datetime.datetime(2015, 6, 23, 0, 0),
             datetime.datetime(2015, 6, 24, 0, 0),
             datetime.datetime(2015, 6, 25, 0, 0),
             datetime.datetime(2015, 6, 26, 0, 0)]

fig, ax = plt.subplots(figsize=(10,3.5))
ax.set_title("Turnstiles Totals for Lexington Station C/A A002 Unit R051 from "
             "6/20/15-6/26-15")

# The only difference is the align kwarg: I've centered the bars on each date
ax.bar(date_list, count_list, align='center', color='wheat',
       edgecolor='#8B7E66', linewidth=4.0)

# This essentially just rotates the x-tick labels. We could have done
# "fig.autofmt_xdate(rotation=315, ha='left')" to match what you had.
fig.autofmt_xdate()

# Add the padding that you're after. This is 5% of the data limits.
ax.margins(x=0.05)

plt.show()

Note that if you wanted to expand the x-limits by exactly 0.6 in each direction, you'd do something like:

xmin, xmax = ax.get_xlim()
ax.set_xlim([xmin - 0.6, xmax + 0.6])

However, ax.margins(percentage) is much easier as long as you're okay with the "padding" being in terms of a ratio of the current axes limits.