Matplotlib cursor position wrong after using asfreq method to change freq of DateTimeIndex from None to something

[willfurnass]

After using the asfreq method to change the frequency of a time-series DataFrame from None to something e.g. 15Min the cursor position in matplotlib graphs of that DataFrame is no longer correct (usually shows a datetime just after the unix epoch). The following demonstrates this (NB dt in df1 is not a constant):

df1 = pandas.read_csv('tseries1.csv', names=['tstamp', 'Q'], parse_dates=True, 
    index_col='tstamp').clip_lower(0).fillna(0)
df1['T'] = pandas.read_csv('tseries2.csv', names=['tstamp', 'T'], parse_dates=True, 
    index_col='tstamp', squeeze=True).clip_lower(0).fillna(0)

df2 = df1.asfreq(freq='15Min', method='ffill')
# NB df1.index.freq is None
# NB df2.index.freq is <15 * Minutes>
df1.plot()
df2.plot()
plt.show()

I find the Matplotlib cursor position to be invaluable when looking for features in very long time-series.

Versions:

• pandas master (commit ID 764b444)
• numpy 1.8
• matplotlib 1.3.0

[jreback]

can post a link or the file for csvs

or post a simple way to recreate them?

[willfurnass]

Hi jreback. Here's a self-contained example:

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

tstamps = np.array([
       datetime(2011, 10,  1,  0,  0),
       datetime(2011, 10, 10,  0, 15),
       datetime(2011, 10, 20,  0, 30), 
       datetime(2012, 11,  6, 23, 15),
       datetime(2012, 11,  8, 23, 30),
       datetime(2012, 11,  9, 23, 45)])

df_orig = pandas.DataFrame(data={'col1':np.linspace(3, 30, 6)}, index=tstamps)
df_orig.plot()
df_new = df_orig.asfreq(freq='15Min', method='ffill')
df_new.plot()
plt.show()

[jreback]

gr8 thanks.....would welcom a PR for this.....!

[willfurnass]

Crude workaround: edit _make_plot method in tools/plotting.py and clobber

if not self.x_compat and self.use_index and self._use_dynamic_x()

so that it always evaluates to False.