Having an issue using matplotlib event.xdata when plotting pandas.Timeseries, I tried to reproduce the answer proposed in a very related question, but get a very strange behavior. Here's the code, adapted to python3 and with a little more stuff in the on_click() function: import numpy as np import pandas as pd import matplotlib.pyplot as plt import matplotlib.dates as mdates def on_click(event): if event.inaxes is not None: # provide raw and converted x data print(f"{event.xdata} --> {mdates.num2date(event.xdata)}") # add a vertical line at clicked location line = ax.axvline(x=event.xdata) plt.draw() t = pd.date_range('2015-11-01', '2016-01-06', freq='H') y = np.random.normal(0, 1, t.size).cumsum() df = pd.DataFrame({'Y':y}, index=t) fig, ax = plt.subplots() line = None df.plot(ax=ax) fig.canvas.mpl_connect('button_press_event', on_click) plt.show() If I launch this, I get the following diagram, with expected date range between Nov. 2015 and Jan. 2016, as is the cursor position information provided in the footer of the window (here 2015-11-01 10:00), and correct location of the vertical lines: However, the command-line output is as follows: C:\Users\me\Documents\code\>python matplotlib_even.xdate_num2date.py 402189.6454115977 --> 1102-02-27 15:29:23.562039+00:00 402907.10400704964 --> 1104-02-15 02:29:46.209088+00:00 Those event.xdata values are clearly out of both input data range and x axis data range, and are unusable for later use (like, try to find the closest y value in the serie). So, does anyone know how I can get a correct xdata? Continue reading...
