python-notes

Python coding tips.

Date Time & Time Zone

There are different types of classes to deal with date time and time zones: datetime, zoneinfo, pytz, pandas.

Use zoneinfo to specify the time zone of an datetime object, do NOT use pytz.

from datetime import datetime
from zoneinfo import ZoneInfo
dt = datetime(2048,1,1,9,30, tzinfo=ZoneInfo('Asia/Hong_Kong'))
dt.timestamp() # convert to unix timestamp

Caution

Below is the wrong way to specify the time zone:

dt = datetime(2048,1,1,9,30, tzinfo=pytz.timezone('Asia/Hong_Kong'))
dt.timestamp() # this will convert to a wrong unix timestamp

Naive datetime vs time-zone-aware datetime

dt_naive = datetime(2048,1,1,9,30) # naive
dt_tz = datetime(2048,1,1,9,30, tzinfo=ZoneInfo('Asia/Hong_Kong')) # tz-aware

# conversion
dt_naive = dt_tz.replace(tzinfo=None)
dt_tz = dt_naive.replace(tzinfo=ZoneInfo('Asia/Hong_Kong'))

Unix timestamp => date time is not that tricky but I suggest still stick with zoneinfo instead of pytz

tstamp = 2461455000
datetime.fromtimestamp(tstamp, tz=ZoneInfo('Asia/Hong_Kong')).strftime('%Y-%m-%d %H:%M:%S')
# datetime.fromtimestamp(tstamp, tz=pytz.timezone('Asia/Hong_Kong')).strftime('%Y-%m-%d %H:%M:%S') # also works but not suggested

For class pandas._libs.tslibs.timestamps.Timestamp object, it can be converted to a tz-aware datetime object

dt_pandas.to_pydatetime()

Note

If you insist to use pytz with datetime, follow the steps below to convert to the correct unix timestamp:

dt = datetime(2048,1,1,9,30, tzinfo=pytz.timezone('Asia/Hong_Kong'))
# dt.timestamp() # this will convert to a wrong unix timestamp
dt_naive = dt.replace(tzinfo=None) # convert back to naive
pytz.timezone('Asia/Hong_Kong').localize(dt_naive).timestamp()

Python Interactive Charts

This issue is found in Python 3.10.12 (main, Nov 20 2023, 15:14:05). In a clean Python3 environment, do pip install matplotlib. Matplotlib will not show the figure. The below codes reproduce the issue.

import matplotlib.pyplot as plt
plt.plot([1,2],[1,2])
plt.show()

Output:

UserWarning: FigureCanvasAgg is non-interactive, and thus cannot be shown

This is because the charting back end is usually "Agg" which is non-interactive. A recommended back end is "TkAgg" by module tkinter which suprisingly is not a dependency of matplotlib. To install tkinter, do

sudo apt install python3-tk

Note

Neither pip install tkinter nor pip install tk works.

Another back end found in some online solutions is PyQt5, which does NOT work for me.

To view the back end used by matplotlib

import matplotlib
matplotlib.get_back()

To change the back end

matplotlib.use('TkAgg')