Format pandas y axis to show time instead of total seconds

Question

i have measurements in a dataframe. Columns are different objects. Index is a datetime64 index. Now for each date I have a measurement in total seconds (int) for each column.

Everything plots quite nice, my only problem instead of showing 6000 seconds on the y axis i want to show 1:40 to indicate 1 hour and 40 minutes.

How can I actually achieve this?

day         Object1  Object2
2017-01-01     6000     1234

I want

day         Object1  Object2
2017-01-01  1:40:00  00:20:34 

Can you hint me on how to do it

Answer 1

It is possible, but ploting timedelta is not supported yet natively.

df['Object1'] = pd.to_timedelta(df['Object1'], unit='s')
df['Object2'] = pd.to_timedelta(df['Object2'], unit='s')

Or:

cols = ['Object1', 'Object2']
df[cols] = df[cols].apply(lambda x: pd.to_timedelta(x, unit='s'))

---

print (df)
          day  Object1  Object2
0  2017-01-01 01:40:00 00:20:34

But is possible it by FuncFormatter:

df = pd.DataFrame({'Object1': [6000, 4000, 3000], 'Object2':[3000,5000,2110]})

import matplotlib.ticker as tkr
import datetime

def func(x, pos):
    return str(datetime.timedelta(seconds=x))
fmt = tkr.FuncFormatter(func)


ax = df.plot(x='Object1', y='Object2', rot=90)
ax.xaxis.set_major_formatter(fmt)
ax.yaxis.set_major_formatter(fmt)