Python Pandas - Perform floor operation on the TimeDeltaIndex with microseconds frequency

The TimeDeltaIndex.floor() method performs a floor operation on TimeDelta values, rounding down to the nearest specified frequency. When working with microseconds frequency, use freq='us' to round down to the nearest microsecond.

Syntax

TimeDeltaIndex.floor(freq)

Parameters

• freq ? The frequency to floor to. For microseconds, use 'us'

Creating a TimeDeltaIndex

First, create a TimeDeltaIndex with various time intervals including nanosecond precision ?

import pandas as pd

# Create a TimeDeltaIndex object with microseconds and nanoseconds
tdIndex = pd.TimedeltaIndex(data=['5 day 8h 20min 35us 45ns', '+17:42:19.999999',
                                  '7 day 3h 08:16:02.000055', '+22:35:25.999999'])

print("Original TimedeltaIndex...")
print(tdIndex)

Original TimedeltaIndex...
TimedeltaIndex(['5 days 08:20:00.000035045', '0 days 17:42:19.999999',
                '7 days 11:16:02.000055', '0 days 22:35:25.999999'],
               dtype='timedelta64[ns]', freq=None)

Performing Floor Operation with Microseconds Frequency

Apply the floor operation to round down to the nearest microsecond ?

import pandas as pd

# Create TimeDeltaIndex with nanosecond precision
tdIndex = pd.TimedeltaIndex(data=['5 day 8h 20min 35us 45ns', '+17:42:19.999999',
                                  '7 day 3h 08:16:02.000055', '+22:35:25.999999'])

# Perform floor operation with microseconds frequency
floored_index = tdIndex.floor(freq='us')

print("After floor operation with microseconds frequency...")
print(floored_index)

After floor operation with microseconds frequency...
TimedeltaIndex(['5 days 08:20:00.000035', '0 days 17:42:19.999999',
                '7 days 11:16:02.000055', '0 days 22:35:25.999999'],
               dtype='timedelta64[ns]', freq=None)

Complete Example

Here's a complete example showing the components breakdown and floor operation ?

import pandas as pd

# Create a TimeDeltaIndex object
tdIndex = pd.TimedeltaIndex(data=['5 day 8h 20min 35us 45ns', '+17:42:19.999999',
                                  '7 day 3h 08:16:02.000055', '+22:35:25.999999'])

# Display original TimedeltaIndex
print("TimedeltaIndex...")
print(tdIndex)

# Show the components breakdown
print("\nThe Dataframe of the components of TimeDeltas...")
print(tdIndex.components)

# Floor operation with microseconds frequency
print("\nPerforming Floor operation with microseconds frequency...")
print(tdIndex.floor(freq='us'))

TimedeltaIndex...
TimedeltaIndex(['5 days 08:20:00.000035045', '0 days 17:42:19.999999',
                '7 days 11:16:02.000055', '0 days 22:35:25.999999'],
               dtype='timedelta64[ns]', freq=None)

The Dataframe of the components of TimeDeltas...
   days  hours  minutes  seconds  milliseconds  microseconds  nanoseconds
0     5      8       20        0             0            35           45
1     0     17       42       19           999           999            0
2     7     11       16        2             0            55            0
3     0     22       35       25           999           999            0

Performing Floor operation with microseconds frequency...
TimedeltaIndex(['5 days 08:20:00.000035', '0 days 17:42:19.999999',
                '7 days 11:16:02.000055', '0 days 22:35:25.999999'],
               dtype='timedelta64[ns]', freq=None)

Key Points

• The floor operation rounds down to the nearest specified frequency unit
• Notice how the nanoseconds (45ns) in the first entry are removed after flooring to microseconds
• Values already at microsecond precision remain unchanged
• The operation returns a new TimeDeltaIndex object

Conclusion

Use TimeDeltaIndex.floor(freq='us') to round TimeDelta values down to the nearest microsecond. This operation removes any nanosecond precision while preserving all higher-level time components.