CODEWITHSUNDEEP
pythonpandasnumpydatetime
ajrlewis
ajrlewis

Reputation: 3058

pandas: Transforming timestamp index to ISO week quickly

I have a function that returns the index value of a pandas Series instance comprised of an array of pd.Timestamp values as an ISO week value.

import numpy as np
import pandas as pd

timestamps = [pd.Timestamp('2014-01-01'), pd.Timestamp('2014-02-01'), pd.Timestamp('2014-02-01')]
quantities = [1.0, 1.0, 1.0]
series = pd.Series(quantities, index=timestamps)

def timestamps_iso() -> np.array:
    timestamps_iso = []
    for timestamp in series.index:
        timestamp_iso = timestamp.isocalendar()[1]
        if timestamp_iso == 53: # fix ISO week number 53.
            timestamp_iso = np.random.choice([52, 1])
        timestamps_iso.append(timestamp_iso)
    return np.array(timestamps_iso)

Can this be made quicker, or done in a more pandas style, please? It takes a lot of time when the index is large to run.

Thanks for any help.

Upvotes: 0

Views: 179

Answers (1)

w-m
w-m

Reputation: 11232

Your question says month value, but the code says week value. I assume you're looking for the week value. You can access the week number on a datetime series with .week:

In [1]: timestamps = [pd.Timestamp('2014-01-01'), pd.Timestamp('2014-02-01'), pd.Timestamp('2014-02-01')] 
   ...: quantities = [1.0, 1.0, 1.0] 
   ...: series = pd.Series(quantities, index=timestamps)                                                                                                                                                                                             

In [2]: series.index.week                                                                                                                                                                                                                            
Out[2]: Int64Index([1, 5, 5], dtype='int64')

Upvotes: 1

Related Questions