[FIXED] Sorting of an index in Datetime format so that the first data of the day is the time 1:00

Issue

I do not understand why this solution does not work to move forward one day in some rows of the series, working on the index.

curva_generacion.loc[curva_generacion.index.hour == 0].index = curva_generacion.loc[curva_generacion.index.hour == 0].index - pd.DateOffset(days=1)

This is my DataFrame, and with the previous solution I am unable to change the data:

                     **SVH.2.1 BIS**
**FECHA**                           
2015-01-01 00:00:00         0.00
2015-01-01 01:00:00         0.00
2015-01-01 02:00:00         0.00
[...]
2015-01-01 21:00:00         0.00
2015-01-01 22:00:00         0.00
2015-01-01 23:00:00         0.00
2015-01-02 00:00:00         0.00
2015-01-02 01:00:00         0.00
2015-01-02 02:00:00         0.00

When it is 0:00, the table includes the data in the following day. To be compatible with the table I already work with, the 0:00 data must belong to the previous day.

I have managed to filter well the data I need to change and subtract a day from it, but the index is not modified.

EDIT

I think I have a conceptual error that I have seen after testing the following code.

auxiliar1 = curva_generacion.loc[curva_generacion.index.hour == 0]
auxiliar1.index = auxiliar1.index - pd.DateOffset(days=1)
auxiliar1.index = auxiliar1.index.map(lambda t: t.replace(year=2015))
auxiliar2 = curva_generacion.loc[curva_generacion.index.hour != 0]
curva_generacion = pd.concat([auxiliar1, auxiliar2])
curva_generacion.sort_index(inplace=True)

With this code I extract the data from which I want to subtract a day, I do the operation, and I concatenate them and one with the data that I have not modified. When I re-sort the index, Python puts back as the first data of the day the one corresponding to 0:00.

What I really need is to sort an index where the first data of the day is the 1:00 a.m. data, and the last is the 0:00 data.

Solution

I can suggest an option with resetting the index, setting the required values and then installing the index.

import pandas as pd

curva_generacion = curva_generacion.reset_index()

ind = curva_generacion['**FECHA**'].dt.hour == 0

curva_generacion.loc[ind, '**FECHA**'] = (curva_generacion.loc[ind, '**FECHA**'] -
                                          pd.DateOffset(days=1))
curva_generacion = curva_generacion.set_index('**FECHA**')

print(curva_generacion)

Output:

                     **SVH.2.1 BIS**
**FECHA**                           
2014-12-31 00:00:00                1
2015-01-01 01:00:00                2
2015-01-01 02:00:00                3
2015-01-01 21:00:00                4
2015-01-01 22:00:00                5
2015-01-01 23:00:00                6
2015-01-01 00:00:00                7
2015-01-02 01:00:00                8
2015-01-02 02:00:00                9

Update: Updating indexes without resetting. And an error when setting individual values: 'TypeError: Index does not support mutable operations'. Rather, it concerns the installation of individual values, since if you reinstall everything, this will complete successfully.

import numpy as np
import pandas as pd
    
ind = curva_generacion.index.hour == 0
val_index = curva_generacion.index.values.copy()
val_index[ind] = val_index[ind] - np.timedelta64(1,'D')

curva_generacion.index = val_index

Answered By - inquirer