ENH: add Series.set_index

[ghost]

Related #21684

update: use set_axis.

import pandas as pd
 
dates=[
"2001-01-01 23:50",
"2001-01-02 14:00", 
"2001-01-03 00:08"] 
 
ser=pd.Series(range(3),pd.DatetimeIndex(dates))

sometimes when writing a sequence of operation, you want to replace the index of the object you're working on. Say, quantizing the dates. I can't find the .method way of doing
ser.index= expr
So that I can't write something like

(ser
      .replace_index(ser.index.floor("D"))    # < this doesn't exist
      .rolling("5d")
       .sum()

Things that don't solve this:

• reindex is aligning, so if your new index permutes the indexes, it isn't suitable.
• set_index takes a column name, not an expression, and isn't available on series. update works for frames, though you must provide an arraylike, not a listlike.
• reindex_like expects a PandasObject instead of a listlike, and is aligning
• reset_index does not take a values argument

In [1]: [_ for _ in dir(ser) if 'index' in _ and not _.startswith("_")]
Out [1]: 
['first_valid_index',
 'index',
 'last_valid_index',
 'reindex',
 'reindex_axis',
 'reindex_like',
 'reset_index',
 'sort_index']

[Liam3851]

You can do this in a chain now by converting the Series to DataFrame, using set_index, and then converting back to Series, since DataFrame.set_index accepts an arraylike to use as the new index:

In [21]: ser.to_frame() \
    ...:    .set_index(ser.index.floor('D')) \
    ...:    .iloc[:, 0] \
    ...:    .rolling('5D') \
    ...:    .sum()
    ...:
Out[21]:
2001-01-01    0.0
2001-01-02    1.0
2001-01-03    3.0
Name: 0, dtype: float64

Perhaps there is a case for set_index on Series that would provide the arraylike functionality of DataFrame.set_index without requiring conversion to/from DataFrame?

[ghost]

Thanks, you're right. I updated the OP.

[ghost]

It turns out that the set_axis method has existed since 0.21 for doing this, but it seems it's not very well-known. Closing for now, as this is a dupe of #21684 and there are larger issues there, with much previous discussion.