pandas series to_dict with repeated indices; make dict with list values

Question

If you have a neat pandas Series object with unique indices, then using pd.Series.to_dict() works as you might expect it: it becomes a Python dict with each index pointing to its respective value.

This gets complicated if you have non-unique indices. My expected behavior is that values with the same index will get grouped together into a list, and the dict will have the index as a key and the list as a value. What I observe instead is a dict with the index as a key and only a single value from the Series as the value in the dict.

Is there a way to achieve my expected behavior, built into pandas, or close to it? Presently, I manually curate values that match each index into the dict in a for loop, looping over the unique index values. Is there a better way to do this?

EDIT: Here's an example:

my_series = pd.Series(['val_1', 'val_2', 'val_3', 'val_4', 'val_5'])
my_series.index = ['1', '1', '2', '2', '2']
my_series

Yields

1    val_1
1    val_2
2    val_3
2    val_4
2    val_5
dtype: object

Now, to_dict() with the 1:1 matching behavior:

my_series.to_dict()
{'1': 'val_2', '2': 'val_5'}

What I would like to see instead is:

{'1': ['val_1', 'val_2'], '2': ['val_3', 'val_4', 'val_5']}

I can achieve this doing

{idx:list(my_series[idx]) for idx in set(my_series.index)}
{'2': ['val_3', 'val_4', 'val_5'], '1': ['val_1', 'val_2']}

What I would like to know is if there is a more native way to do this in pandas, or if this is the best way to handle the problem.

Answer 1

Try this:

my_series.groupby(level=0).agg(list).to_dict()

Out[358]: {'1': ['val_1', 'val_2'], '2': ['val_3', 'val_4', 'val_5']}