Feature request: pandas.Series.query()

I would like to request a pandas.Series.query() method that works identically to pandas.DataFrame.query().

I have a large multi-indexed series that I would like to split into training and testing data for an ML research project. A minimal working example would be:

import pandas as pd

years = range(2002, 2018)
fields = range(1, 5)

index = pd.MultiIndex.from_product(
    [years, fields], names=['year', 'field'])

series = pd.Series(index=index)

What I would like to be able to do is split this series into 2010 data and not 2010 data. Accessing 2010 data is very easy:

test_data = series[2010]

Accessing not 2010 data is very hard. This is the shortest method I’ve found so far:

train_data = series.to_frame().query('year != 2010')[0]

Since pandas.Series doesn’t support query(), I have to convert it to a DataFrame and then back into a Series. Is there any reason why pandas.Series doesn’t support query directly?

[adamjstewart]

I agree that the masking solution works just fine, but I do still think that having Series and DataFrame support different methods is quite non-intuitive. Especially coming from numpy, where there is no distinction between 1D and 2D arrays, they all support identical operations.

[WillAyd]

Perhaps a better solution to your problem would be to use a mask based off of the index values, so:

mask = index.get_level_values('year') == 2010
train = series[mask]
test = series[~mask]