How to plot Pandas .groupby() objects in Seaborn - using method chaining

Question

I would be really grateful for any advice on how to pipe a Pandas .groupby() object into Seaborn

Im trying to plot the .groupby() object but it give me an error message: ValueError: If using all scalar values, you must pass an index

gapminder.\
    groupby('year').\
    agg({'pop' : ['sum'],
         'lifeExp' : ['mean']}).\
    reset_index().\
    pipe((sns.relplot, "data"), x = "pop", y = "lifeExp", kind = "scatter").\
    set(xscale = "log")
plt.show()

Answer 1

Pass a lambda, then you can call the function as you please:

.pipe(lambda d: sns.relplot(data=d, x="pop", y="lifeExp", kind="scatter"))

Where d represents your DataFrame, passed in whole as the sole argument to .pipe .

Answer 2

The answer is helped by a new feature added to Pandas in version 0.25, "Named aggregation with .groupby()". This allows for the creation of variables through the .agg() function and the output into a tidy dataframe with a uniform index, as opposed to the multi-indexed groupby object in Pandas <= version 0.24. Below shows how a query can be chained through Pandas and also seamlessly into Seaborn.

gapminder.\
   groupby('year').\
   agg(pop_sum = ('pop', sum), 
       lifeExp_mean = ('lifeExp', 'mean')).\
   reset_index().\
   pipe((sns.relplot,"data"), x="pop_sum", y="lifeExp_mean", kind="scatter").\
   set(xscale = "log")
plt.show()