allow referring to previously created columns in summarize

[machow]

e.g.

from siuba.data import mtcars
from siuba import *

mtcars >> summarize(avg_mpg = _.mpg.mean(), avg_kpg = _.avg_mpg * 1.6)

[machow]

Just a thought, this is very doable in the experimental fast_summarize. This is because it can compose operations mixing aggregated results (e.g. 1 value per group) and the original data (see ADR-003.

e.g.

mtcars >> group_by(_.cyl) >> summarize(wat = _.mpg.mean() + _.mpg)

Would just be...

• an elementwise operation + over..
• GroupByAgg (mpg.mean())
• Original column mpg