Hygenic Macros

Requesting Hygenic Macros like in macropy: https://github.com/lihaoyi/macropy

Its useful for example to refer to the columns of a dataframe of object in a formula or expression that is later expanded without being eagerly evaluated :

• lm(y~x,df)
• mutate(flights, gain = arr_delay - dep_delay, speed = distance / air_time * 60)

Or write more complex arbitrary domain specific languages for data analysis packages.

Relevant PEP: https://mail.python.org/pipermail/python-ideas/2015-March/032822.html

More use cases: http://multithreaded.stitchfix.com/blog/2015/03/17/grammar-of-data-science/

CC: @shoyer who might be interested in coconut as a Pydata DSL (with piping, and lots of other functional goodness, compiles to valid python 2/3, 100% compatible with python 3 etc) and this issue specifically.

Thanks!

[evhub]

@datnamer @TomAugspurger It doesn’t resolve this issue, but I thought you should know that in v1.1.1, I added a methodcaller implicit partial, which I believe should be a big help in using libraries like pandas with Coconut, since, because Python lacks pipes, pandas uses methods for everything. Thus,

flights |> group_by('year', 'month', 'day')

won’t work in Coconut, since it assumes group_by is a function, not a method. But now, you should be able to do

flights |> .group_by('year', 'month', 'day')

(note that the only change was adding a .) which will allow you to mix and match methods and custom functions when using pipeline-style programming, so that

flights |> .group_by('year', 'month', 'day') |> my_transform_func |> .select('arr_delay', 'dep_delay')

will work, which is much cleaner than in Python, where you’d have to do

my_transform_func(flights.group_by('year', 'month', 'day')).select('arr_delay', 'dep_delay')

which obscures the order in which the functions / methods are called.

[evhub]

Thanks for the issue! I’ll look into this and see what I can do for the next version.