You might like Dplython, pandas-ply, and dfply

koaning / kadro

A friendly pandas wrapper with a more composable grammar support.

MIT License

15 stars 3 forks source link

You might like Dplython, pandas-ply, and dfply #8

Closed data-steve closed 7 years ago

data-steve commented 7 years ago

This is definitely the one thing I miss about R. And Rstudio

koaning commented 7 years ago

I made this kind of as a personal project but also because a few things about those packages did not work well for me;

i am not a big fan of using an infix X. i'd much rather use lambda functions. it feels a bit verbose, but a little bit less 'magical' to me. also ... why not use _ instead of X?
the >> operator from dfply is not a true global operator and will not be very polymorphic. this means that I need to pay attention what functions/methods I can apply.

These preferences are highly personal but reason enough to make a minimum viable library that scratches my own personal itch. I also like being able to use both my own wrapper but still be able to access to good old normal pandas dataframe should I ever need it.

data-steve commented 7 years ago

I agree the X is not ideal. I've seen the _ in scala.

I admire you for doing it. I wish pandas would officially adopt Dplyr-like chaining so I can use those packages at work.

Well done

~ Steve

Sent via telepathy

On Mar 4, 2017, at 3:26 PM, vincent d warmerdam notifications@github.com wrote:

Closed #8.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or mute the thread.

koaning commented 7 years ago

Pandas historically put a bit more focus on performance (and to be frank, pandas is still impressively fast) but it has aquired a lot of technical debt over the years. I imagine there being a fair amount of willingness from some of the Pandas devs to implement this, but isn't super easy to do on such a large project that so many companies rely on. This is why I think a wrapper that demos functionality might make some more sense. You'll lose very little by doing that.

I also don't think pandas is the thing in the way of proper dplyr chaining, it is also the python language that isn't as flexible as R in terms of creating operators.