New family to replace summarise_each/mutate_each

[hadley]

by_cyl <- mtcars[1:5] %>% group_by(cyl)
by_cyl %>% summarise_all(mean)
by_cyl %>% summarise_if(is.numeric, mean)
by_cyl %>% summarise_at(cols(cyl, mpg, wt), funs(mean, median), na.rm = TRUE)

by_cyl %>% mutate_all(scale)
by_cyl %>% mutate_if(is.numeric, scale)
by_cyl %>% mutate_at(cols(cyl, mpg, wt), funs(scale, rescale01))
by_cyl %>% mutate_at(c("cyl", "mpg", "wt"), funs(scale, rescale01))

summarise_each() and mutate_each() would be deprecated in favour of this new approach. And this would replace (e.g) purrrr::dmap().

And cols would be something along the lines of:

cols <- function(...) eval(subsitute(alist(...)))

(but with a special class)

[hadley]

@lionel- would this be a complete replacement for the current purrr functions that work with rows/dataframes? Or would there still be somethings that you couldn't do?

[lionel-]

would this be a complete replacement for the current purrr functions that work with rows/dataframes?

Yes, expect for column collation, which is probably not that important.

And list collation would be handled with tidyr::nest() + dplyr::mutate() + purrr::map() if you want to remove by_slice() from purrr as well.

[lionel-]

do you need me to take a look at this @hadley?

[hadley]

Yes, that would be great!