roll on data frame

[1danjordan]

Currently roll is unpredictable or breaks if passed anything other than a vector

[1danjordan]

need to reliably subset rows of data frames or vectors. Could coerce vectors to data frame, subset, compute function then return to vector. Or could find method that reliably subsets rows and elements of both...

[1danjordan]

Decided roll will only take a vector - data frame or tibbles can be computed on with

df %>% purrr::dmap(roll(.f))

Interactions could be managed by first mutating the df to include the interaction

df %>% mutate(interaction = col_1 * col_2) %>% dmap_at(interaction, roll(.f(.x))