new feature: data_apply()

[DominiqueMakowski]

Would be nice to have a small wrapper for sapply() to run on dataframes to do things like:

mtcars |>
  data_apply(as.factor, exclude="mpg")

[etiennebacher]

Since it has a data_ prefix, I suppose it should only accept functions that return the same number of rows, like as.factor().

[strengejacke]

What would be the use cases? Don't we have a lot of data functions that cover most situations? like to_factor()?

[strengejacke]

Else, we still could do:

data[] <- data |>
  data_select() |>
  lapply(<fun>)

[DominiqueMakowski]

Else, we still could do:

yes but that looks quite hacky

What would be the use cases?

mostly custom functions. basically same as sapply, but with a dataframe output + some selection helpers

[DominiqueMakowski]

Else, we still could do:

Unless you mean that this would be the code of data_apply(), in that case yes

[strengejacke]

yes but that looks quite hacky

Normal pipe workflow? 🤷

[etiennebacher]

Else, we still could do:

data[] <- data |>
  data_select() |>
  lapply(<fun>)

Doesn't work (mpg shouldn't be factor):

library(datawizard)

foo <- mtcars[, 1:5]

foo[] <- foo |> 
  data_select(exclude = "mpg") |>  
  lapply(as.factor)

lapply(foo, class)
#> $mpg
#> [1] "factor"
#> 
#> $cyl
#> [1] "factor"
#> 
#> $disp
#> [1] "factor"
#> 
#> $hp
#> [1] "factor"
#> 
#> $drat
#> [1] "factor"

Can be done quite easily though:

library(datawizard)

foo <- mtcars[, 1:5]

to_convert <- foo |> 
  data_find(exclude = "mpg")

foo[to_convert] <- lapply(foo[to_convert], as.factor)

lapply(foo, class)
#> $mpg
#> [1] "numeric"
#> 
#> $cyl
#> [1] "factor"
#> 
#> $disp
#> [1] "factor"
#> 
#> $hp
#> [1] "factor"
#> 
#> $drat
#> [1] "factor"

[DominiqueMakowski]

data_apply <- function(data, fun, ...) {
  to_convert <- data_find(data, ...)
  data[to_convert] <- as.data.frame(lapply(data[to_convert], fun))
  data
}

It's a small addition to the codebase but quite convenient imo

[etiennebacher]

Should this be an internal function in the package where you want to use it? Personally I'm a bit reluctant because it's getting quite close to dplyr::mutate() and dplyr::summarise() and The alternative takes two lines so I don't really see the point of making another function.

That said, @easystats/core-team what do you think?

[DominiqueMakowski]

I was thinking of an external one as a combination of base's apply + selection

[rempsyc]

I still rely pretty heavily on dplyr personally so I don't really mind. Though, if the code is there already, I see little harm in adding a new function (except associated work with maintenance, testing, documentation).

[bwiernik]

I'm not really understanding the use case here? There's pretty well-known and reasonably convenient base syntax like Daniel showed for developers, and map(), mutate(), and summarize() exist for end users

[bwiernik]

As far as base goes, I don't see why one would use this (or sapply() in general) over apply(., 2) ?