tar_group data.frame helper?

[liutiming]

I was wondering whether a function like the following would be helpful to many people. I can create a PR if so...

#' group a table into n tar_group for use by [targets::tar_group()]
id_group  <- function(id_tbl, n) {
    id_tbl  %>% 
    rowid_to_column  %>% 
    mutate(group_num = rowid %% n + 1) %>% group_by(group_num) %>% tar_group() 
}

[wlandau]

Is the idea to treat each row as a group? If so, iteration = "vector" already covers this, thanks to vctrs magic.

library(targets)
tar_script({
  library(gapminder)
  list(
    tar_target(data, gapminder),
    tar_target(each_row, data, pattern = map(data))
  )
})

tar_make(reporter = "silent")

tar_read(each_row, branches = 1)
#> # A tibble: 1 x 6
#>   country     continent  year lifeExp     pop gdpPercap
#>   <fct>       <fct>     <int>   <dbl>   <int>     <dbl>
#> 1 Afghanistan Asia       1952    28.8 8425333      779.

tar_read(each_row, branches = 2)
#> # A tibble: 1 x 6
#>   country     continent  year lifeExp     pop gdpPercap
#>   <fct>       <fct>     <int>   <dbl>   <int>     <dbl>
#> 1 Afghanistan Asia       1957    30.3 9240934      821.

^{Created on 2021-02-26 by the reprex package (v1.0.0)}

[liutiming]

thanks! not each row, but per n rows -> 1 group

[wlandau]

Interesting. Maybe this would be a good addition to tarchetypes. tar_group_id(name, produce_df(), rows = 10) could be equivalent to:

tar_target(
  name,
  { # best to put this in an exported tarchetypes function with `@keywords internal`
    out <- produce_df()
    out$tar_group <- seq_len(nrow(out)) %% rows + 1L # does the work of tar_group()
    out
  },
  format = "fst_tbl" # default format but could be modified
  iteration = "group" # user should not be able to modify
)

We might also consider a tar_group_by(name, produce_df(), by = c("var1", "var2")) helper equivalent to:

tar_target(
  name, {
    produce_df() %>%
      group_by(var1, var2) %>%
      tar_group() %>%
      ungroup()
  },
  format = "fst_tbl"
  iteration = "group"
)

[wlandau]

Do you have a strong interest in implementation? Otherwise, I will. I feel these are important helpers.

[wlandau]

On reflection, I now have strong opinions on what these functions should be named and how they should be implemented, so I would prefer to implement them myself.

Your idea is super nice, and I think it will make grouped operations a lot easier.

[liutiming]

aha thanks! sure I look forward to it!