data_rename(): "Error: Following variable(s) were not found" in version '0.13.0.17'

[rempsyc]

Yo, trying to integrate the devel version of datawizard '0.13.0.17' I'm facing a breaking change in data_rename() (more info in easystats/report#470). Just wanted to check with you if that's something I should change in report or datawizard. My attempt at a minimal reprex:

datawizard '0.13.0.17'

packageVersion("datawizard")
#> [1] '0.13.0.17'

result <- list(
  setosa.a = structure(data.frame(1)), 
  versicolor.a = structure(data.frame(2)))

new_names <- c("setosa, a", "versicolor, a")

names(result)
#> [1] "setosa.a"     "versicolor.a"
new_names
#> [1] "setosa, a"     "versicolor, a"

datawizard::data_rename(result, select = names(result), replacement = new_names)
#> Error: Following variable(s) were not found: setosa.a, versicolor.a
#>   Possibly misspelled?

^{Created on 2024-12-12 with reprex v2.1.1}

datawizard '0.13.0'

packageVersion("datawizard")
#> [1] '0.13.0'

result <- list(
  setosa.a = structure(data.frame(1)), 
  versicolor.a = structure(data.frame(2)))

new_names <- c("setosa, a", "versicolor, a")

names(result)
#> [1] "setosa.a"     "versicolor.a"
new_names
#> [1] "setosa, a"     "versicolor, a"

datawizard::data_rename(result, select = names(result), replacement = new_names)
#> $`setosa, a`
#>   X1
#> 1  1
#> 
#> $`versicolor, a`
#>   X2
#> 1  2

^{Created on 2024-12-12 with reprex v2.1.1}

The reason seems to be that data_rename() searches for all the variables mentioned in select for each data frame of the list, but each name only appears once in their respective data frame. Therefore, some variables are bound to not be found. So, whereas data_rename() used to work on lists, now it only works on dataframes? This seems to break report::report_sample tests.

[strengejacke]

I'm not sure it was really intended to work on lists, so the revision of data_rename() seems to break this feature now. But we could add it back?

[strengejacke]

Note that you cannot use the safe argument in future versions and have to check if names (in select) really exist.

[etiennebacher]

I'm not sure it was really intended to work on lists, so the revision of data_rename() seems to break this feature now. But we could add it back?

I don't think we should support lists, other data_ functions also don't:

library(datawizard)

d <- data.frame(a = 1:3, b = 4:6)
l <- list(a = 1:3, b = 4:6)

data_select(d, "a")
#>   a
#> 1 1
#> 2 2
#> 3 3
data_select(l, "a")
#> Warning: No column names that matched the required search pattern were found.
#> NULL

It would also be confusing: should data_rename() modify the names of the list or the names of each element in the list?

[strengejacke]

I think so, too. I don't think this was actually intentional. You could use setNames() as a workaround in this particular situation, @rempsyc

[etiennebacher]

Maybe we should change the description of the data argument since it says:

A data frame, or an object that can be coerced to a data frame.

A list is coercible to a dataframe but we don't support it.

[strengejacke]

Yes, thought about that, too. I think this mostly makes sense for data frames. The alternative would be to revise .select_nse(), to it also works with list. Haven't looked into the details whether this can be done easily.

[rempsyc]

Yes, agreed, it would make sense for consistency that it works only on data frames. And, it was only used that way once in report, so I was able to switch it to setNames() like Daniel suggested. Perhaps we should simply have a more informative error on lists, or, as you mention, not attempt to coerce.