as.data.frame.tbl_df() ignores row.names argument

[shivam7898]

Help says that for as.data.frame(), x can be any R object and row.names can be a character vector giving the row names for the data frame.

Although tibble has a valid opinion on rownames, the output is a data.frame. So, I believe that the as.data.frame.tbl_df() method should not ignore row.names argument and set the attribute like as.data.frame.data.frame() does.

Obviously, I am not the expert so I am willing to be corrected. Thank You :)

library(tibble)

# as.data.frame.tbl_df() ignores 'row.names'
as.data.frame(tibble(x = 1:2), row.names = letters[1:2]) 
#  x
#1 1
#2 2

# as.data.frame.data.frame() does not ignore 'row.names'
as.data.frame(data.frame(x = 1:2), row.names = letters[1:2])
#   x
# a 1
# b 2

sessionInfo()

sessionInfo()
# R version 4.2.1 (2022-06-23 ucrt)
# Platform: x86_64-w64-mingw32/x64 (64-bit)
# Running under: Windows 10 x64 (build 19044)
# 
# Matrix products: default
# 
# locale:
# [1] LC_COLLATE=English_India.utf8  LC_CTYPE=English_India.utf8   
# [3] LC_MONETARY=English_India.utf8 LC_NUMERIC=C                  
# [5] LC_TIME=English_India.utf8    
# 
# attached base packages:
# [1] stats     graphics  grDevices datasets  utils     methods   base     
# 
# other attached packages:
# [1] tibble_3.1.7
# 
# loaded via a namespace (and not attached):
#  [1] compiler_4.2.1  magrittr_2.0.3  ellipsis_0.3.2  cli_3.3.0      
#  [5] pillar_1.7.0    glue_1.6.2      crayon_1.5.1    utf8_1.2.2     
#  [9] fansi_1.0.3     vctrs_0.4.1     lifecycle_1.0.1 pkgconfig_2.0.3
# [13] rlang_1.0.3 

[krlmlr]

Thanks. This has been like this from day one, but looks like something we could consider. On the other hand, there's column_to_rownames() that achieves a variant of this today.

@hadley @DavisVaughan: What are your thoughts on supporting as.data.frame.tbl_df(x, row.names = ...) for assigning row names to the resulting data frame?

[hadley]

If it's simple to implement and doesn't affect much downstream code, it seems like it's worth adhering to the as.data.frame API contract (such as it is).

[krlmlr]

Reverted because it's causing downstream failures. Will address in a separate PR after CRAN release.

[hadley]

If it's going to cause revdep failures, that clearly pushes it in to the "not worth it" category, IMO.

[BartJanvanRossum]

I just ran into this as well. Even though I agree that it might not be worth changing, please at least consider documenting this. Now the function arguments suggest functionality that is not implemented which is confusing for me as a user.

[moodymudskipper]

tibbles drop row names by design (we can set some but they're dropped when subsetting rows), however when converting to a native data frame we might want to set those.

as.data.frame.tbl_df() features a row.names arg for compatibility with the generic, but it isn't used.

There were attempts to make it work consistently with the native as.data.frame() but these triggered reverse dependency failures so the idea was shelved. However the behavior is confusing and we wish to at least alleviate the confusion.

We can do so by improving the documentation, and warn when it is used, proposing an alternative using column_to_rownames().