Warn with, or implement, `drop = TRUE` ?

[krlmlr]

Should this give a warning?

x <- tibble::tibble(a = 1:3, b = 2:4)

x[1, , drop = TRUE]
#> # A tibble: 1 × 2
#>       a     b
#>   <int> <int>
#> 1     1     2

^{Created on 2024-03-12 with reprex v2.1.0}

[krlmlr]

Or perhaps just work?

[krlmlr]

There's a case to be made that this should just work:

x <- tibble::tibble(a = 1:3, b = 2:4)

x[, 1, drop = TRUE]
#> [1] 1 2 3

x <- data.frame(a = 1:3, b = 2:4)

x[1, , drop = TRUE]
#> $a
#> [1] 1
#> 
#> $b
#> [1] 2

^{Created on 2025-03-19 with reprex v2.1.1}

[moodymudskipper]

The drop argument is known to create a lot of problems with data frames. In tibble:::[.tbl_dfit is purposefully set toFALSEto avoid those. Howeverdropbehaves differently intibble()anddata.frame()` when the output has a single row.

Currently drop = TRUE in tibble:::[.tbl_dfis equivalent to the default in[.data.frame, while drop = TRUEin[.data.frame` simplifies rows too.

We wish to clarify the situation, by either warning, or enforcing consistency, or maybe well documented new values (NA, "data.frame" ?) to tweak the behavior.

[etiennebacher]

By the way, this passage in the docs seems incorrect:

First the sentence doesn't really match the example, and secondly I understand this as

drop = TRUE returns a list if i corresponds to a single row

which is not what the example shows.

[MichaelChirico]

I think it would help in the description to be a bit more explicit about what's unexpected/desired here, e.g. this example:

There's a case to be made that this should just work:

is using different code for the data.frame vs. tibble examples (in data.frame, i=1, j=<missing>, in tibble, i=<missing>, j=1).

TIL about drop=TRUE behavior on 1-row output & length(j)>1 btw, quite confusing. For others newly initiated:

DF <- data.frame(a = 1:3, b = 4:6)
# "drops" to list
DF[1, , drop=TRUE]
# $a
# [1] 1

# $b
# [1] 4

# drop has no effect
DF[1:2, , drop=TRUE]
#   a b
# 1 1 4
# 2 2 5

# looks like i becomes j?
DF[1, drop=TRUE]
#   a
# 1 1
# 2 2
# 3 3
# Warning message:
# In `[.data.frame`(DF, 1, drop = TRUE) : 'drop' argument will be ignored