Better checking of list arguments to bind_rows()

[krlmlr]

Fixes #3024.

Related: humanpred/pknca#48 (comment)

[lionel-]

Argument 1 must have names

I think it should be: Argument 1 must be named

[lionel-]

i.e. lists are treated as data frames so they should have outer names and the inner names of vectors are irrelevant.

[krlmlr]

To me, in dplyr::bind_rows(a = df), the df is a named argument (because of the name a) but has names itself (the names of the columns. Can you please clarify which call should give which error message?

[lionel-]

If there are unused outer names it should probably be a warning.

[lionel-]

oh I understand now: "Argument 1 must have names". Argument 1 refers to the list, got it.

[lionel-]

If there are unused outer names it should probably be a warning.

Though the outer names can be used as .id column right? So we should allow unused outer names.

[krlmlr]

I'm not sure, this PR is just about fixing an annoyance just before the upcoming dplyr bugfix release, which may or may not be connected with a segfault only CRAN is seeing. Can we discuss wording of error messages after the release?

[hadley]

I think this function would be a bit easier to understand if the first check was to confirm that the input was a vector. But we can do that refactoring later, if needed.

[krlmlr]

Simplified the check.

[lock]

This old issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with reprex) and link to this issue. https://reprex.tidyverse.org/