as_data_frame.NULL

[jennybc]

So that dplyr::tbl_df(NULL) (or tibble::tbl_df(NULL) rather) is NULL instead of an error.

If I understand what's going on, it makes more sense to do this here than in dplyr itself (?).

[codecov-io]

Current coverage is 96.94%

Merging #16 into master will not affect coverage as of b910386

@@            master     #16   diff @@
======================================
  Files           11      11       
  Stmts          393     393       
  Branches         0       0       
  Methods          0       0       
======================================
  Hit            381     381       
  Partial          0       0       
  Missed          12      12       

Review entire Coverage Diff as of b910386

Powered by Codecov. Updated on successful CI builds.

[krlmlr]

Thanks! Eventually, dplyr will use this package (tidyverse/dplyr#1595), but tibble probably needs to be on CRAN first.

[hadley]

I don't think this is quite right - I think it should continue to return a data frame, but it should be a 0-row 0-col data frame, i.e. dplyr::data_frame()

[krlmlr]

Agreed, that's also what as.data.frame() does:

> as.data.frame(NULL)
data frame with 0 columns and 0 rows

@jennybc: Would you like to update the code?

[jennybc]

OK I will change that. I looked for something to emulate and could only find a NULL method for do_(), but I see that it's different.