mutating, arranging, summarizing empty data frames fails with "not compatible with STRSXP"

[richierocks]

To reproduce:

data.frame() %>% mutate()
## Error: not compatible with STRSXP

data.frame() %>% arrange()
## Error: not compatible with STRSXP

data.frame() %>% summarize()
## Error: not compatible with STRSXP

The _ versions also fail.

select, filter, group_by work OK.

Related: [ fails inside mutate

[romainfrancois]

I fixed mutate for the example above. Not sure though how to deal with this:

data.frame() %>% mutate( x = 2L )

[hadley]

@romainfrancois shouldn't that be an error because the output is a different length to the input?

[hadley]

Another good test case would be data.frame() %>% mutate(x = numeric())

[romainfrancois]

Now I get:

> data.frame() %>% mutate(x = 2L)
[1] x
<0 rows> (or 0-length row.names)

Because a length 1 result in mutate is recycled, so here recycled 0 times, but I can make this special case an error if that makes more sense.

[hadley]

Ah, I forgot about that principle. I think either way is probably ok, but recycling to length 0 is a little less frustrating for people.

[romainfrancois]

In that case I think we can close this.

[abresler]

I think this issue has returned with some of the new updates. I have empty data_frames that I am trying to bind and it now returns the dreaded

not compatible with STRSXP 

[hadley]

@abresler could you please file a new issue with a minimal reprex