extract fails on empty data.frame

[antoine-sachet]

Reproducible example using tidyr 0.6.3.9000:

df <- data.frame(col1=character(0))
tidyr::extract(df, col1, into="col2", regex="(whatever)")

Expected behaviour: should return an empty data.frame with a new empty column "col2".

Actual behaviour: fails with error:
Error in names(l) <- enc2utf8(into) : 'names' attribute [1] must be the same length as the vector [0]

Why it should be fixed: this makes an otherwise valid pipeline fail when the result of a filter happens to be empty. This requires boiler-plate to guard against empty data when programming with extract.

[olsgaard]

I am having the same problem. reproducable example using reprex

df <- data.frame(col1 = c("q123", "abc"))
tidyr::extract(df, col1, into = "col2", regex = "[0-9]+", remove = FALSE)
#> Error in names(l) <- enc2utf8(into): 'names' attribute [1] must be the same length as the vector [0]

When I choose "rerun with Debug", the error is in extract_.data.frame, where l is a List of 0

---

I am also running tidyr 0.6.3, in a corporate setting, so updating is non-trivial. Does 0.7.1 fix the issue or are we interfacing with tidyr in the wrong way?

[hadley]

More minimal reprex:

df <- data.frame(x = c("q123", "abc"))
extract(df, x, "y", regex = ".")

I think the problem is that there's no grouping defined in the regular expression and the error message doesn't help you discover that.

[hadley]

Ok, there were actually two problems here. It looks like the motivating issue is actually a small stringi bug; so I'll file over there.