Using 1-based indices in grouping metadata

[romainfrancois]

> tibble( x = 1:4, g = rep(1:2, each = 2)) %>% group_by(g) %>% group_rows()
[[1]]
[1] 1 2

[[2]]
[1] 3 4

[hadley]

Is this going to break tidyr? If so, we need to think a bit about release planning.

[romainfrancois]

I'll have a look at tidyr, how much does it use the internal structure of grouped data frames ?

[romainfrancois]

This is a dangerous off by one kind of PR, and it was a bit too easy so I need to carefully check if there are no other off by one problems lurking around.

[romainfrancois]

tidyr seems immune to these changes:

OK:       323
Failed:   0
Warnings: 0
Skipped:  0

Between #3489 and this, I guess we can do a rev dev check soon, there are probably 📦 that depend on the previous structure of attributes.

[krlmlr]

I could relate segfaults to this pull requests, specifically to c5fc0d2. Will try to come up with a reprex.

[krlmlr]

These segfaults are gone now with 9584596.

[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/