Does this mean copying the bind_rows() implementation from dplyr? This seems to need at least Collecter, SlicingIndex and set_rownames().

Yeah, I'm not sure how to deal with that. One option would be to add a dplyr dependency to tibble (just a suggest), but that seems suboptimal.

I think I'll add rbind.tbl_df() and make sure it behaves like the dplyr equivalent.

I think we can leave this for now.

Can this be reconsidered/reopened? Part of the justification for tibble is so you don't need all of dplyr just to have the data frame goodness. As such tibble should provide a suite of tools for working with tibbles or tbl_dfs.

An rbind.tbl_df() would be a most useful addition to this package

Not in near future - there is a lot of code powering bind rows that we'd need to duplicate. All the variable coercion stuff is complicated.

Is it being considered for the longer term future? Closing this issue gave me the impression this wasn't something you were going to deal with.

I have no doubt this is complicated; if it were easy I suspect it would have been done by now.

In the long long term I'd like to move data frame stuff related here but it's probably at least a year away

I think we could do just what data.table does -- they're patching rbind.data.frame(): tidyverse/dplyr#606 (comment).

Did we agree to not patch fix rbind() and cbind()?

I'd prefer not to mess with base R functions

This old thread has been automatically locked. If you think you have found something related to this, please open a new issue and link to this old issue if necessary.

Let's retry, also with cbind(), now that we have vctrs.

👍

Things appear to have changed for R 4.0.0, the relevant FAQ in data.table was updated. Are we ok with a solution that works only in R 4.0.0 now, should we postpone until the tidyverse requires R 4.0.0?

I think it's fine to add a solution that only works in 4.0.0 and up. That will help the majority of users.

Done. I'll release to CRAN without it, we might want to do revdepchecks because users might rely on downcasting to data frames.

Do we still want this, given the complexities discussed in #909?

@DavisVaughan suggests to

link to vec_rbind() in the tibble::add_row() documentation. i.e. "If you would like to row bind multiple data frames together, see..."

Similar for vec_cbind() and tibble::add_column().