Named append

[hadley]

It would be useful to have some way to add a single column to a specific position in a data frame.

Here's a basic implementation that does what I want, but could clearly be optimised.

insert_col <- function(data, x, after, name = deparse(substitute(x))) {
  as.data.frame(append(data, setNames(list(x), name), after))
}

mtcars %>% insert_col(1:32, 0, "id")

The example gives something I want to do fairly commonly: insert an id column at the front of a data frame. The name of the id column is often given by a variable, so it's useful to have an interface that doesn't use names.

It might also be useful to have insert_cols() which worked like append, but preserved the output class. That would be useful if you had multiple columns you wanted to insert.

[jennybc]

If something like this could be used from dplyr::mutate(), that would really make my day. It is common to have a pipeline that ends with select() just to hitch newly created variables up into position.

[jennybc]

This is very related to the request in #85, but over there it's described as "shelved"? I also need to add a variable to the front of a tibble, which is kind of awkward right now.

[krlmlr]

We have add_row(), so add_column() is a logical extension, and much simpler than a generic cbind() (#85). I'd support both before and after arguments (as name or index, one of them takes precedence).

[hadley]

@krlmlr I like that idea!

[jennybc]

Will the before/after flexibility propagate back to add_row()? That also comes up. Also would make the lack of rbind() slightly less painful (#34).