Implement as_data_frame.tbl() and as.data.frame.tbl()

[krlmlr]

• forwarded to collect() + consider deprecating collect()
• to throw an error asking about explicit collect()

Other data.frame-like accessors like [ and [[ should perhaps throw an error, too.

This can't be in the tibble package, because the collect() method isn't defined there.

[hadley]

I much prefer collect() (even though it does the same thing), because it makes it more clear why you are calling it (in the same way it's useful to have both identity() and force() even though they do the same thing)

[krlmlr]

Okay to forward as_data_frame() to collect() then?

[hadley]

I think so

[krlmlr]

• Default to n=Inf? -> Depends on if interrupt checking is done properly, or if we have a progress bar
• Consider DBI::dbFetchChunks()

[hadley]

Want to do a PR for this?

[krlmlr]

as_data_frame.tbl() will be forwarded to as.data.frame.tbl() by tidyverse/tibble#71, I think -- this should be explicitly tested in dplyr, which I'll do once the tests pass again on Travis.

How about moving the logic for chunked fetches to a separate package/module, e.g. https://github.com/krlmlr/pumpr?