part of issue #446, expand supports list columns but complete does not

[SamanthaToet]

Added a test for type and size of result. complete doesn't work though because it relies on dplyr::left_join which doesn't support list columns currently, @romainfrancois 🦄

[romainfrancois]

I don't think we need the tibble:: prefix here. Also, I'd suggest to create demo only once with a pipe instead of updating it, something like :

demo <- tibble(formulas = formulas, data_sets = data_sets)) %>%
  expand(formulas, data_sets)

[romainfrancois]

I guess the problem you mention about complete() is related to: tidyverse/dplyr#2355 for which we'll need hashing support from vctrs.

This is planned for the 0.9.* series of dplyr.

[SamanthaToet]

Just pushed an update to Romain's comments. Not 100% if I totally fixed the extraneous new line but I used the styler plug in - happy to tweak further.

[hadley]

Thanks @SamanthaToet! I made a few small changes to focus the docs and test on what was actually changed: the implementation of crossing(). This is particularly important for the test — ideally you want to construct a test that exercises only the new lines of code and nothing else.