Fix for dev purrr

[hadley]

Since vec_depth() now does work with captured error objects, because it's now better aligned with what you can pluck().

[Robsteranium]

Thanks for this.

I had actually intended that this test remind me to check whether it was still necessary to replace purrr::vec_depth. I hadn't expected the reminder to work the other way around!

I'll take a look at purrr::pluck_depth and see if that will work better for the recursive csvwr::rmap and csvwr::rlmap variants. It looks like we'll descend into error (as lists) as it stands so I suspect we may want to put a guard to prevent that.

I'd be happy to merge this if it makes reverse-depends tracking/ cran-submission for purrr easier but otherwise I'll leave this open as a reminder to myself!

The above failure on r-devel is unrelated and fixed on the master branch.

[hadley]

FYI purrr is now scheduled for release to CRAN on Dec 19

[hadley]

FYI purrr is now scheduled for release to CRAN on Dec 19

[Robsteranium]

Indeed purrr::pluck_depth has a is_node argument that makes it trivial to add the guard clause. I'll merge this and swap csvw::vec_depth for purrr::pluck_depth on a follow-up branch.

Thanks again.