tidyverse ingest conventions about tidyverse

[hadley]

• It needs to both import tibble and import a function from tibble (otherwise doesn't actually trigger the import). Should always construct tibbles using tibble::as_tibble() (or similar) to ensure that it gets latest updates to spec.

• Something about locale includes: functions should be designed to return identical results when called on different systems. This means not taking deriving default values from the OS.

• If can parse 64 bit integers, should use bit64 package

[jennybc]

It needs to both import tibble and import a function from tibble (otherwise doesn't actually trigger the import).

Maybe tibble should have a function expressly for this purpose? And it could somehow also be used to summon "what is a tibble? go here for a list of need-to-know facts about how it differs from data.frame".

[jennybc]

Recommendation for standard flavour of NA used for missing data?

In base R, it's logical.

In readr, it's character.

In readxl, it's currently numeric. As of v1.0.0, it's logical.

[hadley]

I feel like logical is probably the most reasonable.

[jennybc]

Use a standard @return blurb to document a returned tibble, i.e. don't just say "a data frame".

A good start, for now: @return A [tibble][tibble::tibble-package].

Could also link directly (or indirectly via this tibble link) to a yet-to-be-created explainer in tidyverse.org or tibble's yet-to-be-created website.

This issue was moved by jennybc to tidyverse/design/issues/1.