vec_proxy_missing() adds a new proxy that allows developers to extend the is.na(), na.omit() etc. behavior of their vctrs (as I describe in (#1925). It's useful/needed when vctrs contain special values (or data structures) that represent different types of missingness in their data. In these cases, you need a separate proxy for equal and missing. It also lets you redefine the meaning of complete.
I wanted to take an experimental shot at the implementation here because it didn't seem to hard, and I wanted to play around with the resulting API. I'm extremely satisfied with the result -- this is exactly what I need for implementing a generic vctrs-backed Result<> type in the library I'm working on, and I think could have similar applications elsewhere (like haven::labelled_spss, or declared)
I've added a few tests and documentation updates, but it's only a first-pass. I would be happy to do a deeper & more thorough dive if this is considered for inclusion into vctrs.
vec_proxy_missing()adds a new proxy that allows developers to extend theis.na(),na.omit()etc. behavior of their vctrs (as I describe in (#1925). It's useful/needed when vctrs contain special values (or data structures) that represent different types of missingness in their data. In these cases, you need a separate proxy forequalandmissing. It also lets you redefine the meaning ofcomplete.I wanted to take an experimental shot at the implementation here because it didn't seem to hard, and I wanted to play around with the resulting API. I'm extremely satisfied with the result -- this is exactly what I need for implementing a generic vctrs-backed Result<> type in the library I'm working on, and I think could have similar applications elsewhere (like
haven::labelled_spss, or declared)I've added a few tests and documentation updates, but it's only a first-pass. I would be happy to do a deeper & more thorough dive if this is considered for inclusion into vctrs.
Thank you for your consideration!