Move predicates into own package

[hadley]

Since they're useful in many places

@gaborcsardi do you already have a mini package for predicates? In purrr, we have:

is_atomic <- function(x) {
  typeof(x) %in% c("logical", "integer", "double", "complex", "character", "raw")
}
is_vector <- function(x) {
  is_atomic(x) || is.list(x)
}
is_bare_list <- function(x) {
  !is.object(x) && is_list(x)
}

etc

[gaborcsardi]

Micropackage. :) Not yet. But happy to have one. :)

We just want them to return TRUE/FALSE? I often find it useful if they can also return an error message.

[smbache]

+1

[smbache]

I think true / false is best (more generally applicable). But perhaps as attribute.

[gaborcsardi]

Btw. there are already a lot of assertions in the assertive.* packages, e.g. http://www.r-pkg.org/pkg/assertive.numbers

[jennybc]

@gaborcsardi: following that link made me realize that 14 of the 16 assertive.* packages don't seem to be in the METACRAN mirror?

https://github.com/search?q=user%3Acran+assertive&type=Repositories

[lionel-]

Instead of a micro-package, do we need a "standard lib" package with underscored function names and sane behaviours? This would be a good home for set_names() (#119) etc. Could be called lowliner ;)

[smbache]

Or if only for predicates, "predrcats" ;-)

[smbache]

With a cat hex logo

[gaborcsardi]

@jennybc yeah, the GitHub mirroring is not working now, I'll fix it soon.

[hadley]

I do like predrcats!

[imanuelcostigan]

Is there a reason why is_na is not available ? I am happy to help transport these predicates to a separate package if help is needed.

[smbache]

@imanuelcostigan is_na %>% is_na

[imanuelcostigan]

@smbache ?

[smbache]

Just found it amusing that you point out that is_na is not available (i.e. is_na)... :-)

[imanuelcostigan]

ok LOLz

[imanuelcostigan]

@lionel- i have been also been thinking that something like that would be sweet.

Been reading Swift's API design guidelines (ugh...camelCase) and reflecting on the lack of design principles in R's core packages (not just camelCase vs snake_case). That isn't surprising given its genesis.

[imanuelcostigan]

Starting predrcats package

[hadley]

Will be moving into rlang