Update chk_blank()

[gorcha]

Basically the same as the srcutils change for is_blank() (socialresearchcentre/srcutils#95) - using %in% is unnecessarily slow.

chk_blank <- function(x) {
  if (is.character(x)) {
    is.na(x) | x == ""
  } else {
    is.na(x)
  }
}

I noticed the final is_blank() had a couple of extra bits there, would be good to get your feedback on these @kinto-b:

• There is a stopifnot in the character block to check for residual NA values - I think we have a guarantee of no NA values already because of the way we're checking, so going to leave this out.
• is_blank() checks for "" in factors as well as character variables. I'm not sure how I feel about this for factors since they're conceptually quite different (i.e. I think it's reasonable to assume the user has appropriately checked their values), although this would be a breaking change.

[kinto-b]

Yeah, I don't think the stopifnot() clause is necessary. I'm not sure about factors though. I might put it to the team in an oblique way to see what they find most intuitive. I'll ask them to fill in the blanks and get back to you:

x <- c("cat", "dog", "", "cat", NA)
chk_blank(x)
#> FALSE FALSE TRUE FALSE TRUE

y <- factor(x)
chk_blank(y)
#> FALSE FALSE LGL1 FALSE LGL2

[gorcha]

I'm kind of torn - if factors are treated like character then a "" level on a factor will be treated as a blank. On the one hand this makes sense since the value is "", but on the other it's an assigned level so 🤷

I'm going to treat factors like characters for the moment since that's the current behaviour, but might change my mind at some point 😛

[kinto-b]

Yeah, there weren't any strong intuitions in the team