If and only if checks

[kinto-b]

A relatively common formula is

expect_cond(foo, bar)
expect_cond(bar, foo)

This is a bit verbose, particularly so when foo and bar are complicated expressions. We should allow

expect_iff(foo, bar)

or something like

expect_cond(foo, bar, .cond="if")  # foo --> bar 
expect_cond(foo, bar, .cond="iff") # foo <-> bar

[kinto-b]

Simplest implementation, but perhaps not the best due to generating up to two failures

expect_iff <- function(cond1, cond2, ...) {
  cond1 <- rlang::enquo(cond1)
  cond2 <- rlang::enquo(cond2)
  
  testdat::expect_cond(!!cond1, !!cond2, ...)
  testdat::expect_cond(!!cond2, !!cond1, ...)
}

[kinto-b]

Some other logical relationships we might want to support more directly:

# Not both (often used for not both blank)
expect_cond(foo, !bar)
expect_cond(bar, !foo)

# XOR (mostly covered by `expect_exclusive()`)
expect_cond(foo, !bar)
expect_cond(bar, !foo)
expect_cond(TRUE, foo | bar)

[g-hyo]

Alternative implementation

If we want to avoid generating two failures, we could switch this proposition in expect_cond

testdat/R/expect-conditional.R

Line 50 in 02fca9b

act$result <- !(act$cond1 & !act$cond2)

to

act$result <- (act$cond1 & act$cond2) | (!act$cond1 & !act$cond2)

That being said, in case of failure we probably still want to feedback to the user the cases where cond1 & !cond2 and !cond1 & cond2 separately (A bit like how expect_base gives you cases that are incorrectly missing and cases that are incorrectly non-missing).

API

I think iff is different enough to expect_cond that it would make sense to have it be its own function instead of tacking it on. iff does not really have an antecedent or consequent, only two propositions, and the failure message would be a bit different.

Thoughts @kinto-b?

[kinto-b]

@g-hyo Hey! Yeah agree with everything you said and think modelling this on expect_base() is a very good idea. I guess the same is probably true for expect_xor() and maybe expect_notboth() (bad name but can't think of a better one :/)