Implement expect_unique_combine

[kinto-b]

Currently this is sitting as a TODO in script. Will remove from the script. Need to implement at some point

testdat/R/expect-value.R

Lines 168 to 193 in 9f3ef72

	# TODO - check if any values exist more than once across multiple variables in
	# the entire dataset
	# #' @export
	# #' @rdname value-expectations
	expect_unique_combine <- function(vars, flt = TRUE, data = get_testdata()) {
	browser()
	act <- quasi_label(enquo(data))
	act$var_desc <- as_label_vars(enquo(vars))
	act$flt_desc <- as_label_flt(enquo(flt))
	flt <- enquo(flt)
	act$result <- data %>%
	filter(!!flt) %>%
	select(!!!vars)
	expect_custom(
	all(act$result, na.rm = TRUE),
	glue("{act$lab} has {sum(!act$result, na.rm = TRUE)} records with \\
	duplicates across variables `{act$var_desc}`.
	Filter: {act$flt_desc}"),
	failed_count = sum(!act$result, na.rm = TRUE),
	total_count = sum(!is.na(act$result))
	)
	invisible(act$result)
	}

[kinto-b]

What's the envisioned use-case for this function?

[gorcha]

This is for e.g. checking that no phone number appears more than once in the dataset.

There are 3 kinds of uniqueness check:

• Check that each record is unique for a given set of variables (expect_unique())
• Check that a set of variables each have unique values for each individual record (expect_unique_across())
• Check that a set of variables have unique values across the entire dataset (expect_unique_combined())

A little confusing, but naming things is hard 😛

[kinto-b]

Nah, I think those are good names. I just couldn't think of an instance where you'd want to check that any(duplicated(c(dat$x, dat$y, dat$z))) is false, but the phone numbers check makes sense to me : )