## context("Pairmatch function old")

## test_that("pairmatch", {
##   skip("This fails during R CMD checks, so silencing")
##   ## we could use skip_on_cran() instead but i don't want spurious failures during the local machine check process
##   data(plantdist)

##   stripCall <- function(obj) {
##     attr(obj, "call") <- NULL
##     obj
##   }


##   expect_warning(p1 <- pairmatch(plantdist))
##   expect_warning(f1 <- fullmatch(plantdist, max.controls = 1, min.controls = 1, omit.fraction = 1 - 7/19))
##   expect_equal(stripCall(f1), stripCall(p1))

##   expect_warning(p2 <- pairmatch(plantdist, controls=2))
##   expect_warning(f2 <- fullmatch(plantdist, max.controls = 2, min.controls = 2, omit.fraction = 1 - 14/19))
##   expect_equal(stripCall(f2), stripCall(p2))

##   # plantdist has several 0's, so it won't fail everywhere.
## #  expect_error(pairmatch(plantdist + caliper(plantdist, 1))) # Matching fails everywhere

##   expect_warning(p3 <- pairmatch(plantdist + caliper(plantdist, 5, compare = `<`),
##                                  remove.unmatchables=TRUE)) # Matching works after removing plant 'F'
##   expect_warning(f3 <- fullmatch(plantdist + caliper(plantdist, 5, compare = `<`),
##                                  max.controls = 1, min.controls = 0, omit.fraction = 1 - 6/19))

##   expect_true(all.equal(f3,p3, check.attributes=FALSE))

##   data(nuclearplants)
##   # in both of match_on calls below use sd to maintain backwards compatibility with
##   # pscore.dist, which used sd by default. match_on has used mad as the std. scale
##   # since it was added to the package, so the use of match_on should be consistent
##   # for users going forward.
##   psm <- glm(pr~.-(pr+cost), family=binomial(), data=nuclearplants)
##   psd <- match_on(psm, standardization.scale = sd)
##   pm <- pairmatch(psd, controls=2, data = nuclearplants)

##   # the pm match immediately above was giving slightly different answers in some environment
##   # the problem allowed multiple optimal solutions, and different choices were picked in different environments
##   # the sum of matched distances should be the same across all environments

##   # expect_true(all.equal(summary(pm)$total.distance, 25.83338, tolerance=1e-5))

##   ## We no longer throw an error when a subclass fails
##   # again an error would be thrown (which R CMD CHECK does not like)
##   #expect_error(pairmatch(caliper(match_on(psm, standardization.scale = sd,
##   #   within = exactMatch(pr ~ pt, data =
##   #   nuclearplants)), width=2))) # Fails in subclass '1'
## })