tests/testthat/test-zoo.R

context("zoo")

test_that("we behave similarly to zoo::rollapply", {

  if (!requireNamespace("zoo", quietly = TRUE))
    skip("zoo not installed")

  library(testthat)

  functions <- c("mean", "median", "prod", "min", "max", "sum")
  x <- rnorm(50)
  window <- 5L

  run_tests <- function(data, width, ..., functions, gctorture = FALSE) {
    for (f in functions) {
      RcppRoll <- get(paste("roll", f, sep = "_"), envir = asNamespace("RcppRoll"))
      zoo <- zoo::rollapply(data, width, FUN = get(f), ...)
      if (is.matrix(zoo)) {
        dimnames(zoo) <- NULL
      }
      if (gctorture) gctorture(TRUE)
      RcppRollRes <- RcppRoll(data, width, ...)
      if (gctorture) gctorture(FALSE)
      withCallingHandlers(
        expect_equal(RcppRollRes, zoo),
        error = function(cnd) {
          str(list(fn = f, data = data, width = width, ...))
        }
      )
    }
  }

  run_tests(x, window, functions = functions)

  window <- 50L
  run_tests(x, window, functions = functions)

  window <- 1L
  run_tests(x, window, functions = functions)

  ## test against small numbers
  x <- rnorm(1E3) ^ 100
  run_tests(x, 5L, functions = functions)

  ## and large numbers
  x <- rnorm(1E3, mean = 1E200, sd = 1E201)
  run_tests(x, 5L, functions = functions)

  ## now let's really stress it...
  args <- expand.grid(
    width = list(3L, 10L, 100L),
    fill = list(NA, c(-1, 0, 1)),
    align = list("left", "center", "right"),
    by = c(1L, 2L, 5L),
    na.rm = c(TRUE, FALSE)
  )

  # don't use median here
  data <- rnorm(1E2, 100, 50)
  for (i in 1:nrow(args)) {
    run_tests(data,
              args$width[[i]],
              fill = args$fill[[i]],
              align = args$align[[i]],
              na.rm = args$na.rm[[i]],
              by = args$by[[i]],
              functions = functions)
  }

  data[sample(length(data), length(data) / 3)] <- NA
  for (i in 1:nrow(args)) {
    suppressWarnings(run_tests(data,
                               args$width[[i]],
                               fill = args$fill[[i]],
                               align = args$align[[i]],
                               na.rm = args$na.rm[[i]],
                               by = args$by[[i]],
                               functions = functions))
  }

  data <- matrix(rnorm(2E2, 100, 50), nrow = 100)
  for (i in 1:nrow(args)) {
    run_tests(
      data, args$width[[i]],
      fill = args$fill[[i]],
      align = args$align[[i]],
      by = args$by[[i]],
      functions = functions
    )
  }

})

test_that("we don't segfault when window size > vector size on ops with fill", {

  x <- c(1:5)
  w <- 10
  gctorture(TRUE)
  result <- roll_meanr(x, w)
  gctorture(FALSE)
  expect_identical(
    roll_meanr(x, w),
    rep(NA_real_, length(x))
  )

})

test_that("we handle an empty fill properly", {

  if (!requireNamespace("zoo", quietly = TRUE))
    skip("zoo not installed")

  for (i in 10:100) {
    data <- 1:i
    lhs <- zoo::rollapply(data, 3, mean, by = 3)
    rhs <- roll_mean(data, 3, by = 3, fill = numeric())
    expect_identical(lhs, rhs)
  }
})

test_that("median handles NAs appropriately", {

  y <- c(NA, 1:3, NA)
  expect_equal(
    roll_median(y, n = 3L, na.rm = TRUE),
    c(1.5, 2.0, 2.5)
  )

})

Try the RcppRoll package in your browser

Any scripts or data that you put into this service are public.

RcppRoll documentation built on Sept. 11, 2024, 7:08 p.m.