inst/tests/test-projection.R

In drfinlayscott/FLRcppAdolc: Combine FLR, Rcpp and ADOLC

context("Test projection functions")

test_that("timestep_to_year_season_conversion",{
    flq_in <- random_FLQuant_generator()
    year <- round(runif(1,min=1,max=dim(flq_in)[2]))
    season <- round(runif(1,min=1,max=dim(flq_in)[4]))
    expect_that(test_year_season_to_timestep_FLQuant_double(flq_in, year, season), equals((year-1)*dim(flq_in)[4] + season))
    expect_that(test_year_season_to_timestep_FLQuant_adouble(flq_in, year, season), equals((year-1)*dim(flq_in)[4] + season))
    expect_that(test_year_season_to_timestep(flq_in, year, season), equals((year-1)*dim(flq_in)[4] + season))
    # And the other way
    timestep <- test_year_season_to_timestep(flq_in, year, season)
    expect_that(test_timestep_to_year_season_FLQuant_double(flq_in, timestep), equals(c(year,season)))
    expect_that(test_timestep_to_year_season_FLQuant_adouble(flq_in, timestep), equals(c(year,season)))
    expect_that(test_timestep_to_year_season(flq_in, timestep), equals(c(year,season)))
})

test_that("euclid_norm",{
    size_x <- runif(1, min=5, max = 100) 
    x <- rnorm(size_x)
    expect_that(test_euclid_norm(x), equals(sqrt(sum(x^2))))
})

test_that("Newton-Raphson tests",{
    # Simplest test - solution is 2 or -2/3
    max_iters <- 50
    max_limit <- 100
    tolerance <- 1e-12
    initial <- rnorm(1, sd=1)
    fit <- test_NR1(initial, max_iters, max_limit, tolerance)
    expect_that(((fit$x - 2) < tolerance) | ((fit$x - (-2/3)) < tolerance), is_true())
    expect_that(fit$out, is_identical_to(0L))
    # Check max iters
    fit <- test_NR1(-10, 2, max_limit, tolerance)
    expect_that(fit$out, is_identical_to(1L))
    # Check max limit
    fit <- test_NR1(200, max_iters, max_limit, tolerance)
    expect_that(fit$out, is_identical_to(2L))
    tolerance2 <- 1
    fit <- test_NR1(50, max_iters, max_limit, tolerance2)
    expect_that(((fit$x - 2) < tolerance2) | ((fit$x - (-2/3)) < tolerance2), is_true())
    # 2D test 
    max_iters <- 50
    max_limit <- 1e9 # Crank up the limit - it's OK for this example
    tolerance <- 1e-12
    initial <- abs(rnorm(2, sd=2))
    fit <- test_NR2(initial, max_iters, max_limit, tolerance)
    y1 <- fit$x[1]^2 + fit$x[2]^2 - 4
    y2 <- fit$x[1]^2 - fit$x[2] + 1
    expect_that(sqrt(sum(c(y1,y2)^2)) < tolerance, is_true())
    expect_that(fit$out, is_identical_to(0L))
})