# tests/testthat/test-nonlinear_models.R

# ---- fixtures -------------------------------------------------------------
make_logistic <- function(n = 60, seed = 1) {
  set.seed(seed)
  days <- seq(5, 100, length.out = n)
  bm   <- 500 / (1 + exp((45 - days) / 8)) + rnorm(n, 0, 10)
  data.frame(days = days, biomass_g = pmax(bm, 0))
}

make_asymptotic <- function(n = 50, seed = 2) {
  set.seed(seed)
  n_rate <- seq(0, 200, length.out = n)
  yield  <- 6 * (1 - exp(-0.015 * n_rate)) + rnorm(n, 0, 0.15)
  data.frame(nitrogen = n_rate, yield = pmax(yield, 0))
}

make_quadratic <- function(n = 40, seed = 3) {
  set.seed(seed)
  x <- seq(0, 200, length.out = n)
  y <- 2 + 0.04 * x - 0.0001 * x^2 + rnorm(n, 0, 0.1)
  data.frame(nitrogen = x, yield = y)
}

make_lp <- function(n = 60, seed = 4) {
  set.seed(seed)
  x  <- seq(0, 200, length.out = n)
  cp <- 120
  y  <- ifelse(x < cp, 2 + 0.025 * x, 2 + 0.025 * cp) + rnorm(n, 0, 0.05)
  data.frame(nitrogen = x, yield = y)
}

make_gompertz <- function(n = 60, seed = 5) {
  set.seed(seed)
  days <- seq(1, 100, length.out = n)
  bm   <- 500 * exp(-exp(-0.08 * (days - 40))) + rnorm(n, 0, 5)
  data.frame(days = days, biomass_g = pmax(bm, 0))
}

# ---- fit_nonlinear — basic ------------------------------------------------
test_that("fit_nonlinear logistic returns agriNLS", {
  dat <- make_logistic()
  m   <- fit_nonlinear(dat, "days", "biomass_g", "logistic")
  expect_s3_class(m, "agriNLS")
  expect_equal(m$model, "logistic")
})

test_that("fit_nonlinear asymptotic converges", {
  dat <- make_asymptotic()
  m   <- fit_nonlinear(dat, "nitrogen", "yield", "asymptotic")
  expect_s3_class(m, "agriNLS")
})

test_that("fit_nonlinear quadratic converges", {
  dat <- make_quadratic()
  m   <- fit_nonlinear(dat, "nitrogen", "yield", "quadratic")
  expect_s3_class(m, "agriNLS")
})

test_that("fit_nonlinear linear_plateau converges", {
  dat <- make_lp()
  m   <- fit_nonlinear(dat, "nitrogen", "yield", "linear_plateau",
                       start = list(a = 2, b = 0.025, cp = 120))
  expect_s3_class(m, "agriNLS")
})

test_that("fit_nonlinear gompertz converges", {
  dat <- make_gompertz()
  m   <- tryCatch(
    fit_nonlinear(dat, "days", "biomass_g", "gompertz",
                  start = list(Asym = 500, b2 = -1, b3 = 40)),
    error = function(e) NULL
  )
  skip_if(is.null(m), "Gompertz did not converge on this platform -- skipped")
  expect_s3_class(m, "agriNLS")
})

# ---- coef / fitted / residuals / predict ----------------------------------
test_that("coef.agriNLS returns named vector", {
  dat <- make_logistic()
  m   <- fit_nonlinear(dat, "days", "biomass_g", "logistic")
  cf  <- coef(m)
  expect_true(all(c("Asym", "xmid", "scal") %in% names(cf)))
})

test_that("fitted.agriNLS returns correct length", {
  dat <- make_logistic()
  m   <- fit_nonlinear(dat, "days", "biomass_g", "logistic")
  expect_length(fitted(m), nrow(dat))
})

test_that("residuals.agriNLS returns correct length", {
  dat <- make_logistic()
  m   <- fit_nonlinear(dat, "days", "biomass_g", "logistic")
  expect_length(residuals(m), nrow(dat))
})

test_that("predict.agriNLS works with new data", {
  dat <- make_logistic()
  m   <- fit_nonlinear(dat, "days", "biomass_g", "logistic")
  nd  <- data.frame(days = c(30, 60, 90))
  pr  <- predict(m, newdata = nd)
  expect_length(pr, 3)
})

# ---- print / summary ------------------------------------------------------
test_that("print.agriNLS produces agriReg header", {
  dat <- make_logistic()
  m   <- fit_nonlinear(dat, "days", "biomass_g", "logistic")
  expect_output(print(m), "agriReg")
})

test_that("summary.agriNLS shows goodness-of-fit", {
  dat <- make_logistic()
  m   <- fit_nonlinear(dat, "days", "biomass_g", "logistic")
  expect_output(summary(m), "Goodness-of-fit")
})

# ---- plot -----------------------------------------------------------------
test_that("plot.agriNLS returns ggplot", {
  skip_if_not_installed("ggplot2")
  dat <- make_logistic()
  m   <- fit_nonlinear(dat, "days", "biomass_g", "logistic")
  p   <- plot(m)
  expect_s3_class(p, "gg")
})

test_that("plot.agriNLS with show_residuals=TRUE returns ggplot", {
  skip_if_not_installed("ggplot2")
  dat <- make_logistic()
  m   <- fit_nonlinear(dat, "days", "biomass_g", "logistic")
  p   <- plot(m, show_residuals = TRUE)
  expect_s3_class(p, "gg")
})

# ---- optimum_dose ---------------------------------------------------------
test_that("optimum_dose quadratic returns named vector", {
  dat <- make_quadratic()
  m   <- fit_nonlinear(dat, "nitrogen", "yield", "quadratic")
  opt <- optimum_dose(m)
  expect_named(opt, c("x_opt", "y_max"))
  expect_true(opt["x_opt"] > 0)
})

test_that("optimum_dose linear_plateau returns cp as optimum", {
  dat <- make_lp()
  m   <- fit_nonlinear(dat, "nitrogen", "yield", "linear_plateau",
                       start = list(a = 2, b = 0.025, cp = 120))
  opt <- optimum_dose(m)
  expect_named(opt, c("x_opt", "y_max"))
  # cp should be near 120
  expect_lt(abs(opt["x_opt"] - 120), 20)
})

test_that("optimum_dose errors for unsupported model", {
  dat <- make_logistic()
  m   <- fit_nonlinear(dat, "days", "biomass_g", "logistic")
  expect_error(optimum_dose(m), "only supports")
})

# ---- custom start values --------------------------------------------------
test_that("fit_nonlinear accepts user start values", {
  dat   <- make_logistic()
  start <- list(Asym = 480, xmid = 44, scal = 8)
  m     <- fit_nonlinear(dat, "days", "biomass_g", "logistic", start = start)
  expect_s3_class(m, "agriNLS")
})