test: add failing test for large dataset performance (#420)

tests/testthat/test-performance_large_datasets.R

@@ -0,0 +1,113 @@
+# ==============================================================================
+# ISSUE #420: check_model() performance degradation with large datasets
+# ==============================================================================
+#
+# This test documents the performance issue where check_model() becomes
+# unusably slow (5+ minutes) with datasets >10K observations.
+#
+# The test verifies that plotting completes in reasonable time (<30 seconds)
+# for large datasets.
+#
+# See: https://github.com/easystats/see/issues/420
+# ==============================================================================
+
+test_that("check_model() handles large datasets efficiently", {
+  skip_if_not_installed("performance")
+  skip_if_not_installed("lme4")
+  skip_on_cran()  # Large dataset test
+
+  # Create large dataset (25,000 observations)
+  set.seed(123)
+  large_data <- data.frame(
+    subject = rep(1:500, each = 50),
+    x = rnorm(25000),
+    y = rnorm(25000)
+  )
+
+  # Fit model
+  model <- lme4::lmer(y ~ x + (1 | subject), data = large_data)
+
+  # check_model() should complete in reasonable time (<30 seconds)
+  # Currently this FAILS (takes 5+ minutes)
+  timing <- system.time({
+    result <- performance::check_model(model)
+    # Force plot rendering to measure true performance
+    plot_result <- plot(result)
+  })
+
+  # Test passes if under 30 seconds
+  expect_true(timing["elapsed"] < 30,
+    info = sprintf("check_model() took %.1f seconds (should be <30)", timing["elapsed"])
+  )
+
+  # Result should be valid
+  expect_s3_class(result, "check_model")
+})
+
+test_that("data sampling preserves visual fidelity", {
+  skip_if_not_installed("performance")
+  skip_if_not_installed("lme4")
+
+  # Small dataset should not be sampled
+  small_data <- data.frame(
+    x = rnorm(100),
+    y = rnorm(100)
+  )
+  model_small <- lm(y ~ x, data = small_data)
+
+  # Should complete quickly regardless
+  expect_no_error({
+    result <- performance::check_model(model_small)
+    plot(result)
+  })
+})
+
+test_that("large dataset plot components use sampling", {
+  skip_if_not_installed("performance")
+  skip_on_cran()
+
+  # Create dataset with 15,000 observations (above threshold)
+  set.seed(456)
+  large_data <- data.frame(
+    x = rnorm(15000),
+    y = rnorm(15000) + 0.5 * rnorm(15000)
+  )
+
+  model <- lm(y ~ x, data = large_data)
+
+  # Check that plotting completes quickly
+  timing <- system.time({
+    result <- performance::check_model(model)
+    plot_result <- plot(result)
+  })
+
+  # Should be much faster with sampling
+  expect_true(timing["elapsed"] < 15,
+    info = sprintf("Plotting took %.1f seconds (should be <15 with sampling)", timing["elapsed"])
+  )
+})
+
+test_that("medium datasets complete in reasonable time", {
+  skip_if_not_installed("performance")
+  skip_on_cran()
+
+  # Medium dataset (8,000 observations - below threshold)
+  set.seed(789)
+  medium_data <- data.frame(
+    x = rnorm(8000),
+    y = rnorm(8000) + rnorm(8000)
+  )
+
+  model <- lm(y ~ x, data = medium_data)
+
+  # Should complete in reasonable time even without sampling
+  timing <- system.time({
+    result <- performance::check_model(model)
+    plot_result <- plot(result)
+  })
+
+  # Medium datasets should still be reasonably fast
+  expect_true(timing["elapsed"] < 20,
+    info = sprintf("Medium dataset plotting took %.1f seconds", timing["elapsed"])
+  )
+})