Add tests for grouped history saving

inst/include/dust2/filter.hpp

@@ -45,6 +45,9 @@ class unfilter {
 
   void run(bool set_initial, bool save_history) {
     history_is_current_ = false;
+    if (save_history) {
+      history_.reset();
+    }
     const auto n_times = step_.size();
 
     model.set_time(time_start_);
@@ -145,6 +148,10 @@ class filter {
   }
 
   void run(bool set_initial, bool save_history) {
+    history_is_current_ = false;
+    if (save_history) {
+      history_.reset();
+    }
     const auto n_times = step_.size();
 
     model.set_time(time_start_);

tests/testthat/test-filter.R

@@ -205,6 +205,40 @@ test_that("can run replicated structured unfilter", {
 })
 
 
+test_that("can save history from structured unfilter", {
+  pars <- list(
+    list(beta = 0.1, gamma = 0.2, N = 1000, I0 = 10, exp_noise = 1e6),
+    list(beta = 0.2, gamma = 0.2, N = 1000, I0 = 10, exp_noise = 1e6))
+  time_start <- 0
+  time <- c(4, 8, 12, 16)
+  data <- lapply(1:4, function(i) {
+    lapply(seq_len(2), function(j) list(incidence = 2 * (i - 1) + j))
+  })
+  data1 <- lapply(data, function(x) x[[1]])
+  data2 <- lapply(data, function(x) x[[2]])
+  dt <- 1
+  obj <- dust_unfilter_create(sir(), pars, time_start, time, data,
+                              n_groups = 2)
+  obj1 <- dust_unfilter_create(sir(), pars[[1]], time_start, time, data1)
+  obj2 <- dust_unfilter_create(sir(), pars[[2]], time_start, time, data2)
+
+  ll <- dust_unfilter_run(obj, save_history = TRUE)
+  ll1 <- dust_unfilter_run(obj1, save_history = TRUE)
+  ll2 <- dust_unfilter_run(obj2, save_history = TRUE)
+
+  expect_equal(ll, c(ll1, ll2))
+  h <- dust_unfilter_last_history(obj)
+  h1 <- dust_unfilter_last_history(obj1)
+  h2 <- dust_unfilter_last_history(obj2)
+
+  expect_equal(dim(h), c(5, 1, 2, 4))
+  expect_equal(dim(h1), c(5, 1, 4))
+  expect_equal(dim(h2), c(5, 1, 4))
+  expect_equal(array(h[, , 1, ], dim(h1)), h1)
+  expect_equal(array(h[, , 2, ], dim(h2)), h2)
+})
+
+
 test_that("can run particle filter", {
   pars <- list(beta = 0.1, gamma = 0.2, N = 1000, I0 = 10, exp_noise = 1e6)
 
@@ -309,14 +343,14 @@ test_that("can run a nested particle filter and get the same result", {
   s <- dust_filter_rng_state(obj)
   expect_equal(s, r)
 
-  res <- replicate(20, dust_filter_run(obj))
+  res <- replicate(20, dust_filter_run(obj, save_history = TRUE))
 
   ## now compare:
   data1 <- lapply(data, "[[", 1)
   obj1 <- dust_filter_create(sir(), pars[[1]], time_start, time, data1,
                              n_particles = n_particles, seed = seed)
   s1 <- dust_filter_rng_state(obj1)
-  res1 <- replicate(20, dust_filter_run(obj1))
+  res1 <- replicate(20, dust_filter_run(obj1, save_history = TRUE))
   expect_equal(res1, res[1, ])
   expect_equal(s1, s[1:3232])
 
@@ -325,9 +359,14 @@ test_that("can run a nested particle filter and get the same result", {
   obj2 <- dust_filter_create(sir(), pars[[2]], time_start, time, data2,
                              n_particles = n_particles, seed = seed2)
   s2 <- dust_filter_rng_state(obj2)
-  res2 <- replicate(20, dust_filter_run(obj2))
+  res2 <- replicate(20, dust_filter_run(obj2, save_history = TRUE))
   expect_equal(res2, res[2, ])
   expect_equal(s2, s[3233:6464])
+
+  h <- dust_filter_last_history(obj)
+  expect_equal(dim(h), c(5, 100, 2, 4))
+  expect_equal(h[, , 1, ], dust_filter_last_history(obj1))
+  expect_equal(h[, , 2, ], dust_filter_last_history(obj2))
 })