Merge 1afc7e35ca into 3890e12c9e

Replaced the previous implementation with a more efficient approach using a single loop and pre-allocated vectors. This change reduces the complexity of creating FSRSItems and enhances overall performance, especially for larger datasets.

rslib/src/scheduler/fsrs/params.rs

@@ -481,24 +481,20 @@ pub(crate) fn reviews_for_fsrs(
     let skip = if training { 1 } else { 0 };
     // Convert the remaining entries into separate FSRSItems, where each item
     // contains all reviews done until then.
-    let items: Vec<(RevlogId, FSRSItem)> = entries
+    let mut items = Vec::with_capacity(entries.len());
-        .iter()
+    let mut current_reviews = Vec::with_capacity(entries.len());
-        .enumerate()
+    for (idx, (entry, &delta_t)) in entries.iter().zip(delta_ts.iter()).enumerate() {
-        .skip(skip)
+        current_reviews.push(FSRSReview {
-        .map(|(outer_idx, entry)| {
+            rating: entry.button_chosen as u32,
-            let reviews = entries
+            delta_t,
-                .iter()
+        });
-                .take(outer_idx + 1)
+        if idx >= skip && (!training || current_reviews.last().unwrap().delta_t > 0) {
-                .enumerate()
+            let item = FSRSItem {
-                .map(|(inner_idx, r)| FSRSReview {
+                reviews: current_reviews.clone(),
-                    rating: r.button_chosen as u32,
+            };
-                    delta_t: delta_ts[inner_idx],
+            items.push((entry.id, item));
-                })
+        }
-                .collect();
+    }
-            (entry.id, FSRSItem { reviews })
-        })
-        .filter(|(_, item)| !training || item.reviews.last().unwrap().delta_t > 0)
-        .collect_vec();
     if items.is_empty() {
         None
     } else {