orderin entries by start time is better than id

src/database.rs

@@ -176,7 +176,7 @@ pub trait Database {
     }
 
     fn last_entry_of_sheet(&self, sheet: &str) -> Result<Option<Entry>> {
-        Ok(self.entry_query("select * from entries where sheet=?1 order by id desc limit 1", &[&sheet])?.into_iter().next())
+        Ok(self.entry_query("select * from entries where sheet=?1 order by start desc limit 1", &[&sheet])?.into_iter().next())
     }
 
     fn delete_entry_by_id(&mut self, id: u64) -> Result<()> {
@@ -324,6 +324,7 @@ impl Database for SqliteDatabase {
 #[cfg(test)]
 mod tests {
     use chrono::TimeZone;
+    use pretty_assertions::assert_eq;
 
     use super::*;
 
@@ -490,4 +491,32 @@ mod tests {
             ]
         );
     }
+
+    /// Due to the "archive by time" feature it can happen that an entry is
+    /// split in two: the first (old) entry is archived with an updated end time
+    /// and the second (new) entry is created with the remaining time of the
+    /// original entry. In this case the last entry of the sheet is not the one
+    /// with the largest id, but the one with the latest start time.
+    #[test]
+    fn last_entry_of_sheet_considers_split_entries() {
+        let mut db = SqliteDatabase::from_memory().unwrap();
+        db.init().unwrap();
+
+        let sometime = Utc.ymd(2022, 7, 27);
+
+        db.entry_insert(sometime.and_hms(11, 0, 0), Some(sometime.and_hms(12, 0, 0)), Some("latest".into()), "foo").unwrap();
+        db.entry_insert(sometime.and_hms(10, 0, 0), Some(sometime.and_hms(11, 0, 0)), Some("oldest".into()), "foo").unwrap();
+
+        // filter by start and end
+        assert_eq!(
+            db.last_entry_of_sheet("foo").unwrap().unwrap(),
+            Entry {
+                id: 1,
+                note: Some("latest".into()),
+                start: sometime.and_hms(11, 0, 0),
+                end: Some(sometime.and_hms(12, 0, 0)),
+                sheet: "foo".into(),
+            }
+        );
+    }
 }