Avoid docstore stacking for small segments, fixes #1053 (#1055)

src/indexer/merger.rs

@@ -1004,7 +1004,22 @@ impl IndexMerger {
         } else {
             for reader in &self.readers {
                 let store_reader = reader.get_store_reader()?;
-                if reader.num_deleted_docs() > 0 {
+                if reader.num_deleted_docs() > 0
+                    // If there is not enough data in the store, we avoid stacking in order to
+                    // avoid creating many small blocks in the doc store. Once we have 5 full blocks,
+                    // we start stacking. In the worst case 2/7 of the blocks would be very small. 
+                    // [segment 1 - {1 doc}][segment 2 - {fullblock * 5}{1doc}]  
+                    // => 5 * full blocks, 2 * 1 document blocks
+                    //
+                    // In a more realistic scenario the segments are of the same size, so 1/6 of
+                    // the doc stores would be on average half full, given total randomness (which 
+                    // is not the case here, but not sure how it behaves exactly).
+                    //
+                    // https://github.com/tantivy-search/tantivy/issues/1053
+                    //
+                    // take 7 in order to not walk over all checkpoints.
+                    || store_reader.block_checkpoints().take(7).count() < 6
+                {
                     for doc_bytes_res in store_reader.iter_raw(reader.delete_bitset()) {
                         let doc_bytes = doc_bytes_res?;
                         store_writer.store_bytes(&doc_bytes)?;

src/store/mod.rs

@@ -97,8 +97,10 @@ use self::compression_snap::{compress, decompress};
 #[cfg(test)]
 pub mod tests {
 
+    use futures::executor::block_on;
+
     use super::*;
-    use crate::schema::{self, FieldValue, TextFieldIndexing};
+    use crate::schema::{self, FieldValue, TextFieldIndexing, STORED, TEXT};
     use crate::schema::{Document, TextOptions};
     use crate::{
         directory::{Directory, RamDirectory, WritePtr},
@@ -214,6 +216,47 @@ pub mod tests {
         }
         Ok(())
     }
+    #[test]
+    fn test_merge_of_small_segments() -> crate::Result<()> {
+        let mut schema_builder = schema::Schema::builder();
+
+        let text_field = schema_builder.add_text_field("text_field", TEXT | STORED);
+        let schema = schema_builder.build();
+        let index_builder = Index::builder().schema(schema);
+
+        let index = index_builder.create_in_ram().unwrap();
+
+        {
+            let mut index_writer = index.writer_for_tests().unwrap();
+
+            index_writer.add_document(doc!(text_field=> "1"));
+            assert!(index_writer.commit().is_ok());
+            index_writer.add_document(doc!(text_field=> "2"));
+            assert!(index_writer.commit().is_ok());
+            index_writer.add_document(doc!(text_field=> "3"));
+            assert!(index_writer.commit().is_ok());
+            index_writer.add_document(doc!(text_field=> "4"));
+            assert!(index_writer.commit().is_ok());
+            index_writer.add_document(doc!(text_field=> "5"));
+            assert!(index_writer.commit().is_ok());
+        }
+        // Merging the segments
+        {
+            let segment_ids = index
+                .searchable_segment_ids()
+                .expect("Searchable segments failed.");
+            let mut index_writer = index.writer_for_tests().unwrap();
+            assert!(block_on(index_writer.merge(&segment_ids)).is_ok());
+            assert!(index_writer.wait_merging_threads().is_ok());
+        }
+
+        let searcher = index.reader().unwrap().searcher();
+        assert_eq!(searcher.segment_readers().len(), 1);
+        let reader = searcher.segment_readers().iter().last().unwrap();
+        let store = reader.get_store_reader().unwrap();
+        assert_eq!(store.block_checkpoints().count(), 1);
+        Ok(())
+    }
 }
 
 #[cfg(all(test, feature = "unstable"))]