From 12534e55a341b82e5c27973df9b02656ab9ba097 Mon Sep 17 00:00:00 2001
From: Paul Masurel <paul.masurel@gmail.com>
Date: Wed, 3 Aug 2016 11:13:35 +0900
Subject: [PATCH 1/2] Lighter HeapItem by putting term freq aside, and keeping
 ord as u32

---
 src/postings/union_postings.rs | 33 ++++++++++++++++++++-------------
 1 file changed, 20 insertions(+), 13 deletions(-)

diff --git a/src/postings/union_postings.rs b/src/postings/union_postings.rs
index 2182933b9..a5c4e02a9 100644
--- a/src/postings/union_postings.rs
+++ b/src/postings/union_postings.rs
@@ -6,18 +6,17 @@ use query::MultiTermScorer;
 use postings::ScoredDocSet;
 use query::Scorer;
 use fastfield::U32FastFieldReader;
-
+use std::iter;
 
 #[derive(Eq, PartialEq)]
-struct HeapItem(DocId, usize, u32);
+struct HeapItem(DocId, u32);
 
 impl HeapItem {
     pub fn ord(&self,) -> usize{
-        self.1
+        self.1 as usize
     }
 }
 
-
 impl PartialOrd for HeapItem {
     fn partial_cmp(&self, other:&Self) -> Option<Ordering> {
         Some(self.cmp(&other))
@@ -26,13 +25,14 @@ impl PartialOrd for HeapItem {
 
 impl Ord for HeapItem {
     fn cmp(&self, other:&Self) -> Ordering {
-         (self.0, self.1).cmp(&(other.0, other.1)).reverse()
+         (other.0).cmp(&self.0)
     }
 }
 
 pub struct UnionPostings<TPostings: Postings> {
     fieldnorms_readers: Vec<U32FastFieldReader>,
     postings: Vec<TPostings>,
+    term_frequencies: Vec<u32>,
     queue: BinaryHeap<HeapItem>,
     doc: DocId,
     scorer: MultiTermScorer
@@ -43,23 +43,27 @@ impl<TPostings: Postings> UnionPostings<TPostings> {
     pub fn new(fieldnorms_reader: Vec<U32FastFieldReader>, mut postings: Vec<TPostings>, multi_term_scorer: MultiTermScorer) -> UnionPostings<TPostings> {
         let num_postings = postings.len();
         assert_eq!(fieldnorms_reader.len(), num_postings);
-
+        
         for posting in &mut postings {
             assert!(posting.next());
         }
-        let heap_items: Vec<HeapItem> = postings.iter()
+        let mut term_frequencies: Vec<u32> = iter::repeat(0u32).take(num_postings).collect();
+        let heap_items: Vec<HeapItem> = postings
+            .iter()
             .map(|posting| {
                 (posting.doc(), posting.term_freq())
             })
             .enumerate()
             .map(|(ord, (doc, tf))| {
-                HeapItem(doc, ord, tf)
+                term_frequencies[ord] = tf;
+                HeapItem(doc, ord as u32)
             })
             .collect();
         
         UnionPostings {
             fieldnorms_readers: fieldnorms_reader,
             postings: postings,
+            term_frequencies: term_frequencies,
             queue: BinaryHeap::from(heap_items),
             doc: 0,
             scorer: multi_term_scorer
@@ -71,13 +75,12 @@ impl<TPostings: Postings> UnionPostings<TPostings> {
         let cur_postings = &mut self.postings[ord];
         if cur_postings.next() {
             let doc = cur_postings.doc();
-            let tf = cur_postings.term_freq();
-            self.queue.replace(HeapItem(doc, ord, tf));
+            self.term_frequencies[ord] = cur_postings.term_freq();  
+            self.queue.replace(HeapItem(doc, ord as u32));
         }
         else {
             self.queue.pop();
         }
-        
     }
     
     fn get_field_norm(&self, ord:usize, doc:DocId) -> u32 {
@@ -93,9 +96,11 @@ impl<TPostings: Postings> DocSet for UnionPostings<TPostings> {
         self.scorer.clear();
         let cur_doc: DocId;
         match self.queue.peek() {
-            Some(&HeapItem(doc, ord, tf)) => {
+            Some(&HeapItem(doc, ord)) => {
                 cur_doc = doc;
+                let ord: usize = ord as usize;
                 let fieldnorm = self.get_field_norm(ord, doc);
+                let tf = self.term_frequencies[ord];
                 self.scorer.update(ord, tf, fieldnorm);
                 
             }
@@ -106,11 +111,13 @@ impl<TPostings: Postings> DocSet for UnionPostings<TPostings> {
         self.advance_head();
         loop {
             match self.queue.peek() {
-                Some(&HeapItem(peek_doc, peek_ord, peek_tf))  => {
+                Some(&HeapItem(peek_doc, peek_ord))  => {
                     if peek_doc != cur_doc {
                         break;
                     }
                     else {
+                        let peek_ord: usize = peek_ord as usize;
+                        let peek_tf = self.term_frequencies[peek_ord];
                         let fieldnorm = self.get_field_norm(peek_ord, peek_doc);
                         self.scorer.update(peek_ord, peek_tf, fieldnorm);
                     }

From 44f3f23d481bc88f1f9398b714eb87e34b9fc33d Mon Sep 17 00:00:00 2001
From: Paul Masurel <paul.masurel@gmail.com>
Date: Wed, 3 Aug 2016 15:42:10 +0900
Subject: [PATCH 2/2] Changed block size to 16K

---
 src/store/writer.rs | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/store/writer.rs b/src/store/writer.rs
index 5a968c2bd..281393c7b 100644
--- a/src/store/writer.rs
+++ b/src/store/writer.rs
@@ -9,7 +9,7 @@ use lz4;
 use super::StoreReader;
 use super::OffsetIndex;
 
-const BLOCK_SIZE: usize = 131_072;
+const BLOCK_SIZE: usize = 16_384;
 
 pub struct StoreWriter {
     doc: DocId,