From b2731f0968b1cf538c3d47eee88f63d3b745bc57 Mon Sep 17 00:00:00 2001
From: Danny Robson <danny@nerdcruft.net>
Date: Sun, 6 Dec 2020 06:21:35 +1000
Subject: [PATCH] chunk/map: add a windowed buzhash

---
 emory/chunk/map.cpp    | 53 +++++++++++++++++++++++++++++-------------
 emory/chunk/region.hpp |  6 +++++
 2 files changed, 43 insertions(+), 16 deletions(-)

diff --git a/emory/chunk/map.cpp b/emory/chunk/map.cpp
index 6e3c36c..bdcebf9 100644
--- a/emory/chunk/map.cpp
+++ b/emory/chunk/map.cpp
@@ -17,36 +17,57 @@ using emory::chunk::map;
 
 ///////////////////////////////////////////////////////////////////////////////
 map::map (
-    cruft::view<u08 const *> data,
+    cruft::view<u08 const *> src,
     emory::chunk::params const &p
 ) {
     using hash_type = cruft::hash::buzhash<u64>;
-    hash_type h (p.window, data);
-    auto remain = data.consume (p.window);
+    if (src.size () < p.window)
+        return;
 
     using digest_type = hash_type::digest_type ;
     digest_type const mask = ~digest_type (0) >> (sizeof (digest_type) * 8 - p.bits);
 
-    for (u08 const *cursor = remain.begin (), *start = data.begin (); cursor != remain.end (); cursor++) {
-        auto const digest = h (cursor);
+    u64 hash_state = 0;
+    auto start = src.begin ();
 
-        if (std::distance (start, cursor) < p.minimum)
-            continue;
+    for (auto cursor = src.begin (); cursor < src.end () - p.minimum; ) {
+        hash_state = 0;
+        cursor += p.minimum - p.window;
+        for (std::size_t i = 0; i < p.window; ++i)
+            hash_state = cruft::rotatel (hash_state, 1) ^ *cursor++;
 
-        if (unlikely (digest & mask))
-            continue;
+        for ( ; cursor < src.end () - p.window; ++cursor) {
+            if (likely (hash_state & mask)) {
+                hash_state = cruft::rotatel (hash_state, 1)
+                           ^ cruft::rotatel (u64 (*(cursor - p.window)), p.window)
+                           ^ *cursor;
+                continue;
+            }
 
-        cruft::view<u08 const*> const region { start, cursor };
-        start = cursor + 1;
+            cruft::view<u08 const*> const region { start, cursor };
+
+            elements.push_back ({
+                .offset = {
+                    std::distance (src.begin (), start),
+                    std::distance (src.begin (), cursor)
+                },
+                .digest = static_hash {} (region)
+            });
+
+            start = cursor;
+            break;
+        }
+    }
+
+    if (start != src.end ()) {
+        cruft::view<u08 const*> const region { start, src.end () };
 
         elements.push_back ({
             .offset = {
-                std::pair<std::size_t,std::size_t> {
-                    std::distance (data.begin (), region.begin ()),
-                    std::distance (data.begin (), region.end   ())
-                },
+                std::distance (src.begin (), start),
+                std::distance (src.begin (), src.end ())
             },
-            .digest = static_hash {} (region),
+            .digest = static_hash {} (region)
         });
     }
 }
diff --git a/emory/chunk/region.hpp b/emory/chunk/region.hpp
index 956c0df..1dfc3a3 100644
--- a/emory/chunk/region.hpp
+++ b/emory/chunk/region.hpp
@@ -20,6 +20,12 @@ namespace emory::chunk {
         std::pair<off64_t, off64_t> offset;
         static_hash::digest_t digest;
 
+        off64_t size (void) const
+        {
+            CHECK_LE (offset.first, offset.second);
+            return offset.second - offset.first;
+        }
+
         struct digest_equality {
             bool operator() (region const &a, region const &b) const
             {