PVer4: Keep track of the smallest hashes not their sizes. Replace counters with iterators.

components/safe_browsing_db/v4_store.cc

 // Copyright 2016 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #include "base/base64.h"
 #include "base/bind.h"
 #include "base/files/file_util.h"
 #include "base/metrics/histogram_macros.h"
 #include "base/metrics/sparse_histogram.h"
 #include "base/strings/stringprintf.h"
@@ skipping 174 matching lines @@
   if (lumped_hashes.size() % prefix_size != 0) {
     return ADDITIONS_SIZE_UNEXPECTED_FAILURE;
   }
   // TODO(vakh): Figure out a way to avoid the following copy operation.
   (*additions_map)[prefix_size] = lumped_hashes;
   return APPLY_UPDATE_SUCCESS;
 }
 
 // static
 bool V4Store::GetNextSmallestPrefixSize(const HashPrefixMap& hash_prefix_map,
-                                        const CounterMap& counter_map,
+                                        const IteratorMap& iterator_map,
                                         PrefixSize* smallest_prefix_size) {
   HashPrefix smallest_prefix, current_prefix;
-  for (const auto& counter_pair : counter_map) {
-    PrefixSize prefix_size = counter_pair.first;
-    size_t index = counter_pair.second;
-    size_t sized_index = prefix_size * index;
+  for (const auto& iterator_pair : iterator_map) {
+    PrefixSize prefix_size = iterator_pair.first;
+    HashPrefixes::const_iterator start = iterator_pair.second;
+    HashPrefixes::const_iterator end = start + prefix_size;
 
     const HashPrefixes& hash_prefixes = hash_prefix_map.at(prefix_size);
-    if (sized_index < hash_prefixes.size()) {
-      current_prefix = hash_prefixes.substr(sized_index, prefix_size);
+    if (end <= hash_prefixes.end()) {
+      current_prefix = std::string(start, end);
       if (smallest_prefix.empty() || smallest_prefix > current_prefix) {
         smallest_prefix = current_prefix;

[Nathan Parker, 2016/07/14 22:09:32]

could do
smallest_prefix.swap(current_prefix)

[vakh (use Gerrit instead), 2016/07/14 23:10:41]

Done.

         *smallest_prefix_size = prefix_size;
       }
     }
   }
   return !smallest_prefix.empty();
 }
 
 // static
-void V4Store::InitializeCounterMap(const HashPrefixMap& hash_prefix_map,
-                                   CounterMap* counter_map) {
+void V4Store::InitializeIteratorMap(const HashPrefixMap& hash_prefix_map,
+                                    IteratorMap* iterator_map) {
   for (const auto& map_pair : hash_prefix_map) {
-    (*counter_map)[map_pair.first] = 0;
+    (*iterator_map)[map_pair.first] = map_pair.second.begin();
   }
 }
 
 // static
 HashPrefix V4Store::GetNextUnmergedPrefixForSize(
     PrefixSize prefix_size,
     const HashPrefixMap& hash_prefix_map,
-    const CounterMap& counter_map) {
-  const HashPrefixes& hash_prefixes = hash_prefix_map.at(prefix_size);
-  size_t index_within_list = counter_map.at(prefix_size);
-  size_t sized_index = prefix_size * index_within_list;
-  return hash_prefixes.substr(sized_index, prefix_size);
+    const IteratorMap& iterator_map) {
+  HashPrefixes::const_iterator start = iterator_map.at(prefix_size);
+  HashPrefixes::const_iterator end = start + prefix_size;
+  return std::string(start, end);

[Nathan Parker, 2016/07/14 22:09:32]

could add a dcheck that end <= hash_prefix_map.at(prefix_size).end().  Or...
remove hash_prefix_map from the function altogether.

[vakh (use Gerrit instead), 2016/07/14 23:10:41]

Even better. Removed the method entirely and made the MergeUpdate method
simpler.

 }
 
 // static
 void V4Store::ReserveSpaceInPrefixMap(const HashPrefixMap& other_prefixes_map,
                                       HashPrefixMap* prefix_map_to_update) {
   for (const auto& pair : other_prefixes_map) {
     PrefixSize prefix_size = pair.first;
     size_t prefix_length_to_add = pair.second.length();
 
     const HashPrefixes& existing_prefixes =
         (*prefix_map_to_update)[prefix_size];
     size_t existing_capacity = existing_prefixes.capacity();
 
     (*prefix_map_to_update)[prefix_size].reserve(existing_capacity +
                                                  prefix_length_to_add);
   }
 }
 
 ApplyUpdateResult V4Store::MergeUpdate(const HashPrefixMap& old_prefixes_map,
                                        const HashPrefixMap& additions_map) {
   DCHECK(hash_prefix_map_.empty());
   hash_prefix_map_.clear();
   ReserveSpaceInPrefixMap(old_prefixes_map, &hash_prefix_map_);
   ReserveSpaceInPrefixMap(additions_map, &hash_prefix_map_);
 
   PrefixSize next_size_for_old;
-  CounterMap old_counter_map;
-  InitializeCounterMap(old_prefixes_map, &old_counter_map);
+  IteratorMap old_iterator_map;
+  InitializeIteratorMap(old_prefixes_map, &old_iterator_map);
   bool old_has_unmerged = GetNextSmallestPrefixSize(
-      old_prefixes_map, old_counter_map, &next_size_for_old);
+      old_prefixes_map, old_iterator_map, &next_size_for_old);
 
   PrefixSize next_size_for_additions;
-  CounterMap additions_counter_map;
-  InitializeCounterMap(additions_map, &additions_counter_map);
+  IteratorMap additions_iterator_map;
+  InitializeIteratorMap(additions_map, &additions_iterator_map);
   bool additions_has_unmerged = GetNextSmallestPrefixSize(
-      additions_map, additions_counter_map, &next_size_for_additions);
+      additions_map, additions_iterator_map, &next_size_for_additions);
 
   // Classical merge sort.
   // The two constructs to merge are maps: old_prefixes_map, additions_map.
 
   HashPrefix next_smallest_prefix_old, next_smallest_prefix_additions;
   // At least one of the maps still has elements that need to be merged into the
   // new store.
   while (old_has_unmerged || additions_has_unmerged) {
     // Old map still has elements that need to be merged.
     if (old_has_unmerged) {
       // Get the lexographically smallest hash prefix from the old store.
       next_smallest_prefix_old = GetNextUnmergedPrefixForSize(
-          next_size_for_old, old_prefixes_map, old_counter_map);
+          next_size_for_old, old_prefixes_map, old_iterator_map);
     }
 
     // Additions map still has elements that need to be merged.
     if (additions_has_unmerged) {
       // Get the lexographically smallest hash prefix from the additions in the
       // latest update from the server.
       next_smallest_prefix_additions = GetNextUnmergedPrefixForSize(
-          next_size_for_additions, additions_map, additions_counter_map);
+          next_size_for_additions, additions_map, additions_iterator_map);
     }
 
     // If the same hash prefix appears in the existing store and the additions
     // list, something is clearly wrong. Discard the update.
     if (old_has_unmerged && additions_has_unmerged &&
         next_smallest_prefix_old == next_smallest_prefix_additions) {
       return ADDITIONS_HAS_EXISTING_PREFIX_FAILURE;
     }
 
     // Select which map to pick the next hash prefix from to keep the result in
     // lexographically sorted order.
     bool pick_from_old =
         old_has_unmerged &&
         (!additions_has_unmerged ||
          (next_smallest_prefix_old < next_smallest_prefix_additions));
 
     if (pick_from_old) {
       hash_prefix_map_[next_size_for_old] += next_smallest_prefix_old;
 
-      // Update the counter map, which means that we have merged one hash
+      // Update the iterator map, which means that we have merged one hash
       // prefix of size |next_size_for_old| from the old store.
-      old_counter_map[next_size_for_old]++;
+      old_iterator_map[next_size_for_old] += next_size_for_old;
 
       // Find the next smallest unmerged element in the old store's map.
       old_has_unmerged = GetNextSmallestPrefixSize(
-          old_prefixes_map, old_counter_map, &next_size_for_old);
+          old_prefixes_map, old_iterator_map, &next_size_for_old);
     } else {
       hash_prefix_map_[next_size_for_additions] +=
           next_smallest_prefix_additions;
 
-      // Update the counter map, which means that we have merged one hash
+      // Update the iterator map, which means that we have merged one hash
       // prefix of size |next_size_for_additions| from the update.
-      additions_counter_map[next_size_for_additions]++;
+      additions_iterator_map[next_size_for_additions] +=
+          next_size_for_additions;
 
       // Find the next smallest unmerged element in the additions map.
       additions_has_unmerged = GetNextSmallestPrefixSize(
-          additions_map, additions_counter_map, &next_size_for_additions);
+          additions_map, additions_iterator_map, &next_size_for_additions);
     }
   }
 
   return APPLY_UPDATE_SUCCESS;
 }
 
 StoreReadResult V4Store::ReadFromDisk() {
   DCHECK(task_runner_->RunsTasksOnCurrentThread());
 
   std::string contents;
@@ skipping 73 matching lines @@
 
   if (!base::Move(new_filename, store_path_)) {
     DVLOG(1) << "store_path_: " << store_path_.value();
     return UNABLE_TO_RENAME_FAILURE;
   }
 
   return WRITE_SUCCESS;
 }
 
 }  // namespace safe_browsing