Fix an issue with unnecessary capture in lambda expressions

Closes https://github.com/facebook/rocksdb/issues/3900

CMakeLists.txt

@@ -845,6 +845,7 @@ if(WITH_TESTS)
         db/db_inplace_update_test.cc
         db/db_io_failure_test.cc
         db/db_iter_test.cc
+        db/db_iter_stress_test.cc
         db/db_iterator_test.cc
         db/db_log_iter_test.cc
         db/db_memtable_test.cc

HISTORY.md

@@ -1,12 +1,19 @@
 # Rocksdb Change Log
 ## Unreleased
 ### Public API Change
+* For users of `Statistics` objects created via `CreateDBStatistics()`, the format of the string returned by its `ToString()` method has changed.
+
+## 5.14.0 (5/16/2018)
+### Public API Change
 * Add a BlockBasedTableOption to align uncompressed data blocks on the smaller of block size or page size boundary, to reduce flash reads by avoiding reads spanning 4K pages.
 * The background thread naming convention changed (on supporting platforms) to "rocksdb:<thread pool priority><thread number>", e.g., "rocksdb:low0".
 * Add a new ticker stat rocksdb.number.multiget.keys.found to count number of keys successfully read in MultiGet calls
 * Touch-up to write-related counters in PerfContext. New counters added: write_scheduling_flushes_compactions_time, write_thread_wait_nanos. Counters whose behavior was fixed or modified: write_memtable_time, write_pre_and_post_process_time, write_delay_time.
 * Posix Env's NewRandomRWFile() will fail if the file doesn't exist.
 * Now, `DBOptions::use_direct_io_for_flush_and_compaction` only applies to background writes, and `DBOptions::use_direct_reads` applies to both user reads and background reads. This conforms with Linux's `open(2)` manpage, which advises against simultaneously reading a file in buffered and direct modes, due to possibly undefined behavior and degraded performance.
+* Iterator::Valid() always returns false if !status().ok(). So, now when doing a Seek() followed by some Next()s, there's no need to check status() after every operation.
+* Iterator::Seek()/SeekForPrev()/SeekToFirst()/SeekToLast() always resets status().
+* Introduced `CompressionOptions::kDefaultCompressionLevel`, which is a generic way to tell RocksDB to use the compression library's default level. It is now the default value for `CompressionOptions::level`. Previously the level defaulted to -1, which gave poor compression ratios in ZSTD.
 
 ### New Features
 * Introduce TTL for level compaction so that all files older than ttl go through the compaction process to get rid of old data.
@@ -14,6 +21,8 @@
 * Add DB properties "rocksdb.block-cache-capacity", "rocksdb.block-cache-usage", "rocksdb.block-cache-pinned-usage" to show block cache usage.
 * Add `Env::LowerThreadPoolCPUPriority(Priority)` method, which lowers the CPU priority of background (esp. compaction) threads to minimize interference with foreground tasks.
 * Fsync parent directory after deleting a file in delete scheduler.
+* In level-based compaction, if bottom-pri thread pool was setup via `Env::SetBackgroundThreads()`, compactions to the bottom level will be delegated to that thread pool.
+* `prefix_extractor` has been moved from ImmutableCFOptions to MutableCFOptions, meaning it can be dynamically changed without a DB restart.
 
 ### Bug Fixes
 * Fsync after writing global seq number to the ingestion file in ExternalSstFileIngestionJob.
@@ -21,6 +30,7 @@
 * Fix `BackupableDBOptions::max_valid_backups_to_open` to not delete backup files when refcount cannot be accurately determined.
 * Fix memory leak when pin_l0_filter_and_index_blocks_in_cache is used with partitioned filters
 * Disable rollback of merge operands in WritePrepared transactions to work around an issue in MyRocks. It can be enabled back by setting TransactionDBOptions::rollback_merge_operands to true.
+* Fix wrong results by ReverseBytewiseComparator::FindShortSuccessor()
 
 ### Java API Changes
 * Add `BlockBasedTableConfig.setBlockCache` to allow sharing a block cache across DB instances.

Makefile

@@ -402,6 +402,7 @@ TESTS = \
 	db_blob_index_test \
 	db_bloom_filter_test \
 	db_iter_test \
+	db_iter_stress_test \
 	db_log_iter_test \
 	db_compaction_filter_test \
 	db_compaction_test \
@@ -1195,6 +1196,9 @@ db_tailing_iter_test: db/db_tailing_iter_test.o db/db_test_util.o $(LIBOBJECTS)
 db_iter_test: db/db_iter_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
+db_iter_stress_test: db/db_iter_stress_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
 db_universal_compaction_test: db/db_universal_compaction_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 

TARGETS

@@ -560,6 +560,11 @@ ROCKS_TESTS = [
         "db/db_iter_test.cc",
         "serial",
     ],
+    [
+        "db_iter_stress_test",
+        "db/db_iter_stress_test.cc",
+        "serial",
+    ],
     [
         "db_iterator_test",
         "db/db_iterator_test.cc",

cache/lru_cache.cc

@@ -99,14 +99,20 @@ void LRUHandleTable::Resize() {
   length_ = new_length;
 }
 
-LRUCacheShard::LRUCacheShard()
-    : capacity_(0), high_pri_pool_usage_(0), strict_capacity_limit_(false),
-      high_pri_pool_ratio_(0), high_pri_pool_capacity_(0), usage_(0),
+LRUCacheShard::LRUCacheShard(size_t capacity, bool strict_capacity_limit,
+                             double high_pri_pool_ratio)
+    : capacity_(0),
+      high_pri_pool_usage_(0),
+      strict_capacity_limit_(strict_capacity_limit),
+      high_pri_pool_ratio_(high_pri_pool_ratio),
+      high_pri_pool_capacity_(0),
+      usage_(0),
       lru_usage_(0) {
   // Make empty circular linked list
   lru_.next = &lru_;
   lru_.prev = &lru_;
   lru_low_pri_ = &lru_;
+  SetCapacity(capacity);
 }
 
 LRUCacheShard::~LRUCacheShard() {}
@@ -244,17 +250,13 @@ void* LRUCacheShard::operator new(size_t size) {
   return port::cacheline_aligned_alloc(size);
 }
 
-void* LRUCacheShard::operator new[](size_t size) {
-  return port::cacheline_aligned_alloc(size);
-}
+void* LRUCacheShard::operator new(size_t /*size*/, void* ptr) { return ptr; }
 
 void LRUCacheShard::operator delete(void *memblock) {
   port::cacheline_aligned_free(memblock);
 }
 
-void LRUCacheShard::operator delete[](void* memblock) {
-  port::cacheline_aligned_free(memblock);
-}
+void LRUCacheShard::operator delete(void* /*memblock*/, void* /*ptr*/) {}
 
 void LRUCacheShard::SetCapacity(size_t capacity) {
   autovector<LRUHandle*> last_reference_list;
@@ -473,15 +475,21 @@ LRUCache::LRUCache(size_t capacity, int num_shard_bits,
                    bool strict_capacity_limit, double high_pri_pool_ratio)
     : ShardedCache(capacity, num_shard_bits, strict_capacity_limit) {
   num_shards_ = 1 << num_shard_bits;
-  shards_ = new LRUCacheShard[num_shards_];
-  SetCapacity(capacity);
-  SetStrictCapacityLimit(strict_capacity_limit);
+  shards_ = reinterpret_cast<LRUCacheShard*>(
+      port::cacheline_aligned_alloc(sizeof(LRUCacheShard) * num_shards_));
+  size_t per_shard = (capacity + (num_shards_ - 1)) / num_shards_;
   for (int i = 0; i < num_shards_; i++) {
-    shards_[i].SetHighPriorityPoolRatio(high_pri_pool_ratio);
+    new (&shards_[i])
+        LRUCacheShard(per_shard, strict_capacity_limit, high_pri_pool_ratio);
   }
 }
 
-LRUCache::~LRUCache() { delete[] shards_; }
+LRUCache::~LRUCache() {
+  for (int i = 0; i < num_shards_; i++) {
+    shards_[i].~LRUCacheShard();
+  }
+  port::cacheline_aligned_free(shards_);
+}
 
 CacheShard* LRUCache::GetShard(int shard) {
   return reinterpret_cast<CacheShard*>(&shards_[shard]);

cache/lru_cache.h

@@ -156,7 +156,8 @@ class LRUHandleTable {
 // A single shard of sharded cache.
 class ALIGN_AS(CACHE_LINE_SIZE) LRUCacheShard : public CacheShard {
  public:
-  LRUCacheShard();
+  LRUCacheShard(size_t capacity, bool strict_capacity_limit,
+                double high_pri_pool_ratio);
   virtual ~LRUCacheShard();
 
   // Separate from constructor so caller can easily make an array of LRUCache
@@ -206,13 +207,16 @@ class ALIGN_AS(CACHE_LINE_SIZE) LRUCacheShard : public CacheShard {
   double GetHighPriPoolRatio();
 
   // Overloading to aligned it to cache line size
+  // They are used by tests.
   void* operator new(size_t);
 
-  void* operator new[](size_t);
+  // placement new
+  void* operator new(size_t, void*);
 
   void operator delete(void *);
 
-  void operator delete[](void*);
+  // placement delete, does nothing.
+  void operator delete(void*, void*);
 
  private:
   void LRU_Remove(LRUHandle* e);

cache/lru_cache_test.cc

@@ -7,6 +7,7 @@
 
 #include <string>
 #include <vector>
+#include "port/port.h"
 #include "util/testharness.h"
 
 namespace rocksdb {
@@ -17,19 +18,8 @@ class LRUCacheTest : public testing::Test {
   ~LRUCacheTest() {}
 
   void NewCache(size_t capacity, double high_pri_pool_ratio = 0.0) {
-    cache_.reset(
-#if defined(_MSC_VER)
-#pragma warning(push)
-#pragma warning(disable: 4316) // We've validated the alignment with the new operators
-#endif
-      new LRUCacheShard()
-#if defined(_MSC_VER)
-#pragma warning(pop)
-#endif
-    );
-    cache_->SetCapacity(capacity);
-    cache_->SetStrictCapacityLimit(false);
-    cache_->SetHighPriorityPoolRatio(high_pri_pool_ratio);
+    cache_.reset(new LRUCacheShard(capacity, false /*strict_capcity_limit*/,
+                                   high_pri_pool_ratio));
   }
 
   void Insert(const std::string& key,
@@ -114,7 +104,7 @@ TEST_F(LRUCacheTest, BasicLRU) {
   ValidateLRUList({"e", "z", "d", "u", "v"});
 }
 
-TEST_F(LRUCacheTest, MidPointInsertion) {
+TEST_F(LRUCacheTest, EntriesWithPriority) {
   // Allocate 2 cache entries to high-pri pool.
   NewCache(5, 0.45);
 

db/builder.cc

@@ -39,7 +39,7 @@ namespace rocksdb {
 class TableFactory;
 
 TableBuilder* NewTableBuilder(
-    const ImmutableCFOptions& ioptions,
+    const ImmutableCFOptions& ioptions, const MutableCFOptions& moptions,
     const InternalKeyComparator& internal_comparator,
     const std::vector<std::unique_ptr<IntTblPropCollectorFactory>>*
         int_tbl_prop_collector_factories,
@@ -52,19 +52,20 @@ TableBuilder* NewTableBuilder(
           TablePropertiesCollectorFactory::Context::kUnknownColumnFamily) ==
          column_family_name.empty());
   return ioptions.table_factory->NewTableBuilder(
-      TableBuilderOptions(
-          ioptions, internal_comparator, int_tbl_prop_collector_factories,
-          compression_type, compression_opts, compression_dict, skip_filters,
-          column_family_name, level, creation_time, oldest_key_time),
+      TableBuilderOptions(ioptions, moptions, internal_comparator,
+                          int_tbl_prop_collector_factories, compression_type,
+                          compression_opts, compression_dict, skip_filters,
+                          column_family_name, level, creation_time,
+                          oldest_key_time),
       column_family_id, file);
 }
 
 Status BuildTable(
     const std::string& dbname, Env* env, const ImmutableCFOptions& ioptions,
-    const MutableCFOptions& /*mutable_cf_options*/,
-    const EnvOptions& env_options, TableCache* table_cache,
-    InternalIterator* iter, std::unique_ptr<InternalIterator> range_del_iter,
-    FileMetaData* meta, const InternalKeyComparator& internal_comparator,
+    const MutableCFOptions& mutable_cf_options, const EnvOptions& env_options,
+    TableCache* table_cache, InternalIterator* iter,
+    std::unique_ptr<InternalIterator> range_del_iter, FileMetaData* meta,
+    const InternalKeyComparator& internal_comparator,
     const std::vector<std::unique_ptr<IntTblPropCollectorFactory>>*
         int_tbl_prop_collector_factories,
     uint32_t column_family_id, const std::string& column_family_name,
@@ -122,10 +123,11 @@ Status BuildTable(
       file_writer.reset(new WritableFileWriter(std::move(file), env_options,
                                                ioptions.statistics));
       builder = NewTableBuilder(
-          ioptions, internal_comparator, int_tbl_prop_collector_factories,
-          column_family_id, column_family_name, file_writer.get(), compression,
-          compression_opts, level, nullptr /* compression_dict */,
-          false /* skip_filters */, creation_time, oldest_key_time);
+          ioptions, mutable_cf_options, internal_comparator,
+          int_tbl_prop_collector_factories, column_family_id,
+          column_family_name, file_writer.get(), compression, compression_opts,
+          level, nullptr /* compression_dict */, false /* skip_filters */,
+          creation_time, oldest_key_time);
     }
 
     MergeHelper merge(env, internal_comparator.user_comparator(),
@@ -195,7 +197,8 @@ Status BuildTable(
       // to cache it here for further user reads
       std::unique_ptr<InternalIterator> it(table_cache->NewIterator(
           ReadOptions(), env_options, internal_comparator, meta->fd,
-          nullptr /* range_del_agg */, nullptr,
+          nullptr /* range_del_agg */,
+          mutable_cf_options.prefix_extractor.get(), nullptr,
           (internal_stats == nullptr) ? nullptr
                                       : internal_stats->GetFileReadHist(0),
           false /* for_compaction */, nullptr /* arena */,

db/builder.h

@@ -43,7 +43,7 @@ class InternalIterator;
 // @param compression_dict Data for presetting the compression library's
 //    dictionary, or nullptr.
 TableBuilder* NewTableBuilder(
-    const ImmutableCFOptions& options,
+    const ImmutableCFOptions& options, const MutableCFOptions& moptions,
     const InternalKeyComparator& internal_comparator,
     const std::vector<std::unique_ptr<IntTblPropCollectorFactory>>*
         int_tbl_prop_collector_factories,

db/column_family_test.cc

@@ -299,6 +299,9 @@ class ColumnFamilyTest : public testing::Test {
   }
 
   void PutRandomData(int cf, int num, int key_value_size, bool save = false) {
+    if (cf >= static_cast<int>(keys_.size())) {
+      keys_.resize(cf + 1);
+    }
     for (int i = 0; i < num; ++i) {
       // 10 bytes for key, rest is value
       if (!save) {
@@ -306,7 +309,7 @@ class ColumnFamilyTest : public testing::Test {
                       RandomString(&rnd_, key_value_size - 10)));
       } else {
         std::string key = test::RandomKey(&rnd_, 11);
-        keys_.insert(key);
+        keys_[cf].insert(key);
         ASSERT_OK(Put(cf, key, RandomString(&rnd_, key_value_size - 10)));
       }
     }
@@ -506,7 +509,7 @@ class ColumnFamilyTest : public testing::Test {
 
   std::vector<ColumnFamilyHandle*> handles_;
   std::vector<std::string> names_;
-  std::set<std::string> keys_;
+  std::vector<std::set<std::string>> keys_;
   ColumnFamilyOptions column_family_options_;
   DBOptions db_options_;
   std::string dbname_;
@@ -1414,8 +1417,8 @@ TEST_F(ColumnFamilyTest, MultipleManualCompactions) {
   CompactAll(2);
   AssertFilesPerLevel("0,1", 2);
   // Compare against saved keys
-  std::set<std::string>::iterator key_iter = keys_.begin();
-  while (key_iter != keys_.end()) {
+  std::set<std::string>::iterator key_iter = keys_[1].begin();
+  while (key_iter != keys_[1].end()) {
     ASSERT_NE("NOT_FOUND", Get(1, *key_iter));
     key_iter++;
   }
@@ -1508,8 +1511,8 @@ TEST_F(ColumnFamilyTest, AutomaticAndManualCompactions) {
   CompactAll(2);
   AssertFilesPerLevel("0,1", 2);
   // Compare against saved keys
-  std::set<std::string>::iterator key_iter = keys_.begin();
-  while (key_iter != keys_.end()) {
+  std::set<std::string>::iterator key_iter = keys_[1].begin();
+  while (key_iter != keys_[1].end()) {
     ASSERT_NE("NOT_FOUND", Get(1, *key_iter));
     key_iter++;
   }
@@ -1604,8 +1607,8 @@ TEST_F(ColumnFamilyTest, ManualAndAutomaticCompactions) {
   CompactAll(2);
   AssertFilesPerLevel("0,1", 2);
   // Compare against saved keys
-  std::set<std::string>::iterator key_iter = keys_.begin();
-  while (key_iter != keys_.end()) {
+  std::set<std::string>::iterator key_iter = keys_[1].begin();
+  while (key_iter != keys_[1].end()) {
     ASSERT_NE("NOT_FOUND", Get(1, *key_iter));
     key_iter++;
   }
@@ -1703,8 +1706,8 @@ TEST_F(ColumnFamilyTest, SameCFManualManualCompactions) {
   ASSERT_LE(NumTableFilesAtLevel(0, 1), 2);
 
   // Compare against saved keys
-  std::set<std::string>::iterator key_iter = keys_.begin();
-  while (key_iter != keys_.end()) {
+  std::set<std::string>::iterator key_iter = keys_[1].begin();
+  while (key_iter != keys_[1].end()) {
     ASSERT_NE("NOT_FOUND", Get(1, *key_iter));
     key_iter++;
   }
@@ -1793,8 +1796,8 @@ TEST_F(ColumnFamilyTest, SameCFManualAutomaticCompactions) {
   ASSERT_LE(NumTableFilesAtLevel(0, 1), 2);
 
   // Compare against saved keys
-  std::set<std::string>::iterator key_iter = keys_.begin();
-  while (key_iter != keys_.end()) {
+  std::set<std::string>::iterator key_iter = keys_[1].begin();
+  while (key_iter != keys_[1].end()) {
     ASSERT_NE("NOT_FOUND", Get(1, *key_iter));
     key_iter++;
   }
@@ -1883,8 +1886,8 @@ TEST_F(ColumnFamilyTest, SameCFManualAutomaticCompactionsLevel) {
   AssertFilesPerLevel("0,1", 1);
 
   // Compare against saved keys
-  std::set<std::string>::iterator key_iter = keys_.begin();
-  while (key_iter != keys_.end()) {
+  std::set<std::string>::iterator key_iter = keys_[1].begin();
+  while (key_iter != keys_[1].end()) {
     ASSERT_NE("NOT_FOUND", Get(1, *key_iter));
     key_iter++;
   }
@@ -1972,8 +1975,8 @@ TEST_F(ColumnFamilyTest, SameCFAutomaticManualCompactions) {
   // VERIFY compaction "one"
   AssertFilesPerLevel("1", 1);
   // Compare against saved keys
-  std::set<std::string>::iterator key_iter = keys_.begin();
-  while (key_iter != keys_.end()) {
+  std::set<std::string>::iterator key_iter = keys_[1].begin();
+  while (key_iter != keys_[1].end()) {
     ASSERT_NE("NOT_FOUND", Get(1, *key_iter));
     key_iter++;
   }
@@ -3189,19 +3192,38 @@ TEST_F(ColumnFamilyTest, MultipleCFPathsTest) {
   CreateColumnFamilies({"one", "two"}, {cf_opt1, cf_opt2});
   Reopen({ColumnFamilyOptions(), cf_opt1, cf_opt2});
 
-  PutRandomData(1, 100, 100);
+  PutRandomData(1, 100, 100, true /* save */);
   Flush(1);
 
   // Check that files are generated in appropriate paths.
   ASSERT_EQ(1, GetSstFileCount(cf_opt1.cf_paths[0].path));
   ASSERT_EQ(0, GetSstFileCount(dbname_));
 
-  PutRandomData(2, 100, 100);
+  PutRandomData(2, 100, 100, true /* save */);
   Flush(2);
 
   ASSERT_EQ(1, GetSstFileCount(cf_opt2.cf_paths[0].path));
   ASSERT_EQ(0, GetSstFileCount(dbname_));
+
+  // Re-open and verify the keys.
+  Reopen({ColumnFamilyOptions(), cf_opt1, cf_opt2});
+  DBImpl* dbi = reinterpret_cast<DBImpl*>(db_);
+  for (int cf = 1; cf != 3; ++cf) {
+    ReadOptions read_options;
+    read_options.readahead_size = 0;
+    auto it = dbi->NewIterator(read_options, handles_[cf]);
+    for (it->SeekToFirst(); it->Valid(); it->Next()) {
+      Slice key(it->key());
+      ASSERT_NE(keys_[cf].end(), keys_[cf].find(key.ToString()));
+    }
+    delete it;
+
+    for (const auto& key : keys_[cf]) {
+      ASSERT_NE("NOT_FOUND", Get(cf, key));
+    }
+  }
 }
+
 }  // namespace rocksdb
 
 int main(int argc, char** argv) {

db/compacted_db_impl.cc

@@ -49,8 +49,8 @@ Status CompactedDBImpl::Get(const ReadOptions& options, ColumnFamilyHandle*,
                          GetContext::kNotFound, key, value, nullptr, nullptr,
                          nullptr, nullptr);
   LookupKey lkey(key, kMaxSequenceNumber);
-  files_.files[FindFile(key)].fd.table_reader->Get(
-      options, lkey.internal_key(), &get_context);
+  files_.files[FindFile(key)].fd.table_reader->Get(options, lkey.internal_key(),
+                                                   &get_context, nullptr);
   if (get_context.State() == GetContext::kFound) {
     return Status::OK();
   }
@@ -82,7 +82,7 @@ std::vector<Status> CompactedDBImpl::MultiGet(const ReadOptions& options,
                              GetContext::kNotFound, keys[idx], &pinnable_val,
                              nullptr, nullptr, nullptr, nullptr);
       LookupKey lkey(keys[idx], kMaxSequenceNumber);
-      r->Get(options, lkey.internal_key(), &get_context);
+      r->Get(options, lkey.internal_key(), &get_context, nullptr);
       value.assign(pinnable_val.data(), pinnable_val.size());
       if (get_context.State() == GetContext::kFound) {
         statuses[idx] = Status::OK();

db/compaction_job.cc

@@ -1183,7 +1183,9 @@ Status CompactionJob::FinishCompactionOutputFile(
     // to cache it here for further user reads
     InternalIterator* iter = cfd->table_cache()->NewIterator(
         ReadOptions(), env_options_, cfd->internal_comparator(), meta->fd,
-        nullptr /* range_del_agg */, nullptr,
+        nullptr /* range_del_agg */,
+        sub_compact->compaction->mutable_cf_options()->prefix_extractor.get(),
+        nullptr,
         cfd->internal_stats()->GetFileReadHist(
             compact_->compaction->output_level()),
         false, nullptr /* arena */, false /* skip_filters */,
@@ -1386,9 +1388,10 @@ Status CompactionJob::OpenCompactionOutputFile(
   }
 
   sub_compact->builder.reset(NewTableBuilder(
-      *cfd->ioptions(), cfd->internal_comparator(),
-      cfd->int_tbl_prop_collector_factories(), cfd->GetID(), cfd->GetName(),
-      sub_compact->outfile.get(), sub_compact->compaction->output_compression(),
+      *cfd->ioptions(), *(sub_compact->compaction->mutable_cf_options()),
+      cfd->internal_comparator(), cfd->int_tbl_prop_collector_factories(),
+      cfd->GetID(), cfd->GetName(), sub_compact->outfile.get(),
+      sub_compact->compaction->output_compression(),
       cfd->ioptions()->compression_opts,
       sub_compact->compaction->output_level(), &sub_compact->compression_dict,
       skip_filters, output_file_creation_time));

db/compaction_picker.cc

@@ -477,7 +477,7 @@ bool CompactionPicker::SetupOtherInputs(
       ROCKS_LOG_INFO(ioptions_.info_log,
                      "[%s] Expanding@%d %" ROCKSDB_PRIszt "+%" ROCKSDB_PRIszt
                      "(%" PRIu64 "+%" PRIu64 " bytes) to %" ROCKSDB_PRIszt
-                     "+%" ROCKSDB_PRIszt " (%" PRIu64 "+%" PRIu64 "bytes)\n",
+                     "+%" ROCKSDB_PRIszt " (%" PRIu64 "+%" PRIu64 " bytes)\n",
                      cf_name.c_str(), input_level, inputs->size(),
                      output_level_inputs->size(), inputs_size,
                      output_level_inputs_size, expanded_inputs.size(),

db/comparator_db_test.cc

@@ -2,6 +2,7 @@
 //  This source code is licensed under both the GPLv2 (found in the
 //  COPYING file in the root directory) and Apache 2.0 License
 //  (found in the LICENSE.Apache file in the root directory).
+#include <array>
 #include <map>
 #include <string>
 
@@ -433,6 +434,145 @@ TEST_F(ComparatorDBTest, TwoStrComparator) {
   }
 }
 
+TEST_F(ComparatorDBTest, FindShortestSeparator) {
+  std::string s1 = "abc1xyz";
+  std::string s2 = "abc3xy";
+
+  BytewiseComparator()->FindShortestSeparator(&s1, s2);
+  ASSERT_EQ("abc2", s1);
+
+  s1 = "abc5xyztt";
+
+  ReverseBytewiseComparator()->FindShortestSeparator(&s1, s2);
+  ASSERT_EQ("abc5", s1);
+
+  s1 = "abc3";
+  s2 = "abc2xy";
+  ReverseBytewiseComparator()->FindShortestSeparator(&s1, s2);
+  ASSERT_EQ("abc3", s1);
+
+  s1 = "abc3xyz";
+  s2 = "abc2xy";
+  ReverseBytewiseComparator()->FindShortestSeparator(&s1, s2);
+  ASSERT_EQ("abc3", s1);
+
+  s1 = "abc3xyz";
+  s2 = "abc2";
+  ReverseBytewiseComparator()->FindShortestSeparator(&s1, s2);
+  ASSERT_EQ("abc3", s1);
+
+  std::string old_s1 = s1 = "abc2xy";
+  s2 = "abc2";
+  ReverseBytewiseComparator()->FindShortestSeparator(&s1, s2);
+  ASSERT_TRUE(old_s1 >= s1);
+  ASSERT_TRUE(s1 > s2);
+}
+
+TEST_F(ComparatorDBTest, SeparatorSuccessorRandomizeTest) {
+  // Char list for boundary cases.
+  std::array<unsigned char, 6> char_list{{0, 1, 2, 253, 254, 255}};
+  Random rnd(301);
+
+  for (int attempts = 0; attempts < 1000; attempts++) {
+    uint32_t size1 = rnd.Skewed(4);
+    uint32_t size2;
+
+    if (rnd.OneIn(2)) {
+      // size2 to be random size
+      size2 = rnd.Skewed(4);
+    } else {
+      // size1 is within [-2, +2] of size1
+      int diff = static_cast<int>(rnd.Uniform(5)) - 2;
+      int tmp_size2 = static_cast<int>(size1) + diff;
+      if (tmp_size2 < 0) {
+        tmp_size2 = 0;
+      }
+      size2 = static_cast<uint32_t>(tmp_size2);
+    }
+
+    std::string s1;
+    std::string s2;
+    for (uint32_t i = 0; i < size1; i++) {
+      if (rnd.OneIn(2)) {
+        // Use random byte
+        s1 += static_cast<char>(rnd.Uniform(256));
+      } else {
+        // Use one byte in char_list
+        char c = static_cast<char>(char_list[rnd.Uniform(sizeof(char_list))]);
+        s1 += c;
+      }
+    }
+
+    // First set s2 to be the same as s1, and then modify s2.
+    s2 = s1;
+    s2.resize(size2);
+    // We start from the back of the string
+    if (size2 > 0) {
+      uint32_t pos = size2 - 1;
+      do {
+        if (pos >= size1 || rnd.OneIn(4)) {
+          // For 1/4 chance, use random byte
+          s2[pos] = static_cast<char>(rnd.Uniform(256));
+        } else if (rnd.OneIn(4)) {
+          // In 1/4 chance, stop here.
+          break;
+        } else {
+          // Create a char within [-2, +2] of the matching char of s1.
+          int diff = static_cast<int>(rnd.Uniform(5)) - 2;
+          // char may be signed or unsigned based on platform.
+          int s1_char = static_cast<int>(static_cast<unsigned char>(s1[pos]));
+          int s2_char = s1_char + diff;
+          if (s2_char < 0) {
+            s2_char = 0;
+          }
+          if (s2_char > 255) {
+            s2_char = 255;
+          }
+          s2[pos] = static_cast<char>(s2_char);
+        }
+      } while (pos-- != 0);
+    }
+
+    // Test separators
+    for (int rev = 0; rev < 2; rev++) {
+      if (rev == 1) {
+        // switch s1 and s2
+        std::string t = s1;
+        s1 = s2;
+        s2 = t;
+      }
+      std::string separator = s1;
+      BytewiseComparator()->FindShortestSeparator(&separator, s2);
+      std::string rev_separator = s1;
+      ReverseBytewiseComparator()->FindShortestSeparator(&rev_separator, s2);
+
+      if (s1 == s2) {
+        ASSERT_EQ(s1, separator);
+        ASSERT_EQ(s2, rev_separator);
+      } else if (s1 < s2) {
+        ASSERT_TRUE(s1 <= separator);
+        ASSERT_TRUE(s2 > separator);
+        ASSERT_LE(separator.size(), std::max(s1.size(), s2.size()));
+        ASSERT_EQ(s1, rev_separator);
+      } else {
+        ASSERT_TRUE(s1 >= rev_separator);
+        ASSERT_TRUE(s2 < rev_separator);
+        ASSERT_LE(rev_separator.size(), std::max(s1.size(), s2.size()));
+        ASSERT_EQ(s1, separator);
+      }
+    }
+
+    // Test successors
+    std::string succ = s1;
+    BytewiseComparator()->FindShortSuccessor(&succ);
+    ASSERT_TRUE(succ >= s1);
+
+    succ = s1;
+    ReverseBytewiseComparator()->FindShortSuccessor(&succ);
+    ASSERT_TRUE(succ <= s1);
+  }
+}
+
 }  // namespace rocksdb
 
 int main(int argc, char** argv) {

db/convenience.cc

@@ -50,8 +50,9 @@ Status VerifySstFileChecksum(const Options& options,
   std::unique_ptr<RandomAccessFileReader> file_reader(
       new RandomAccessFileReader(std::move(file), file_path));
   s = ioptions.table_factory->NewTableReader(
-      TableReaderOptions(ioptions, env_options, internal_comparator,
-                         false /* skip_filters */, -1 /* level */),
+      TableReaderOptions(ioptions, options.prefix_extractor.get(), env_options,
+                         internal_comparator, false /* skip_filters */,
+                         -1 /* level */),
       std::move(file_reader), file_size, &table_reader,
       false /* prefetch_index_and_filter_in_cache */);
   if (!s.ok()) {

db/corruption_test.cc

@@ -333,9 +333,9 @@ TEST_F(CorruptionTest, TableFileIndexData) {
   Corrupt(kTableFile, -2000, 500);
   Reopen();
   dbi = reinterpret_cast<DBImpl*>(db_);
-  // one full file should be readable, since only one was corrupted
+  // one full file may be readable, since only one was corrupted
   // the other file should be fully non-readable, since index was corrupted
-  Check(5000, 5000);
+  Check(0, 5000);
   ASSERT_NOK(dbi->VerifyChecksum());
 }
 

db/db_compaction_test.cc

@@ -3128,6 +3128,48 @@ TEST_P(DBCompactionTestWithParam, IntraL0CompactionDoesNotObsoleteDeletions) {
   ASSERT_TRUE(db_->Get(roptions, Key(0), &result).IsNotFound());
 }
 
+TEST_P(DBCompactionTestWithParam, FullCompactionInBottomPriThreadPool) {
+  const int kNumFilesTrigger = 3;
+  Env::Default()->SetBackgroundThreads(1, Env::Priority::BOTTOM);
+  for (bool use_universal_compaction : {false, true}) {
+    Options options = CurrentOptions();
+    if (use_universal_compaction) {
+      options.compaction_style = kCompactionStyleUniversal;
+    } else {
+      options.compaction_style = kCompactionStyleLevel;
+      options.level_compaction_dynamic_level_bytes = true;
+    }
+    options.num_levels = 4;
+    options.write_buffer_size = 100 << 10;     // 100KB
+    options.target_file_size_base = 32 << 10;  // 32KB
+    options.level0_file_num_compaction_trigger = kNumFilesTrigger;
+    // Trigger compaction if size amplification exceeds 110%
+    options.compaction_options_universal.max_size_amplification_percent = 110;
+    DestroyAndReopen(options);
+
+    int num_bottom_pri_compactions = 0;
+    SyncPoint::GetInstance()->SetCallBack(
+        "DBImpl::BGWorkBottomCompaction",
+        [&](void* /*arg*/) { ++num_bottom_pri_compactions; });
+    SyncPoint::GetInstance()->EnableProcessing();
+
+    Random rnd(301);
+    for (int num = 0; num < kNumFilesTrigger; num++) {
+      ASSERT_EQ(NumSortedRuns(), num);
+      int key_idx = 0;
+      GenerateNewFile(&rnd, &key_idx);
+    }
+    dbfull()->TEST_WaitForCompact();
+
+    ASSERT_EQ(1, num_bottom_pri_compactions);
+
+    // Verify that size amplification did occur
+    ASSERT_EQ(NumSortedRuns(), 1);
+    rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+  }
+  Env::Default()->SetBackgroundThreads(0, Env::Priority::BOTTOM);
+}
+
 TEST_F(DBCompactionTest, OptimizedDeletionObsoleting) {
   // Deletions can be dropped when compacted to non-last level if they fall
   // outside the lower-level files' key-ranges.

db/db_impl.cc

@@ -718,6 +718,11 @@ Status DBImpl::FlushWAL(bool sync) {
     if (!s.ok()) {
       ROCKS_LOG_ERROR(immutable_db_options_.info_log, "WAL flush error %s",
                       s.ToString().c_str());
+      // In case there is a fs error we should set it globally to prevent the
+      // future writes
+      WriteStatusCheck(s);
+      // whether sync or not, we should abort the rest of function upon error
+      return s;
     }
     if (!sync) {
       ROCKS_LOG_DEBUG(immutable_db_options_.info_log, "FlushWAL sync=false");
@@ -805,6 +810,8 @@ void DBImpl::MarkLogsSynced(uint64_t up_to, bool synced_dir,
     assert(log.getting_synced);
     if (status.ok() && logs_.size() > 1) {
       logs_to_free_.push_back(log.ReleaseWriter());
+      // To modify logs_ both mutex_ and log_write_mutex_ must be held
+      InstrumentedMutexLock l(&log_write_mutex_);
       it = logs_.erase(it);
     } else {
       log.getting_synced = false;
@@ -962,7 +969,7 @@ InternalIterator* DBImpl::NewInternalIterator(
   MergeIteratorBuilder merge_iter_builder(
       &cfd->internal_comparator(), arena,
       !read_options.total_order_seek &&
-          cfd->ioptions()->prefix_extractor != nullptr);
+          super_version->mutable_cf_options.prefix_extractor != nullptr);
   // Collect iterator for mutable mem
   merge_iter_builder.AddIterator(
       super_version->mem->NewIterator(read_options, arena));
@@ -1561,11 +1568,11 @@ Iterator* DBImpl::NewIterator(const ReadOptions& read_options,
 #else
     SuperVersion* sv = cfd->GetReferencedSuperVersion(&mutex_);
     auto iter = new ForwardIterator(this, read_options, cfd, sv);
-    result =
-        NewDBIterator(env_, read_options, *cfd->ioptions(),
-                      cfd->user_comparator(), iter, kMaxSequenceNumber,
-                      sv->mutable_cf_options.max_sequential_skip_in_iterations,
-                      read_callback);
+    result = NewDBIterator(
+        env_, read_options, *cfd->ioptions(), sv->mutable_cf_options,
+        cfd->user_comparator(), iter, kMaxSequenceNumber,
+        sv->mutable_cf_options.max_sequential_skip_in_iterations,
+        read_callback);
 #endif
   } else {
     // Note: no need to consider the special case of
@@ -1630,7 +1637,7 @@ ArenaWrappedDBIter* DBImpl::NewIteratorImpl(const ReadOptions& read_options,
   // likely that any iterator pointer is close to the iterator it points to so
   // that they are likely to be in the same cache line and/or page.
   ArenaWrappedDBIter* db_iter = NewArenaWrappedDbIterator(
-      env_, read_options, *cfd->ioptions(), snapshot,
+      env_, read_options, *cfd->ioptions(), sv->mutable_cf_options, snapshot,
       sv->mutable_cf_options.max_sequential_skip_in_iterations,
       sv->version_number, read_callback,
       ((read_options.snapshot != nullptr) ? nullptr : this), cfd, allow_blob,
@@ -1681,8 +1688,8 @@ Status DBImpl::NewIterators(
       SuperVersion* sv = cfd->GetReferencedSuperVersion(&mutex_);
       auto iter = new ForwardIterator(this, read_options, cfd, sv);
       iterators->push_back(NewDBIterator(
-          env_, read_options, *cfd->ioptions(), cfd->user_comparator(), iter,
-          kMaxSequenceNumber,
+          env_, read_options, *cfd->ioptions(), sv->mutable_cf_options,
+          cfd->user_comparator(), iter, kMaxSequenceNumber,
           sv->mutable_cf_options.max_sequential_skip_in_iterations,
           read_callback));
     }
@@ -2856,7 +2863,9 @@ Status DBImpl::IngestExternalFile(
     pending_output_elem = CaptureCurrentFileNumberInPendingOutputs();
   }
 
-  status = ingestion_job.Prepare(external_files);
+  SuperVersion* super_version = cfd->GetReferencedSuperVersion(&mutex_);
+  status = ingestion_job.Prepare(external_files, super_version);
+  CleanupSuperVersion(super_version);
   if (!status.ok()) {
     return status;
   }

db/db_impl.h

@@ -221,6 +221,7 @@ class DBImpl : public DB {
   virtual Status Flush(const FlushOptions& options,
                        ColumnFamilyHandle* column_family) override;
   virtual Status FlushWAL(bool sync) override;
+  bool TEST_WALBufferIsEmpty();
   virtual Status SyncWAL() override;
 
   virtual SequenceNumber GetLatestSequenceNumber() const override;

db/db_impl_compaction_flush.cc

@@ -1866,9 +1866,7 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
 
     // Clear Instrument
     ThreadStatusUtil::ResetThreadStatus();
-  } else if (c->column_family_data()->ioptions()->compaction_style ==
-                 kCompactionStyleUniversal &&
-             !is_prepicked && c->output_level() > 0 &&
+  } else if (!is_prepicked && c->output_level() > 0 &&
              c->output_level() ==
                  c->column_family_data()
                      ->current()
@@ -1876,9 +1874,9 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
                      ->MaxOutputLevel(
                          immutable_db_options_.allow_ingest_behind) &&
              env_->GetBackgroundThreads(Env::Priority::BOTTOM) > 0) {
-    // Forward universal compactions involving last level to the bottom pool
-    // if it exists, such that long-running compactions can't block short-
-    // lived ones, like L0->L0s.
+    // Forward compactions involving last level to the bottom pool if it exists,
+    // such that compactions unlikely to contribute to write stalls can be
+    // delayed or deprioritized.
     TEST_SYNC_POINT("DBImpl::BackgroundCompaction:ForwardToBottomPriPool");
     CompactionArg* ca = new CompactionArg;
     ca->db = this;

db/db_impl_debug.cc

@@ -25,6 +25,12 @@ void DBImpl::TEST_SwitchWAL() {
   SwitchWAL(&write_context);
 }
 
+bool DBImpl::TEST_WALBufferIsEmpty() {
+  InstrumentedMutexLock wl(&log_write_mutex_);
+  log::Writer* cur_log_writer = logs_.back().writer;
+  return cur_log_writer->TEST_BufferIsEmpty();
+}
+
 int64_t DBImpl::TEST_MaxNextLevelOverlappingBytes(
     ColumnFamilyHandle* column_family) {
   ColumnFamilyData* cfd;

db/db_impl_open.cc

@@ -1090,7 +1090,8 @@ Status DBImpl::Open(const DBOptions& db_options, const std::string& dbname,
             new_log_number,
             new log::Writer(
                 std::move(file_writer), new_log_number,
-                impl->immutable_db_options_.recycle_log_file_num > 0));
+                impl->immutable_db_options_.recycle_log_file_num > 0,
+                impl->immutable_db_options_.manual_wal_flush));
       }
 
       // set column family handles
@@ -1217,6 +1218,9 @@ Status DBImpl::Open(const DBOptions& db_options, const std::string& dbname,
   if (s.ok()) {
     ROCKS_LOG_INFO(impl->immutable_db_options_.info_log, "DB pointer %p", impl);
     LogFlush(impl->immutable_db_options_.info_log);
+    assert(impl->TEST_WALBufferIsEmpty());
+    // If the assert above fails then we need to FlushWAL before returning
+    // control back to the user.
     if (!persist_options_status.ok()) {
       s = Status::IOError(
           "DB::Open() failed --- Unable to persist Options file",

db/db_impl_readonly.cc

@@ -58,7 +58,7 @@ Iterator* DBImplReadOnly::NewIterator(const ReadOptions& read_options,
   SequenceNumber latest_snapshot = versions_->LastSequence();
   ReadCallback* read_callback = nullptr;  // No read callback provided.
   auto db_iter = NewArenaWrappedDbIterator(
-      env_, read_options, *cfd->ioptions(),
+      env_, read_options, *cfd->ioptions(), super_version->mutable_cf_options,
       (read_options.snapshot != nullptr
            ? reinterpret_cast<const SnapshotImpl*>(read_options.snapshot)
                  ->number_
@@ -88,7 +88,7 @@ Status DBImplReadOnly::NewIterators(
     auto* cfd = reinterpret_cast<ColumnFamilyHandleImpl*>(cfh)->cfd();
     auto* sv = cfd->GetSuperVersion()->Ref();
     auto* db_iter = NewArenaWrappedDbIterator(
-        env_, read_options, *cfd->ioptions(),
+        env_, read_options, *cfd->ioptions(), sv->mutable_cf_options,
         (read_options.snapshot != nullptr
              ? reinterpret_cast<const SnapshotImpl*>(read_options.snapshot)
                    ->number_

db/db_impl_write.cc

@@ -211,6 +211,7 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
   // and protects against concurrent loggers and concurrent writes
   // into memtables
 
+  TEST_SYNC_POINT("DBImpl::WriteImpl:BeforeLeaderEnters");
   last_batch_group_size_ =
       write_thread_.EnterAsBatchGroupLeader(&w, &write_group);
 

db/db_iter.h

@@ -30,6 +30,7 @@ class InternalIterator;
 // into appropriate user keys.
 extern Iterator* NewDBIterator(Env* env, const ReadOptions& read_options,
                                const ImmutableCFOptions& cf_options,
+                               const MutableCFOptions& mutable_cf_options,
                                const Comparator* user_key_comparator,
                                InternalIterator* internal_iter,
                                const SequenceNumber& sequence,
@@ -71,6 +72,7 @@ class ArenaWrappedDBIter : public Iterator {
 
   void Init(Env* env, const ReadOptions& read_options,
             const ImmutableCFOptions& cf_options,
+            const MutableCFOptions& mutable_cf_options,
             const SequenceNumber& sequence,
             uint64_t max_sequential_skip_in_iterations, uint64_t version_number,
             ReadCallback* read_callback, bool allow_blob, bool allow_refresh);
@@ -102,7 +104,8 @@ class ArenaWrappedDBIter : public Iterator {
 // be supported.
 extern ArenaWrappedDBIter* NewArenaWrappedDbIterator(
     Env* env, const ReadOptions& read_options,
-    const ImmutableCFOptions& cf_options, const SequenceNumber& sequence,
+    const ImmutableCFOptions& cf_options,
+    const MutableCFOptions& mutable_cf_options, const SequenceNumber& sequence,
     uint64_t max_sequential_skip_in_iterations, uint64_t version_number,
     ReadCallback* read_callback, DBImpl* db_impl = nullptr,
     ColumnFamilyData* cfd = nullptr, bool allow_blob = false,

db/db_iter_stress_test.cc

@@ -0,0 +1,654 @@
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
+//  This source code is licensed under both the GPLv2 (found in the
+//  COPYING file in the root directory) and Apache 2.0 License
+//  (found in the LICENSE.Apache file in the root directory).
+
+#include "db/db_iter.h"
+#include "db/dbformat.h"
+#include "rocksdb/comparator.h"
+#include "rocksdb/options.h"
+#include "rocksdb/slice.h"
+#include "util/random.h"
+#include "util/string_util.h"
+#include "util/testharness.h"
+#include "utilities/merge_operators.h"
+
+#ifdef GFLAGS
+
+#include "util/gflags_compat.h"
+
+using GFLAGS_NAMESPACE::ParseCommandLineFlags;
+
+DEFINE_bool(verbose, false,
+            "Print huge, detailed trace. Intended for debugging failures.");
+
+#else
+
+void ParseCommandLineFlags(int*, char***, bool) {}
+bool FLAGS_verbose = false;
+
+#endif
+
+namespace rocksdb {
+
+class DBIteratorStressTest : public testing::Test {
+ public:
+  Env* env_;
+
+  DBIteratorStressTest() : env_(Env::Default()) {}
+};
+
+namespace {
+
+struct Entry {
+  std::string key;
+  ValueType type;  // kTypeValue, kTypeDeletion, kTypeMerge
+  uint64_t sequence;
+  std::string ikey;  // internal key, made from `key`, `sequence` and `type`
+  std::string value;
+  // If false, we'll pretend that this entry doesn't exist.
+  bool visible = true;
+
+  bool operator<(const Entry& e) const {
+    if (key != e.key) return key < e.key;
+    return std::tie(sequence, type) > std::tie(e.sequence, e.type);
+  }
+};
+
+struct Data {
+  std::vector<Entry> entries;
+
+  // Indices in `entries` with `visible` = false.
+  std::vector<size_t> hidden;
+  // Keys of entries whose `visible` changed since the last seek of iterators.
+  std::set<std::string> recently_touched_keys;
+};
+
+struct StressTestIterator : public InternalIterator {
+  Data* data;
+  Random64* rnd;
+  InternalKeyComparator cmp;
+
+  // Each operation will return error with this probability...
+  double error_probability = 0;
+  // ... and add/remove entries with this probability.
+  double mutation_probability = 0;
+  // The probability of adding vs removing entries will be chosen so that the
+  // amount of removed entries stays somewhat close to this number.
+  double target_hidden_fraction = 0;
+  // If true, print all mutations to stdout for debugging.
+  bool trace = false;
+
+  int iter = -1;
+  Status status_;
+
+  StressTestIterator(Data* _data, Random64* _rnd, const Comparator* _cmp)
+      : data(_data), rnd(_rnd), cmp(_cmp) {}
+
+  bool Valid() const override {
+    if (iter >= 0 && iter < (int)data->entries.size()) {
+      assert(status_.ok());
+      return true;
+    }
+    return false;
+  }
+
+  Status status() const override { return status_; }
+
+  bool MaybeFail() {
+    if (rnd->Next() >=
+        std::numeric_limits<uint64_t>::max() * error_probability) {
+      return false;
+    }
+    if (rnd->Next() % 2) {
+      status_ = Status::Incomplete("test");
+    } else {
+      status_ = Status::IOError("test");
+    }
+    if (trace) {
+      std::cout << "injecting " << status_.ToString() << std::endl;
+    }
+    iter = -1;
+    return true;
+  }
+
+  void MaybeMutate() {
+    if (rnd->Next() >=
+        std::numeric_limits<uint64_t>::max() * mutation_probability) {
+      return;
+    }
+    do {
+      // If too many entries are hidden, hide less, otherwise hide more.
+      double hide_probability =
+          data->hidden.size() > data->entries.size() * target_hidden_fraction
+              ? 1. / 3
+              : 2. / 3;
+      if (data->hidden.empty()) {
+        hide_probability = 1;
+      }
+      bool do_hide =
+          rnd->Next() < std::numeric_limits<uint64_t>::max() * hide_probability;
+      if (do_hide) {
+        // Hide a random entry.
+        size_t idx = rnd->Next() % data->entries.size();
+        Entry& e = data->entries[idx];
+        if (e.visible) {
+          if (trace) {
+            std::cout << "hiding idx " << idx << std::endl;
+          }
+          e.visible = false;
+          data->hidden.push_back(idx);
+          data->recently_touched_keys.insert(e.key);
+        } else {
+          // Already hidden. Let's go unhide something instead, just because
+          // it's easy and it doesn't really matter what we do.
+          do_hide = false;
+        }
+      }
+      if (!do_hide) {
+        // Unhide a random entry.
+        size_t hi = rnd->Next() % data->hidden.size();
+        size_t idx = data->hidden[hi];
+        if (trace) {
+          std::cout << "unhiding idx " << idx << std::endl;
+        }
+        Entry& e = data->entries[idx];
+        assert(!e.visible);
+        e.visible = true;
+        data->hidden[hi] = data->hidden.back();
+        data->hidden.pop_back();
+        data->recently_touched_keys.insert(e.key);
+      }
+    } while (rnd->Next() % 3 != 0);  // do 3 mutations on average
+  }
+
+  void SkipForward() {
+    while (iter < (int)data->entries.size() && !data->entries[iter].visible) {
+      ++iter;
+    }
+  }
+  void SkipBackward() {
+    while (iter >= 0 && !data->entries[iter].visible) {
+      --iter;
+    }
+  }
+
+  void SeekToFirst() override {
+    if (MaybeFail()) return;
+    MaybeMutate();
+
+    status_ = Status::OK();
+    iter = 0;
+    SkipForward();
+  }
+  void SeekToLast() override {
+    if (MaybeFail()) return;
+    MaybeMutate();
+
+    status_ = Status::OK();
+    iter = (int)data->entries.size() - 1;
+    SkipBackward();
+  }
+
+  void Seek(const Slice& target) override {
+    if (MaybeFail()) return;
+    MaybeMutate();
+
+    status_ = Status::OK();
+    // Binary search.
+    auto it = std::partition_point(
+        data->entries.begin(), data->entries.end(),
+        [&](const Entry& e) { return cmp.Compare(e.ikey, target) < 0; });
+    iter = (int)(it - data->entries.begin());
+    SkipForward();
+  }
+  void SeekForPrev(const Slice& target) override {
+    if (MaybeFail()) return;
+    MaybeMutate();
+
+    status_ = Status::OK();
+    // Binary search.
+    auto it = std::partition_point(
+        data->entries.begin(), data->entries.end(),
+        [&](const Entry& e) { return cmp.Compare(e.ikey, target) <= 0; });
+    iter = (int)(it - data->entries.begin());
+    --iter;
+    SkipBackward();
+  }
+
+  void Next() override {
+    assert(Valid());
+    if (MaybeFail()) return;
+    MaybeMutate();
+    ++iter;
+    SkipForward();
+  }
+  void Prev() override {
+    assert(Valid());
+    if (MaybeFail()) return;
+    MaybeMutate();
+    --iter;
+    SkipBackward();
+  }
+
+  Slice key() const override {
+    assert(Valid());
+    return data->entries[iter].ikey;
+  }
+  Slice value() const override {
+    assert(Valid());
+    return data->entries[iter].value;
+  }
+
+  bool IsKeyPinned() const override { return true; }
+  bool IsValuePinned() const override { return true; }
+};
+
+// A small reimplementation of DBIter, supporting only some of the features,
+// and doing everything in O(log n).
+// Skips all keys that are in recently_touched_keys.
+struct ReferenceIterator {
+  Data* data;
+  uint64_t sequence;  // ignore entries with sequence number below this
+
+  bool valid = false;
+  std::string key;
+  std::string value;
+
+  ReferenceIterator(Data* _data, uint64_t _sequence)
+      : data(_data), sequence(_sequence) {}
+
+  bool Valid() const { return valid; }
+
+  // Finds the first entry with key
+  // greater/less/greater-or-equal/less-or-equal than `key`, depending on
+  // arguments: if `skip`, inequality is strict; if `forward`, it's
+  // greater/greater-or-equal, otherwise less/less-or-equal.
+  // Sets `key` to the result.
+  // If no such key exists, returns false. Doesn't check `visible`.
+  bool FindNextKey(bool skip, bool forward) {
+    valid = false;
+    auto it = std::partition_point(data->entries.begin(), data->entries.end(),
+                                   [&](const Entry& e) {
+                                     if (forward != skip) {
+                                       return e.key < key;
+                                     } else {
+                                       return e.key <= key;
+                                     }
+                                   });
+    if (forward) {
+      if (it != data->entries.end()) {
+        key = it->key;
+        return true;
+      }
+    } else {
+      if (it != data->entries.begin()) {
+        --it;
+        key = it->key;
+        return true;
+      }
+    }
+    return false;
+  }
+
+  bool FindValueForCurrentKey() {
+    if (data->recently_touched_keys.count(key)) {
+      return false;
+    }
+
+    // Find the first entry for the key. The caller promises that it exists.
+    auto it = std::partition_point(data->entries.begin(), data->entries.end(),
+                                   [&](const Entry& e) {
+                                     if (e.key != key) {
+                                       return e.key < key;
+                                     }
+                                     return e.sequence > sequence;
+                                   });
+
+    // Find the first visible entry.
+    for (;; ++it) {
+      if (it == data->entries.end()) {
+        return false;
+      }
+      Entry& e = *it;
+      if (e.key != key) {
+        return false;
+      }
+      assert(e.sequence <= sequence);
+      if (!e.visible) continue;
+      if (e.type == kTypeDeletion) {
+        return false;
+      }
+      if (e.type == kTypeValue) {
+        value = e.value;
+        valid = true;
+        return true;
+      }
+      assert(e.type == kTypeMerge);
+      break;
+    }
+
+    // Collect merge operands.
+    std::vector<Slice> operands;
+    for (; it != data->entries.end(); ++it) {
+      Entry& e = *it;
+      if (e.key != key) {
+        break;
+      }
+      assert(e.sequence <= sequence);
+      if (!e.visible) continue;
+      if (e.type == kTypeDeletion) {
+        break;
+      }
+      operands.push_back(e.value);
+      if (e.type == kTypeValue) {
+        break;
+      }
+    }
+
+    // Do a merge.
+    value = operands.back().ToString();
+    for (int i = (int)operands.size() - 2; i >= 0; --i) {
+      value.append(",");
+      value.append(operands[i].data(), operands[i].size());
+    }
+
+    valid = true;
+    return true;
+  }
+
+  // Start at `key` and move until we encounter a valid value.
+  // `forward` defines the direction of movement.
+  // If `skip` is true, we're looking for key not equal to `key`.
+  void DoTheThing(bool skip, bool forward) {
+    while (FindNextKey(skip, forward) && !FindValueForCurrentKey()) {
+      skip = true;
+    }
+  }
+
+  void Seek(const Slice& target) {
+    key = target.ToString();
+    DoTheThing(false, true);
+  }
+  void SeekForPrev(const Slice& target) {
+    key = target.ToString();
+    DoTheThing(false, false);
+  }
+  void SeekToFirst() { Seek(""); }
+  void SeekToLast() {
+    key = data->entries.back().key;
+    DoTheThing(false, false);
+  }
+  void Next() {
+    assert(Valid());
+    DoTheThing(true, true);
+  }
+  void Prev() {
+    assert(Valid());
+    DoTheThing(true, false);
+  }
+};
+
+}  // namespace
+
+// Use an internal iterator that sometimes returns errors and sometimes
+// adds/removes entries on the fly. Do random operations on a DBIter and
+// check results.
+// TODO: can be improved for more coverage:
+//   * Override IsKeyPinned() and IsValuePinned() to actually use
+//     PinnedIteratorManager and check that there's no use-after free.
+//   * Try different combinations of prefix_extractor, total_order_seek,
+//     prefix_same_as_start, iterate_lower_bound, iterate_upper_bound.
+TEST_F(DBIteratorStressTest, StressTest) {
+  // We use a deterministic RNG, and everything happens in a single thread.
+  Random64 rnd(826909345792864532ll);
+
+  auto gen_key = [&](int max_key) {
+    assert(max_key > 0);
+    int len = 0;
+    int a = max_key;
+    while (a) {
+      a /= 10;
+      ++len;
+    }
+    std::string s = ToString(rnd.Next() % static_cast<uint64_t>(max_key));
+    s.insert(0, len - (int)s.size(), '0');
+    return s;
+  };
+
+  Options options;
+  options.merge_operator = MergeOperators::CreateFromStringId("stringappend");
+  ReadOptions ropt;
+
+  size_t num_matching = 0;
+  size_t num_at_end = 0;
+  size_t num_not_ok = 0;
+  size_t num_recently_removed = 0;
+
+  // Number of iterations for each combination of parameters
+  // (there are ~250 of those).
+  // Tweak this to change the test run time.
+  // As of the time of writing, the test takes ~4 seconds for value of 5000.
+  const int num_iterations = 5000;
+  // Enable this to print all the operations for debugging.
+  bool trace = FLAGS_verbose;
+
+  for (int num_entries : {5, 10, 100}) {
+    for (double key_space : {0.1, 1.0, 3.0}) {
+      for (ValueType prevalent_entry_type :
+           {kTypeValue, kTypeDeletion, kTypeMerge}) {
+        for (double error_probability : {0.01, 0.1}) {
+          for (double mutation_probability : {0.01, 0.5}) {
+            for (double target_hidden_fraction : {0.1, 0.5}) {
+              std::string trace_str =
+                  "entries: " + ToString(num_entries) +
+                  ", key_space: " + ToString(key_space) +
+                  ", error_probability: " + ToString(error_probability) +
+                  ", mutation_probability: " + ToString(mutation_probability) +
+                  ", target_hidden_fraction: " +
+                  ToString(target_hidden_fraction);
+              SCOPED_TRACE(trace_str);
+              if (trace) {
+                std::cout << trace_str << std::endl;
+              }
+
+              // Generate data.
+              Data data;
+              int max_key = (int)(num_entries * key_space) + 1;
+              for (int i = 0; i < num_entries; ++i) {
+                Entry e;
+                e.key = gen_key(max_key);
+                if (rnd.Next() % 10 != 0) {
+                  e.type = prevalent_entry_type;
+                } else {
+                  const ValueType types[] = {kTypeValue, kTypeDeletion,
+                                             kTypeMerge};
+                  e.type =
+                      types[rnd.Next() % (sizeof(types) / sizeof(types[0]))];
+                }
+                e.sequence = i;
+                e.value = "v" + ToString(i);
+                ParsedInternalKey internal_key(e.key, e.sequence, e.type);
+                AppendInternalKey(&e.ikey, internal_key);
+
+                data.entries.push_back(e);
+              }
+              std::sort(data.entries.begin(), data.entries.end());
+              if (trace) {
+                std::cout << "entries:";
+                for (size_t i = 0; i < data.entries.size(); ++i) {
+                  Entry& e = data.entries[i];
+                  std::cout
+                      << "\n  idx " << i << ": \"" << e.key << "\": \""
+                      << e.value << "\" seq: " << e.sequence << " type: "
+                      << (e.type == kTypeValue
+                              ? "val"
+                              : e.type == kTypeDeletion ? "del" : "merge");
+                }
+                std::cout << std::endl;
+              }
+
+              std::unique_ptr<Iterator> db_iter;
+              std::unique_ptr<ReferenceIterator> ref_iter;
+              for (int iteration = 0; iteration < num_iterations; ++iteration) {
+                SCOPED_TRACE(iteration);
+                // Create a new iterator every ~30 operations.
+                if (db_iter == nullptr || rnd.Next() % 30 == 0) {
+                  uint64_t sequence = rnd.Next() % (data.entries.size() + 2);
+                  ref_iter.reset(new ReferenceIterator(&data, sequence));
+                  if (trace) {
+                    std::cout << "new iterator, seq: " << sequence << std::endl;
+                  }
+
+                  auto internal_iter =
+                      new StressTestIterator(&data, &rnd, BytewiseComparator());
+                  internal_iter->error_probability = error_probability;
+                  internal_iter->mutation_probability = mutation_probability;
+                  internal_iter->target_hidden_fraction =
+                      target_hidden_fraction;
+                  internal_iter->trace = trace;
+                  db_iter.reset(NewDBIterator(
+                      env_, ropt, ImmutableCFOptions(options),
+                      MutableCFOptions(options), BytewiseComparator(),
+                      internal_iter, sequence,
+                      options.max_sequential_skip_in_iterations,
+                      nullptr /*read_callback*/));
+                }
+
+                // Do a random operation. It's important to do it on ref_it
+                // later than on db_iter to make sure ref_it sees the correct
+                // recently_touched_keys.
+                std::string old_key;
+                bool forward = rnd.Next() % 2 > 0;
+                // Do Next()/Prev() ~90% of the time.
+                bool seek = !ref_iter->Valid() || rnd.Next() % 10 == 0;
+                if (trace) {
+                  std::cout << iteration << ": ";
+                }
+
+                if (!seek) {
+                  assert(db_iter->Valid());
+                  old_key = ref_iter->key;
+                  if (trace) {
+                    std::cout << (forward ? "Next" : "Prev") << std::endl;
+                  }
+
+                  if (forward) {
+                    db_iter->Next();
+                    ref_iter->Next();
+                  } else {
+                    db_iter->Prev();
+                    ref_iter->Prev();
+                  }
+                } else {
+                  data.recently_touched_keys.clear();
+                  // Do SeekToFirst less often than Seek.
+                  if (rnd.Next() % 4 == 0) {
+                    if (trace) {
+                      std::cout << (forward ? "SeekToFirst" : "SeekToLast")
+                                << std::endl;
+                    }
+
+                    if (forward) {
+                      old_key = "";
+                      db_iter->SeekToFirst();
+                      ref_iter->SeekToFirst();
+                    } else {
+                      old_key = data.entries.back().key;
+                      db_iter->SeekToLast();
+                      ref_iter->SeekToLast();
+                    }
+                  } else {
+                    old_key = gen_key(max_key);
+                    if (trace) {
+                      std::cout << (forward ? "Seek" : "SeekForPrev") << " \""
+                                << old_key << '"' << std::endl;
+                    }
+                    if (forward) {
+                      db_iter->Seek(old_key);
+                      ref_iter->Seek(old_key);
+                    } else {
+                      db_iter->SeekForPrev(old_key);
+                      ref_iter->SeekForPrev(old_key);
+                    }
+                  }
+                }
+
+                // Check the result.
+                if (db_iter->Valid()) {
+                  ASSERT_TRUE(db_iter->status().ok());
+                  if (data.recently_touched_keys.count(
+                          db_iter->key().ToString())) {
+                    // Ended on a key that may have been mutated during the
+                    // operation. Reference iterator skips such keys, so we
+                    // can't check the exact result.
+
+                    // Check that the key moved in the right direction.
+                    if (forward) {
+                      if (seek)
+                        ASSERT_GE(db_iter->key().ToString(), old_key);
+                      else
+                        ASSERT_GT(db_iter->key().ToString(), old_key);
+                    } else {
+                      if (seek)
+                        ASSERT_LE(db_iter->key().ToString(), old_key);
+                      else
+                        ASSERT_LT(db_iter->key().ToString(), old_key);
+                    }
+
+                    if (ref_iter->Valid()) {
+                      // Check that DBIter didn't miss any non-mutated key.
+                      if (forward) {
+                        ASSERT_LT(db_iter->key().ToString(), ref_iter->key);
+                      } else {
+                        ASSERT_GT(db_iter->key().ToString(), ref_iter->key);
+                      }
+                    }
+                    // Tell the next iteration of the loop to reseek the
+                    // iterators.
+                    ref_iter->valid = false;
+
+                    ++num_recently_removed;
+                  } else {
+                    ASSERT_TRUE(ref_iter->Valid());
+                    ASSERT_EQ(ref_iter->key, db_iter->key().ToString());
+                    ASSERT_EQ(ref_iter->value, db_iter->value());
+                    ++num_matching;
+                  }
+                } else if (db_iter->status().ok()) {
+                  ASSERT_FALSE(ref_iter->Valid());
+                  ++num_at_end;
+                } else {
+                  // Non-ok status. Nothing to check here.
+                  // Tell the next iteration of the loop to reseek the
+                  // iterators.
+                  ref_iter->valid = false;
+                  ++num_not_ok;
+                }
+              }
+            }
+          }
+        }
+      }
+    }
+  }
+
+  // Check that all cases were hit many times.
+  EXPECT_GT(num_matching, 10000);
+  EXPECT_GT(num_at_end, 10000);
+  EXPECT_GT(num_not_ok, 10000);
+  EXPECT_GT(num_recently_removed, 10000);
+
+  std::cout << "stats:\n  exact matches: " << num_matching
+            << "\n  end reached: " << num_at_end
+            << "\n  non-ok status: " << num_not_ok
+            << "\n  mutated on the fly: " << num_recently_removed << std::endl;
+}
+
+}  // namespace rocksdb
+
+int main(int argc, char** argv) {
+  ::testing::InitGoogleTest(&argc, argv);
+  ParseCommandLineFlags(&argc, &argv, true);
+  return RUN_ALL_TESTS();
+}

db/db_iterator_test.cc

@@ -862,6 +862,8 @@ TEST_P(DBIteratorTest, IteratorPinsRef) {
   } while (ChangeCompactOptions());
 }
 
+// SetOptions not defined in ROCKSDB LITE
+#ifndef ROCKSDB_LITE
 TEST_P(DBIteratorTest, DBIteratorBoundTest) {
   Options options = CurrentOptions();
   options.env = env_;
@@ -946,9 +948,7 @@ TEST_P(DBIteratorTest, DBIteratorBoundTest) {
   }
 
   // prefix is the first letter of the key
-  options.prefix_extractor.reset(NewFixedPrefixTransform(1));
-
-  DestroyAndReopen(options);
+  ASSERT_OK(dbfull()->SetOptions({{"prefix_extractor", "fixed:1"}}));
   ASSERT_OK(Put("a", "0"));
   ASSERT_OK(Put("foo", "bar"));
   ASSERT_OK(Put("foo1", "bar1"));
@@ -1035,6 +1035,7 @@ TEST_P(DBIteratorTest, DBIteratorBoundTest) {
     ASSERT_EQ(static_cast<int>(get_perf_context()->internal_delete_skipped_count), 0);
   }
 }
+#endif
 
 TEST_P(DBIteratorTest, DBIteratorBoundOptimizationTest) {
   int upper_bound_hits = 0;
@@ -2232,6 +2233,46 @@ TEST_P(DBIteratorTest, SeekAfterHittingManyInternalKeys) {
   ASSERT_EQ(iter2->value().ToString(), "val_6");
 }
 
+// Reproduces a former bug where iterator would skip some records when DBIter
+// re-seeks subiterator with Incomplete status.
+TEST_P(DBIteratorTest, NonBlockingIterationBugRepro) {
+  Options options = CurrentOptions();
+  BlockBasedTableOptions table_options;
+  // Make sure the sst file has more than one block.
+  table_options.flush_block_policy_factory =
+      std::make_shared<FlushBlockEveryKeyPolicyFactory>();
+  options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+  DestroyAndReopen(options);
+
+  // Two records in sst file, each in its own block.
+  Put("b", "");
+  Put("d", "");
+  Flush();
+
+  // Create a nonblocking iterator before writing to memtable.
+  ReadOptions ropt;
+  ropt.read_tier = kBlockCacheTier;
+  unique_ptr<Iterator> iter(NewIterator(ropt));
+
+  // Overwrite a key in memtable many times to hit
+  // max_sequential_skip_in_iterations (which is 8 by default).
+  for (int i = 0; i < 20; ++i) {
+    Put("c", "");
+  }
+
+  // Load the second block in sst file into the block cache.
+  {
+    unique_ptr<Iterator> iter2(NewIterator(ReadOptions()));
+    iter2->Seek("d");
+  }
+
+  // Finally seek the nonblocking iterator.
+  iter->Seek("a");
+  // With the bug, the status used to be OK, and the iterator used to point to
+  // "d".
+  EXPECT_TRUE(iter->status().IsIncomplete());
+}
+
 INSTANTIATE_TEST_CASE_P(DBIteratorTestInstance, DBIteratorTest,
                         testing::Values(true, false));
 

db/db_options_test.cc

@@ -242,11 +242,12 @@ TEST_F(DBOptionsTest, WritableFileMaxBufferSize) {
   ASSERT_EQ(unmatch_cnt, 0);
   ASSERT_GE(match_cnt, 11);
 
-  buffer_size = 512 * 1024;
-  match_cnt = 0;
-  unmatch_cnt = 0;
   ASSERT_OK(
       dbfull()->SetDBOptions({{"writable_file_max_buffer_size", "524288"}}));
+  buffer_size = 512 * 1024;
+  match_cnt = 0;
+  unmatch_cnt = 0;  // SetDBOptions() will create a WriteableFileWriter
+
   ASSERT_EQ(buffer_size,
             dbfull()->GetDBOptions().writable_file_max_buffer_size);
   i = 0;

db/db_test.cc

@@ -2153,10 +2153,14 @@ TEST_F(DBTest, GroupCommitTest) {
   do {
     Options options = CurrentOptions();
     options.env = env_;
-    env_->log_write_slowdown_.store(100);
     options.statistics = rocksdb::CreateDBStatistics();
     Reopen(options);
 
+    rocksdb::SyncPoint::GetInstance()->LoadDependency(
+        {{"WriteThread::JoinBatchGroup:BeganWaiting",
+          "DBImpl::WriteImpl:BeforeLeaderEnters"}});
+    rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
     // Start threads
     GCThread thread[kGCNumThreads];
     for (int id = 0; id < kGCNumThreads; id++) {
@@ -2165,13 +2169,7 @@ TEST_F(DBTest, GroupCommitTest) {
       thread[id].done = false;
       env_->StartThread(GCThreadBody, &thread[id]);
     }
-
-    for (int id = 0; id < kGCNumThreads; id++) {
-      while (thread[id].done == false) {
-        env_->SleepForMicroseconds(100000);
-      }
-    }
-    env_->log_write_slowdown_.store(0);
+    env_->WaitForJoin();
 
     ASSERT_GT(TestGetTickerCount(options, WRITE_DONE_BY_OTHER), 0);
 
@@ -4607,6 +4605,181 @@ TEST_F(DBTest, FileCreationRandomFailure) {
 }
 
 #ifndef ROCKSDB_LITE
+int CountIter(Iterator* iter, const Slice& key) {
+  int count = 0;
+  for (iter->Seek(key); iter->Valid() && iter->status() == Status::OK();
+       iter->Next()) {
+    count++;
+  }
+  return count;
+}
+
+// Create multiple SST files each with a different prefix_extractor config,
+// verify iterators can read all SST files using the latest config.
+TEST_F(DBTest, DynamicBloomFilterMultipleSST) {
+  Options options;
+  options.create_if_missing = true;
+  options.prefix_extractor.reset(NewFixedPrefixTransform(1));
+  options.disable_auto_compactions = true;
+  // Enable prefix bloom for SST files
+  BlockBasedTableOptions table_options;
+  table_options.filter_policy.reset(NewBloomFilterPolicy(10, true));
+  options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+  DestroyAndReopen(options);
+
+  ReadOptions read_options;
+  read_options.prefix_same_as_start = true;
+
+  // first SST with fixed:1 BF
+  ASSERT_OK(Put("foo2", "bar2"));
+  ASSERT_OK(Put("foo", "bar"));
+  ASSERT_OK(Put("foq1", "bar1"));
+  ASSERT_OK(Put("fpa", "0"));
+  dbfull()->Flush(FlushOptions());
+  Iterator* iter_old = db_->NewIterator(read_options);
+  ASSERT_EQ(CountIter(iter_old, "foo"), 4);
+
+  ASSERT_OK(dbfull()->SetOptions({{"prefix_extractor", "capped:3"}}));
+  ASSERT_EQ(0, strcmp(dbfull()->GetOptions().prefix_extractor->Name(),
+                      "rocksdb.CappedPrefix.3"));
+  Iterator* iter = db_->NewIterator(read_options);
+  ASSERT_EQ(CountIter(iter, "foo"), 2);
+
+  // second SST with capped:3 BF
+  ASSERT_OK(Put("foo3", "bar3"));
+  ASSERT_OK(Put("foo4", "bar4"));
+  ASSERT_OK(Put("foq5", "bar5"));
+  ASSERT_OK(Put("fpb", "1"));
+  dbfull()->Flush(FlushOptions());
+  // BF is cappped:3 now
+  Iterator* iter_tmp = db_->NewIterator(read_options);
+  ASSERT_EQ(CountIter(iter_tmp, "foo"), 4);
+  delete iter_tmp;
+
+  ASSERT_OK(dbfull()->SetOptions({{"prefix_extractor", "fixed:2"}}));
+  ASSERT_EQ(0, strcmp(dbfull()->GetOptions().prefix_extractor->Name(),
+                      "rocksdb.FixedPrefix.2"));
+  // third SST with fixed:2 BF
+  ASSERT_OK(Put("foo6", "bar6"));
+  ASSERT_OK(Put("foo7", "bar7"));
+  ASSERT_OK(Put("foq8", "bar8"));
+  ASSERT_OK(Put("fpc", "2"));
+  dbfull()->Flush(FlushOptions());
+  // BF is fixed:2 now
+  iter_tmp = db_->NewIterator(read_options);
+  ASSERT_EQ(CountIter(iter_tmp, "foo"), 9);
+  delete iter_tmp;
+
+  // TODO(Zhongyi): verify existing iterator cannot see newly inserted keys
+  ASSERT_EQ(CountIter(iter_old, "foo"), 4);
+  ASSERT_EQ(CountIter(iter, "foo"), 2);
+  delete iter;
+  delete iter_old;
+
+  // keys in all three SSTs are visible to iterator
+  Iterator* iter_all = db_->NewIterator(read_options);
+  ASSERT_EQ(CountIter(iter_all, "foo"), 9);
+  delete iter_all;
+  ASSERT_OK(dbfull()->SetOptions({{"prefix_extractor", "capped:3"}}));
+  ASSERT_EQ(0, strcmp(dbfull()->GetOptions().prefix_extractor->Name(),
+                      "rocksdb.CappedPrefix.3"));
+  iter_all = db_->NewIterator(read_options);
+  ASSERT_EQ(CountIter(iter_all, "foo"), 6);
+  delete iter_all;
+  // TODO(Zhongyi): add test for cases where certain SST are skipped
+  // Also verify BF related counters like BLOOM_FILTER_USEFUL
+}
+
+// Create a new column family in a running DB, change prefix_extractor
+// dynamically, verify the iterator created on the new column family behaves
+// as expected
+TEST_F(DBTest, DynamicBloomFilterNewColumnFamily) {
+  Options options = CurrentOptions();
+  options.create_if_missing = true;
+  options.prefix_extractor.reset(NewFixedPrefixTransform(1));
+  options.disable_auto_compactions = true;
+  // Enable prefix bloom for SST files
+  BlockBasedTableOptions table_options;
+  table_options.filter_policy.reset(NewBloomFilterPolicy(10, true));
+  options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+  CreateAndReopenWithCF({"pikachu"}, options);
+  ReadOptions read_options;
+  read_options.prefix_same_as_start = true;
+  // create a new CF and set prefix_extractor dynamically
+  options.prefix_extractor.reset(NewCappedPrefixTransform(3));
+  CreateColumnFamilies({"ramen_dojo"}, options);
+  ASSERT_EQ(0,
+            strcmp(dbfull()->GetOptions(handles_[2]).prefix_extractor->Name(),
+                   "rocksdb.CappedPrefix.3"));
+  ASSERT_OK(Put(2, "foo3", "bar3"));
+  ASSERT_OK(Put(2, "foo4", "bar4"));
+  ASSERT_OK(Put(2, "foo5", "bar5"));
+  ASSERT_OK(Put(2, "foq6", "bar6"));
+  ASSERT_OK(Put(2, "fpq7", "bar7"));
+  dbfull()->Flush(FlushOptions());
+  Iterator* iter = db_->NewIterator(read_options, handles_[2]);
+  ASSERT_EQ(CountIter(iter, "foo"), 3);
+  delete iter;
+  ASSERT_OK(
+      dbfull()->SetOptions(handles_[2], {{"prefix_extractor", "fixed:2"}}));
+  ASSERT_EQ(0,
+            strcmp(dbfull()->GetOptions(handles_[2]).prefix_extractor->Name(),
+                   "rocksdb.FixedPrefix.2"));
+  iter = db_->NewIterator(read_options, handles_[2]);
+  ASSERT_EQ(CountIter(iter, "foo"), 4);
+  delete iter;
+}
+
+// Verify it's possible to change prefix_extractor at runtime and iterators
+// behaves as expected
+TEST_F(DBTest, DynamicBloomFilterOptions) {
+  Options options;
+  options.create_if_missing = true;
+  options.prefix_extractor.reset(NewFixedPrefixTransform(1));
+  options.disable_auto_compactions = true;
+  // Enable prefix bloom for SST files
+  BlockBasedTableOptions table_options;
+  table_options.filter_policy.reset(NewBloomFilterPolicy(10, true));
+  options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+  DestroyAndReopen(options);
+
+  ASSERT_OK(Put("foo2", "bar2"));
+  ASSERT_OK(Put("foo", "bar"));
+  ASSERT_OK(Put("foo1", "bar1"));
+  ASSERT_OK(Put("fpa", "0"));
+  dbfull()->Flush(FlushOptions());
+  ASSERT_OK(Put("foo3", "bar3"));
+  ASSERT_OK(Put("foo4", "bar4"));
+  ASSERT_OK(Put("foo5", "bar5"));
+  ASSERT_OK(Put("fpb", "1"));
+  dbfull()->Flush(FlushOptions());
+  ASSERT_OK(Put("foo6", "bar6"));
+  ASSERT_OK(Put("foo7", "bar7"));
+  ASSERT_OK(Put("foo8", "bar8"));
+  ASSERT_OK(Put("fpc", "2"));
+  dbfull()->Flush(FlushOptions());
+
+  ReadOptions read_options;
+  read_options.prefix_same_as_start = true;
+  Iterator* iter = db_->NewIterator(read_options);
+  ASSERT_EQ(CountIter(iter, "foo"), 12);
+  delete iter;
+  Iterator* iter_old = db_->NewIterator(read_options);
+  ASSERT_EQ(CountIter(iter_old, "foo"), 12);
+
+  ASSERT_OK(dbfull()->SetOptions({{"prefix_extractor", "capped:3"}}));
+  ASSERT_EQ(0, strcmp(dbfull()->GetOptions().prefix_extractor->Name(),
+                      "rocksdb.CappedPrefix.3"));
+  iter = db_->NewIterator(read_options);
+  // "fp*" should be skipped
+  ASSERT_EQ(CountIter(iter, "foo"), 9);
+  delete iter;
+
+  // iterator created before should not be affected and see all keys
+  ASSERT_EQ(CountIter(iter_old, "foo"), 12);
+  delete iter_old;
+}
+
 TEST_F(DBTest, DynamicMiscOptions) {
   // Test max_sequential_skip_in_iterations
   Options options;
@@ -5445,18 +5618,17 @@ TEST_F(DBTest, HardLimit) {
 #if !defined(ROCKSDB_LITE) && !defined(ROCKSDB_DISABLE_STALL_NOTIFICATION)
 class WriteStallListener : public EventListener {
  public:
-  WriteStallListener() : cond_(&mutex_), 
-    condition_(WriteStallCondition::kNormal),
-    expected_(WriteStallCondition::kNormal),
-    expected_set_(false) 
-  {}
+  WriteStallListener()
+      : cond_(&mutex_),
+        condition_(WriteStallCondition::kNormal),
+        expected_(WriteStallCondition::kNormal),
+        expected_set_(false) {}
   void OnStallConditionsChanged(const WriteStallInfo& info) override {
     MutexLock l(&mutex_);
     condition_ = info.condition.cur;
-    if (expected_set_ && 
-      condition_ == expected_) {
-        cond_.Signal();
-        expected_set_ = false;
+    if (expected_set_ && condition_ == expected_) {
+      cond_.Signal();
+      expected_set_ = false;
     }
   }
   bool CheckCondition(WriteStallCondition expected) {

db/db_universal_compaction_test.cc

@@ -1680,50 +1680,6 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionSecondPathRatio) {
   Destroy(options);
 }
 
-TEST_P(DBTestUniversalCompaction, FullCompactionInBottomPriThreadPool) {
-  const int kNumFilesTrigger = 3;
-  Env::Default()->SetBackgroundThreads(1, Env::Priority::BOTTOM);
-  for (bool allow_ingest_behind : {false, true}) {
-    Options options = CurrentOptions();
-    options.allow_ingest_behind = allow_ingest_behind;
-    options.compaction_style = kCompactionStyleUniversal;
-    options.num_levels = num_levels_;
-    options.write_buffer_size = 100 << 10;     // 100KB
-    options.target_file_size_base = 32 << 10;  // 32KB
-    options.level0_file_num_compaction_trigger = kNumFilesTrigger;
-    // Trigger compaction if size amplification exceeds 110%
-    options.compaction_options_universal.max_size_amplification_percent = 110;
-    DestroyAndReopen(options);
-
-    int num_bottom_pri_compactions = 0;
-    SyncPoint::GetInstance()->SetCallBack(
-        "DBImpl::BGWorkBottomCompaction",
-        [&](void* /*arg*/) { ++num_bottom_pri_compactions; });
-    SyncPoint::GetInstance()->EnableProcessing();
-
-    Random rnd(301);
-    for (int num = 0; num < kNumFilesTrigger; num++) {
-      ASSERT_EQ(NumSortedRuns(), num);
-      int key_idx = 0;
-      GenerateNewFile(&rnd, &key_idx);
-    }
-    dbfull()->TEST_WaitForCompact();
-
-    if (allow_ingest_behind || num_levels_ > 1) {
-      // allow_ingest_behind increases number of levels while sanitizing.
-      ASSERT_EQ(1, num_bottom_pri_compactions);
-    } else {
-      // for single-level universal, everything's bottom level so nothing should
-      // be executed in bottom-pri thread pool.
-      ASSERT_EQ(0, num_bottom_pri_compactions);
-    }
-    // Verify that size amplification did occur
-    ASSERT_EQ(NumSortedRuns(), 1);
-    rocksdb::SyncPoint::GetInstance()->DisableProcessing();
-  }
-  Env::Default()->SetBackgroundThreads(0, Env::Priority::BOTTOM);
-}
-
 TEST_P(DBTestUniversalCompaction, ConcurrentBottomPriLowPriCompactions) {
   if (num_levels_ == 1) {
     // for single-level universal, everything's bottom level so nothing should

db/db_write_test.cc

@@ -50,6 +50,7 @@ TEST_P(DBWriteTest, IOErrorOnWALWritePropagateToWriteThreadFollower) {
   std::atomic<int> leader_count{0};
   std::vector<port::Thread> threads;
   mock_env->SetFilesystemActive(false);
+
   // Wait until all threads linked to write threads, to make sure
   // all threads join the same batch group.
   SyncPoint::GetInstance()->SetCallBack(
@@ -68,7 +69,19 @@ TEST_P(DBWriteTest, IOErrorOnWALWritePropagateToWriteThreadFollower) {
     threads.push_back(port::Thread(
         [&](int index) {
           // All threads should fail.
-          ASSERT_FALSE(Put("key" + ToString(index), "value").ok());
+          auto res = Put("key" + ToString(index), "value");
+          if (options.manual_wal_flush) {
+            ASSERT_TRUE(res.ok());
+            // we should see fs error when we do the flush
+
+            // TSAN reports a false alarm for lock-order-inversion but Open and
+            // FlushWAL are not run concurrently. Disabling this until TSAN is
+            // fixed.
+            // res = dbfull()->FlushWAL(false);
+            // ASSERT_FALSE(res.ok());
+          } else {
+            ASSERT_FALSE(res.ok());
+          }
         },
         i));
   }
@@ -80,6 +93,22 @@ TEST_P(DBWriteTest, IOErrorOnWALWritePropagateToWriteThreadFollower) {
   Close();
 }
 
+TEST_P(DBWriteTest, ManualWalFlushInEffect) {
+  Options options = GetOptions();
+  Reopen(options);
+  // try the 1st WAL created during open
+  ASSERT_TRUE(Put("key" + ToString(0), "value").ok());
+  ASSERT_TRUE(options.manual_wal_flush != dbfull()->TEST_WALBufferIsEmpty());
+  ASSERT_TRUE(dbfull()->FlushWAL(false).ok());
+  ASSERT_TRUE(dbfull()->TEST_WALBufferIsEmpty());
+  // try the 2nd wal created during SwitchWAL
+  dbfull()->TEST_SwitchWAL();
+  ASSERT_TRUE(Put("key" + ToString(0), "value").ok());
+  ASSERT_TRUE(options.manual_wal_flush != dbfull()->TEST_WALBufferIsEmpty());
+  ASSERT_TRUE(dbfull()->FlushWAL(false).ok());
+  ASSERT_TRUE(dbfull()->TEST_WALBufferIsEmpty());
+}
+
 TEST_P(DBWriteTest, IOErrorOnWALWriteTriggersReadOnlyMode) {
   std::unique_ptr<FaultInjectionTestEnv> mock_env(
       new FaultInjectionTestEnv(Env::Default()));
@@ -90,7 +119,22 @@ TEST_P(DBWriteTest, IOErrorOnWALWriteTriggersReadOnlyMode) {
     // Forcibly fail WAL write for the first Put only. Subsequent Puts should
     // fail due to read-only mode
     mock_env->SetFilesystemActive(i != 0);
-    ASSERT_FALSE(Put("key" + ToString(i), "value").ok());
+    auto res = Put("key" + ToString(i), "value");
+    // TSAN reports a false alarm for lock-order-inversion but Open and
+    // FlushWAL are not run concurrently. Disabling this until TSAN is
+    // fixed.
+    /*
+    if (options.manual_wal_flush && i == 0) {
+      // even with manual_wal_flush the 2nd Put should return error because of
+      // the read-only mode
+      ASSERT_TRUE(res.ok());
+      // we should see fs error when we do the flush
+      res = dbfull()->FlushWAL(false);
+    }
+    */
+    if (!options.manual_wal_flush) {
+      ASSERT_FALSE(res.ok());
+    }
   }
   // Close before mock_env destruct.
   Close();

db/external_sst_file_ingestion_job.cc

@@ -29,13 +29,13 @@
 namespace rocksdb {
 
 Status ExternalSstFileIngestionJob::Prepare(
-    const std::vector<std::string>& external_files_paths) {
+    const std::vector<std::string>& external_files_paths, SuperVersion* sv) {
   Status status;
 
   // Read the information of files we are ingesting
   for (const std::string& file_path : external_files_paths) {
     IngestedFileInfo file_to_ingest;
-    status = GetIngestedFileInfo(file_path, &file_to_ingest);
+    status = GetIngestedFileInfo(file_path, &file_to_ingest, sv);
     if (!status.ok()) {
       return status;
     }
@@ -284,7 +284,8 @@ void ExternalSstFileIngestionJob::Cleanup(const Status& status) {
 }
 
 Status ExternalSstFileIngestionJob::GetIngestedFileInfo(
-    const std::string& external_file, IngestedFileInfo* file_to_ingest) {
+    const std::string& external_file, IngestedFileInfo* file_to_ingest,
+    SuperVersion* sv) {
   file_to_ingest->external_file_path = external_file;
 
   // Get external file size
@@ -306,8 +307,9 @@ Status ExternalSstFileIngestionJob::GetIngestedFileInfo(
                                                    external_file));
 
   status = cfd_->ioptions()->table_factory->NewTableReader(
-      TableReaderOptions(*cfd_->ioptions(), env_options_,
-                         cfd_->internal_comparator()),
+      TableReaderOptions(*cfd_->ioptions(),
+                         sv->mutable_cf_options.prefix_extractor.get(),
+                         env_options_, cfd_->internal_comparator()),
       std::move(sst_file_reader), file_to_ingest->file_size, &table_reader);
   if (!status.ok()) {
     return status;
@@ -363,7 +365,8 @@ Status ExternalSstFileIngestionJob::GetIngestedFileInfo(
   // We need to disable fill_cache so that we read from the file without
   // updating the block cache.
   ro.fill_cache = false;
-  std::unique_ptr<InternalIterator> iter(table_reader->NewIterator(ro));
+  std::unique_ptr<InternalIterator> iter(table_reader->NewIterator(
+      ro, sv->mutable_cf_options.prefix_extractor.get()));
 
   // Get first (smallest) key from file
   iter->SeekToFirst();

db/external_sst_file_ingestion_job.h

@@ -86,7 +86,8 @@ class ExternalSstFileIngestionJob {
         job_start_time_(env_->NowMicros()) {}
 
   // Prepare the job by copying external files into the DB.
-  Status Prepare(const std::vector<std::string>& external_files_paths);
+  Status Prepare(const std::vector<std::string>& external_files_paths,
+                 SuperVersion* sv);
 
   // Check if we need to flush the memtable before running the ingestion job
   // This will be true if the files we are ingesting are overlapping with any
@@ -119,7 +120,8 @@ class ExternalSstFileIngestionJob {
   // Open the external file and populate `file_to_ingest` with all the
   // external information we need to ingest this file.
   Status GetIngestedFileInfo(const std::string& external_file,
-                             IngestedFileInfo* file_to_ingest);
+                             IngestedFileInfo* file_to_ingest,
+                             SuperVersion* sv);
 
   // Assign `file_to_ingest` the appropriate sequence number and  the lowest
   // possible level that it can be ingested to according to compaction_style.

db/forward_iterator.cc

@@ -27,20 +27,22 @@ namespace rocksdb {
 // Usage:
 //     ForwardLevelIterator iter;
 //     iter.SetFileIndex(file_index);
-//     iter.Seek(target);
+//     iter.Seek(target); // or iter.SeekToFirst();
 //     iter.Next()
 class ForwardLevelIterator : public InternalIterator {
  public:
   ForwardLevelIterator(const ColumnFamilyData* const cfd,
                        const ReadOptions& read_options,
-                       const std::vector<FileMetaData*>& files)
+                       const std::vector<FileMetaData*>& files,
+                       const SliceTransform* prefix_extractor)
       : cfd_(cfd),
         read_options_(read_options),
         files_(files),
         valid_(false),
         file_index_(std::numeric_limits<uint32_t>::max()),
         file_iter_(nullptr),
-        pinned_iters_mgr_(nullptr) {}
+        pinned_iters_mgr_(nullptr),
+        prefix_extractor_(prefix_extractor) {}
 
   ~ForwardLevelIterator() {
     // Reset current pointer
@@ -53,11 +55,11 @@ class ForwardLevelIterator : public InternalIterator {
 
   void SetFileIndex(uint32_t file_index) {
     assert(file_index < files_.size());
+    status_ = Status::OK();
     if (file_index != file_index_) {
       file_index_ = file_index;
       Reset();
     }
-    valid_ = false;
   }
   void Reset() {
     assert(file_index_ < files_.size());
@@ -75,12 +77,12 @@ class ForwardLevelIterator : public InternalIterator {
         read_options_, *(cfd_->soptions()), cfd_->internal_comparator(),
         files_[file_index_]->fd,
         read_options_.ignore_range_deletions ? nullptr : &range_del_agg,
-        nullptr /* table_reader_ptr */, nullptr, false);
+        prefix_extractor_, nullptr /* table_reader_ptr */, nullptr, false);
     file_iter_->SetPinnedItersMgr(pinned_iters_mgr_);
+    valid_ = false;
     if (!range_del_agg.IsEmpty()) {
       status_ = Status::NotSupported(
           "Range tombstones unsupported with ForwardIterator");
-      valid_ = false;
     }
   }
   void SeekToLast() override {
@@ -95,12 +97,27 @@ class ForwardLevelIterator : public InternalIterator {
     return valid_;
   }
   void SeekToFirst() override {
-    SetFileIndex(0);
+    assert(file_iter_ != nullptr);
+    if (!status_.ok()) {
+      assert(!valid_);
+      return;
+    }
     file_iter_->SeekToFirst();
     valid_ = file_iter_->Valid();
   }
   void Seek(const Slice& internal_key) override {
     assert(file_iter_ != nullptr);
+
+    // This deviates from the usual convention for InternalIterator::Seek() in
+    // that it doesn't discard pre-existing error status. That's because this
+    // Seek() is only supposed to be called immediately after SetFileIndex()
+    // (which discards pre-existing error status), and SetFileIndex() may set
+    // an error status, which we shouldn't discard.
+    if (!status_.ok()) {
+      assert(!valid_);
+      return;
+    }
+
     file_iter_->Seek(internal_key);
     valid_ = file_iter_->Valid();
   }
@@ -112,8 +129,12 @@ class ForwardLevelIterator : public InternalIterator {
     assert(valid_);
     file_iter_->Next();
     for (;;) {
-      if (file_iter_->status().IsIncomplete() || file_iter_->Valid()) {
-        valid_ = !file_iter_->status().IsIncomplete();
+      valid_ = file_iter_->Valid();
+      if (!file_iter_->status().ok()) {
+        assert(!valid_);
+        return;
+      }
+      if (valid_) {
         return;
       }
       if (file_index_ + 1 >= files_.size()) {
@@ -121,6 +142,10 @@ class ForwardLevelIterator : public InternalIterator {
         return;
       }
       SetFileIndex(file_index_ + 1);
+      if (!status_.ok()) {
+        assert(!valid_);
+        return;
+      }
       file_iter_->SeekToFirst();
     }
   }
@@ -135,7 +160,7 @@ class ForwardLevelIterator : public InternalIterator {
   Status status() const override {
     if (!status_.ok()) {
       return status_;
-    } else if (file_iter_ && !file_iter_->status().ok()) {
+    } else if (file_iter_) {
       return file_iter_->status();
     }
     return Status::OK();
@@ -165,6 +190,7 @@ class ForwardLevelIterator : public InternalIterator {
   Status status_;
   InternalIterator* file_iter_;
   PinnedIteratorsManager* pinned_iters_mgr_;
+  const SliceTransform* prefix_extractor_;
 };
 
 ForwardIterator::ForwardIterator(DBImpl* db, const ReadOptions& read_options,
@@ -173,7 +199,7 @@ ForwardIterator::ForwardIterator(DBImpl* db, const ReadOptions& read_options,
     : db_(db),
       read_options_(read_options),
       cfd_(cfd),
-      prefix_extractor_(cfd->ioptions()->prefix_extractor),
+      prefix_extractor_(current_sv->mutable_cf_options.prefix_extractor.get()),
       user_comparator_(cfd->user_comparator()),
       immutable_min_heap_(MinIterComparator(&cfd_->internal_comparator())),
       sv_(current_sv),
@@ -299,9 +325,6 @@ bool ForwardIterator::IsOverUpperBound(const Slice& internal_key) const {
 }
 
 void ForwardIterator::Seek(const Slice& internal_key) {
-  if (IsOverUpperBound(internal_key)) {
-    valid_ = false;
-  }
   if (sv_ == nullptr) {
     RebuildIterators(true);
   } else if (sv_->version_number != cfd_->GetSuperVersionNumber()) {
@@ -605,13 +628,16 @@ void ForwardIterator::RebuildIterators(bool refresh_sv) {
     if ((read_options_.iterate_upper_bound != nullptr) &&
         cfd_->internal_comparator().user_comparator()->Compare(
             l0->smallest.user_key(), *read_options_.iterate_upper_bound) > 0) {
-      has_iter_trimmed_for_upper_bound_ = true;
+      // No need to set has_iter_trimmed_for_upper_bound_: this ForwardIterator
+      // will never be interested in files with smallest key above
+      // iterate_upper_bound, since iterate_upper_bound can't be changed.
       l0_iters_.push_back(nullptr);
       continue;
     }
     l0_iters_.push_back(cfd_->table_cache()->NewIterator(
         read_options_, *cfd_->soptions(), cfd_->internal_comparator(), l0->fd,
-        read_options_.ignore_range_deletions ? nullptr : &range_del_agg));
+        read_options_.ignore_range_deletions ? nullptr : &range_del_agg,
+        sv_->mutable_cf_options.prefix_extractor.get()));
   }
   BuildLevelIterators(vstorage);
   current_ = nullptr;
@@ -681,7 +707,8 @@ void ForwardIterator::RenewIterators() {
     l0_iters_new.push_back(cfd_->table_cache()->NewIterator(
         read_options_, *cfd_->soptions(), cfd_->internal_comparator(),
         l0_files_new[inew]->fd,
-        read_options_.ignore_range_deletions ? nullptr : &range_del_agg));
+        read_options_.ignore_range_deletions ? nullptr : &range_del_agg,
+        svnew->mutable_cf_options.prefix_extractor.get()));
   }
 
   for (auto* f : l0_iters_) {
@@ -722,8 +749,9 @@ void ForwardIterator::BuildLevelIterators(const VersionStorageInfo* vstorage) {
         has_iter_trimmed_for_upper_bound_ = true;
       }
     } else {
-      level_iters_.push_back(
-          new ForwardLevelIterator(cfd_, read_options_, level_files));
+      level_iters_.push_back(new ForwardLevelIterator(
+          cfd_, read_options_, level_files,
+          sv_->mutable_cf_options.prefix_extractor.get()));
     }
   }
 }
@@ -738,7 +766,8 @@ void ForwardIterator::ResetIncompleteIterators() {
     DeleteIterator(l0_iters_[i]);
     l0_iters_[i] = cfd_->table_cache()->NewIterator(
         read_options_, *cfd_->soptions(), cfd_->internal_comparator(),
-        l0_files[i]->fd, nullptr /* range_del_agg */);
+        l0_files[i]->fd, nullptr /* range_del_agg */,
+        sv_->mutable_cf_options.prefix_extractor.get());
     l0_iters_[i]->SetPinnedItersMgr(pinned_iters_mgr_);
   }
 
@@ -773,7 +802,7 @@ void ForwardIterator::UpdateCurrent() {
       current_ = mutable_iter_;
     }
   }
-  valid_ = (current_ != nullptr);
+  valid_ = current_ != nullptr && immutable_status_.ok();
   if (!status_.ok()) {
     status_ = Status::OK();
   }

db/log_writer.cc

@@ -92,6 +92,8 @@ Status Writer::AddRecord(const Slice& slice) {
   return s;
 }
 
+bool Writer::TEST_BufferIsEmpty() { return dest_->TEST_BufferIsEmpty(); }
+
 Status Writer::EmitPhysicalRecord(RecordType t, const char* ptr, size_t n) {
   assert(n <= 0xffff);  // Must fit in two bytes
 

db/log_writer.h

@@ -85,6 +85,8 @@ class Writer {
 
   Status WriteBuffer();
 
+  bool TEST_BufferIsEmpty();
+
  private:
   unique_ptr<WritableFileWriter> dest_;
   size_t block_offset_;       // Current offset in block

db/managed_iterator.cc

@@ -101,9 +101,7 @@ void ManagedIterator::SeekToLast() {
   }
   assert(mutable_iter_ != nullptr);
   mutable_iter_->SeekToLast();
-  if (mutable_iter_->status().ok()) {
-    UpdateCurrent();
-  }
+  UpdateCurrent();
 }
 
 void ManagedIterator::SeekToFirst() {
@@ -146,27 +144,13 @@ void ManagedIterator::Prev() {
   }
   MILock l(&in_use_, this);
   if (NeedToRebuild()) {
-    std::string current_key = key().ToString();
-    Slice old_key(current_key);
-    RebuildIterator();
-    SeekInternal(old_key, false);
-    UpdateCurrent();
+    RebuildIterator(true);
     if (!valid_) {
       return;
     }
-    if (key().compare(old_key) != 0) {
-      valid_ = false;
-      status_ = Status::Incomplete("Cannot do Prev now");
-      return;
-    }
   }
   mutable_iter_->Prev();
-  if (mutable_iter_->status().ok()) {
-    UpdateCurrent();
-    status_ = Status::OK();
-  } else {
-    status_ = mutable_iter_->status();
-  }
+  UpdateCurrent();
 }
 
 void ManagedIterator::Next() {
@@ -176,19 +160,10 @@ void ManagedIterator::Next() {
   }
   MILock l(&in_use_, this);
   if (NeedToRebuild()) {
-    std::string current_key = key().ToString();
-    Slice old_key(current_key.data(), cached_key_.Size());
-    RebuildIterator();
-    SeekInternal(old_key, false);
-    UpdateCurrent();
+    RebuildIterator(true);
     if (!valid_) {
       return;
     }
-    if (key().compare(old_key) != 0) {
-      valid_ = false;
-      status_ = Status::Incomplete("Cannot do Next now");
-      return;
-    }
   }
   mutable_iter_->Next();
   UpdateCurrent();
@@ -206,21 +181,38 @@ Slice ManagedIterator::value() const {
 
 Status ManagedIterator::status() const { return status_; }
 
-void ManagedIterator::RebuildIterator() {
+void ManagedIterator::RebuildIterator(bool reseek) {
+  std::string current_key;
+  if (reseek) {
+    current_key = key().ToString();
+  }
+
   svnum_ = cfd_->GetSuperVersionNumber();
   mutable_iter_ = unique_ptr<Iterator>(db_->NewIterator(read_options_, &cfh_));
+
+  if (reseek) {
+    Slice old_key(current_key.data(), current_key.size());
+    SeekInternal(old_key, false);
+    UpdateCurrent();
+    if (!valid_ || key().compare(old_key) != 0) {
+      valid_ = false;
+      status_ = Status::Incomplete(
+          "Next/Prev failed because current key has "
+          "been removed");
+    }
+  }
 }
 
 void ManagedIterator::UpdateCurrent() {
   assert(mutable_iter_ != nullptr);
 
   valid_ = mutable_iter_->Valid();
+  status_ = mutable_iter_->status();
+
   if (!valid_) {
-    status_ = mutable_iter_->status();
     return;
   }
 
-  status_ = Status::OK();
   cached_key_.SetUserKey(mutable_iter_->key());
   cached_value_.SetUserKey(mutable_iter_->value());
 }

db/managed_iterator.h

@@ -54,7 +54,7 @@ class ManagedIterator : public Iterator {
   }
 
  private:
-  void RebuildIterator();
+  void RebuildIterator(bool reseek = false);
   void UpdateCurrent();
   void SeekInternal(const Slice& user_key, bool seek_to_first);
   bool NeedToRebuild();

db/memtable.cc

@@ -74,8 +74,8 @@ MemTable::MemTable(const InternalKeyComparator& cmp,
               : nullptr,
           mutable_cf_options.memtable_huge_page_size),
       table_(ioptions.memtable_factory->CreateMemTableRep(
-          comparator_, &arena_, ioptions.prefix_extractor, ioptions.info_log,
-          column_family_id)),
+          comparator_, &arena_, mutable_cf_options.prefix_extractor.get(),
+          ioptions.info_log, column_family_id)),
       range_del_table_(SkipListFactory().CreateMemTableRep(
           comparator_, &arena_, nullptr /* transform */, ioptions.info_log,
           column_family_id)),
@@ -95,7 +95,7 @@ MemTable::MemTable(const InternalKeyComparator& cmp,
       locks_(moptions_.inplace_update_support
                  ? moptions_.inplace_update_num_locks
                  : 0),
-      prefix_extractor_(ioptions.prefix_extractor),
+      prefix_extractor_(mutable_cf_options.prefix_extractor.get()),
       flush_state_(FLUSH_NOT_REQUESTED),
       env_(ioptions.env),
       insert_with_hint_prefix_extractor_(

db/plain_table_db_test.cc

@@ -262,11 +262,13 @@ class TestPlainTableReader : public PlainTableReader {
                        const TableProperties* table_properties,
                        unique_ptr<RandomAccessFileReader>&& file,
                        const ImmutableCFOptions& ioptions,
+                       const SliceTransform* prefix_extractor,
                        bool* expect_bloom_not_match, bool store_index_in_file,
                        uint32_t column_family_id,
                        const std::string& column_family_name)
       : PlainTableReader(ioptions, std::move(file), env_options, icomparator,
-                         encoding_type, file_size, table_properties),
+                         encoding_type, file_size, table_properties,
+                         prefix_extractor),
         expect_bloom_not_match_(expect_bloom_not_match) {
     Status s = MmapDataIfNeeded();
     EXPECT_TRUE(s.ok());
@@ -360,7 +362,8 @@ class TestPlainTableFactory : public PlainTableFactory {
         table_reader_options.env_options,
         table_reader_options.internal_comparator, encoding_type, file_size,
         bloom_bits_per_key_, hash_table_ratio_, index_sparseness_, props,
-        std::move(file), table_reader_options.ioptions, expect_bloom_not_match_,
+        std::move(file), table_reader_options.ioptions,
+        table_reader_options.prefix_extractor, expect_bloom_not_match_,
         store_index_in_file_, column_family_id_, column_family_name_));
 
     *table = std::move(new_reader);

db/range_del_aggregator.cc

@@ -195,6 +195,10 @@ Status RangeDelAggregator::AddTombstones(
   input->SeekToFirst();
   bool first_iter = true;
   while (input->Valid()) {
+    // The tombstone map holds slices into the iterator's memory. This assert
+    // ensures pinning the iterator also pins the keys/values.
+    assert(input->IsKeyPinned() && input->IsValuePinned());
+
     if (first_iter) {
       if (rep_ == nullptr) {
         InitRep({upper_bound_});

db/repair.cc

@@ -205,7 +205,7 @@ class Repairer {
       ROCKS_LOG_WARN(db_options_.info_log,
                      "**** Repaired rocksdb %s; "
                      "recovered %" ROCKSDB_PRIszt " files; %" PRIu64
-                     "bytes. "
+                     " bytes. "
                      "Some data may have been lost. "
                      "****",
                      dbname_.c_str(), tables_.size(), bytes);
@@ -501,7 +501,8 @@ class Repairer {
     if (status.ok()) {
       InternalIterator* iter = table_cache_->NewIterator(
           ReadOptions(), env_options_, cfd->internal_comparator(), t->meta.fd,
-          nullptr /* range_del_agg */);
+          nullptr /* range_del_agg */,
+          cfd->GetLatestMutableCFOptions()->prefix_extractor.get());
       bool empty = true;
       ParsedInternalKey parsed;
       t->min_sequence = 0;

db/table_cache.cc

@@ -89,8 +89,8 @@ Status TableCache::GetTableReader(
     const InternalKeyComparator& internal_comparator, const FileDescriptor& fd,
     bool sequential_mode, size_t readahead, bool record_read_stats,
     HistogramImpl* file_read_hist, unique_ptr<TableReader>* table_reader,
-    bool skip_filters, int level, bool prefetch_index_and_filter_in_cache,
-    bool for_compaction) {
+    const SliceTransform* prefix_extractor, bool skip_filters, int level,
+    bool prefetch_index_and_filter_in_cache, bool for_compaction) {
   std::string fname =
       TableFileName(ioptions_.cf_paths, fd.GetNumber(), fd.GetPathId());
   unique_ptr<RandomAccessFile> file;
@@ -115,8 +115,8 @@ Status TableCache::GetTableReader(
             record_read_stats ? ioptions_.statistics : nullptr, SST_READ_MICROS,
             file_read_hist, ioptions_.rate_limiter, for_compaction));
     s = ioptions_.table_factory->NewTableReader(
-        TableReaderOptions(ioptions_, env_options, internal_comparator,
-                           skip_filters, level),
+        TableReaderOptions(ioptions_, prefix_extractor, env_options,
+                           internal_comparator, skip_filters, level),
         std::move(file_reader), fd.GetFileSize(), table_reader,
         prefetch_index_and_filter_in_cache);
     TEST_SYNC_POINT("TableCache::GetTableReader:0");
@@ -134,6 +134,7 @@ void TableCache::EraseHandle(const FileDescriptor& fd, Cache::Handle* handle) {
 Status TableCache::FindTable(const EnvOptions& env_options,
                              const InternalKeyComparator& internal_comparator,
                              const FileDescriptor& fd, Cache::Handle** handle,
+                             const SliceTransform* prefix_extractor,
                              const bool no_io, bool record_read_stats,
                              HistogramImpl* file_read_hist, bool skip_filters,
                              int level,
@@ -154,7 +155,8 @@ Status TableCache::FindTable(const EnvOptions& env_options,
     s = GetTableReader(env_options, internal_comparator, fd,
                        false /* sequential mode */, 0 /* readahead */,
                        record_read_stats, file_read_hist, &table_reader,
-                       skip_filters, level, prefetch_index_and_filter_in_cache);
+                       prefix_extractor, skip_filters, level,
+                       prefetch_index_and_filter_in_cache);
     if (!s.ok()) {
       assert(table_reader == nullptr);
       RecordTick(ioptions_.statistics, NO_FILE_ERRORS);
@@ -175,9 +177,9 @@ Status TableCache::FindTable(const EnvOptions& env_options,
 InternalIterator* TableCache::NewIterator(
     const ReadOptions& options, const EnvOptions& env_options,
     const InternalKeyComparator& icomparator, const FileDescriptor& fd,
-    RangeDelAggregator* range_del_agg, TableReader** table_reader_ptr,
-    HistogramImpl* file_read_hist, bool for_compaction, Arena* arena,
-    bool skip_filters, int level) {
+    RangeDelAggregator* range_del_agg, const SliceTransform* prefix_extractor,
+    TableReader** table_reader_ptr, HistogramImpl* file_read_hist,
+    bool for_compaction, Arena* arena, bool skip_filters, int level) {
   PERF_TIMER_GUARD(new_table_iterator_nanos);
 
   Status s;
@@ -210,7 +212,7 @@ InternalIterator* TableCache::NewIterator(
     s = GetTableReader(
         env_options, icomparator, fd, true /* sequential_mode */, readahead,
         !for_compaction /* record stats */, nullptr, &table_reader_unique_ptr,
-        false /* skip_filters */, level,
+        prefix_extractor, false /* skip_filters */, level,
         true /* prefetch_index_and_filter_in_cache */, for_compaction);
     if (s.ok()) {
       table_reader = table_reader_unique_ptr.release();
@@ -218,7 +220,7 @@ InternalIterator* TableCache::NewIterator(
   } else {
     table_reader = fd.table_reader;
     if (table_reader == nullptr) {
-      s = FindTable(env_options, icomparator, fd, &handle,
+      s = FindTable(env_options, icomparator, fd, &handle, prefix_extractor,
                     options.read_tier == kBlockCacheTier /* no_io */,
                     !for_compaction /* record read_stats */, file_read_hist,
                     skip_filters, level);
@@ -233,7 +235,8 @@ InternalIterator* TableCache::NewIterator(
         !options.table_filter(*table_reader->GetTableProperties())) {
       result = NewEmptyInternalIterator(arena);
     } else {
-      result = table_reader->NewIterator(options, arena, skip_filters);
+      result = table_reader->NewIterator(options, prefix_extractor, arena,
+                                         skip_filters);
     }
     if (create_new_table_reader) {
       assert(handle == nullptr);
@@ -276,12 +279,13 @@ InternalIterator* TableCache::NewIterator(
 InternalIterator* TableCache::NewRangeTombstoneIterator(
     const ReadOptions& options, const EnvOptions& env_options,
     const InternalKeyComparator& icomparator, const FileDescriptor& fd,
-    HistogramImpl* file_read_hist, bool skip_filters, int level) {
+    HistogramImpl* file_read_hist, bool skip_filters, int level,
+    const SliceTransform* prefix_extractor) {
   Status s;
   Cache::Handle* handle = nullptr;
   TableReader* table_reader = fd.table_reader;
   if (table_reader == nullptr) {
-    s = FindTable(env_options, icomparator, fd, &handle,
+    s = FindTable(env_options, icomparator, fd, &handle, prefix_extractor,
                   options.read_tier == kBlockCacheTier /* no_io */,
                   true /* record read_stats */, file_read_hist, skip_filters,
                   level);
@@ -313,8 +317,10 @@ InternalIterator* TableCache::NewRangeTombstoneIterator(
 Status TableCache::Get(const ReadOptions& options,
                        const InternalKeyComparator& internal_comparator,
                        const FileDescriptor& fd, const Slice& k,
-                       GetContext* get_context, HistogramImpl* file_read_hist,
-                       bool skip_filters, int level) {
+                       GetContext* get_context,
+                       const SliceTransform* prefix_extractor,
+                       HistogramImpl* file_read_hist, bool skip_filters,
+                       int level) {
   std::string* row_cache_entry = nullptr;
   bool done = false;
 #ifndef ROCKSDB_LITE
@@ -378,10 +384,10 @@ Status TableCache::Get(const ReadOptions& options,
   Cache::Handle* handle = nullptr;
   if (!done && s.ok()) {
     if (t == nullptr) {
-      s = FindTable(env_options_, internal_comparator, fd, &handle,
-                    options.read_tier == kBlockCacheTier /* no_io */,
-                    true /* record_read_stats */, file_read_hist, skip_filters,
-                    level);
+      s = FindTable(
+          env_options_, internal_comparator, fd, &handle, prefix_extractor,
+          options.read_tier == kBlockCacheTier /* no_io */,
+          true /* record_read_stats */, file_read_hist, skip_filters, level);
       if (s.ok()) {
         t = GetTableReaderFromHandle(handle);
       }
@@ -400,7 +406,7 @@ Status TableCache::Get(const ReadOptions& options,
     }
     if (s.ok()) {
       get_context->SetReplayLog(row_cache_entry);  // nullptr if no cache.
-      s = t->Get(options, k, get_context, skip_filters);
+      s = t->Get(options, k, get_context, prefix_extractor, skip_filters);
       get_context->SetReplayLog(nullptr);
     } else if (options.read_tier == kBlockCacheTier && s.IsIncomplete()) {
       // Couldn't find Table in cache but treat as kFound if no_io set
@@ -430,7 +436,8 @@ Status TableCache::Get(const ReadOptions& options,
 Status TableCache::GetTableProperties(
     const EnvOptions& env_options,
     const InternalKeyComparator& internal_comparator, const FileDescriptor& fd,
-    std::shared_ptr<const TableProperties>* properties, bool no_io) {
+    std::shared_ptr<const TableProperties>* properties,
+    const SliceTransform* prefix_extractor, bool no_io) {
   Status s;
   auto table_reader = fd.table_reader;
   // table already been pre-loaded?
@@ -441,7 +448,8 @@ Status TableCache::GetTableProperties(
   }
 
   Cache::Handle* table_handle = nullptr;
-  s = FindTable(env_options, internal_comparator, fd, &table_handle, no_io);
+  s = FindTable(env_options, internal_comparator, fd, &table_handle,
+                prefix_extractor, no_io);
   if (!s.ok()) {
     return s;
   }
@@ -454,8 +462,8 @@ Status TableCache::GetTableProperties(
 
 size_t TableCache::GetMemoryUsageByTableReader(
     const EnvOptions& env_options,
-    const InternalKeyComparator& internal_comparator,
-    const FileDescriptor& fd) {
+    const InternalKeyComparator& internal_comparator, const FileDescriptor& fd,
+    const SliceTransform* prefix_extractor) {
   Status s;
   auto table_reader = fd.table_reader;
   // table already been pre-loaded?
@@ -464,7 +472,8 @@ size_t TableCache::GetMemoryUsageByTableReader(
   }
 
   Cache::Handle* table_handle = nullptr;
-  s = FindTable(env_options, internal_comparator, fd, &table_handle, true);
+  s = FindTable(env_options, internal_comparator, fd, &table_handle,
+                prefix_extractor, true);
   if (!s.ok()) {
     return 0;
   }

db/table_cache.h

@@ -54,6 +54,7 @@ class TableCache {
       const ReadOptions& options, const EnvOptions& toptions,
       const InternalKeyComparator& internal_comparator,
       const FileDescriptor& file_fd, RangeDelAggregator* range_del_agg,
+      const SliceTransform* prefix_extractor = nullptr,
       TableReader** table_reader_ptr = nullptr,
       HistogramImpl* file_read_hist = nullptr, bool for_compaction = false,
       Arena* arena = nullptr, bool skip_filters = false, int level = -1);
@@ -62,7 +63,8 @@ class TableCache {
       const ReadOptions& options, const EnvOptions& toptions,
       const InternalKeyComparator& internal_comparator,
       const FileDescriptor& file_fd, HistogramImpl* file_read_hist,
-      bool skip_filters, int level);
+      bool skip_filters, int level,
+      const SliceTransform* prefix_extractor = nullptr);
 
   // If a seek to internal key "k" in specified file finds an entry,
   // call (*handle_result)(arg, found_key, found_value) repeatedly until
@@ -75,8 +77,10 @@ class TableCache {
   Status Get(const ReadOptions& options,
              const InternalKeyComparator& internal_comparator,
              const FileDescriptor& file_fd, const Slice& k,
-             GetContext* get_context, HistogramImpl* file_read_hist = nullptr,
-             bool skip_filters = false, int level = -1);
+             GetContext* get_context,
+             const SliceTransform* prefix_extractor = nullptr,
+             HistogramImpl* file_read_hist = nullptr, bool skip_filters = false,
+             int level = -1);
 
   // Evict any entry for the specified file number
   static void Evict(Cache* cache, uint64_t file_number);
@@ -91,6 +95,7 @@ class TableCache {
   Status FindTable(const EnvOptions& toptions,
                    const InternalKeyComparator& internal_comparator,
                    const FileDescriptor& file_fd, Cache::Handle**,
+                   const SliceTransform* prefix_extractor = nullptr,
                    const bool no_io = false, bool record_read_stats = true,
                    HistogramImpl* file_read_hist = nullptr,
                    bool skip_filters = false, int level = -1,
@@ -109,6 +114,7 @@ class TableCache {
                             const InternalKeyComparator& internal_comparator,
                             const FileDescriptor& file_meta,
                             std::shared_ptr<const TableProperties>* properties,
+                            const SliceTransform* prefix_extractor = nullptr,
                             bool no_io = false);
 
   // Return total memory usage of the table reader of the file.
@@ -116,7 +122,8 @@ class TableCache {
   size_t GetMemoryUsageByTableReader(
       const EnvOptions& toptions,
       const InternalKeyComparator& internal_comparator,
-      const FileDescriptor& fd);
+      const FileDescriptor& fd,
+      const SliceTransform* prefix_extractor = nullptr);
 
   // Release the handle from a cache
   void ReleaseHandle(Cache::Handle* handle);
@@ -133,6 +140,7 @@ class TableCache {
                         size_t readahead, bool record_read_stats,
                         HistogramImpl* file_read_hist,
                         unique_ptr<TableReader>* table_reader,
+                        const SliceTransform* prefix_extractor = nullptr,
                         bool skip_filters = false, int level = -1,
                         bool prefetch_index_and_filter_in_cache = true,
                         bool for_compaction = false);

db/table_properties_collector_test.cc

@@ -39,6 +39,7 @@ static const uint32_t kTestColumnFamilyId = 66;
 static const std::string kTestColumnFamilyName = "test_column_fam";
 
 void MakeBuilder(const Options& options, const ImmutableCFOptions& ioptions,
+                 const MutableCFOptions& moptions,
                  const InternalKeyComparator& internal_comparator,
                  const std::vector<std::unique_ptr<IntTblPropCollectorFactory>>*
                      int_tbl_prop_collector_factories,
@@ -48,10 +49,9 @@ void MakeBuilder(const Options& options, const ImmutableCFOptions& ioptions,
   writable->reset(new WritableFileWriter(std::move(wf), EnvOptions()));
   int unknown_level = -1;
   builder->reset(NewTableBuilder(
-      ioptions, internal_comparator, int_tbl_prop_collector_factories,
-      kTestColumnFamilyId, kTestColumnFamilyName,
-      writable->get(), options.compression, options.compression_opts,
-      unknown_level));
+      ioptions, moptions, internal_comparator, int_tbl_prop_collector_factories,
+      kTestColumnFamilyId, kTestColumnFamilyName, writable->get(),
+      options.compression, options.compression_opts, unknown_level));
 }
 }  // namespace
 
@@ -251,6 +251,7 @@ void TestCustomizedTablePropertiesCollector(
   std::unique_ptr<TableBuilder> builder;
   std::unique_ptr<WritableFileWriter> writer;
   const ImmutableCFOptions ioptions(options);
+  const MutableCFOptions moptions(options);
   std::vector<std::unique_ptr<IntTblPropCollectorFactory>>
       int_tbl_prop_collector_factories;
   if (test_int_tbl_prop_collector) {
@@ -259,7 +260,7 @@ void TestCustomizedTablePropertiesCollector(
   } else {
     GetIntTblPropCollectorFactory(ioptions, &int_tbl_prop_collector_factories);
   }
-  MakeBuilder(options, ioptions, internal_comparator,
+  MakeBuilder(options, ioptions, moptions, internal_comparator,
               &int_tbl_prop_collector_factories, &writer, &builder);
 
   SequenceNumber seqNum = 0U;
@@ -401,10 +402,11 @@ void TestInternalKeyPropertiesCollector(
         new InternalKeyPropertiesCollectorFactory);
   }
   const ImmutableCFOptions ioptions(options);
+  MutableCFOptions moptions(options);
 
   for (int iter = 0; iter < 2; ++iter) {
-    MakeBuilder(options, ioptions, pikc, &int_tbl_prop_collector_factories,
-                &writable, &builder);
+    MakeBuilder(options, ioptions, moptions, pikc,
+                &int_tbl_prop_collector_factories, &writable, &builder);
     for (const auto& k : keys) {
       builder->Add(k.Encode(), "val");
     }

db/version_builder.cc

@@ -368,7 +368,8 @@ class VersionBuilder::Rep {
   }
 
   void LoadTableHandlers(InternalStats* internal_stats, int max_threads,
-                         bool prefetch_index_and_filter_in_cache) {
+                         bool prefetch_index_and_filter_in_cache,
+                         const SliceTransform* prefix_extractor) {
     assert(table_cache_ != nullptr);
     // <file metadata, level>
     std::vector<std::pair<FileMetaData*, int>> files_meta;
@@ -390,12 +391,12 @@ class VersionBuilder::Rep {
 
         auto* file_meta = files_meta[file_idx].first;
         int level = files_meta[file_idx].second;
-        table_cache_->FindTable(env_options_,
-                                *(base_vstorage_->InternalComparator()),
-                                file_meta->fd, &file_meta->table_reader_handle,
-                                false /*no_io */, true /* record_read_stats */,
-                                internal_stats->GetFileReadHist(level), false,
-                                level, prefetch_index_and_filter_in_cache);
+        table_cache_->FindTable(
+            env_options_, *(base_vstorage_->InternalComparator()),
+            file_meta->fd, &file_meta->table_reader_handle, prefix_extractor,
+            false /*no_io */, true /* record_read_stats */,
+            internal_stats->GetFileReadHist(level), false, level,
+            prefetch_index_and_filter_in_cache);
         if (file_meta->table_reader_handle != nullptr) {
           // Load table_reader
           file_meta->fd.table_reader = table_cache_->GetTableReaderFromHandle(
@@ -455,11 +456,12 @@ void VersionBuilder::SaveTo(VersionStorageInfo* vstorage) {
   rep_->SaveTo(vstorage);
 }
 
-void VersionBuilder::LoadTableHandlers(
-    InternalStats* internal_stats, int max_threads,
-    bool prefetch_index_and_filter_in_cache) {
+void VersionBuilder::LoadTableHandlers(InternalStats* internal_stats,
+                                       int max_threads,
+                                       bool prefetch_index_and_filter_in_cache,
+                                       const SliceTransform* prefix_extractor) {
   rep_->LoadTableHandlers(internal_stats, max_threads,
-                          prefetch_index_and_filter_in_cache);
+                          prefetch_index_and_filter_in_cache, prefix_extractor);
 }
 
 void VersionBuilder::MaybeAddFile(VersionStorageInfo* vstorage, int level,

db/version_builder.h

@@ -9,6 +9,7 @@
 //
 #pragma once
 #include "rocksdb/env.h"
+#include "rocksdb/slice_transform.h"
 
 namespace rocksdb {
 
@@ -33,7 +34,8 @@ class VersionBuilder {
   void Apply(VersionEdit* edit);
   void SaveTo(VersionStorageInfo* vstorage);
   void LoadTableHandlers(InternalStats* internal_stats, int max_threads,
-                         bool prefetch_index_and_filter_in_cache);
+                         bool prefetch_index_and_filter_in_cache,
+                         const SliceTransform* prefix_extractor);
   void MaybeAddFile(VersionStorageInfo* vstorage, int level, FileMetaData* f);
 
  private:

db/version_edit.cc

@@ -20,7 +20,7 @@ namespace rocksdb {
 
 // Tag numbers for serialized VersionEdit.  These numbers are written to
 // disk and should not be changed.
-enum Tag {
+enum Tag : uint32_t {
   kComparator = 1,
   kLogNumber = 2,
   kNextFileNumber = 3,
@@ -42,7 +42,7 @@ enum Tag {
   kMaxColumnFamily = 203,
 };
 
-enum CustomTag {
+enum CustomTag : uint32_t {
   kTerminate = 1,  // The end of customized fields
   kNeedCompaction = 2,
   // Since Manifest is not entirely currently forward-compatible, and the only

db/version_set.cc

@@ -463,7 +463,8 @@ class LevelIterator final : public InternalIterator {
   LevelIterator(TableCache* table_cache, const ReadOptions& read_options,
                 const EnvOptions& env_options,
                 const InternalKeyComparator& icomparator,
-                const LevelFilesBrief* flevel, bool should_sample,
+                const LevelFilesBrief* flevel,
+                const SliceTransform* prefix_extractor, bool should_sample,
                 HistogramImpl* file_read_hist, bool for_compaction,
                 bool skip_filters, int level, RangeDelAggregator* range_del_agg)
       : table_cache_(table_cache),
@@ -471,6 +472,7 @@ class LevelIterator final : public InternalIterator {
         env_options_(env_options),
         icomparator_(icomparator),
         flevel_(flevel),
+        prefix_extractor_(prefix_extractor),
         file_read_hist_(file_read_hist),
         should_sample_(should_sample),
         for_compaction_(for_compaction),
@@ -502,13 +504,7 @@ class LevelIterator final : public InternalIterator {
     return file_iter_.value();
   }
   virtual Status status() const override {
-    // It'd be nice if status() returned a const Status& instead of a Status
-    if (!status_.ok()) {
-      return status_;
-    } else if (file_iter_.iter() != nullptr) {
-      return file_iter_.status();
-    }
-    return Status::OK();
+    return file_iter_.iter() ? file_iter_.status() : Status::OK();
   }
   virtual void SetPinnedItersMgr(
       PinnedIteratorsManager* pinned_iters_mgr) override {
@@ -553,8 +549,9 @@ class LevelIterator final : public InternalIterator {
 
     return table_cache_->NewIterator(
         read_options_, env_options_, icomparator_, file_meta.fd, range_del_agg_,
-        nullptr /* don't need reference to table */, file_read_hist_,
-        for_compaction_, nullptr /* arena */, skip_filters_, level_);
+        prefix_extractor_, nullptr /* don't need reference to table */,
+        file_read_hist_, for_compaction_, nullptr /* arena */, skip_filters_,
+        level_);
   }
 
   TableCache* table_cache_;
@@ -563,6 +560,7 @@ class LevelIterator final : public InternalIterator {
   const InternalKeyComparator& icomparator_;
   const LevelFilesBrief* flevel_;
   mutable FileDescriptor current_value_;
+  const SliceTransform* prefix_extractor_;
 
   HistogramImpl* file_read_hist_;
   bool should_sample_;
@@ -573,7 +571,6 @@ class LevelIterator final : public InternalIterator {
   RangeDelAggregator* range_del_agg_;
   IteratorWrapper file_iter_;  // May be nullptr
   PinnedIteratorsManager* pinned_iters_mgr_;
-  Status status_;
 };
 
 void LevelIterator::Seek(const Slice& target) {
@@ -628,16 +625,9 @@ void LevelIterator::Prev() {
 }
 
 void LevelIterator::SkipEmptyFileForward() {
-  // For an error (IO error, checksum mismatch, etc), we skip the file
-  // and move to the next one and continue reading data.
-  // TODO this behavior is from LevelDB. We should revisit it.
   while (file_iter_.iter() == nullptr ||
-         (!file_iter_.Valid() && !file_iter_.status().IsIncomplete())) {
-    if (file_iter_.iter() != nullptr && !file_iter_.Valid() &&
-        file_iter_.iter()->IsOutOfBound()) {
-      return;
-    }
-
+         (!file_iter_.Valid() && file_iter_.status().ok() &&
+          !file_iter_.iter()->IsOutOfBound())) {
     // Move to next file
     if (file_index_ >= flevel_->num_files - 1) {
       // Already at the last file
@@ -657,7 +647,7 @@ void LevelIterator::SkipEmptyFileForward() {
 
 void LevelIterator::SkipEmptyFileBackward() {
   while (file_iter_.iter() == nullptr ||
-         (!file_iter_.Valid() && !file_iter_.status().IsIncomplete())) {
+         (!file_iter_.Valid() && file_iter_.status().ok())) {
     // Move to previous file
     if (file_index_ == 0) {
       // Already the first file
@@ -672,13 +662,6 @@ void LevelIterator::SkipEmptyFileBackward() {
 }
 
 void LevelIterator::SetFileIterator(InternalIterator* iter) {
-  if (file_iter_.iter() != nullptr && status_.ok()) {
-    // TODO right now we don't invalidate the iterator even if the status is
-    // not OK. We should consider to do that so that it is harder for users to
-    // skip errors.
-    status_ = file_iter_.status();
-  }
-
   if (pinned_iters_mgr_ && iter) {
     iter->SetPinnedItersMgr(pinned_iters_mgr_);
   }
@@ -743,8 +726,8 @@ Status Version::GetTableProperties(std::shared_ptr<const TableProperties>* tp,
   auto table_cache = cfd_->table_cache();
   auto ioptions = cfd_->ioptions();
   Status s = table_cache->GetTableProperties(
-      env_options_, cfd_->internal_comparator(), file_meta->fd,
-      tp, true /* no io */);
+      env_options_, cfd_->internal_comparator(), file_meta->fd, tp,
+      mutable_cf_options_.prefix_extractor.get(), true /* no io */);
   if (s.ok()) {
     return s;
   }
@@ -878,8 +861,8 @@ size_t Version::GetMemoryUsageByTableReaders() {
   for (auto& file_level : storage_info_.level_files_brief_) {
     for (size_t i = 0; i < file_level.num_files; i++) {
       total_usage += cfd_->table_cache()->GetMemoryUsageByTableReader(
-          env_options_, cfd_->internal_comparator(),
-          file_level.files[i].fd);
+          env_options_, cfd_->internal_comparator(), file_level.files[i].fd,
+          mutable_cf_options_.prefix_extractor.get());
     }
   }
   return total_usage;
@@ -1017,8 +1000,9 @@ void Version::AddIteratorsForLevel(const ReadOptions& read_options,
       const auto& file = storage_info_.LevelFilesBrief(0).files[i];
       merge_iter_builder->AddIterator(cfd_->table_cache()->NewIterator(
           read_options, soptions, cfd_->internal_comparator(), file.fd,
-          range_del_agg, nullptr, cfd_->internal_stats()->GetFileReadHist(0),
-          false, arena, false /* skip_filters */, 0 /* level */));
+          range_del_agg, mutable_cf_options_.prefix_extractor.get(), nullptr,
+          cfd_->internal_stats()->GetFileReadHist(0), false, arena,
+          false /* skip_filters */, 0 /* level */));
     }
     if (should_sample) {
       // Count ones for every L0 files. This is done per iterator creation
@@ -1037,7 +1021,7 @@ void Version::AddIteratorsForLevel(const ReadOptions& read_options,
     merge_iter_builder->AddIterator(new (mem) LevelIterator(
         cfd_->table_cache(), read_options, soptions,
         cfd_->internal_comparator(), &storage_info_.LevelFilesBrief(level),
-        should_sample_file_read(),
+        mutable_cf_options_.prefix_extractor.get(), should_sample_file_read(),
         cfd_->internal_stats()->GetFileReadHist(level),
         false /* for_compaction */, IsFilterSkipped(level), level,
         range_del_agg));
@@ -1069,8 +1053,9 @@ Status Version::OverlapWithLevelIterator(const ReadOptions& read_options,
       }
       ScopedArenaIterator iter(cfd_->table_cache()->NewIterator(
           read_options, env_options, cfd_->internal_comparator(), file->fd,
-          &range_del_agg, nullptr, cfd_->internal_stats()->GetFileReadHist(0),
-          false, &arena, false /* skip_filters */, 0 /* level */));
+          &range_del_agg, mutable_cf_options_.prefix_extractor.get(), nullptr,
+          cfd_->internal_stats()->GetFileReadHist(0), false, &arena,
+          false /* skip_filters */, 0 /* level */));
       status = OverlapWithIterator(
           ucmp, smallest_user_key, largest_user_key, iter.get(), overlap);
       if (!status.ok() || *overlap) {
@@ -1082,7 +1067,7 @@ Status Version::OverlapWithLevelIterator(const ReadOptions& read_options,
     ScopedArenaIterator iter(new (mem) LevelIterator(
         cfd_->table_cache(), read_options, env_options,
         cfd_->internal_comparator(), &storage_info_.LevelFilesBrief(level),
-        should_sample_file_read(),
+        mutable_cf_options_.prefix_extractor.get(), should_sample_file_read(),
         cfd_->internal_stats()->GetFileReadHist(level),
         false /* for_compaction */, IsFilterSkipped(level), level,
         &range_del_agg));
@@ -1143,7 +1128,9 @@ VersionStorageInfo::VersionStorageInfo(
 }
 
 Version::Version(ColumnFamilyData* column_family_data, VersionSet* vset,
-                 const EnvOptions& env_opt, uint64_t version_number)
+                 const EnvOptions& env_opt,
+                 const MutableCFOptions mutable_cf_options,
+                 uint64_t version_number)
     : env_(vset->env_),
       cfd_(column_family_data),
       info_log_((cfd_ == nullptr) ? nullptr : cfd_->ioptions()->info_log),
@@ -1167,6 +1154,7 @@ Version::Version(ColumnFamilyData* column_family_data, VersionSet* vset,
       prev_(this),
       refs_(0),
       env_options_(env_opt),
+      mutable_cf_options_(mutable_cf_options),
       version_number_(version_number) {}
 
 void Version::Get(const ReadOptions& read_options, const LookupKey& k,
@@ -1210,6 +1198,7 @@ void Version::Get(const ReadOptions& read_options, const LookupKey& k,
 
     *status = table_cache_->Get(
         read_options, *internal_comparator(), f->fd, ikey, &get_context,
+        mutable_cf_options_.prefix_extractor.get(),
         cfd_->internal_stats()->GetFileReadHist(fp.GetHitFileLevel()),
         IsFilterSkipped(static_cast<int>(fp.GetHitFileLevel()),
                         fp.IsHitFileLastInLevel()),
@@ -2796,7 +2785,7 @@ Status VersionSet::LogAndApply(ColumnFamilyData* column_family_data,
     LogAndApplyCFHelper(w.edit_list.front());
     batch_edits.push_back(w.edit_list.front());
   } else {
-    v = new Version(column_family_data, this, env_options_,
+    v = new Version(column_family_data, this, env_options_, mutable_cf_options,
                     current_version_number_++);
     builder_guard.reset(new BaseReferencedVersionBuilder(column_family_data));
     auto* builder = builder_guard->version_builder();
@@ -2857,7 +2846,8 @@ Status VersionSet::LogAndApply(ColumnFamilyData* column_family_data,
       builder_guard->version_builder()->LoadTableHandlers(
           column_family_data->internal_stats(),
           column_family_data->ioptions()->optimize_filters_for_hits,
-          true /* prefetch_index_and_filter_in_cache */);
+          true /* prefetch_index_and_filter_in_cache */,
+          mutable_cf_options.prefix_extractor.get());
     }
 
     // This is fine because everything inside of this block is serialized --
@@ -3348,11 +3338,13 @@ Status VersionSet::Recover(
         // Need to do it out of the mutex.
         builder->LoadTableHandlers(
             cfd->internal_stats(), db_options_->max_file_opening_threads,
-            false /* prefetch_index_and_filter_in_cache */);
+            false /* prefetch_index_and_filter_in_cache */,
+            cfd->GetLatestMutableCFOptions()->prefix_extractor.get());
       }
 
-      Version* v =
-          new Version(cfd, this, env_options_, current_version_number_++);
+      Version* v = new Version(cfd, this, env_options_,
+                               *cfd->GetLatestMutableCFOptions(),
+                               current_version_number_++);
       builder->SaveTo(v->storage_info());
 
       // Install recovered version
@@ -3717,8 +3709,9 @@ Status VersionSet::DumpManifest(Options& options, std::string& dscname,
       assert(builders_iter != builders.end());
       auto builder = builders_iter->second->version_builder();
 
-      Version* v =
-          new Version(cfd, this, env_options_, current_version_number_++);
+      Version* v = new Version(cfd, this, env_options_,
+                               *cfd->GetLatestMutableCFOptions(),
+                               current_version_number_++);
       builder->SaveTo(v->storage_info());
       v->PrepareApply(*cfd->GetLatestMutableCFOptions(), false);
 
@@ -3941,7 +3934,8 @@ uint64_t VersionSet::ApproximateSize(Version* v, const FdWithKeyRange& f,
     TableReader* table_reader_ptr;
     InternalIterator* iter = v->cfd_->table_cache()->NewIterator(
         ReadOptions(), v->env_options_, v->cfd_->internal_comparator(), f.fd,
-        nullptr /* range_del_agg */, &table_reader_ptr);
+        nullptr /* range_del_agg */,
+        v->GetMutableCFOptions().prefix_extractor.get(), &table_reader_ptr);
     if (table_reader_ptr != nullptr) {
       result = table_reader_ptr->ApproximateOffsetOf(key);
     }
@@ -4021,6 +4015,7 @@ InternalIterator* VersionSet::MakeInputIterator(
           list[num++] = cfd->table_cache()->NewIterator(
               read_options, env_options_compactions, cfd->internal_comparator(),
               flevel->files[i].fd, range_del_agg,
+              c->mutable_cf_options()->prefix_extractor.get(),
               nullptr /* table_reader_ptr */,
               nullptr /* no per level latency histogram */,
               true /* for_compaction */, nullptr /* arena */,
@@ -4031,6 +4026,7 @@ InternalIterator* VersionSet::MakeInputIterator(
         list[num++] = new LevelIterator(
             cfd->table_cache(), read_options, env_options_compactions,
             cfd->internal_comparator(), c->input_levels(which),
+            c->mutable_cf_options()->prefix_extractor.get(),
             false /* should_sample */,
             nullptr /* no per level latency histogram */,
             true /* for_compaction */, false /* skip_filters */,
@@ -4170,7 +4166,9 @@ ColumnFamilyData* VersionSet::CreateColumnFamily(
     const ColumnFamilyOptions& cf_options, VersionEdit* edit) {
   assert(edit->is_column_family_add_);
 
-  Version* dummy_versions = new Version(nullptr, this, env_options_);
+  MutableCFOptions dummy_cf_options;
+  Version* dummy_versions =
+      new Version(nullptr, this, env_options_, dummy_cf_options);
   // Ref() dummy version once so that later we can call Unref() to delete it
   // by avoiding calling "delete" explicitly (~Version is private)
   dummy_versions->Ref();
@@ -4178,8 +4176,9 @@ ColumnFamilyData* VersionSet::CreateColumnFamily(
       edit->column_family_name_, edit->column_family_, dummy_versions,
       cf_options);
 
-  Version* v =
-      new Version(new_cfd, this, env_options_, current_version_number_++);
+  Version* v = new Version(new_cfd, this, env_options_,
+                           *new_cfd->GetLatestMutableCFOptions(),
+                           current_version_number_++);
 
   // Fill level target base information.
   v->storage_info()->CalculateBaseBytes(*new_cfd->ioptions(),

db/version_set.h

@@ -633,6 +633,8 @@ class Version {
 
   uint64_t GetSstFilesSize();
 
+  MutableCFOptions GetMutableCFOptions() { return mutable_cf_options_; }
+
  private:
   Env* env_;
   friend class VersionSet;
@@ -680,13 +682,14 @@ class Version {
   Version* prev_;               // Previous version in linked list
   int refs_;                    // Number of live refs to this version
   const EnvOptions env_options_;
+  const MutableCFOptions mutable_cf_options_;
 
   // A version number that uniquely represents this version. This is
   // used for debugging and logging purposes only.
   uint64_t version_number_;
 
   Version(ColumnFamilyData* cfd, VersionSet* vset, const EnvOptions& env_opt,
-          uint64_t version_number = 0);
+          MutableCFOptions mutable_cf_options, uint64_t version_number = 0);
 
   ~Version();
 

db/write_thread.cc

@@ -318,6 +318,7 @@ void WriteThread::JoinBatchGroup(Writer* w) {
      * 3.2) an existing memtable writer group leader tell us to finish memtable
      *      writes in parallel.
      */
+    TEST_SYNC_POINT_CALLBACK("WriteThread::JoinBatchGroup:BeganWaiting", w);
     AwaitState(w, STATE_GROUP_LEADER | STATE_MEMTABLE_WRITER_LEADER |
                       STATE_PARALLEL_MEMTABLE_WRITER | STATE_COMPLETED,
                &jbg_ctx);

include/rocksdb/advanced_options.h

@@ -87,6 +87,14 @@ struct CompactionOptionsFIFO {
 
 // Compression options for different compression algorithms like Zlib
 struct CompressionOptions {
+  // RocksDB's generic default compression level. Internally it'll be translated
+  // to the default compression level specific to the library being used (see
+  // comment above `ColumnFamilyOptions::compression`).
+  //
+  // The default value is the max 16-bit int as it'll be written out in OPTIONS
+  // file, which should be portable.
+  const static int kDefaultCompressionLevel = 32767;
+
   int window_bits;
   int level;
   int strategy;
@@ -120,7 +128,7 @@ struct CompressionOptions {
 
   CompressionOptions()
       : window_bits(-14),
-        level(-1),
+        level(kDefaultCompressionLevel),
         strategy(0),
         max_dict_bytes(0),
         zstd_max_train_bytes(0) {}

include/rocksdb/iterator.h

@@ -33,6 +33,7 @@ class Iterator : public Cleanable {
 
   // An iterator is either positioned at a key/value pair, or
   // not valid.  This method returns true iff the iterator is valid.
+  // Always returns false if !status().ok().
   virtual bool Valid() const = 0;
 
   // Position at the first key in the source.  The iterator is Valid()
@@ -46,6 +47,9 @@ class Iterator : public Cleanable {
   // Position at the first key in the source that at or past target.
   // The iterator is Valid() after this call iff the source contains
   // an entry that comes at or past target.
+  // All Seek*() methods clear any error status() that the iterator had prior to
+  // the call; after the seek, status() indicates only the error (if any) that
+  // happened during the seek, not any past errors.
   virtual void Seek(const Slice& target) = 0;
 
   // Position at the last key in the source that at or before target.
@@ -72,7 +76,7 @@ class Iterator : public Cleanable {
   // Return the value for the current entry.  The underlying storage for
   // the returned slice is valid only until the next modification of
   // the iterator.
-  // REQUIRES: !AtEnd() && !AtStart()
+  // REQUIRES: Valid()
   virtual Slice value() const = 0;
 
   // If an error has occurred, return it.  Else return an ok status.

include/rocksdb/options.h

@@ -198,11 +198,21 @@ struct ColumnFamilyOptions : public AdvancedColumnFamilyOptions {
   // Typical speeds of kSnappyCompression on an Intel(R) Core(TM)2 2.4GHz:
   //    ~200-500MB/s compression
   //    ~400-800MB/s decompression
+  //
   // Note that these speeds are significantly faster than most
   // persistent storage speeds, and therefore it is typically never
   // worth switching to kNoCompression.  Even if the input data is
   // incompressible, the kSnappyCompression implementation will
   // efficiently detect that and will switch to uncompressed mode.
+  //
+  // If you do not set `compression_opts.level`, or set it to
+  // `CompressionOptions::kDefaultCompressionLevel`, we will attempt to pick the
+  // default corresponding to `compression` as follows:
+  //
+  // - kZSTD: 3
+  // - kZlibCompression: Z_DEFAULT_COMPRESSION (currently -1)
+  // - kLZ4HCCompression: 0
+  // - For all others, we do not specify a compression level
   CompressionType compression;
 
   // Compression algorithm that will be used for the bottommost level that
@@ -271,7 +281,7 @@ struct ColumnFamilyOptions : public AdvancedColumnFamilyOptions {
   // later in the vector.
   // Note that, if a path is supplied to multiple column
   // families, it would have files and total size from all
-  // the column families combined. User should privision for the
+  // the column families combined. User should provision for the
   // total size(from all the column families) in such cases.
   //
   // If left empty, db_paths will be used.
@@ -558,8 +568,9 @@ struct DBOptions {
 
   // manifest file is rolled over on reaching this limit.
   // The older manifest file be deleted.
-  // The default value is MAX_INT so that roll-over does not take place.
-  uint64_t max_manifest_file_size = std::numeric_limits<uint64_t>::max();
+  // The default value is 1GB so that the manifest file can grow, but not
+  // reach the limit of storage capacity.
+  uint64_t max_manifest_file_size = 1024 * 1024 * 1024;
 
   // Number of shards used for table cache.
   int table_cache_numshardbits = 6;

include/rocksdb/slice.h

@@ -43,7 +43,9 @@ class Slice {
 
   // Create a slice that refers to s[0,strlen(s)-1]
   /* implicit */
-  Slice(const char* s) : data_(s), size_(strlen(s)) { }
+  Slice(const char* s) : data_(s) {
+    size_ = (s == nullptr) ? 0 : strlen(s);
+  }
 
   // Create a single slice from SliceParts using buf as storage.
   // buf must exist as long as the returned Slice exists.

include/rocksdb/statistics.h

@@ -619,6 +619,8 @@ struct HistogramData {
   // zero-initialize new members since old Statistics::histogramData()
   // implementations won't write them.
   double max = 0.0;
+  uint64_t count = 0;
+  uint64_t sum = 0;
 };
 
 enum StatsLevel {

include/rocksdb/utilities/transaction_db.h

@@ -193,6 +193,7 @@ class TransactionDB : public StackableDB {
   }
   // Open a TransactionDB similar to DB::Open().
   // Internally call PrepareWrap() and WrapDB()
+  // If the return status is not ok, then dbptr is set to nullptr.
   static Status Open(const Options& options,
                      const TransactionDBOptions& txn_db_options,
                      const std::string& dbname, TransactionDB** dbptr);
@@ -203,27 +204,29 @@ class TransactionDB : public StackableDB {
                      const std::vector<ColumnFamilyDescriptor>& column_families,
                      std::vector<ColumnFamilyHandle*>* handles,
                      TransactionDB** dbptr);
-  // The following functions are used to open a TransactionDB internally using
-  // an opened DB or StackableDB.
-  // 1. Call prepareWrap(), passing an empty std::vector<size_t> to
-  // compaction_enabled_cf_indices.
-  // 2. Open DB or Stackable DB with db_options and column_families passed to
-  // prepareWrap()
   // Note: PrepareWrap() may change parameters, make copies before the
   // invocation if needed.
-  // 3. Call Wrap*DB() with compaction_enabled_cf_indices in step 1 and handles
-  // of the opened DB/StackableDB in step 2
   static void PrepareWrap(DBOptions* db_options,
                           std::vector<ColumnFamilyDescriptor>* column_families,
                           std::vector<size_t>* compaction_enabled_cf_indices);
+  // If the return status is not ok, then dbptr will bet set to nullptr. The
+  // input db parameter might or might not be deleted as a result of the
+  // failure. If it is properly deleted it will be set to nullptr. If the return
+  // status is ok, the ownership of db is transferred to dbptr.
   static Status WrapDB(DB* db, const TransactionDBOptions& txn_db_options,
                        const std::vector<size_t>& compaction_enabled_cf_indices,
                        const std::vector<ColumnFamilyHandle*>& handles,
                        TransactionDB** dbptr);
+  // If the return status is not ok, then dbptr will bet set to nullptr. The
+  // input db parameter might or might not be deleted as a result of the
+  // failure. If it is properly deleted it will be set to nullptr. If the return
+  // status is ok, the ownership of db is transferred to dbptr.
   static Status WrapStackableDB(
       StackableDB* db, const TransactionDBOptions& txn_db_options,
       const std::vector<size_t>& compaction_enabled_cf_indices,
       const std::vector<ColumnFamilyHandle*>& handles, TransactionDB** dbptr);
+  // Since the destructor in StackableDB is virtual, this destructor is virtual
+  // too. The root db will be deleted by the base's destructor.
   ~TransactionDB() override {}
 
   // Starts a new Transaction.
@@ -252,6 +255,7 @@ class TransactionDB : public StackableDB {
 
  protected:
   // To Create an TransactionDB, call Open()
+  // The ownership of db is transferred to the base StackableDB
   explicit TransactionDB(DB* db) : StackableDB(db) {}
 
  private:

include/rocksdb/version.h

@@ -5,7 +5,7 @@
 #pragma once
 
 #define ROCKSDB_MAJOR 5
-#define ROCKSDB_MINOR 13
+#define ROCKSDB_MINOR 14
 #define ROCKSDB_PATCH 0
 
 // Do not use these. We made the mistake of declaring macros starting with

java/rocksjni/transaction_db.cc

@@ -307,6 +307,7 @@ jobject Java_org_rocksdb_TransactionDB_getLockStatusData(JNIEnv* env,
     return nullptr;
   }
 
+<<<<<<< Updated upstream
   const rocksdb::HashMapJni::FnMapKV<const int32_t, const rocksdb::KeyLockInfo>
       fn_map_kv =
           [env, txn_db, &lock_status_data](
@@ -315,6 +316,13 @@ jobject Java_org_rocksdb_TransactionDB_getLockStatusData(JNIEnv* env,
             const jobject jlong_column_family_id =
                 rocksdb::LongJni::valueOf(env, pair.first);
             if (jlong_column_family_id == nullptr) {
+=======
+  const rocksdb::HashMapJni::FnMapKV<const int32_t, const rocksdb::KeyLockInfo> fn_map_kv =
+      [env](const std::pair<const int32_t, const rocksdb::KeyLockInfo>& pair) {
+          const jobject jlong_column_family_id =
+              rocksdb::LongJni::valueOf(env, pair.first);
+          if (jlong_column_family_id == nullptr) {
+>>>>>>> Stashed changes
               // an error occurred
               return std::unique_ptr<std::pair<jobject, jobject>>(nullptr);
             }

memtable/inlineskiplist.h

@@ -184,10 +184,9 @@ class InlineSkipList {
   const uint16_t kBranching_;
   const uint32_t kScaledInverseBranching_;
 
+  Allocator* const allocator_;  // Allocator used for allocations of nodes
   // Immutable after construction
   Comparator const compare_;
-  Allocator* const allocator_;  // Allocator used for allocations of nodes
-
   Node* const head_;
 
   // Modified only by Insert().  Read racily by readers, but stale
@@ -587,8 +586,8 @@ InlineSkipList<Comparator>::InlineSkipList(const Comparator cmp,
     : kMaxHeight_(static_cast<uint16_t>(max_height)),
       kBranching_(static_cast<uint16_t>(branching_factor)),
       kScaledInverseBranching_((Random::kMaxNext + 1) / kBranching_),
-      compare_(cmp),
       allocator_(allocator),
+      compare_(cmp),
       head_(AllocateNode(0, max_height)),
       max_height_(1),
       seq_splice_(AllocateSplice()) {

monitoring/histogram.cc

@@ -235,6 +235,8 @@ void HistogramStat::Data(HistogramData * const data) const {
   data->max = static_cast<double>(max());
   data->average = Average();
   data->standard_deviation = StandardDeviation();
+  data->count = num();
+  data->sum = sum();
 }
 
 void HistogramImpl::Clear() {

monitoring/instrumented_mutex.cc

@@ -10,25 +10,21 @@
 
 namespace rocksdb {
 namespace {
-bool ShouldReportToStats(Env* env, Statistics* stats) {
-  return env != nullptr && stats != nullptr &&
-          stats->stats_level_ > kExceptTimeForMutex;
+Statistics* stats_for_report(Env* env, Statistics* stats) {
+  if (env != nullptr && stats != nullptr &&
+      stats->stats_level_ > kExceptTimeForMutex) {
+    return stats;
+  } else {
+    return nullptr;
+  }
 }
 }  // namespace
 
 void InstrumentedMutex::Lock() {
-  PERF_CONDITIONAL_TIMER_FOR_MUTEX_GUARD(db_mutex_lock_nanos,
-                                         stats_code_ == DB_MUTEX_WAIT_MICROS);
-  uint64_t wait_time_micros = 0;
-  if (ShouldReportToStats(env_, stats_)) {
-    {
-      StopWatch sw(env_, nullptr, 0, &wait_time_micros);
-      LockInternal();
-    }
-    RecordTick(stats_, stats_code_, wait_time_micros);
-  } else {
-    LockInternal();
-  }
+  PERF_CONDITIONAL_TIMER_FOR_MUTEX_GUARD(
+      db_mutex_lock_nanos, stats_code_ == DB_MUTEX_WAIT_MICROS,
+      stats_for_report(env_, stats_), stats_code_);
+  LockInternal();
 }
 
 void InstrumentedMutex::LockInternal() {
@@ -39,18 +35,10 @@ void InstrumentedMutex::LockInternal() {
 }
 
 void InstrumentedCondVar::Wait() {
-  PERF_CONDITIONAL_TIMER_FOR_MUTEX_GUARD(db_condition_wait_nanos,
-                                         stats_code_ == DB_MUTEX_WAIT_MICROS);
-  uint64_t wait_time_micros = 0;
-  if (ShouldReportToStats(env_, stats_)) {
-    {
-      StopWatch sw(env_, nullptr, 0, &wait_time_micros);
-      WaitInternal();
-    }
-    RecordTick(stats_, stats_code_, wait_time_micros);
-  } else {
-    WaitInternal();
-  }
+  PERF_CONDITIONAL_TIMER_FOR_MUTEX_GUARD(
+      db_condition_wait_nanos, stats_code_ == DB_MUTEX_WAIT_MICROS,
+      stats_for_report(env_, stats_), stats_code_);
+  WaitInternal();
 }
 
 void InstrumentedCondVar::WaitInternal() {
@@ -61,20 +49,10 @@ void InstrumentedCondVar::WaitInternal() {
 }
 
 bool InstrumentedCondVar::TimedWait(uint64_t abs_time_us) {
-  PERF_CONDITIONAL_TIMER_FOR_MUTEX_GUARD(db_condition_wait_nanos,
-                                         stats_code_ == DB_MUTEX_WAIT_MICROS);
-  uint64_t wait_time_micros = 0;
-  bool result = false;
-  if (ShouldReportToStats(env_, stats_)) {
-    {
-      StopWatch sw(env_, nullptr, 0, &wait_time_micros);
-      result = TimedWaitInternal(abs_time_us);
-    }
-    RecordTick(stats_, stats_code_, wait_time_micros);
-  } else {
-    result = TimedWaitInternal(abs_time_us);
-  }
-  return result;
+  PERF_CONDITIONAL_TIMER_FOR_MUTEX_GUARD(
+      db_condition_wait_nanos, stats_code_ == DB_MUTEX_WAIT_MICROS,
+      stats_for_report(env_, stats_), stats_code_);
+  return TimedWaitInternal(abs_time_us);
 }
 
 bool InstrumentedCondVar::TimedWaitInternal(uint64_t abs_time_us) {

monitoring/perf_context_imp.h

@@ -41,10 +41,12 @@ extern __thread PerfContext perf_context;
   PerfStepTimer perf_step_timer_##metric(&(perf_context.metric)); \
   perf_step_timer_##metric.Start();
 
-#define PERF_CONDITIONAL_TIMER_FOR_MUTEX_GUARD(metric, condition)       \
-  PerfStepTimer perf_step_timer_##metric(&(perf_context.metric), true); \
-  if (condition) {                                                      \
-    perf_step_timer_##metric.Start();                                   \
+#define PERF_CONDITIONAL_TIMER_FOR_MUTEX_GUARD(metric, condition, stats,      \
+                                               ticker_type)                   \
+  PerfStepTimer perf_step_timer_##metric(&(perf_context.metric), true, stats, \
+                                         ticker_type);                        \
+  if (condition) {                                                            \
+    perf_step_timer_##metric.Start();                                         \
   }
 
 // Update metric with time elapsed since last START. start time is reset

monitoring/perf_step_timer.h

@@ -12,19 +12,25 @@ namespace rocksdb {
 
 class PerfStepTimer {
  public:
-  explicit PerfStepTimer(uint64_t* metric, bool for_mutex = false)
-      : enabled_(perf_level >= PerfLevel::kEnableTime ||
-                 (!for_mutex && perf_level >= kEnableTimeExceptForMutex)),
-        env_(enabled_ ? Env::Default() : nullptr),
+  explicit PerfStepTimer(uint64_t* metric, bool for_mutex = false,
+                         Statistics* statistics = nullptr,
+                         uint32_t ticker_type = 0)
+      : perf_counter_enabled_(
+            perf_level >= PerfLevel::kEnableTime ||
+            (!for_mutex && perf_level >= kEnableTimeExceptForMutex)),
+        env_((perf_counter_enabled_ || statistics != nullptr) ? Env::Default()
+                                                              : nullptr),
         start_(0),
-        metric_(metric) {}
+        metric_(metric),
+        statistics_(statistics),
+        ticker_type_(ticker_type) {}
 
   ~PerfStepTimer() {
     Stop();
   }
 
   void Start() {
-    if (enabled_) {
+    if (perf_counter_enabled_ || statistics_ != nullptr) {
       start_ = env_->NowNanos();
     }
   }
@@ -39,16 +45,25 @@ class PerfStepTimer {
 
   void Stop() {
     if (start_) {
-      *metric_ += env_->NowNanos() - start_;
+      uint64_t duration = env_->NowNanos() - start_;
+      if (perf_counter_enabled_) {
+        *metric_ += duration;
+      }
+
+      if (statistics_ != nullptr) {
+        RecordTick(statistics_, ticker_type_, duration);
+      }
       start_ = 0;
     }
   }
 
  private:
-  const bool enabled_;
+  const bool perf_counter_enabled_;
   Env* const env_;
   uint64_t start_;
   uint64_t* metric_;
+  Statistics* statistics_;
+  uint32_t ticker_type_;
 };
 
 }  // namespace rocksdb

monitoring/statistics.cc

@@ -169,11 +169,17 @@ std::string StatisticsImpl::ToString() const {
       char buffer[kTmpStrBufferSize];
       HistogramData hData;
       getHistogramImplLocked(h.first)->Data(&hData);
-      snprintf(
+      // don't handle failures - buffer should always be big enough and arguments
+      // should be provided correctly
+      int ret = snprintf(
           buffer, kTmpStrBufferSize,
-          "%s statistics Percentiles :=> 50 : %f 95 : %f 99 : %f 100 : %f\n",
-          h.second.c_str(), hData.median, hData.percentile95,
-          hData.percentile99, hData.max);
+          "%s P50 : %f P95 : %f P99 : %f P100 : %f COUNT : %" PRIu64 " SUM : %"
+          PRIu64 "\n", h.second.c_str(), hData.median, hData.percentile95,
+          hData.percentile99, hData.max, hData.count, hData.sum);
+      if (ret < 0 || ret >= kTmpStrBufferSize) {
+        assert(false);
+        continue;
+      }
       res.append(buffer);
     }
   }

options/cf_options.cc

@@ -27,7 +27,6 @@ ImmutableCFOptions::ImmutableCFOptions(const ImmutableDBOptions& db_options,
                                        const ColumnFamilyOptions& cf_options)
     : compaction_style(cf_options.compaction_style),
       compaction_pri(cf_options.compaction_pri),
-      prefix_extractor(cf_options.prefix_extractor.get()),
       user_comparator(cf_options.comparator),
       internal_comparator(InternalKeyComparator(cf_options.comparator)),
       merge_operator(cf_options.merge_operator.get()),
@@ -143,6 +142,9 @@ void MutableCFOptions::Dump(Logger* log) const {
   ROCKS_LOG_INFO(log,
                  "                 inplace_update_num_locks: %" ROCKSDB_PRIszt,
                  inplace_update_num_locks);
+  ROCKS_LOG_INFO(
+      log, "                         prefix_extractor: %s",
+      prefix_extractor == nullptr ? "nullptr" : prefix_extractor->Name());
   ROCKS_LOG_INFO(log, "                 disable_auto_compactions: %d",
                  disable_auto_compactions);
   ROCKS_LOG_INFO(log, "      soft_pending_compaction_bytes_limit: %" PRIu64,
@@ -189,4 +191,7 @@ void MutableCFOptions::Dump(Logger* log) const {
                  static_cast<int>(compression));
 }
 
+MutableCFOptions::MutableCFOptions(const Options& options)
+    : MutableCFOptions(ColumnFamilyOptions(options)) {}
+
 }  // namespace rocksdb

options/cf_options.h

@@ -30,8 +30,6 @@ struct ImmutableCFOptions {
 
   CompactionPri compaction_pri;
 
-  const SliceTransform* prefix_extractor;
-
   const Comparator* user_comparator;
   InternalKeyComparator internal_comparator;
 
@@ -134,6 +132,7 @@ struct MutableCFOptions {
         memtable_huge_page_size(options.memtable_huge_page_size),
         max_successive_merges(options.max_successive_merges),
         inplace_update_num_locks(options.inplace_update_num_locks),
+        prefix_extractor(options.prefix_extractor),
         disable_auto_compactions(options.disable_auto_compactions),
         soft_pending_compaction_bytes_limit(
             options.soft_pending_compaction_bytes_limit),
@@ -168,6 +167,7 @@ struct MutableCFOptions {
         memtable_huge_page_size(0),
         max_successive_merges(0),
         inplace_update_num_locks(0),
+        prefix_extractor(nullptr),
         disable_auto_compactions(false),
         soft_pending_compaction_bytes_limit(0),
         hard_pending_compaction_bytes_limit(0),
@@ -185,6 +185,8 @@ struct MutableCFOptions {
         report_bg_io_stats(false),
         compression(Snappy_Supported() ? kSnappyCompression : kNoCompression) {}
 
+  explicit MutableCFOptions(const Options& options);
+
   // Must be called after any change to MutableCFOptions
   void RefreshDerivedOptions(int num_levels, CompactionStyle compaction_style);
 
@@ -210,6 +212,7 @@ struct MutableCFOptions {
   size_t memtable_huge_page_size;
   size_t max_successive_merges;
   size_t inplace_update_num_locks;
+  std::shared_ptr<const SliceTransform> prefix_extractor;
 
   // Compaction related options
   bool disable_auto_compactions;

options/options_helper.cc

@@ -124,6 +124,8 @@ DBOptions BuildDBOptions(const ImmutableDBOptions& immutable_db_options,
       immutable_db_options.allow_ingest_behind;
   options.preserve_deletes =
       immutable_db_options.preserve_deletes;
+  options.two_write_queues = immutable_db_options.two_write_queues;
+  options.manual_wal_flush = immutable_db_options.manual_wal_flush;
 
   return options;
 }
@@ -143,6 +145,7 @@ ColumnFamilyOptions BuildColumnFamilyOptions(
   cf_opts.max_successive_merges = mutable_cf_options.max_successive_merges;
   cf_opts.inplace_update_num_locks =
       mutable_cf_options.inplace_update_num_locks;
+  cf_opts.prefix_extractor = mutable_cf_options.prefix_extractor;
 
   // Compaction related options
   cf_opts.disable_auto_compactions =
@@ -381,7 +384,8 @@ bool ParseSliceTransformHelper(
     const std::string& kFixedPrefixName, const std::string& kCappedPrefixName,
     const std::string& value,
     std::shared_ptr<const SliceTransform>* slice_transform) {
-
+  const char* no_op_name = "rocksdb.Noop";
+  size_t no_op_length = strlen(no_op_name);
   auto& pe_value = value;
   if (pe_value.size() > kFixedPrefixName.size() &&
       pe_value.compare(0, kFixedPrefixName.size(), kFixedPrefixName) == 0) {
@@ -393,6 +397,10 @@ bool ParseSliceTransformHelper(
     int prefix_length =
         ParseInt(trim(pe_value.substr(kCappedPrefixName.size())));
     slice_transform->reset(NewCappedPrefixTransform(prefix_length));
+  } else if (pe_value.size() == no_op_length &&
+             pe_value.compare(0, no_op_length, no_op_name) == 0) {
+    const SliceTransform* no_op_transform = NewNoopTransform();
+    slice_transform->reset(no_op_transform);
   } else if (value == kNullptrString) {
     slice_transform->reset();
   } else {
@@ -1789,7 +1797,7 @@ std::unordered_map<std::string, OptionTypeInfo>
         {"prefix_extractor",
          {offset_of(&ColumnFamilyOptions::prefix_extractor),
           OptionType::kSliceTransform, OptionVerificationType::kByNameAllowNull,
-          false, 0}},
+          true, offsetof(struct MutableCFOptions, prefix_extractor)}},
         {"memtable_insert_with_hint_prefix_extractor",
          {offset_of(
               &ColumnFamilyOptions::memtable_insert_with_hint_prefix_extractor),

options/options_parser.cc

@@ -17,6 +17,7 @@
 #include "rocksdb/convenience.h"
 #include "rocksdb/db.h"
 #include "util/cast_util.h"
+#include "util/file_reader_writer.h"
 #include "util/string_util.h"
 #include "util/sync_point.h"
 
@@ -41,12 +42,16 @@ Status PersistRocksDBOptions(const DBOptions& db_opt,
     return Status::InvalidArgument(
         "cf_names.size() and cf_opts.size() must be the same");
   }
-  std::unique_ptr<WritableFile> writable;
+  std::unique_ptr<WritableFile> wf;
 
-  Status s = env->NewWritableFile(file_name, &writable, EnvOptions());
+  Status s = env->NewWritableFile(file_name, &wf, EnvOptions());
   if (!s.ok()) {
     return s;
   }
+  unique_ptr<WritableFileWriter> writable;
+  writable.reset(new WritableFileWriter(std::move(wf), EnvOptions(),
+                                        nullptr /* statistics */));
+
   std::string options_file_content;
 
   writable->Append(option_file_header + "[" +
@@ -93,8 +98,7 @@ Status PersistRocksDBOptions(const DBOptions& db_opt,
       writable->Append(options_file_content + "\n");
     }
   }
-  writable->Flush();
-  writable->Fsync();
+  writable->Sync(true /* use_fsync */);
   writable->Close();
 
   return RocksDBOptionsParser::VerifyRocksDBOptionsFromFile(

src.mk

@@ -250,6 +250,8 @@ MAIN_SOURCES =                                                          \
   cache/cache_bench.cc                                                  \
   cache/cache_test.cc                                                   \
   db/column_family_test.cc                                              \
+  db/compact_files_test.cc                                              \
+  db/compaction_iterator_test.cc                                        \
   db/compaction_job_stats_test.cc                                       \
   db/compaction_job_test.cc                                             \
   db/compaction_picker_test.cc                                          \
@@ -257,6 +259,7 @@ MAIN_SOURCES =                                                          \
   db/corruption_test.cc                                                 \
   db/cuckoo_table_db_test.cc                                            \
   db/db_basic_test.cc                                                   \
+  db/db_blob_index_test.cc                                              \
   db/db_block_cache_test.cc                                             \
   db/db_bloom_filter_test.cc                                            \
   db/db_compaction_filter_test.cc                                       \
@@ -267,38 +270,56 @@ MAIN_SOURCES =                                                          \
   db/db_inplace_update_test.cc                                          \
   db/db_io_failure_test.cc                                              \
   db/db_iter_test.cc                                                    \
+  db/db_iter_stress_test.cc                                             \
   db/db_iterator_test.cc                                                \
   db/db_log_iter_test.cc                                                \
   db/db_memtable_test.cc                                                \
   db/db_merge_operator_test.cc                                          \
   db/db_options_test.cc                                                 \
+  db/db_properties_test.cc                                              \
   db/db_range_del_test.cc                                               \
   db/db_sst_test.cc                                                     \
   db/db_statistics_test.cc                                              \
   db/db_table_properties_test.cc                                        \
   db/db_tailing_iter_test.cc                                            \
   db/db_test.cc                                                         \
+  db/db_test2.cc                                                        \
   db/db_universal_compaction_test.cc                                    \
   db/db_wal_test.cc                                                     \
   db/db_write_test.cc                                                   \
   db/dbformat_test.cc                                                   \
   db/deletefile_test.cc                                                 \
-  db/obsolete_files_test.cc						\
+  db/env_timed_test.cc                                                  \
   db/external_sst_file_basic_test.cc                                    \
   db/external_sst_file_test.cc                                          \
   db/fault_injection_test.cc                                            \
   db/file_indexer_test.cc                                               \
+  db/file_reader_writer_test.cc                                         \
   db/filename_test.cc                                                   \
   db/flush_job_test.cc                                                  \
+  db/hash_table_test.cc                                                 \
+  db/hash_test.cc                                                       \
+  db/heap_test.cc                                                       \
   db/listener_test.cc                                                   \
   db/log_test.cc                                                        \
+  db/lru_cache_test.cc                                                  \
   db/manual_compaction_test.cc                                          \
+  db/memtable_list_test.cc                                              \
+  db/merge_helper_test.cc                                               \
   db/merge_test.cc                                                      \
+  db/obsolete_files_test.cc						\
+  db/options_settable_test.cc                                           \
   db/options_file_test.cc                                               \
+  db/partitioned_filter_block_test.cc                                   \
   db/perf_context_test.cc                                               \
+  db/persistent_cache_test.cc                                           \
   db/plain_table_db_test.cc                                             \
   db/prefix_test.cc                                                     \
+  db/redis_test.cc                                                      \
+  db/repair_test.cc                                                     \
+  db/range_del_aggregator_test.cc                                       \
   db/table_properties_collector_test.cc                                 \
+  db/util_merge_operators_test.cc                                       \
   db/version_builder_test.cc                                            \
   db/version_edit_test.cc                                               \
   db/version_set_test.cc                                                \

table/block.cc

@@ -402,12 +402,14 @@ Block::Block(BlockContents&& contents, SequenceNumber _global_seqno,
       data_(contents_.data.data()),
       size_(contents_.data.size()),
       restart_offset_(0),
+      num_restarts_(0),
       global_seqno_(_global_seqno) {
   if (size_ < sizeof(uint32_t)) {
     size_ = 0;  // Error marker
   } else {
+    num_restarts_ = NumRestarts();
     restart_offset_ =
-        static_cast<uint32_t>(size_) - (1 + NumRestarts()) * sizeof(uint32_t);
+        static_cast<uint32_t>(size_) - (1 + num_restarts_) * sizeof(uint32_t);
     if (restart_offset_ > size_ - sizeof(uint32_t)) {
       // The size is too small for NumRestarts() and therefore
       // restart_offset_ wrapped around.
@@ -429,17 +431,17 @@ BlockIter* Block::NewIterator(const Comparator* cmp, BlockIter* iter,
     ret_iter = new BlockIter;
   }
   if (size_ < 2*sizeof(uint32_t)) {
-    ret_iter->SetStatus(Status::Corruption("bad block contents"));
+    ret_iter->Invalidate(Status::Corruption("bad block contents"));
     return ret_iter;
   }
-  const uint32_t num_restarts = NumRestarts();
-  if (num_restarts == 0) {
-    ret_iter->SetStatus(Status::OK());
+  if (num_restarts_ == 0) {
+    // Empty block.
+    ret_iter->Invalidate(Status::OK());
     return ret_iter;
   } else {
     BlockPrefixIndex* prefix_index_ptr =
         total_order_seek ? nullptr : prefix_index_.get();
-    ret_iter->Initialize(cmp, data_, restart_offset_, num_restarts,
+    ret_iter->Initialize(cmp, data_, restart_offset_, num_restarts_,
                          prefix_index_ptr, global_seqno_,
                          read_amp_bitmap_.get());
 

table/block.h

@@ -184,6 +184,7 @@ class Block {
   const char* data_;            // contents_.data.data()
   size_t size_;                 // contents_.data.size()
   uint32_t restart_offset_;     // Offset in data_ of restart array
+  uint32_t num_restarts_;
   std::unique_ptr<BlockPrefixIndex> prefix_index_;
   std::unique_ptr<BlockReadAmpBitmap> read_amp_bitmap_;
   // All keys in the block will have seqno = global_seqno_, regardless of
@@ -241,8 +242,21 @@ class BlockIter final : public InternalIterator {
     last_bitmap_offset_ = current_ + 1;
   }
 
-  void SetStatus(Status s) {
+  // Makes Valid() return false, status() return `s`, and Seek()/Prev()/etc do
+  // nothing.
+  void Invalidate(Status s) {
+    // Assert that the BlockIter is never deleted while Pinning is Enabled.
+    assert(!pinned_iters_mgr_ ||
+           (pinned_iters_mgr_ && !pinned_iters_mgr_->PinningEnabled()));
+
+    data_ = nullptr;
+    current_ = restarts_;
     status_ = s;
+
+    // Clear prev entries cache.
+    prev_entries_keys_buff_.clear();
+    prev_entries_.clear();
+    prev_entries_idx_ = -1;
   }
 
   virtual bool Valid() const override { return current_ < restarts_; }

table/block_based_filter_block.cc

@@ -186,7 +186,8 @@ BlockBasedFilterBlockReader::BlockBasedFilterBlockReader(
 }
 
 bool BlockBasedFilterBlockReader::KeyMayMatch(
-    const Slice& key, uint64_t block_offset, const bool /*no_io*/,
+    const Slice& key, const SliceTransform* /* prefix_extractor */,
+    uint64_t block_offset, const bool /*no_io*/,
     const Slice* const /*const_ikey_ptr*/) {
   assert(block_offset != kNotValid);
   if (!whole_key_filtering_) {
@@ -196,7 +197,8 @@ bool BlockBasedFilterBlockReader::KeyMayMatch(
 }
 
 bool BlockBasedFilterBlockReader::PrefixMayMatch(
-    const Slice& prefix, uint64_t block_offset, const bool /*no_io*/,
+    const Slice& prefix, const SliceTransform* /* prefix_extractor */,
+    uint64_t block_offset, const bool /*no_io*/,
     const Slice* const /*const_ikey_ptr*/) {
   assert(block_offset != kNotValid);
   if (!prefix_extractor_) {

table/block_based_filter_block.h

@@ -83,13 +83,14 @@ class BlockBasedFilterBlockReader : public FilterBlockReader {
                               bool whole_key_filtering,
                               BlockContents&& contents, Statistics* statistics);
   virtual bool IsBlockBased() override { return true; }
+
   virtual bool KeyMayMatch(
-      const Slice& key, uint64_t block_offset = kNotValid,
-      const bool no_io = false,
+      const Slice& key, const SliceTransform* prefix_extractor,
+      uint64_t block_offset = kNotValid, const bool no_io = false,
       const Slice* const const_ikey_ptr = nullptr) override;
   virtual bool PrefixMayMatch(
-      const Slice& prefix, uint64_t block_offset = kNotValid,
-      const bool no_io = false,
+      const Slice& prefix, const SliceTransform* prefix_extractor,
+      uint64_t block_offset = kNotValid, const bool no_io = false,
       const Slice* const const_ikey_ptr = nullptr) override;
   virtual size_t ApproximateMemoryUsage() const override;
 

table/block_based_filter_block_test.cc

@@ -59,8 +59,8 @@ TEST_F(FilterBlockTest, EmptyBuilder) {
   ASSERT_EQ("\\x00\\x00\\x00\\x00\\x0b", EscapeString(block.data));
   BlockBasedFilterBlockReader reader(nullptr, table_options_, true,
                                      std::move(block), nullptr);
-  ASSERT_TRUE(reader.KeyMayMatch("foo", 0));
-  ASSERT_TRUE(reader.KeyMayMatch("foo", 100000));
+  ASSERT_TRUE(reader.KeyMayMatch("foo", nullptr, uint64_t{0}));
+  ASSERT_TRUE(reader.KeyMayMatch("foo", nullptr, 100000));
 }
 
 TEST_F(FilterBlockTest, SingleChunk) {
@@ -78,13 +78,13 @@ TEST_F(FilterBlockTest, SingleChunk) {
   BlockContents block(builder.Finish(), false, kNoCompression);
   BlockBasedFilterBlockReader reader(nullptr, table_options_, true,
                                      std::move(block), nullptr);
-  ASSERT_TRUE(reader.KeyMayMatch("foo", 100));
-  ASSERT_TRUE(reader.KeyMayMatch("bar", 100));
-  ASSERT_TRUE(reader.KeyMayMatch("box", 100));
-  ASSERT_TRUE(reader.KeyMayMatch("hello", 100));
-  ASSERT_TRUE(reader.KeyMayMatch("foo", 100));
-  ASSERT_TRUE(!reader.KeyMayMatch("missing", 100));
-  ASSERT_TRUE(!reader.KeyMayMatch("other", 100));
+  ASSERT_TRUE(reader.KeyMayMatch("foo", nullptr, 100));
+  ASSERT_TRUE(reader.KeyMayMatch("bar", nullptr, 100));
+  ASSERT_TRUE(reader.KeyMayMatch("box", nullptr, 100));
+  ASSERT_TRUE(reader.KeyMayMatch("hello", nullptr, 100));
+  ASSERT_TRUE(reader.KeyMayMatch("foo", nullptr, 100));
+  ASSERT_TRUE(!reader.KeyMayMatch("missing", nullptr, 100));
+  ASSERT_TRUE(!reader.KeyMayMatch("other", nullptr, 100));
 }
 
 TEST_F(FilterBlockTest, MultiChunk) {
@@ -112,28 +112,28 @@ TEST_F(FilterBlockTest, MultiChunk) {
                                      std::move(block), nullptr);
 
   // Check first filter
-  ASSERT_TRUE(reader.KeyMayMatch("foo", 0));
-  ASSERT_TRUE(reader.KeyMayMatch("bar", 2000));
-  ASSERT_TRUE(!reader.KeyMayMatch("box", 0));
-  ASSERT_TRUE(!reader.KeyMayMatch("hello", 0));
+  ASSERT_TRUE(reader.KeyMayMatch("foo", nullptr, uint64_t{0}));
+  ASSERT_TRUE(reader.KeyMayMatch("bar", nullptr, 2000));
+  ASSERT_TRUE(!reader.KeyMayMatch("box", nullptr, uint64_t{0}));
+  ASSERT_TRUE(!reader.KeyMayMatch("hello", nullptr, uint64_t{0}));
 
   // Check second filter
-  ASSERT_TRUE(reader.KeyMayMatch("box", 3100));
-  ASSERT_TRUE(!reader.KeyMayMatch("foo", 3100));
-  ASSERT_TRUE(!reader.KeyMayMatch("bar", 3100));
-  ASSERT_TRUE(!reader.KeyMayMatch("hello", 3100));
+  ASSERT_TRUE(reader.KeyMayMatch("box", nullptr, 3100));
+  ASSERT_TRUE(!reader.KeyMayMatch("foo", nullptr, 3100));
+  ASSERT_TRUE(!reader.KeyMayMatch("bar", nullptr, 3100));
+  ASSERT_TRUE(!reader.KeyMayMatch("hello", nullptr, 3100));
 
   // Check third filter (empty)
-  ASSERT_TRUE(!reader.KeyMayMatch("foo", 4100));
-  ASSERT_TRUE(!reader.KeyMayMatch("bar", 4100));
-  ASSERT_TRUE(!reader.KeyMayMatch("box", 4100));
-  ASSERT_TRUE(!reader.KeyMayMatch("hello", 4100));
+  ASSERT_TRUE(!reader.KeyMayMatch("foo", nullptr, 4100));
+  ASSERT_TRUE(!reader.KeyMayMatch("bar", nullptr, 4100));
+  ASSERT_TRUE(!reader.KeyMayMatch("box", nullptr, 4100));
+  ASSERT_TRUE(!reader.KeyMayMatch("hello", nullptr, 4100));
 
   // Check last filter
-  ASSERT_TRUE(reader.KeyMayMatch("box", 9000));
-  ASSERT_TRUE(reader.KeyMayMatch("hello", 9000));
-  ASSERT_TRUE(!reader.KeyMayMatch("foo", 9000));
-  ASSERT_TRUE(!reader.KeyMayMatch("bar", 9000));
+  ASSERT_TRUE(reader.KeyMayMatch("box", nullptr, 9000));
+  ASSERT_TRUE(reader.KeyMayMatch("hello", nullptr, 9000));
+  ASSERT_TRUE(!reader.KeyMayMatch("foo", nullptr, 9000));
+  ASSERT_TRUE(!reader.KeyMayMatch("bar", nullptr, 9000));
 }
 
 // Test for block based filter block
@@ -156,8 +156,8 @@ TEST_F(BlockBasedFilterBlockTest, BlockBasedEmptyBuilder) {
   ASSERT_EQ("\\x00\\x00\\x00\\x00\\x0b", EscapeString(block.data));
   FilterBlockReader* reader = new BlockBasedFilterBlockReader(
       nullptr, table_options_, true, std::move(block), nullptr);
-  ASSERT_TRUE(reader->KeyMayMatch("foo", 0));
-  ASSERT_TRUE(reader->KeyMayMatch("foo", 100000));
+  ASSERT_TRUE(reader->KeyMayMatch("foo", nullptr, uint64_t{0}));
+  ASSERT_TRUE(reader->KeyMayMatch("foo", nullptr, 100000));
 
   delete builder;
   delete reader;
@@ -177,13 +177,13 @@ TEST_F(BlockBasedFilterBlockTest, BlockBasedSingleChunk) {
   BlockContents block(builder->Finish(), false, kNoCompression);
   FilterBlockReader* reader = new BlockBasedFilterBlockReader(
       nullptr, table_options_, true, std::move(block), nullptr);
-  ASSERT_TRUE(reader->KeyMayMatch("foo", 100));
-  ASSERT_TRUE(reader->KeyMayMatch("bar", 100));
-  ASSERT_TRUE(reader->KeyMayMatch("box", 100));
-  ASSERT_TRUE(reader->KeyMayMatch("hello", 100));
-  ASSERT_TRUE(reader->KeyMayMatch("foo", 100));
-  ASSERT_TRUE(!reader->KeyMayMatch("missing", 100));
-  ASSERT_TRUE(!reader->KeyMayMatch("other", 100));
+  ASSERT_TRUE(reader->KeyMayMatch("foo", nullptr, 100));
+  ASSERT_TRUE(reader->KeyMayMatch("bar", nullptr, 100));
+  ASSERT_TRUE(reader->KeyMayMatch("box", nullptr, 100));
+  ASSERT_TRUE(reader->KeyMayMatch("hello", nullptr, 100));
+  ASSERT_TRUE(reader->KeyMayMatch("foo", nullptr, 100));
+  ASSERT_TRUE(!reader->KeyMayMatch("missing", nullptr, 100));
+  ASSERT_TRUE(!reader->KeyMayMatch("other", nullptr, 100));
 
   delete builder;
   delete reader;
@@ -215,28 +215,28 @@ TEST_F(BlockBasedFilterBlockTest, BlockBasedMultiChunk) {
       nullptr, table_options_, true, std::move(block), nullptr);
 
   // Check first filter
-  ASSERT_TRUE(reader->KeyMayMatch("foo", 0));
-  ASSERT_TRUE(reader->KeyMayMatch("bar", 2000));
-  ASSERT_TRUE(!reader->KeyMayMatch("box", 0));
-  ASSERT_TRUE(!reader->KeyMayMatch("hello", 0));
+  ASSERT_TRUE(reader->KeyMayMatch("foo", nullptr, uint64_t{0}));
+  ASSERT_TRUE(reader->KeyMayMatch("bar", nullptr, 2000));
+  ASSERT_TRUE(!reader->KeyMayMatch("box", nullptr, uint64_t{0}));
+  ASSERT_TRUE(!reader->KeyMayMatch("hello", nullptr, uint64_t{0}));
 
   // Check second filter
-  ASSERT_TRUE(reader->KeyMayMatch("box", 3100));
-  ASSERT_TRUE(!reader->KeyMayMatch("foo", 3100));
-  ASSERT_TRUE(!reader->KeyMayMatch("bar", 3100));
-  ASSERT_TRUE(!reader->KeyMayMatch("hello", 3100));
+  ASSERT_TRUE(reader->KeyMayMatch("box", nullptr, 3100));
+  ASSERT_TRUE(!reader->KeyMayMatch("foo", nullptr, 3100));
+  ASSERT_TRUE(!reader->KeyMayMatch("bar", nullptr, 3100));
+  ASSERT_TRUE(!reader->KeyMayMatch("hello", nullptr, 3100));
 
   // Check third filter (empty)
-  ASSERT_TRUE(!reader->KeyMayMatch("foo", 4100));
-  ASSERT_TRUE(!reader->KeyMayMatch("bar", 4100));
-  ASSERT_TRUE(!reader->KeyMayMatch("box", 4100));
-  ASSERT_TRUE(!reader->KeyMayMatch("hello", 4100));
+  ASSERT_TRUE(!reader->KeyMayMatch("foo", nullptr, 4100));
+  ASSERT_TRUE(!reader->KeyMayMatch("bar", nullptr, 4100));
+  ASSERT_TRUE(!reader->KeyMayMatch("box", nullptr, 4100));
+  ASSERT_TRUE(!reader->KeyMayMatch("hello", nullptr, 4100));
 
   // Check last filter
-  ASSERT_TRUE(reader->KeyMayMatch("box", 9000));
-  ASSERT_TRUE(reader->KeyMayMatch("hello", 9000));
-  ASSERT_TRUE(!reader->KeyMayMatch("foo", 9000));
-  ASSERT_TRUE(!reader->KeyMayMatch("bar", 9000));
+  ASSERT_TRUE(reader->KeyMayMatch("box", nullptr, 9000));
+  ASSERT_TRUE(reader->KeyMayMatch("hello", nullptr, 9000));
+  ASSERT_TRUE(!reader->KeyMayMatch("foo", nullptr, 9000));
+  ASSERT_TRUE(!reader->KeyMayMatch("bar", nullptr, 9000));
 
   delete builder;
   delete reader;

table/block_based_table_builder.cc

@@ -62,14 +62,16 @@ namespace {
 
 // Create a filter block builder based on its type.
 FilterBlockBuilder* CreateFilterBlockBuilder(
-    const ImmutableCFOptions& opt, const BlockBasedTableOptions& table_opt,
+    const ImmutableCFOptions& /*opt*/, const MutableCFOptions& mopt,
+    const BlockBasedTableOptions& table_opt,
     PartitionedIndexBuilder* const p_index_builder) {
   if (table_opt.filter_policy == nullptr) return nullptr;
 
   FilterBitsBuilder* filter_bits_builder =
       table_opt.filter_policy->GetFilterBitsBuilder();
   if (filter_bits_builder == nullptr) {
-    return new BlockBasedFilterBlockBuilder(opt.prefix_extractor, table_opt);
+    return new BlockBasedFilterBlockBuilder(mopt.prefix_extractor.get(),
+                                            table_opt);
   } else {
     if (table_opt.partition_filters) {
       assert(p_index_builder != nullptr);
@@ -82,11 +84,11 @@ FilterBlockBuilder* CreateFilterBlockBuilder(
           (100 - table_opt.block_size_deviation)) + 99) / 100);
       partition_size = std::max(partition_size, static_cast<uint32_t>(1));
       return new PartitionedFilterBlockBuilder(
-          opt.prefix_extractor, table_opt.whole_key_filtering,
+          mopt.prefix_extractor.get(), table_opt.whole_key_filtering,
           filter_bits_builder, table_opt.index_block_restart_interval,
           p_index_builder, partition_size);
     } else {
-      return new FullFilterBlockBuilder(opt.prefix_extractor,
+      return new FullFilterBlockBuilder(mopt.prefix_extractor.get(),
                                         table_opt.whole_key_filtering,
                                         filter_bits_builder);
     }
@@ -244,6 +246,7 @@ class BlockBasedTableBuilder::BlockBasedTablePropertiesCollector
 
 struct BlockBasedTableBuilder::Rep {
   const ImmutableCFOptions ioptions;
+  const MutableCFOptions moptions;
   const BlockBasedTableOptions table_options;
   const InternalKeyComparator& internal_comparator;
   WritableFileWriter* file;
@@ -280,7 +283,7 @@ struct BlockBasedTableBuilder::Rep {
 
   std::vector<std::unique_ptr<IntTblPropCollector>> table_properties_collectors;
 
-  Rep(const ImmutableCFOptions& _ioptions,
+  Rep(const ImmutableCFOptions& _ioptions, const MutableCFOptions& _moptions,
       const BlockBasedTableOptions& table_opt,
       const InternalKeyComparator& icomparator,
       const std::vector<std::unique_ptr<IntTblPropCollectorFactory>>*
@@ -292,6 +295,7 @@ struct BlockBasedTableBuilder::Rep {
       const std::string& _column_family_name, const uint64_t _creation_time,
       const uint64_t _oldest_key_time)
       : ioptions(_ioptions),
+        moptions(_moptions),
         table_options(table_opt),
         internal_comparator(icomparator),
         file(f),
@@ -300,8 +304,8 @@ struct BlockBasedTableBuilder::Rep {
                       : 0),
         data_block(table_options.block_restart_interval,
                    table_options.use_delta_encoding),
-        range_del_block(1),  // TODO(andrewkr): restart_interval unnecessary
-        internal_prefix_transform(_ioptions.prefix_extractor),
+        range_del_block(1 /* block_restart_interval */),
+        internal_prefix_transform(_moptions.prefix_extractor.get()),
         compression_type(_compression_type),
         compression_opts(_compression_opts),
         compression_dict(_compression_dict),
@@ -326,8 +330,8 @@ struct BlockBasedTableBuilder::Rep {
     if (skip_filters) {
       filter_builder = nullptr;
     } else {
-      filter_builder.reset(
-          CreateFilterBlockBuilder(_ioptions, table_options, p_index_builder_));
+      filter_builder.reset(CreateFilterBlockBuilder(
+          _ioptions, _moptions, table_options, p_index_builder_));
     }
 
     for (auto& collector_factories : *int_tbl_prop_collector_factories) {
@@ -337,12 +341,12 @@ struct BlockBasedTableBuilder::Rep {
     table_properties_collectors.emplace_back(
         new BlockBasedTablePropertiesCollector(
             table_options.index_type, table_options.whole_key_filtering,
-            _ioptions.prefix_extractor != nullptr));
+            _moptions.prefix_extractor != nullptr));
   }
 };
 
 BlockBasedTableBuilder::BlockBasedTableBuilder(
-    const ImmutableCFOptions& ioptions,
+    const ImmutableCFOptions& ioptions, const MutableCFOptions& moptions,
     const BlockBasedTableOptions& table_options,
     const InternalKeyComparator& internal_comparator,
     const std::vector<std::unique_ptr<IntTblPropCollectorFactory>>*
@@ -365,11 +369,11 @@ BlockBasedTableBuilder::BlockBasedTableBuilder(
     sanitized_table_options.format_version = 1;
   }
 
-  rep_ = new Rep(ioptions, sanitized_table_options, internal_comparator,
-                 int_tbl_prop_collector_factories, column_family_id, file,
-                 compression_type, compression_opts, compression_dict,
-                 skip_filters, column_family_name, creation_time,
-                 oldest_key_time);
+  rep_ =
+      new Rep(ioptions, moptions, sanitized_table_options, internal_comparator,
+              int_tbl_prop_collector_factories, column_family_id, file,
+              compression_type, compression_opts, compression_dict,
+              skip_filters, column_family_name, creation_time, oldest_key_time);
 
   if (rep_->filter_builder != nullptr) {
     rep_->filter_builder->StartBlock(0);
@@ -737,8 +741,8 @@ Status BlockBasedTableBuilder::Finish() {
                                          : "nullptr";
       r->props.compression_name = CompressionTypeToString(r->compression_type);
       r->props.prefix_extractor_name =
-          r->ioptions.prefix_extractor != nullptr
-              ? r->ioptions.prefix_extractor->Name()
+          r->moptions.prefix_extractor != nullptr
+              ? r->moptions.prefix_extractor->Name()
               : "nullptr";
 
       std::string property_collectors_names = "[";

table/block_based_table_builder.h

@@ -38,7 +38,7 @@ class BlockBasedTableBuilder : public TableBuilder {
   // @param compression_dict Data for presetting the compression library's
   //    dictionary, or nullptr.
   BlockBasedTableBuilder(
-      const ImmutableCFOptions& ioptions,
+      const ImmutableCFOptions& ioptions, const MutableCFOptions& moptions,
       const BlockBasedTableOptions& table_options,
       const InternalKeyComparator& internal_comparator,
       const std::vector<std::unique_ptr<IntTblPropCollectorFactory>>*

table/block_based_table_factory.cc

@@ -69,16 +69,17 @@ Status BlockBasedTableFactory::NewTableReader(
   return BlockBasedTable::Open(
       table_reader_options.ioptions, table_reader_options.env_options,
       table_options_, table_reader_options.internal_comparator, std::move(file),
-      file_size, table_reader, prefetch_index_and_filter_in_cache,
-      table_reader_options.skip_filters, table_reader_options.level);
+      file_size, table_reader, table_reader_options.prefix_extractor,
+      prefetch_index_and_filter_in_cache, table_reader_options.skip_filters,
+      table_reader_options.level);
 }
 
 TableBuilder* BlockBasedTableFactory::NewTableBuilder(
     const TableBuilderOptions& table_builder_options, uint32_t column_family_id,
     WritableFileWriter* file) const {
   auto table_builder = new BlockBasedTableBuilder(
-      table_builder_options.ioptions, table_options_,
-      table_builder_options.internal_comparator,
+      table_builder_options.ioptions, table_builder_options.moptions,
+      table_options_, table_builder_options.internal_comparator,
       table_builder_options.int_tbl_prop_collector_factories, column_family_id,
       file, table_builder_options.compression_type,
       table_builder_options.compression_opts,

table/block_based_table_reader.cc

@@ -173,6 +173,29 @@ Cache::Handle* GetEntryFromCache(Cache* block_cache, const Slice& key,
   return cache_handle;
 }
 
+// For hash based index, return true if prefix_extractor and
+// prefix_extractor_block mismatch, false otherwise. This flag will be used
+// as total_order_seek via NewIndexIterator
+bool PrefixExtractorChanged(
+    std::shared_ptr<const TableProperties> table_properties,
+    const SliceTransform* prefix_extractor) {
+  // BlockBasedTableOptions::kHashSearch requires prefix_extractor to be set.
+  // Turn off hash index in prefix_extractor is not set; if  prefix_extractor
+  // is set but prefix_extractor_block is not set, also disable hash index
+  if (prefix_extractor == nullptr || table_properties == nullptr ||
+      table_properties->prefix_extractor_name.empty()) {
+    return true;
+  }
+
+  // prefix_extractor and prefix_extractor_block are both non-empty
+  if (table_properties->prefix_extractor_name.compare(
+          prefix_extractor->Name()) != 0) {
+    return true;
+  } else {
+    return false;
+  }
+}
+
 }  // namespace
 
 // Index that allows binary search lookup in a two-level index structure.
@@ -221,7 +244,8 @@ class PartitionIndexReader : public IndexReader, public Cleanable {
       ro.fill_cache = fill_cache;
       return new BlockBasedTableIterator(
           table_, ro, *icomparator_,
-          index_block_->NewIterator(icomparator_, nullptr, true), false);
+          index_block_->NewIterator(icomparator_, nullptr, true), false,
+          /* prefix_extractor */ nullptr);
     }
     // TODO(myabandeh): Update TwoLevelIterator to be able to make use of
     // on-stack BlockIter while the state is on heap. Currentlly it assumes
@@ -646,6 +670,7 @@ Status BlockBasedTable::Open(const ImmutableCFOptions& ioptions,
                              unique_ptr<RandomAccessFileReader>&& file,
                              uint64_t file_size,
                              unique_ptr<TableReader>* table_reader,
+                             const SliceTransform* prefix_extractor,
                              const bool prefetch_index_and_filter_in_cache,
                              const bool skip_filters, const int level) {
   table_reader->reset();
@@ -697,7 +722,7 @@ Status BlockBasedTable::Open(const ImmutableCFOptions& ioptions,
   // We need to wrap data with internal_prefix_transform to make sure it can
   // handle prefix correctly.
   rep->internal_prefix_transform.reset(
-      new InternalKeySliceTransform(rep->ioptions.prefix_extractor));
+      new InternalKeySliceTransform(prefix_extractor));
   SetupCacheKeyPrefix(rep, file_size);
   unique_ptr<BlockBasedTable> new_table(new BlockBasedTable(rep));
 
@@ -863,8 +888,14 @@ Status BlockBasedTable::Open(const ImmutableCFOptions& ioptions,
       // block_cache
 
       CachableEntry<IndexReader> index_entry;
-      unique_ptr<InternalIterator> iter(
-          new_table->NewIndexIterator(ReadOptions(), nullptr, &index_entry));
+      bool prefix_extractor_changed = false;
+      // check prefix_extractor match only if hash based index is used
+      if (rep->index_type == BlockBasedTableOptions::kHashSearch) {
+        prefix_extractor_changed =
+            PrefixExtractorChanged(rep->table_properties, prefix_extractor);
+      }
+      unique_ptr<InternalIterator> iter(new_table->NewIndexIterator(
+          ReadOptions(), prefix_extractor_changed, nullptr, &index_entry));
       s = iter->status();
       if (s.ok()) {
         // This is the first call to NewIndexIterator() since we're in Open().
@@ -879,9 +910,9 @@ Status BlockBasedTable::Open(const ImmutableCFOptions& ioptions,
         }
 
         // Hack: Call GetFilter() to implicitly add filter to the block_cache
-        auto filter_entry = new_table->GetFilter();
+        auto filter_entry = new_table->GetFilter(prefix_extractor);
         if (filter_entry.value != nullptr) {
-          filter_entry.value->CacheDependencies(pin);
+          filter_entry.value->CacheDependencies(pin, prefix_extractor);
         }
         // if pin_l0_filter_and_index_blocks_in_cache is true, and this is
         // a level0 file, then save it in rep_->filter_entry; it will be
@@ -913,13 +944,14 @@ Status BlockBasedTable::Open(const ImmutableCFOptions& ioptions,
       // Set filter block
       if (rep->filter_policy) {
         const bool is_a_filter_partition = true;
-        auto filter = new_table->ReadFilter(
-            prefetch_buffer.get(), rep->filter_handle, !is_a_filter_partition);
+        auto filter =
+            new_table->ReadFilter(prefetch_buffer.get(), rep->filter_handle,
+                                  !is_a_filter_partition, prefix_extractor);
         rep->filter.reset(filter);
         // Refer to the comment above about paritioned indexes always being
         // cached
         if (filter && (prefetch_index_and_filter_in_cache || level == 0)) {
-          filter->CacheDependencies(pin);
+          filter->CacheDependencies(pin, prefix_extractor);
         }
       }
     } else {
@@ -1215,7 +1247,8 @@ Status BlockBasedTable::PutDataBlockToCache(
 
 FilterBlockReader* BlockBasedTable::ReadFilter(
     FilePrefetchBuffer* prefetch_buffer, const BlockHandle& filter_handle,
-    const bool is_a_filter_partition) const {
+    const bool is_a_filter_partition,
+    const SliceTransform* prefix_extractor) const {
   auto& rep = rep_;
   // TODO: We might want to unify with ReadBlockFromFile() if we start
   // requiring checksum verification in Table::Open.
@@ -1248,14 +1281,14 @@ FilterBlockReader* BlockBasedTable::ReadFilter(
   switch (filter_type) {
     case Rep::FilterType::kPartitionedFilter: {
       return new PartitionedFilterBlockReader(
-          rep->prefix_filtering ? rep->ioptions.prefix_extractor : nullptr,
+          rep->prefix_filtering ? prefix_extractor : nullptr,
           rep->whole_key_filtering, std::move(block), nullptr,
           rep->ioptions.statistics, rep->internal_comparator, this);
     }
 
     case Rep::FilterType::kBlockFilter:
       return new BlockBasedFilterBlockReader(
-          rep->prefix_filtering ? rep->ioptions.prefix_extractor : nullptr,
+          rep->prefix_filtering ? prefix_extractor : nullptr,
           rep->table_options, rep->whole_key_filtering, std::move(block),
           rep->ioptions.statistics);
 
@@ -1264,7 +1297,7 @@ FilterBlockReader* BlockBasedTable::ReadFilter(
           rep->filter_policy->GetFilterBitsReader(block.data);
       assert(filter_bits_reader != nullptr);
       return new FullFilterBlockReader(
-          rep->prefix_filtering ? rep->ioptions.prefix_extractor : nullptr,
+          rep->prefix_filtering ? prefix_extractor : nullptr,
           rep->whole_key_filtering, std::move(block), filter_bits_reader,
           rep->ioptions.statistics);
     }
@@ -1278,18 +1311,18 @@ FilterBlockReader* BlockBasedTable::ReadFilter(
 }
 
 BlockBasedTable::CachableEntry<FilterBlockReader> BlockBasedTable::GetFilter(
-    FilePrefetchBuffer* prefetch_buffer, bool no_io,
-    GetContext* get_context) const {
+    const SliceTransform* prefix_extractor, FilePrefetchBuffer* prefetch_buffer,
+    bool no_io, GetContext* get_context) const {
   const BlockHandle& filter_blk_handle = rep_->filter_handle;
   const bool is_a_filter_partition = true;
   return GetFilter(prefetch_buffer, filter_blk_handle, !is_a_filter_partition,
-                   no_io, get_context);
+                   no_io, get_context, prefix_extractor);
 }
 
 BlockBasedTable::CachableEntry<FilterBlockReader> BlockBasedTable::GetFilter(
     FilePrefetchBuffer* prefetch_buffer, const BlockHandle& filter_blk_handle,
-    const bool is_a_filter_partition, bool no_io,
-    GetContext* get_context) const {
+    const bool is_a_filter_partition, bool no_io, GetContext* get_context,
+    const SliceTransform* prefix_extractor) const {
   // If cache_index_and_filter_blocks is false, filter should be pre-populated.
   // We will return rep_->filter anyway. rep_->filter can be nullptr if filter
   // read fails at Open() time. We don't want to reload again since it will
@@ -1329,8 +1362,8 @@ BlockBasedTable::CachableEntry<FilterBlockReader> BlockBasedTable::GetFilter(
     // Do not invoke any io.
     return CachableEntry<FilterBlockReader>();
   } else {
-    filter =
-        ReadFilter(prefetch_buffer, filter_blk_handle, is_a_filter_partition);
+    filter = ReadFilter(prefetch_buffer, filter_blk_handle,
+                        is_a_filter_partition, prefix_extractor);
     if (filter != nullptr) {
       Status s = block_cache->Insert(
           key, filter, filter->size(), &DeleteCachedFilterEntry, &cache_handle,
@@ -1362,18 +1395,23 @@ BlockBasedTable::CachableEntry<FilterBlockReader> BlockBasedTable::GetFilter(
   return { filter, cache_handle };
 }
 
+// disable_prefix_seek should be set to true when prefix_extractor found in SST
+// differs from the one in mutable_cf_options and index type is HashBasedIndex
 InternalIterator* BlockBasedTable::NewIndexIterator(
-    const ReadOptions& read_options, BlockIter* input_iter,
-    CachableEntry<IndexReader>* index_entry, GetContext* get_context) {
+    const ReadOptions& read_options, bool disable_prefix_seek,
+    BlockIter* input_iter, CachableEntry<IndexReader>* index_entry,
+    GetContext* get_context) {
   // index reader has already been pre-populated.
   if (rep_->index_reader) {
     return rep_->index_reader->NewIterator(
-        input_iter, read_options.total_order_seek, read_options.fill_cache);
+        input_iter, read_options.total_order_seek || disable_prefix_seek,
+        read_options.fill_cache);
   }
   // we have a pinned index block
   if (rep_->index_entry.IsSet()) {
     return rep_->index_entry.value->NewIterator(
-        input_iter, read_options.total_order_seek, read_options.fill_cache);
+        input_iter, read_options.total_order_seek || disable_prefix_seek,
+        read_options.fill_cache);
   }
 
   PERF_TIMER_GUARD(read_index_block_nanos);
@@ -1391,7 +1429,7 @@ InternalIterator* BlockBasedTable::NewIndexIterator(
 
   if (cache_handle == nullptr && no_io) {
     if (input_iter != nullptr) {
-      input_iter->SetStatus(Status::Incomplete("no blocking io"));
+      input_iter->Invalidate(Status::Incomplete("no blocking io"));
       return input_iter;
     } else {
       return NewErrorInternalIterator(Status::Incomplete("no blocking io"));
@@ -1438,7 +1476,7 @@ InternalIterator* BlockBasedTable::NewIndexIterator(
       RecordTick(statistics, BLOCK_CACHE_ADD_FAILURES);
       // make sure if something goes wrong, index_reader shall remain intact.
       if (input_iter != nullptr) {
-        input_iter->SetStatus(s);
+        input_iter->Invalidate(s);
         return input_iter;
       } else {
         return NewErrorInternalIterator(s);
@@ -1449,7 +1487,7 @@ InternalIterator* BlockBasedTable::NewIndexIterator(
 
   assert(cache_handle);
   auto* iter = index_reader->NewIterator(
-      input_iter, read_options.total_order_seek);
+      input_iter, read_options.total_order_seek || disable_prefix_seek);
 
   // the caller would like to take ownership of the index block
   // don't call RegisterCleanup() in this case, the caller will take care of it
@@ -1506,7 +1544,7 @@ BlockIter* BlockBasedTable::NewDataBlockIterator(
   if (s.ok() && block.value == nullptr) {
     if (no_io) {
       // Could not read from block_cache and can't do IO
-      iter->SetStatus(Status::Incomplete("no blocking io"));
+      iter->Invalidate(Status::Incomplete("no blocking io"));
       return iter;
     }
     std::unique_ptr<Block> block_value;
@@ -1566,7 +1604,7 @@ BlockIter* BlockBasedTable::NewDataBlockIterator(
     }
   } else {
     assert(block.value == nullptr);
-    iter->SetStatus(s);
+    iter->Invalidate(s);
   }
   return iter;
 }
@@ -1682,31 +1720,32 @@ BlockBasedTable::PartitionedIndexIteratorState::NewSecondaryIterator(
 // Otherwise, this method guarantees no I/O will be incurred.
 //
 // REQUIRES: this method shouldn't be called while the DB lock is held.
-bool BlockBasedTable::PrefixMayMatch(const Slice& internal_key) {
+bool BlockBasedTable::PrefixMayMatch(const Slice& internal_key,
+                                     const SliceTransform* prefix_extractor) {
   if (!rep_->filter_policy) {
     return true;
   }
 
-  assert(rep_->ioptions.prefix_extractor != nullptr);
+  assert(prefix_extractor != nullptr);
   auto user_key = ExtractUserKey(internal_key);
-  if (!rep_->ioptions.prefix_extractor->InDomain(user_key) ||
-      rep_->table_properties->prefix_extractor_name.compare(
-          rep_->ioptions.prefix_extractor->Name()) != 0) {
+  if (!prefix_extractor->InDomain(user_key)) {
     return true;
   }
-  auto prefix = rep_->ioptions.prefix_extractor->Transform(user_key);
+  assert(rep_->table_properties->prefix_extractor_name.compare(
+             prefix_extractor->Name()) == 0);
+  auto prefix = prefix_extractor->Transform(user_key);
 
   bool may_match = true;
   Status s;
 
   // First, try check with full filter
-  auto filter_entry = GetFilter();
+  auto filter_entry = GetFilter(prefix_extractor);
   FilterBlockReader* filter = filter_entry.value;
   if (filter != nullptr) {
     if (!filter->IsBlockBased()) {
       const Slice* const const_ikey_ptr = &internal_key;
-      may_match =
-          filter->PrefixMayMatch(prefix, kNotValid, false, const_ikey_ptr);
+      may_match = filter->PrefixMayMatch(prefix, prefix_extractor, kNotValid,
+                                         false, const_ikey_ptr);
     } else {
       InternalKey internal_key_prefix(prefix, kMaxSequenceNumber, kTypeValue);
       auto internal_prefix = internal_key_prefix.Encode();
@@ -1718,7 +1757,11 @@ bool BlockBasedTable::PrefixMayMatch(const Slice& internal_key) {
       no_io_read_options.read_tier = kBlockCacheTier;
 
       // Then, try find it within each block
-      unique_ptr<InternalIterator> iiter(NewIndexIterator(no_io_read_options));
+      // we already know prefix_extractor and prefix_extractor_name must match
+      // because `CheckPrefixMayMatch` first checks `check_filter_ == true`
+      bool prefix_extractor_changed = false;
+      unique_ptr<InternalIterator> iiter(
+          NewIndexIterator(no_io_read_options, prefix_extractor_changed));
       iiter->Seek(internal_prefix);
 
       if (!iiter->Valid()) {
@@ -1750,7 +1793,8 @@ bool BlockBasedTable::PrefixMayMatch(const Slice& internal_key) {
         BlockHandle handle;
         s = handle.DecodeFrom(&handle_value);
         assert(s.ok());
-        may_match = filter->PrefixMayMatch(prefix, handle.offset());
+        may_match =
+            filter->PrefixMayMatch(prefix, prefix_extractor, handle.offset());
       }
     }
   }
@@ -1772,7 +1816,7 @@ bool BlockBasedTable::PrefixMayMatch(const Slice& internal_key) {
 }
 
 void BlockBasedTableIterator::Seek(const Slice& target) {
-  if (!CheckPrefixMayMatch(target)) {
+  if (!CheckPrefixMayMatch(target, prefix_extractor_)) {
     ResetDataIter();
     return;
   }
@@ -1796,7 +1840,7 @@ void BlockBasedTableIterator::Seek(const Slice& target) {
 }
 
 void BlockBasedTableIterator::SeekForPrev(const Slice& target) {
-  if (!CheckPrefixMayMatch(target)) {
+  if (!CheckPrefixMayMatch(target, prefix_extractor_)) {
     ResetDataIter();
     return;
   }
@@ -1876,7 +1920,9 @@ void BlockBasedTableIterator::InitDataBlock() {
   BlockHandle data_block_handle;
   Slice handle_slice = index_iter_->value();
   if (!block_iter_points_to_real_block_ ||
-      handle_slice.compare(prev_index_value_) != 0) {
+      handle_slice.compare(prev_index_value_) != 0 ||
+      // if previous attempt of reading the block missed cache, try again
+      data_block_iter_.status().IsIncomplete()) {
     if (block_iter_points_to_real_block_) {
       ResetDataIter();
     }
@@ -1973,22 +2019,29 @@ void BlockBasedTableIterator::FindKeyBackward() {
   // code simplicity.
 }
 
-InternalIterator* BlockBasedTable::NewIterator(const ReadOptions& read_options,
-                                               Arena* arena,
-                                               bool skip_filters) {
+InternalIterator* BlockBasedTable::NewIterator(
+    const ReadOptions& read_options, const SliceTransform* prefix_extractor,
+    Arena* arena, bool skip_filters) {
+  bool prefix_extractor_changed =
+      PrefixExtractorChanged(rep_->table_properties, prefix_extractor);
   if (arena == nullptr) {
     return new BlockBasedTableIterator(
         this, read_options, rep_->internal_comparator,
-        NewIndexIterator(read_options),
+        NewIndexIterator(
+            read_options,
+            prefix_extractor_changed &&
+            rep_->index_type == BlockBasedTableOptions::kHashSearch),
         !skip_filters && !read_options.total_order_seek &&
-            rep_->ioptions.prefix_extractor != nullptr);
+        prefix_extractor != nullptr && !prefix_extractor_changed,
+        prefix_extractor);
   } else {
     auto* mem = arena->AllocateAligned(sizeof(BlockBasedTableIterator));
     return new (mem) BlockBasedTableIterator(
         this, read_options, rep_->internal_comparator,
-        NewIndexIterator(read_options),
+        NewIndexIterator(read_options, prefix_extractor_changed),
         !skip_filters && !read_options.total_order_seek &&
-            rep_->ioptions.prefix_extractor != nullptr);
+        prefix_extractor != nullptr && !prefix_extractor_changed,
+        prefix_extractor);
   }
 }
 
@@ -2022,10 +2075,10 @@ InternalIterator* BlockBasedTable::NewRangeTombstoneIterator(
   return NewDataBlockIterator(rep_, read_options, Slice(str));
 }
 
-bool BlockBasedTable::FullFilterKeyMayMatch(const ReadOptions& read_options,
-                                            FilterBlockReader* filter,
-                                            const Slice& internal_key,
-                                            const bool no_io) const {
+bool BlockBasedTable::FullFilterKeyMayMatch(
+    const ReadOptions& read_options, FilterBlockReader* filter,
+    const Slice& internal_key, const bool no_io,
+    const SliceTransform* prefix_extractor) const {
   if (filter == nullptr || filter->IsBlockBased()) {
     return true;
   }
@@ -2033,15 +2086,15 @@ bool BlockBasedTable::FullFilterKeyMayMatch(const ReadOptions& read_options,
   const Slice* const const_ikey_ptr = &internal_key;
   bool may_match = true;
   if (filter->whole_key_filtering()) {
-    may_match = filter->KeyMayMatch(user_key, kNotValid, no_io, const_ikey_ptr);
-  } else if (!read_options.total_order_seek &&
-             rep_->ioptions.prefix_extractor &&
+    may_match = filter->KeyMayMatch(user_key, prefix_extractor, kNotValid,
+                                    no_io, const_ikey_ptr);
+  } else if (!read_options.total_order_seek && prefix_extractor &&
              rep_->table_properties->prefix_extractor_name.compare(
-                 rep_->ioptions.prefix_extractor->Name()) == 0 &&
-             rep_->ioptions.prefix_extractor->InDomain(user_key) &&
-             !filter->PrefixMayMatch(
-                 rep_->ioptions.prefix_extractor->Transform(user_key),
-                 kNotValid, false, const_ikey_ptr)) {
+                 prefix_extractor->Name()) == 0 &&
+             prefix_extractor->InDomain(user_key) &&
+             !filter->PrefixMayMatch(prefix_extractor->Transform(user_key),
+                                     prefix_extractor, kNotValid, false,
+                                     const_ikey_ptr)) {
     may_match = false;
   }
   if (may_match) {
@@ -2051,25 +2104,36 @@ bool BlockBasedTable::FullFilterKeyMayMatch(const ReadOptions& read_options,
 }
 
 Status BlockBasedTable::Get(const ReadOptions& read_options, const Slice& key,
-                            GetContext* get_context, bool skip_filters) {
+                            GetContext* get_context,
+                            const SliceTransform* prefix_extractor,
+                            bool skip_filters) {
   Status s;
   const bool no_io = read_options.read_tier == kBlockCacheTier;
   CachableEntry<FilterBlockReader> filter_entry;
   if (!skip_filters) {
     filter_entry =
-        GetFilter(/*prefetch_buffer*/ nullptr,
+        GetFilter(prefix_extractor, /*prefetch_buffer*/ nullptr,
                   read_options.read_tier == kBlockCacheTier, get_context);
   }
   FilterBlockReader* filter = filter_entry.value;
 
   // First check the full filter
   // If full filter not useful, Then go into each block
-  if (!FullFilterKeyMayMatch(read_options, filter, key, no_io)) {
+  if (!FullFilterKeyMayMatch(read_options, filter, key, no_io,
+                             prefix_extractor)) {
     RecordTick(rep_->ioptions.statistics, BLOOM_FILTER_USEFUL);
   } else {
     BlockIter iiter_on_stack;
-    auto iiter = NewIndexIterator(read_options, &iiter_on_stack,
-                                  /* index_entry */ nullptr, get_context);
+    // if prefix_extractor found in block differs from options, disable
+    // BlockPrefixIndex. Only do this check when index_type is kHashSearch.
+    bool prefix_extractor_changed = false;
+    if (rep_->index_type == BlockBasedTableOptions::kHashSearch) {
+      prefix_extractor_changed =
+          PrefixExtractorChanged(rep_->table_properties, prefix_extractor);
+    }
+    auto iiter = NewIndexIterator(read_options, prefix_extractor_changed,
+                                  &iiter_on_stack, /* index_entry */ nullptr,
+                                  get_context);
     std::unique_ptr<InternalIterator> iiter_unique_ptr;
     if (iiter != &iiter_on_stack) {
       iiter_unique_ptr.reset(iiter);
@@ -2084,7 +2148,8 @@ Status BlockBasedTable::Get(const ReadOptions& read_options, const Slice& key,
       bool not_exist_in_filter =
           filter != nullptr && filter->IsBlockBased() == true &&
           handle.DecodeFrom(&handle_value).ok() &&
-          !filter->KeyMayMatch(ExtractUserKey(key), handle.offset(), no_io);
+          !filter->KeyMayMatch(ExtractUserKey(key), prefix_extractor,
+                               handle.offset(), no_io);
 
       if (not_exist_in_filter) {
         // Not found
@@ -2156,7 +2221,7 @@ Status BlockBasedTable::Prefetch(const Slice* const begin,
   }
 
   BlockIter iiter_on_stack;
-  auto iiter = NewIndexIterator(ReadOptions(), &iiter_on_stack);
+  auto iiter = NewIndexIterator(ReadOptions(), false, &iiter_on_stack);
   std::unique_ptr<InternalIterator> iiter_unique_ptr;
   if (iiter != &iiter_on_stack) {
     iiter_unique_ptr = std::unique_ptr<InternalIterator>(iiter);
@@ -2213,7 +2278,8 @@ Status BlockBasedTable::VerifyChecksum() {
   }
   // Check Data blocks
   BlockIter iiter_on_stack;
-  InternalIterator* iiter = NewIndexIterator(ReadOptions(), &iiter_on_stack);
+  InternalIterator* iiter =
+      NewIndexIterator(ReadOptions(), false, &iiter_on_stack);
   std::unique_ptr<InternalIterator> iiter_unique_ptr;
   if (iiter != &iiter_on_stack) {
     iiter_unique_ptr = std::unique_ptr<InternalIterator>(iiter);
@@ -2306,20 +2372,20 @@ Status BlockBasedTable::CreateIndexReader(
     if (pos != props.end()) {
       index_type_on_file = static_cast<BlockBasedTableOptions::IndexType>(
           DecodeFixed32(pos->second.c_str()));
+      // update index_type with the true type
+      rep_->index_type = index_type_on_file;
     }
   }
 
   auto file = rep_->file.get();
   const InternalKeyComparator* icomparator = &rep_->internal_comparator;
   const Footer& footer = rep_->footer;
-  if (index_type_on_file == BlockBasedTableOptions::kHashSearch &&
-      rep_->ioptions.prefix_extractor == nullptr) {
-    ROCKS_LOG_WARN(rep_->ioptions.info_log,
-                   "BlockBasedTableOptions::kHashSearch requires "
-                   "options.prefix_extractor to be set."
-                   " Fall back to binary search index.");
-    index_type_on_file = BlockBasedTableOptions::kBinarySearch;
-  }
+
+  // kHashSearch requires non-empty prefix_extractor but bypass checking
+  // prefix_extractor here since we have no access to MutableCFOptions.
+  // Add prefix_extractor_changed flag in  BlockBasedTable::NewIndexIterator.
+  // If prefix_extractor does not match prefix_extractor_name from table
+  // properties, turn off Hash Index by setting total_order_seek to true
 
   switch (index_type_on_file) {
     case BlockBasedTableOptions::kTwoLevelIndexSearch: {
@@ -2459,7 +2525,8 @@ Status BlockBasedTable::GetKVPairsFromDataBlocks(
   return Status::OK();
 }
 
-Status BlockBasedTable::DumpTable(WritableFile* out_file) {
+Status BlockBasedTable::DumpTable(WritableFile* out_file,
+                                  const SliceTransform* prefix_extractor) {
   // Output Footer
   out_file->Append(
       "Footer Details:\n"
@@ -2540,7 +2607,7 @@ Status BlockBasedTable::DumpTable(WritableFile* out_file) {
         s = block_fetcher.ReadBlockContents();
         if (!s.ok()) {
           rep_->filter.reset(new BlockBasedFilterBlockReader(
-              rep_->ioptions.prefix_extractor, table_options,
+              prefix_extractor, table_options,
               table_options.whole_key_filtering, std::move(block),
               rep_->ioptions.statistics));
         }
@@ -2625,7 +2692,6 @@ Status BlockBasedTable::DumpIndexBlock(WritableFile* out_file) {
   out_file->Append(
       "Index Details:\n"
       "--------------------------------------\n");
-
   std::unique_ptr<InternalIterator> blockhandles_iter(
       NewIndexIterator(ReadOptions()));
   Status s = blockhandles_iter->status();

table/block_based_table_reader.h

@@ -90,25 +90,29 @@ class BlockBasedTable : public TableReader {
                      const InternalKeyComparator& internal_key_comparator,
                      unique_ptr<RandomAccessFileReader>&& file,
                      uint64_t file_size, unique_ptr<TableReader>* table_reader,
+                     const SliceTransform* prefix_extractor = nullptr,
                      bool prefetch_index_and_filter_in_cache = true,
                      bool skip_filters = false, int level = -1);
 
-  bool PrefixMayMatch(const Slice& internal_key);
+  bool PrefixMayMatch(const Slice& internal_key,
+                      const SliceTransform* prefix_extractor = nullptr);
 
   // Returns a new iterator over the table contents.
   // The result of NewIterator() is initially invalid (caller must
   // call one of the Seek methods on the iterator before using it).
   // @param skip_filters Disables loading/accessing the filter block
-  InternalIterator* NewIterator(
-      const ReadOptions&, Arena* arena = nullptr,
-      bool skip_filters = false) override;
+  InternalIterator* NewIterator(const ReadOptions&,
+                                const SliceTransform* prefix_extractor,
+                                Arena* arena = nullptr,
+                                bool skip_filters = false) override;
 
   InternalIterator* NewRangeTombstoneIterator(
       const ReadOptions& read_options) override;
 
   // @param skip_filters Disables loading/accessing the filter block
   Status Get(const ReadOptions& readOptions, const Slice& key,
-             GetContext* get_context, bool skip_filters = false) override;
+             GetContext* get_context, const SliceTransform* prefix_extractor,
+             bool skip_filters = false) override;
 
   // Pre-fetch the disk blocks that correspond to the key range specified by
   // (kbegin, kend). The call will return error status in the event of
@@ -136,7 +140,8 @@ class BlockBasedTable : public TableReader {
   size_t ApproximateMemoryUsage() const override;
 
   // convert SST file to a human readable form
-  Status DumpTable(WritableFile* out_file) override;
+  Status DumpTable(WritableFile* out_file,
+                   const SliceTransform* prefix_extractor = nullptr) override;
 
   Status VerifyChecksum() override;
 
@@ -253,12 +258,13 @@ class BlockBasedTable : public TableReader {
   // if `no_io == true`, we will not try to read filter/index from sst file
   // were they not present in cache yet.
   CachableEntry<FilterBlockReader> GetFilter(
+      const SliceTransform* prefix_extractor = nullptr,
       FilePrefetchBuffer* prefetch_buffer = nullptr, bool no_io = false,
       GetContext* get_context = nullptr) const;
   virtual CachableEntry<FilterBlockReader> GetFilter(
       FilePrefetchBuffer* prefetch_buffer, const BlockHandle& filter_blk_handle,
-      const bool is_a_filter_partition, bool no_io,
-      GetContext* get_context) const;
+      const bool is_a_filter_partition, bool no_io, GetContext* get_context,
+      const SliceTransform* prefix_extractor = nullptr) const;
 
   // Get the iterator from the index reader.
   // If input_iter is not set, return new Iterator
@@ -271,7 +277,8 @@ class BlockBasedTable : public TableReader {
   //  3. We disallowed any io to be performed, that is, read_options ==
   //     kBlockCacheTier
   InternalIterator* NewIndexIterator(
-      const ReadOptions& read_options, BlockIter* input_iter = nullptr,
+      const ReadOptions& read_options, bool prefix_extractor_changed = false,
+      BlockIter* input_iter = nullptr,
       CachableEntry<IndexReader>* index_entry = nullptr,
       GetContext* get_context = nullptr);
 
@@ -325,9 +332,10 @@ class BlockBasedTable : public TableReader {
       InternalIterator* preloaded_meta_index_iter = nullptr,
       const int level = -1);
 
-  bool FullFilterKeyMayMatch(const ReadOptions& read_options,
-                             FilterBlockReader* filter, const Slice& user_key,
-                             const bool no_io) const;
+  bool FullFilterKeyMayMatch(
+      const ReadOptions& read_options, FilterBlockReader* filter,
+      const Slice& user_key, const bool no_io,
+      const SliceTransform* prefix_extractor = nullptr) const;
 
   // Read the meta block from sst.
   static Status ReadMetaBlock(Rep* rep, FilePrefetchBuffer* prefetch_buffer,
@@ -337,9 +345,10 @@ class BlockBasedTable : public TableReader {
   Status VerifyChecksumInBlocks(InternalIterator* index_iter);
 
   // Create the filter from the filter block.
-  FilterBlockReader* ReadFilter(FilePrefetchBuffer* prefetch_buffer,
-                                const BlockHandle& filter_handle,
-                                const bool is_a_filter_partition) const;
+  FilterBlockReader* ReadFilter(
+      FilePrefetchBuffer* prefetch_buffer, const BlockHandle& filter_handle,
+      const bool is_a_filter_partition,
+      const SliceTransform* prefix_extractor = nullptr) const;
 
   static void SetupCacheKeyPrefix(Rep* rep, uint64_t file_size);
 
@@ -499,14 +508,16 @@ class BlockBasedTableIterator : public InternalIterator {
   BlockBasedTableIterator(BlockBasedTable* table,
                           const ReadOptions& read_options,
                           const InternalKeyComparator& icomp,
-                          InternalIterator* index_iter, bool check_filter)
+                          InternalIterator* index_iter, bool check_filter,
+                          const SliceTransform* prefix_extractor)
       : table_(table),
         read_options_(read_options),
         icomp_(icomp),
         index_iter_(index_iter),
         pinned_iters_mgr_(nullptr),
         block_iter_points_to_real_block_(false),
-        check_filter_(check_filter) {}
+        check_filter_(check_filter),
+        prefix_extractor_(prefix_extractor) {}
 
   ~BlockBasedTableIterator() { delete index_iter_; }
 
@@ -528,11 +539,9 @@ class BlockBasedTableIterator : public InternalIterator {
     return data_block_iter_.value();
   }
   Status status() const override {
-    // It'd be nice if status() returned a const Status& instead of a Status
     if (!index_iter_->status().ok()) {
       return index_iter_->status();
-    } else if (block_iter_points_to_real_block_ &&
-               !data_block_iter_.status().ok()) {
+    } else if (block_iter_points_to_real_block_) {
       return data_block_iter_.status();
     } else {
       return Status::OK();
@@ -554,8 +563,9 @@ class BlockBasedTableIterator : public InternalIterator {
            block_iter_points_to_real_block_;
   }
 
-  bool CheckPrefixMayMatch(const Slice& ikey) {
-    if (check_filter_ && !table_->PrefixMayMatch(ikey)) {
+  bool CheckPrefixMayMatch(const Slice& ikey,
+                           const SliceTransform* prefix_extractor = nullptr) {
+    if (check_filter_ && !table_->PrefixMayMatch(ikey, prefix_extractor)) {
       // TODO remember the iterator is invalidated because of prefix
       // match. This can avoid the upper level file iterator to falsely
       // believe the position is the end of the SST file and move to
@@ -571,8 +581,7 @@ class BlockBasedTableIterator : public InternalIterator {
       if (pinned_iters_mgr_ != nullptr && pinned_iters_mgr_->PinningEnabled()) {
         data_block_iter_.DelegateCleanupsTo(pinned_iters_mgr_);
       }
-      data_block_iter_.~BlockIter();
-      new (&data_block_iter_) BlockIter();
+      data_block_iter_.Invalidate(Status::OK());
       block_iter_points_to_real_block_ = false;
     }
   }
@@ -602,6 +611,7 @@ class BlockBasedTableIterator : public InternalIterator {
   bool check_filter_;
   // TODO use block offset instead
   std::string prev_index_value_;
+  const SliceTransform* prefix_extractor_;
 
   static const size_t kInitReadaheadSize = 8 * 1024;
   // Found that 256 KB readahead size provides the best performance, based on

table/cuckoo_table_reader.cc

@@ -141,6 +141,7 @@ CuckooTableReader::CuckooTableReader(
 
 Status CuckooTableReader::Get(const ReadOptions& /*readOptions*/,
                               const Slice& key, GetContext* get_context,
+                              const SliceTransform* /* prefix_extractor */,
                               bool /*skip_filters*/) {
   assert(key.size() == key_length_ + (is_last_level_ ? 8 : 0));
   Slice user_key = ExtractUserKey(key);
@@ -206,7 +207,7 @@ class CuckooTableIterator : public InternalIterator {
   void Prev() override;
   Slice key() const override;
   Slice value() const override;
-  Status status() const override { return status_; }
+  Status status() const override { return Status::OK(); }
   void InitIfNeeded();
 
  private:
@@ -241,7 +242,6 @@ class CuckooTableIterator : public InternalIterator {
   void PrepareKVAtCurrIdx();
   CuckooTableReader* reader_;
   bool initialized_;
-  Status status_;
   // Contains a map of keys to bucket_id sorted in key order.
   std::vector<uint32_t> sorted_bucket_ids_;
   // We assume that the number of items can be stored in uint32 (4 Billion).
@@ -378,7 +378,9 @@ extern InternalIterator* NewErrorInternalIterator(const Status& status,
                                                   Arena* arena);
 
 InternalIterator* CuckooTableReader::NewIterator(
-    const ReadOptions& /*read_options*/, Arena* arena, bool /*skip_filters*/) {
+    const ReadOptions& /*read_options*/,
+    const SliceTransform* /* prefix_extractor */, Arena* arena,
+    bool /*skip_filters*/) {
   if (!status().ok()) {
     return NewErrorInternalIterator(
         Status::Corruption("CuckooTableReader status is not okay."), arena);

table/cuckoo_table_reader.h

@@ -42,12 +42,14 @@ class CuckooTableReader: public TableReader {
 
   Status status() const { return status_; }
 
-  Status Get(const ReadOptions& read_options, const Slice& key,
-             GetContext* get_context, bool skip_filters = false) override;
+  Status Get(const ReadOptions& readOptions, const Slice& key,
+             GetContext* get_context, const SliceTransform* prefix_extractor,
+             bool skip_filters = false) override;
 
-  InternalIterator* NewIterator(
-      const ReadOptions&, Arena* arena = nullptr,
-      bool skip_filters = false) override;
+  InternalIterator* NewIterator(const ReadOptions&,
+                                const SliceTransform* prefix_extractor,
+                                Arena* arena = nullptr,
+                                bool skip_filters = false) override;
   void Prepare(const Slice& target) override;
 
   // Report an approximation of how much memory has been used.

table/cuckoo_table_reader_test.cc

@@ -127,7 +127,8 @@ class CuckooReaderTest : public testing::Test {
       GetContext get_context(ucomp, nullptr, nullptr, nullptr,
                              GetContext::kNotFound, Slice(user_keys[i]), &value,
                              nullptr, nullptr, nullptr, nullptr);
-      ASSERT_OK(reader.Get(ReadOptions(), Slice(keys[i]), &get_context));
+      ASSERT_OK(
+          reader.Get(ReadOptions(), Slice(keys[i]), &get_context, nullptr));
       ASSERT_STREQ(values[i].c_str(), value.data());
     }
   }
@@ -149,7 +150,8 @@ class CuckooReaderTest : public testing::Test {
     CuckooTableReader reader(ioptions, std::move(file_reader), file_size, ucomp,
                              GetSliceHash);
     ASSERT_OK(reader.status());
-    InternalIterator* it = reader.NewIterator(ReadOptions(), nullptr);
+    InternalIterator* it =
+        reader.NewIterator(ReadOptions(), nullptr, nullptr, false);
     ASSERT_OK(it->status());
     ASSERT_TRUE(!it->Valid());
     it->SeekToFirst();
@@ -188,7 +190,7 @@ class CuckooReaderTest : public testing::Test {
     delete it;
 
     Arena arena;
-    it = reader.NewIterator(ReadOptions(), &arena);
+    it = reader.NewIterator(ReadOptions(), nullptr, &arena);
     ASSERT_OK(it->status());
     ASSERT_TRUE(!it->Valid());
     it->Seek(keys[num_items/2]);
@@ -337,7 +339,8 @@ TEST_F(CuckooReaderTest, WhenKeyNotFound) {
   GetContext get_context(ucmp, nullptr, nullptr, nullptr, GetContext::kNotFound,
                          Slice(not_found_key), &value, nullptr, nullptr,
                          nullptr, nullptr);
-  ASSERT_OK(reader.Get(ReadOptions(), Slice(not_found_key), &get_context));
+  ASSERT_OK(
+      reader.Get(ReadOptions(), Slice(not_found_key), &get_context, nullptr));
   ASSERT_TRUE(value.empty());
   ASSERT_OK(reader.status());
   // Search for a key with an independent hash value.
@@ -350,7 +353,8 @@ TEST_F(CuckooReaderTest, WhenKeyNotFound) {
   GetContext get_context2(ucmp, nullptr, nullptr, nullptr,
                           GetContext::kNotFound, Slice(not_found_key2), &value,
                           nullptr, nullptr, nullptr, nullptr);
-  ASSERT_OK(reader.Get(ReadOptions(), Slice(not_found_key2), &get_context2));
+  ASSERT_OK(
+      reader.Get(ReadOptions(), Slice(not_found_key2), &get_context2, nullptr));
   ASSERT_TRUE(value.empty());
   ASSERT_OK(reader.status());
 
@@ -365,7 +369,8 @@ TEST_F(CuckooReaderTest, WhenKeyNotFound) {
   GetContext get_context3(ucmp, nullptr, nullptr, nullptr,
                           GetContext::kNotFound, Slice(unused_key), &value,
                           nullptr, nullptr, nullptr, nullptr);
-  ASSERT_OK(reader.Get(ReadOptions(), Slice(unused_key), &get_context3));
+  ASSERT_OK(
+      reader.Get(ReadOptions(), Slice(unused_key), &get_context3, nullptr));
   ASSERT_TRUE(value.empty());
   ASSERT_OK(reader.status());
 }
@@ -443,7 +448,7 @@ void WriteFile(const std::vector<std::string>& keys,
   for (uint64_t i = 0; i < num; ++i) {
     value.Reset();
     value.clear();
-    ASSERT_OK(reader.Get(r_options, Slice(keys[i]), &get_context));
+    ASSERT_OK(reader.Get(r_options, Slice(keys[i]), &get_context, nullptr));
     ASSERT_TRUE(Slice(keys[i]) == Slice(&keys[i][0], 4));
   }
 }
@@ -496,13 +501,13 @@ void ReadKeys(uint64_t num, uint32_t batch_size) {
       }
       for (uint64_t j = i; j < i+batch_size && j < num; ++j) {
         reader.Get(r_options, Slice(reinterpret_cast<char*>(&keys[j]), 16),
-                   &get_context);
+                   &get_context, nullptr);
       }
     }
   } else {
     for (uint64_t i = 0; i < num; i++) {
       reader.Get(r_options, Slice(reinterpret_cast<char*>(&keys[i]), 16),
-                 &get_context);
+                 &get_context, nullptr);
     }
   }
   float time_per_op = (env->NowMicros() - start_time) * 1.0f / num;

table/filter_block.h

@@ -93,16 +93,21 @@ class FilterBlockReader {
    * built upon InternalKey and must be provided via const_ikey_ptr when running
    * queries.
    */
-  virtual bool KeyMayMatch(const Slice& key, uint64_t block_offset = kNotValid,
+  virtual bool KeyMayMatch(const Slice& key,
+                           const SliceTransform* prefix_extractor,
+                           uint64_t block_offset = kNotValid,
                            const bool no_io = false,
                            const Slice* const const_ikey_ptr = nullptr) = 0;
+
   /**
    * no_io and const_ikey_ptr here means the same as in KeyMayMatch
    */
   virtual bool PrefixMayMatch(const Slice& prefix,
+                              const SliceTransform* prefix_extractor,
                               uint64_t block_offset = kNotValid,
                               const bool no_io = false,
                               const Slice* const const_ikey_ptr = nullptr) = 0;
+
   virtual size_t ApproximateMemoryUsage() const = 0;
   virtual size_t size() const { return size_; }
   virtual Statistics* statistics() const { return statistics_; }
@@ -115,7 +120,8 @@ class FilterBlockReader {
     return error_msg;
   }
 
-  virtual void CacheDependencies(bool /*pin*/) {}
+  virtual void CacheDependencies(bool /*pin*/,
+                                 const SliceTransform* /*prefix_extractor*/) {}
 
  protected:
   bool whole_key_filtering_;

table/full_filter_block.cc

@@ -103,9 +103,10 @@ FullFilterBlockReader::FullFilterBlockReader(
   block_contents_ = std::move(contents);
 }
 
-bool FullFilterBlockReader::KeyMayMatch(const Slice& key, uint64_t block_offset,
-                                        const bool /*no_io*/,
-                                        const Slice* const /*const_ikey_ptr*/) {
+bool FullFilterBlockReader::KeyMayMatch(
+    const Slice& key, const SliceTransform* /*prefix_extractor*/,
+    uint64_t block_offset, const bool /*no_io*/,
+    const Slice* const /*const_ikey_ptr*/) {
 #ifdef NDEBUG
   (void)block_offset;
 #endif
@@ -117,7 +118,8 @@ bool FullFilterBlockReader::KeyMayMatch(const Slice& key, uint64_t block_offset,
 }
 
 bool FullFilterBlockReader::PrefixMayMatch(
-    const Slice& prefix, uint64_t block_offset, const bool /*no_io*/,
+    const Slice& prefix, const SliceTransform* /* prefix_extractor */,
+    uint64_t block_offset, const bool /*no_io*/,
     const Slice* const /*const_ikey_ptr*/) {
 #ifdef NDEBUG
   (void)block_offset;

table/full_filter_block.h

@@ -96,13 +96,15 @@ class FullFilterBlockReader : public FilterBlockReader {
   ~FullFilterBlockReader() {}
 
   virtual bool IsBlockBased() override { return false; }
+
   virtual bool KeyMayMatch(
-      const Slice& key, uint64_t block_offset = kNotValid,
-      const bool no_io = false,
+      const Slice& key, const SliceTransform* prefix_extractor,
+      uint64_t block_offset = kNotValid, const bool no_io = false,
       const Slice* const const_ikey_ptr = nullptr) override;
+
   virtual bool PrefixMayMatch(
-      const Slice& prefix, uint64_t block_offset = kNotValid,
-      const bool no_io = false,
+      const Slice& prefix, const SliceTransform* prefix_extractor,
+      uint64_t block_offset = kNotValid, const bool no_io = false,
       const Slice* const const_ikey_ptr = nullptr) override;
   virtual size_t ApproximateMemoryUsage() const override;
 

table/full_filter_block_test.cc

@@ -113,7 +113,7 @@ TEST_F(PluginFullFilterBlockTest, PluginEmptyBuilder) {
       nullptr, true, block,
       table_options_.filter_policy->GetFilterBitsReader(block), nullptr);
   // Remain same symantic with blockbased filter
-  ASSERT_TRUE(reader.KeyMayMatch("foo"));
+  ASSERT_TRUE(reader.KeyMayMatch("foo", nullptr));
 }
 
 TEST_F(PluginFullFilterBlockTest, PluginSingleChunk) {
@@ -128,13 +128,13 @@ TEST_F(PluginFullFilterBlockTest, PluginSingleChunk) {
   FullFilterBlockReader reader(
       nullptr, true, block,
       table_options_.filter_policy->GetFilterBitsReader(block), nullptr);
-  ASSERT_TRUE(reader.KeyMayMatch("foo"));
-  ASSERT_TRUE(reader.KeyMayMatch("bar"));
-  ASSERT_TRUE(reader.KeyMayMatch("box"));
-  ASSERT_TRUE(reader.KeyMayMatch("hello"));
-  ASSERT_TRUE(reader.KeyMayMatch("foo"));
-  ASSERT_TRUE(!reader.KeyMayMatch("missing"));
-  ASSERT_TRUE(!reader.KeyMayMatch("other"));
+  ASSERT_TRUE(reader.KeyMayMatch("foo", nullptr));
+  ASSERT_TRUE(reader.KeyMayMatch("bar", nullptr));
+  ASSERT_TRUE(reader.KeyMayMatch("box", nullptr));
+  ASSERT_TRUE(reader.KeyMayMatch("hello", nullptr));
+  ASSERT_TRUE(reader.KeyMayMatch("foo", nullptr));
+  ASSERT_TRUE(!reader.KeyMayMatch("missing", nullptr));
+  ASSERT_TRUE(!reader.KeyMayMatch("other", nullptr));
 }
 
 class FullFilterBlockTest : public testing::Test {
@@ -158,7 +158,7 @@ TEST_F(FullFilterBlockTest, EmptyBuilder) {
       nullptr, true, block,
       table_options_.filter_policy->GetFilterBitsReader(block), nullptr);
   // Remain same symantic with blockbased filter
-  ASSERT_TRUE(reader.KeyMayMatch("foo"));
+  ASSERT_TRUE(reader.KeyMayMatch("foo", nullptr));
 }
 
 TEST_F(FullFilterBlockTest, DuplicateEntries) {
@@ -208,13 +208,13 @@ TEST_F(FullFilterBlockTest, SingleChunk) {
   FullFilterBlockReader reader(
       nullptr, true, block,
       table_options_.filter_policy->GetFilterBitsReader(block), nullptr);
-  ASSERT_TRUE(reader.KeyMayMatch("foo"));
-  ASSERT_TRUE(reader.KeyMayMatch("bar"));
-  ASSERT_TRUE(reader.KeyMayMatch("box"));
-  ASSERT_TRUE(reader.KeyMayMatch("hello"));
-  ASSERT_TRUE(reader.KeyMayMatch("foo"));
-  ASSERT_TRUE(!reader.KeyMayMatch("missing"));
-  ASSERT_TRUE(!reader.KeyMayMatch("other"));
+  ASSERT_TRUE(reader.KeyMayMatch("foo", nullptr));
+  ASSERT_TRUE(reader.KeyMayMatch("bar", nullptr));
+  ASSERT_TRUE(reader.KeyMayMatch("box", nullptr));
+  ASSERT_TRUE(reader.KeyMayMatch("hello", nullptr));
+  ASSERT_TRUE(reader.KeyMayMatch("foo", nullptr));
+  ASSERT_TRUE(!reader.KeyMayMatch("missing", nullptr));
+  ASSERT_TRUE(!reader.KeyMayMatch("other", nullptr));
 }
 
 }  // namespace rocksdb

table/internal_iterator.h

@@ -22,6 +22,7 @@ class InternalIterator : public Cleanable {
 
   // An iterator is either positioned at a key/value pair, or
   // not valid.  This method returns true iff the iterator is valid.
+  // Always returns false if !status().ok().
   virtual bool Valid() const = 0;
 
   // Position at the first key in the source.  The iterator is Valid()
@@ -35,6 +36,9 @@ class InternalIterator : public Cleanable {
   // Position at the first key in the source that at or past target
   // The iterator is Valid() after this call iff the source contains
   // an entry that comes at or past target.
+  // All Seek*() methods clear any error status() that the iterator had prior to
+  // the call; after the seek, status() indicates only the error (if any) that
+  // happened during the seek, not any past errors.
   virtual void Seek(const Slice& target) = 0;
 
   // Position at the first key in the source that at or before target
@@ -61,7 +65,7 @@ class InternalIterator : public Cleanable {
   // Return the value for the current entry.  The underlying storage for
   // the returned slice is valid only until the next modification of
   // the iterator.
-  // REQUIRES: !AtEnd() && !AtStart()
+  // REQUIRES: Valid()
   virtual Slice value() const = 0;
 
   // If an error has occurred, return it.  Else return an ok status.

table/iterator_wrapper.h

@@ -87,6 +87,7 @@ class IteratorWrapper {
     valid_ = iter_->Valid();
     if (valid_) {
       key_ = iter_->key();
+      assert(iter_->status().ok());
     }
   }
 

table/merging_iterator.cc

@@ -52,12 +52,21 @@ class MergingIterator : public InternalIterator {
     }
     for (auto& child : children_) {
       if (child.Valid()) {
+        assert(child.status().ok());
         minHeap_.push(&child);
+      } else {
+        considerStatus(child.status());
       }
     }
     current_ = CurrentForward();
   }
 
+  void considerStatus(Status s) {
+    if (!s.ok() && status_.ok()) {
+      status_ = s;
+    }
+  }
+
   virtual void AddIterator(InternalIterator* iter) {
     assert(direction_ == kForward);
     children_.emplace_back(iter);
@@ -66,8 +75,11 @@ class MergingIterator : public InternalIterator {
     }
     auto new_wrapper = children_.back();
     if (new_wrapper.Valid()) {
+      assert(new_wrapper.status().ok());
       minHeap_.push(&new_wrapper);
       current_ = CurrentForward();
+    } else {
+      considerStatus(new_wrapper.status());
     }
   }
 
@@ -77,14 +89,22 @@ class MergingIterator : public InternalIterator {
     }
   }
 
-  virtual bool Valid() const override { return (current_ != nullptr); }
+  virtual bool Valid() const override {
+    return current_ != nullptr && status_.ok();
+  }
+
+  virtual Status status() const override { return status_; }
 
   virtual void SeekToFirst() override {
     ClearHeaps();
+    status_ = Status::OK();
     for (auto& child : children_) {
       child.SeekToFirst();
       if (child.Valid()) {
+        assert(child.status().ok());
         minHeap_.push(&child);
+      } else {
+        considerStatus(child.status());
       }
     }
     direction_ = kForward;
@@ -94,10 +114,14 @@ class MergingIterator : public InternalIterator {
   virtual void SeekToLast() override {
     ClearHeaps();
     InitMaxHeap();
+    status_ = Status::OK();
     for (auto& child : children_) {
       child.SeekToLast();
       if (child.Valid()) {
+        assert(child.status().ok());
         maxHeap_->push(&child);
+      } else {
+        considerStatus(child.status());
       }
     }
     direction_ = kReverse;
@@ -106,6 +130,7 @@ class MergingIterator : public InternalIterator {
 
   virtual void Seek(const Slice& target) override {
     ClearHeaps();
+    status_ = Status::OK();
     for (auto& child : children_) {
       {
         PERF_TIMER_GUARD(seek_child_seek_time);
@@ -114,8 +139,11 @@ class MergingIterator : public InternalIterator {
       PERF_COUNTER_ADD(seek_child_seek_count, 1);
 
       if (child.Valid()) {
+        assert(child.status().ok());
         PERF_TIMER_GUARD(seek_min_heap_time);
         minHeap_.push(&child);
+      } else {
+        considerStatus(child.status());
       }
     }
     direction_ = kForward;
@@ -128,6 +156,7 @@ class MergingIterator : public InternalIterator {
   virtual void SeekForPrev(const Slice& target) override {
     ClearHeaps();
     InitMaxHeap();
+    status_ = Status::OK();
 
     for (auto& child : children_) {
       {
@@ -137,8 +166,11 @@ class MergingIterator : public InternalIterator {
       PERF_COUNTER_ADD(seek_child_seek_count, 1);
 
       if (child.Valid()) {
+        assert(child.status().ok());
         PERF_TIMER_GUARD(seek_max_heap_time);
         maxHeap_->push(&child);
+      } else {
+        considerStatus(child.status());
       }
     }
     direction_ = kReverse;
@@ -172,9 +204,11 @@ class MergingIterator : public InternalIterator {
       // current is still valid after the Next() call above.  Call
       // replace_top() to restore the heap property.  When the same child
       // iterator yields a sequence of keys, this is cheap.
+      assert(current_->status().ok());
       minHeap_.replace_top(current_);
     } else {
       // current stopped being valid, remove it from the heap.
+      considerStatus(current_->status());
       minHeap_.pop();
     }
     current_ = CurrentForward();
@@ -191,28 +225,35 @@ class MergingIterator : public InternalIterator {
       // just after the if-block.
       ClearHeaps();
       InitMaxHeap();
+      Slice target = key();
       for (auto& child : children_) {
         if (&child != current_) {
           if (!prefix_seek_mode_) {
-            child.Seek(key());
+            child.Seek(target);
             if (child.Valid()) {
               // Child is at first entry >= key().  Step back one to be < key()
               TEST_SYNC_POINT_CALLBACK("MergeIterator::Prev:BeforePrev",
                                        &child);
+              assert(child.status().ok());
               child.Prev();
             } else {
               // Child has no entries >= key().  Position at last entry.
               TEST_SYNC_POINT("MergeIterator::Prev:BeforeSeekToLast");
+              considerStatus(child.status());
               child.SeekToLast();
             }
+            considerStatus(child.status());
           } else {
-            child.SeekForPrev(key());
-            if (child.Valid() && comparator_->Equal(key(), child.key())) {
+            child.SeekForPrev(target);
+            considerStatus(child.status());
+            if (child.Valid() && comparator_->Equal(target, child.key())) {
               child.Prev();
+              considerStatus(child.status());
             }
           }
         }
         if (child.Valid()) {
+          assert(child.status().ok());
           maxHeap_->push(&child);
         }
       }
@@ -238,9 +279,11 @@ class MergingIterator : public InternalIterator {
       // current is still valid after the Prev() call above.  Call
       // replace_top() to restore the heap property.  When the same child
       // iterator yields a sequence of keys, this is cheap.
+      assert(current_->status().ok());
       maxHeap_->replace_top(current_);
     } else {
       // current stopped being valid, remove it from the heap.
+      considerStatus(current_->status());
       maxHeap_->pop();
     }
     current_ = CurrentReverse();
@@ -256,17 +299,6 @@ class MergingIterator : public InternalIterator {
     return current_->value();
   }
 
-  virtual Status status() const override {
-    Status s;
-    for (auto& child : children_) {
-      s = child.status();
-      if (!s.ok()) {
-        break;
-      }
-    }
-    return s;
-  }
-
   virtual void SetPinnedItersMgr(
       PinnedIteratorsManager* pinned_iters_mgr) override {
     pinned_iters_mgr_ = pinned_iters_mgr;
@@ -302,6 +334,8 @@ class MergingIterator : public InternalIterator {
   // child iterators are valid.  This is the top of minHeap_ or maxHeap_
   // depending on the direction.
   IteratorWrapper* current_;
+  // If any of the children have non-ok status, this is one of them.
+  Status status_;
   // Which direction is the iterator moving?
   enum Direction {
     kForward,
@@ -334,11 +368,14 @@ void MergingIterator::SwitchToForward() {
   // Otherwise, advance the non-current children.  We advance current_
   // just after the if-block.
   ClearHeaps();
+  Slice target = key();
   for (auto& child : children_) {
     if (&child != current_) {
-      child.Seek(key());
-      if (child.Valid() && comparator_->Equal(key(), child.key())) {
+      child.Seek(target);
+      considerStatus(child.status());
+      if (child.Valid() && comparator_->Equal(target, child.key())) {
         child.Next();
+        considerStatus(child.status());
       }
     }
     if (child.Valid()) {

table/mock_table.cc

@@ -26,14 +26,16 @@ stl_wrappers::KVMap MakeMockFile(
   return stl_wrappers::KVMap(l, stl_wrappers::LessOfComparator(&icmp_));
 }
 
-InternalIterator* MockTableReader::NewIterator(const ReadOptions&,
-                                               Arena* /*arena*/,
-                                               bool /*skip_filters*/) {
+InternalIterator* MockTableReader::NewIterator(
+    const ReadOptions&, const SliceTransform* /* prefix_extractor */,
+    Arena* /*arena*/, bool /*skip_filters*/) {
   return new MockTableIterator(table_);
 }
 
 Status MockTableReader::Get(const ReadOptions&, const Slice& key,
-                            GetContext* get_context, bool /*skip_filters*/) {
+                            GetContext* get_context,
+                            const SliceTransform* /*prefix_extractor*/,
+                            bool /*skip_filters*/) {
   std::unique_ptr<MockTableIterator> iter(new MockTableIterator(table_));
   for (iter->Seek(key); iter->Valid(); iter->Next()) {
     ParsedInternalKey parsed_key;

table/mock_table.h

@@ -39,10 +39,12 @@ class MockTableReader : public TableReader {
   explicit MockTableReader(const stl_wrappers::KVMap& table) : table_(table) {}
 
   InternalIterator* NewIterator(const ReadOptions&,
-                                Arena* arena,
+                                const SliceTransform* prefix_extractor,
+                                Arena* arena = nullptr,
                                 bool skip_filters = false) override;
 
-  Status Get(const ReadOptions&, const Slice& key, GetContext* get_context,
+  Status Get(const ReadOptions& readOptions, const Slice& key,
+             GetContext* get_context, const SliceTransform* prefix_extractor,
              bool skip_filters = false) override;
 
   uint64_t ApproximateOffsetOf(const Slice& /*key*/) override { return 0; }

table/partitioned_filter_block.cc

@@ -130,7 +130,8 @@ PartitionedFilterBlockReader::~PartitionedFilterBlockReader() {
 }
 
 bool PartitionedFilterBlockReader::KeyMayMatch(
-    const Slice& key, uint64_t block_offset, const bool no_io,
+    const Slice& key, const SliceTransform* prefix_extractor,
+    uint64_t block_offset, const bool no_io,
     const Slice* const const_ikey_ptr) {
   assert(const_ikey_ptr != nullptr);
   assert(block_offset == kNotValid);
@@ -145,12 +146,14 @@ bool PartitionedFilterBlockReader::KeyMayMatch(
     return false;
   }
   bool cached = false;
-  auto filter_partition = GetFilterPartition(nullptr /* prefetch_buffer */,
-                                             &filter_handle, no_io, &cached);
+  auto filter_partition =
+      GetFilterPartition(nullptr /* prefetch_buffer */, &filter_handle, no_io,
+                         &cached, prefix_extractor);
   if (UNLIKELY(!filter_partition.value)) {
     return true;
   }
-  auto res = filter_partition.value->KeyMayMatch(key, block_offset, no_io);
+  auto res = filter_partition.value->KeyMayMatch(key, prefix_extractor,
+                                                 block_offset, no_io);
   if (cached) {
     return res;
   }
@@ -163,7 +166,8 @@ bool PartitionedFilterBlockReader::KeyMayMatch(
 }
 
 bool PartitionedFilterBlockReader::PrefixMayMatch(
-    const Slice& prefix, uint64_t block_offset, const bool no_io,
+    const Slice& prefix, const SliceTransform* prefix_extractor,
+    uint64_t block_offset, const bool no_io,
     const Slice* const const_ikey_ptr) {
 #ifdef NDEBUG
   (void)block_offset;
@@ -181,12 +185,14 @@ bool PartitionedFilterBlockReader::PrefixMayMatch(
     return false;
   }
   bool cached = false;
-  auto filter_partition = GetFilterPartition(nullptr /* prefetch_buffer */,
-                                             &filter_handle, no_io, &cached);
+  auto filter_partition =
+      GetFilterPartition(nullptr /* prefetch_buffer */, &filter_handle, no_io,
+                         &cached, prefix_extractor);
   if (UNLIKELY(!filter_partition.value)) {
     return true;
   }
-  auto res = filter_partition.value->PrefixMayMatch(prefix, kNotValid, no_io);
+  auto res = filter_partition.value->PrefixMayMatch(prefix, prefix_extractor,
+                                                    kNotValid, no_io);
   if (cached) {
     return res;
   }
@@ -214,7 +220,7 @@ Slice PartitionedFilterBlockReader::GetFilterPartitionHandle(
 BlockBasedTable::CachableEntry<FilterBlockReader>
 PartitionedFilterBlockReader::GetFilterPartition(
     FilePrefetchBuffer* prefetch_buffer, Slice* handle_value, const bool no_io,
-    bool* cached) {
+    bool* cached, const SliceTransform* prefix_extractor) {
   BlockHandle fltr_blk_handle;
   auto s = fltr_blk_handle.DecodeFrom(handle_value);
   assert(s.ok());
@@ -237,10 +243,10 @@ PartitionedFilterBlockReader::GetFilterPartition(
     }
     return table_->GetFilter(/*prefetch_buffer*/ nullptr, fltr_blk_handle,
                              is_a_filter_partition, no_io,
-                             /* get_context */ nullptr);
+                             /* get_context */ nullptr, prefix_extractor);
   } else {
     auto filter = table_->ReadFilter(prefetch_buffer, fltr_blk_handle,
-                                     is_a_filter_partition);
+                                     is_a_filter_partition, prefix_extractor);
     return {filter, nullptr};
   }
 }
@@ -257,7 +263,8 @@ void ReleaseFilterCachedEntry(void* arg, void* h) {
 }
 
 // TODO(myabandeh): merge this with the same function in IndexReader
-void PartitionedFilterBlockReader::CacheDependencies(bool pin) {
+void PartitionedFilterBlockReader::CacheDependencies(
+    bool pin, const SliceTransform* prefix_extractor) {
   // Before read partitions, prefetch them to avoid lots of IOs
   auto rep = table_->rep_;
   BlockIter biter;
@@ -308,9 +315,9 @@ void PartitionedFilterBlockReader::CacheDependencies(bool pin) {
 
     const bool no_io = true;
     const bool is_a_filter_partition = true;
-    auto filter = table_->GetFilter(prefetch_buffer.get(), handle,
-                                    is_a_filter_partition, !no_io,
-                                    /* get_context */ nullptr);
+    auto filter = table_->GetFilter(
+        prefetch_buffer.get(), handle, is_a_filter_partition, !no_io,
+        /* get_context */ nullptr, prefix_extractor);
     if (LIKELY(filter.IsSet())) {
       if (pin) {
         filter_map_[handle.offset()] = std::move(filter);

table/partitioned_filter_block.h

@@ -79,12 +79,12 @@ class PartitionedFilterBlockReader : public FilterBlockReader,
 
   virtual bool IsBlockBased() override { return false; }
   virtual bool KeyMayMatch(
-      const Slice& key, uint64_t block_offset = kNotValid,
-      const bool no_io = false,
+      const Slice& key, const SliceTransform* prefix_extractor,
+      uint64_t block_offset = kNotValid, const bool no_io = false,
       const Slice* const const_ikey_ptr = nullptr) override;
   virtual bool PrefixMayMatch(
-      const Slice& prefix, uint64_t block_offset = kNotValid,
-      const bool no_io = false,
+      const Slice& prefix, const SliceTransform* prefix_extractor,
+      uint64_t block_offset = kNotValid, const bool no_io = false,
       const Slice* const const_ikey_ptr = nullptr) override;
   virtual size_t ApproximateMemoryUsage() const override;
 
@@ -92,8 +92,9 @@ class PartitionedFilterBlockReader : public FilterBlockReader,
   Slice GetFilterPartitionHandle(const Slice& entry);
   BlockBasedTable::CachableEntry<FilterBlockReader> GetFilterPartition(
       FilePrefetchBuffer* prefetch_buffer, Slice* handle, const bool no_io,
-      bool* cached);
-  virtual void CacheDependencies(bool pin) override;
+      bool* cached, const SliceTransform* prefix_extractor = nullptr);
+  virtual void CacheDependencies(
+      bool bin, const SliceTransform* prefix_extractor) override;
 
   const SliceTransform* prefix_extractor_;
   std::unique_ptr<Block> idx_on_fltr_blk_;

table/partitioned_filter_block_test.cc

@@ -29,8 +29,8 @@ class MockedBlockBasedTable : public BlockBasedTable {
 
   virtual CachableEntry<FilterBlockReader> GetFilter(
       FilePrefetchBuffer*, const BlockHandle& filter_blk_handle,
-      const bool /* unused */, bool /* unused */,
-      GetContext* /* unused */) const override {
+      const bool /* unused */, bool /* unused */, GetContext* /* unused */,
+      const SliceTransform* /* unused */) const override {
     Slice slice = slices[filter_blk_handle.offset()];
     auto obj = new FullFilterBlockReader(
         nullptr, true, BlockContents(slice, false, kNoCompression),
@@ -121,6 +121,7 @@ class PartitionedFilterBlockTest : public testing::Test {
     } while (status.IsIncomplete());
     const Options options;
     const ImmutableCFOptions ioptions(options);
+    const MutableCFOptions moptions(options);
     const EnvOptions env_options;
     table.reset(new MockedBlockBasedTable(new BlockBasedTable::Rep(
         ioptions, env_options, table_options_, icomp, false)));
@@ -138,23 +139,27 @@ class PartitionedFilterBlockTest : public testing::Test {
     for (auto key : keys) {
       auto ikey = InternalKey(key, 0, ValueType::kTypeValue);
       const Slice ikey_slice = Slice(*ikey.rep());
-      ASSERT_TRUE(reader->KeyMayMatch(key, kNotValid, !no_io, &ikey_slice));
+      ASSERT_TRUE(
+          reader->KeyMayMatch(key, nullptr, kNotValid, !no_io, &ikey_slice));
     }
     {
       // querying a key twice
       auto ikey = InternalKey(keys[0], 0, ValueType::kTypeValue);
       const Slice ikey_slice = Slice(*ikey.rep());
-      ASSERT_TRUE(reader->KeyMayMatch(keys[0], kNotValid, !no_io, &ikey_slice));
+      ASSERT_TRUE(reader->KeyMayMatch(keys[0], nullptr, kNotValid, !no_io,
+                                      &ikey_slice));
     }
     // querying missing keys
     for (auto key : missing_keys) {
       auto ikey = InternalKey(key, 0, ValueType::kTypeValue);
       const Slice ikey_slice = Slice(*ikey.rep());
       if (empty) {
-        ASSERT_TRUE(reader->KeyMayMatch(key, kNotValid, !no_io, &ikey_slice));
+        ASSERT_TRUE(
+            reader->KeyMayMatch(key, nullptr, kNotValid, !no_io, &ikey_slice));
       } else {
         // assuming a good hash function
-        ASSERT_FALSE(reader->KeyMayMatch(key, kNotValid, !no_io, &ikey_slice));
+        ASSERT_FALSE(
+            reader->KeyMayMatch(key, nullptr, kNotValid, !no_io, &ikey_slice));
       }
     }
   }

table/plain_table_builder.cc

@@ -57,7 +57,7 @@ extern const uint64_t kPlainTableMagicNumber = 0x8242229663bf9564ull;
 extern const uint64_t kLegacyPlainTableMagicNumber = 0x4f3418eb7a8f13b8ull;
 
 PlainTableBuilder::PlainTableBuilder(
-    const ImmutableCFOptions& ioptions,
+    const ImmutableCFOptions& ioptions, const MutableCFOptions& moptions,
     const std::vector<std::unique_ptr<IntTblPropCollectorFactory>>*
         int_tbl_prop_collector_factories,
     uint32_t column_family_id, WritableFileWriter* file, uint32_t user_key_len,
@@ -66,20 +66,21 @@ PlainTableBuilder::PlainTableBuilder(
     uint32_t num_probes, size_t huge_page_tlb_size, double hash_table_ratio,
     bool store_index_in_file)
     : ioptions_(ioptions),
+      moptions_(moptions),
       bloom_block_(num_probes),
       file_(file),
       bloom_bits_per_key_(bloom_bits_per_key),
       huge_page_tlb_size_(huge_page_tlb_size),
-      encoder_(encoding_type, user_key_len, ioptions.prefix_extractor,
+      encoder_(encoding_type, user_key_len, moptions.prefix_extractor.get(),
                index_sparseness),
       store_index_in_file_(store_index_in_file),
-      prefix_extractor_(ioptions.prefix_extractor) {
+      prefix_extractor_(moptions.prefix_extractor.get()) {
   // Build index block and save it in the file if hash_table_ratio > 0
   if (store_index_in_file_) {
     assert(hash_table_ratio > 0 || IsTotalOrderMode());
-    index_builder_.reset(
-        new PlainTableIndexBuilder(&arena_, ioptions, index_sparseness,
-                                   hash_table_ratio, huge_page_tlb_size_));
+    index_builder_.reset(new PlainTableIndexBuilder(
+        &arena_, ioptions, moptions.prefix_extractor.get(), index_sparseness,
+        hash_table_ratio, huge_page_tlb_size_));
     properties_.user_collected_properties
         [PlainTablePropertyNames::kBloomVersion] = "1";  // For future use
   }
@@ -96,8 +97,8 @@ PlainTableBuilder::PlainTableBuilder(
   properties_.format_version = (encoding_type == kPlain) ? 0 : 1;
   properties_.column_family_id = column_family_id;
   properties_.column_family_name = column_family_name;
-  properties_.prefix_extractor_name = ioptions_.prefix_extractor != nullptr
-                                          ? ioptions_.prefix_extractor->Name()
+  properties_.prefix_extractor_name = moptions_.prefix_extractor != nullptr
+                                          ? moptions_.prefix_extractor->Name()
                                           : "nullptr";
 
   std::string val;
@@ -131,11 +132,11 @@ void PlainTableBuilder::Add(const Slice& key, const Slice& value) {
 
   // Store key hash
   if (store_index_in_file_) {
-    if (ioptions_.prefix_extractor == nullptr) {
+    if (moptions_.prefix_extractor == nullptr) {
       keys_or_prefixes_hashes_.push_back(GetSliceHash(internal_key.user_key));
     } else {
       Slice prefix =
-          ioptions_.prefix_extractor->Transform(internal_key.user_key);
+          moptions_.prefix_extractor->Transform(internal_key.user_key);
       keys_or_prefixes_hashes_.push_back(GetSliceHash(prefix));
     }
   }

table/plain_table_builder.h

@@ -32,7 +32,7 @@ class PlainTableBuilder: public TableBuilder {
   // will be part of level specified by 'level'.  A value of -1 means
   // that the caller does not know which level the output file will reside.
   PlainTableBuilder(
-      const ImmutableCFOptions& ioptions,
+      const ImmutableCFOptions& ioptions, const MutableCFOptions& moptions,
       const std::vector<std::unique_ptr<IntTblPropCollectorFactory>>*
           int_tbl_prop_collector_factories,
       uint32_t column_family_id, WritableFileWriter* file,
@@ -79,6 +79,7 @@ class PlainTableBuilder: public TableBuilder {
  private:
   Arena arena_;
   const ImmutableCFOptions& ioptions_;
+  const MutableCFOptions& moptions_;
   std::vector<std::unique_ptr<IntTblPropCollector>>
       table_properties_collectors_;
 

table/plain_table_factory.cc

@@ -27,7 +27,7 @@ Status PlainTableFactory::NewTableReader(
       table_reader_options.internal_comparator, std::move(file), file_size,
       table, table_options_.bloom_bits_per_key, table_options_.hash_table_ratio,
       table_options_.index_sparseness, table_options_.huge_page_tlb_size,
-      table_options_.full_scan_mode);
+      table_options_.full_scan_mode, table_reader_options.prefix_extractor);
 }
 
 TableBuilder* PlainTableFactory::NewTableBuilder(
@@ -38,7 +38,7 @@ TableBuilder* PlainTableFactory::NewTableBuilder(
   // tables
   //
   return new PlainTableBuilder(
-      table_builder_options.ioptions,
+      table_builder_options.ioptions, table_builder_options.moptions,
       table_builder_options.int_tbl_prop_collector_factories, column_family_id,
       file, table_options_.user_key_len, table_options_.encoding_type,
       table_options_.index_sparseness, table_options_.bloom_bits_per_key,