rocksdb-native 2.2.0 → 2.3.0

package/vendor/librocksdb/src/rocksdb.cc

@@ -17,6 +17,8 @@ using namespace rocksdb;
 
 static_assert(sizeof(Slice) == sizeof(rocksdb_slice_t));
 
+namespace {
+
 static const rocksdb_options_t rocksdb__default_options = {
   .version = 0,
   .read_only = false,
@@ -33,12 +35,35 @@ static const rocksdb_options_t rocksdb__default_options = {
   .table_format_version = 6,
 };
 
+static const rocksdb_read_options_t rocksdb__default_read_options = {
+  .version = 0,
+  .snapshot = nullptr,
+};
+
+static const rocksdb_write_options_t rocksdb__default_write_options = {
+  .version = 0,
+};
+
 template <auto rocksdb_options_t::*P, typename T>
 static inline T
 rocksdb__option (const rocksdb_options_t *options, int min_version, T fallback = T(rocksdb__default_options.*P)) {
   return options->version >= min_version ? T(options->*P) : fallback;
 }
 
+template <auto rocksdb_read_options_t::*P, typename T>
+static inline T
+rocksdb__option (const rocksdb_read_options_t *options, int min_version, T fallback = T(rocksdb__default_read_options.*P)) {
+  return options->version >= min_version ? T(options->*P) : fallback;
+}
+
+template <auto rocksdb_write_options_t::*P, typename T>
+static inline T
+rocksdb__option (const rocksdb_write_options_t *options, int min_version, T fallback = T(rocksdb__default_write_options.*P)) {
+  return options->version >= min_version ? T(options->*P) : fallback;
+}
+
+} // namespace
+
 extern "C" int
 rocksdb_init (uv_loop_t *loop, rocksdb_t *db) {
   db->loop = loop;
@@ -47,6 +72,8 @@ rocksdb_init (uv_loop_t *loop, rocksdb_t *db) {
   return 0;
 }
 
+namespace {
+
 template <typename T>
 static inline void
 rocksdb__on_status (uv_work_t *handle, int status) {
@@ -59,6 +86,8 @@ rocksdb__on_status (uv_work_t *handle, int status) {
   if (error) free(error);
 }
 
+} // namespace
+
 static void
 rocksdb__on_open (uv_work_t *handle) {
   auto req = reinterpret_cast<rocksdb_open_t *>(handle->data);
@@ -146,6 +175,8 @@ rocksdb_open (rocksdb_t *db, rocksdb_open_t *req, const char *path, const rocksd
   return uv_queue_work(db->loop, &req->worker, rocksdb__on_open, rocksdb__on_status<rocksdb_open_t>);
 }
 
+namespace {
+
 static void
 rocksdb__on_close (uv_work_t *handle) {
   auto req = reinterpret_cast<rocksdb_close_t *>(handle->data);
@@ -165,6 +196,8 @@ rocksdb__on_close (uv_work_t *handle) {
   delete rocks;
 }
 
+} // namespace
+
 extern "C" int
 rocksdb_close (rocksdb_t *db, rocksdb_close_t *req, rocksdb_close_cb cb) {
   req->db = db;
@@ -191,6 +224,8 @@ rocksdb_slice_empty (void) {
   return {.data = nullptr, .len = 0};
 }
 
+namespace {
+
 static inline rocksdb_slice_t
 rocksdb__slice_copy (const Slice &slice) {
   auto len = slice.size();
@@ -323,8 +358,8 @@ rocksdb__iterator_valid (Iterator *iterator, T *req) {
 
 template <bool reverse = false>
 static inline Iterator *
-rocksdb__iterator_open (DB *db, const rocksdb_range_t &range) {
-  auto iterator = db->NewIterator(ReadOptions());
+rocksdb__iterator_open (DB *db, const ReadOptions &options, const rocksdb_range_t &range) {
+  auto iterator = db->NewIterator(options);
 
   rocksdb__iterator_seek<reverse>(iterator, range);
 
@@ -338,10 +373,18 @@ rocksdb__iterator_open (T *req) {
 
   const auto &range = req->range;
 
+  auto snapshot = rocksdb__option<&rocksdb_read_options_t::snapshot, rocksdb_snapshot_t *>(
+    &req->options, 0
+  );
+
+  ReadOptions options;
+
+  if (snapshot) options.snapshot = reinterpret_cast<const Snapshot *>(snapshot->handle);
+
   if (req->reverse) {
-    return rocksdb__iterator_open<true>(db, range);
+    return rocksdb__iterator_open<true>(db, options, range);
   } else {
-    return rocksdb__iterator_open<false>(db, range);
+    return rocksdb__iterator_open<false>(db, options, range);
   }
 }
 
@@ -366,6 +409,10 @@ rocksdb__iterator_refresh (Iterator *iterator, T *req) {
   rocksdb__iterator_seek(iterator, req);
 }
 
+} // namespace
+
+namespace {
+
 static void
 rocksdb__on_after_iterator (uv_work_t *handle, int status) {
   auto req = reinterpret_cast<rocksdb_iterator_t *>(handle->data);
@@ -394,9 +441,12 @@ rocksdb__on_iterator_open (uv_work_t *handle) {
   }
 }
 
+} // namespace
+
 extern "C" int
-rocksdb_iterator_open (rocksdb_t *db, rocksdb_iterator_t *iterator, rocksdb_range_t range, bool reverse, rocksdb_iterator_cb cb) {
+rocksdb_iterator_open (rocksdb_t *db, rocksdb_iterator_t *iterator, rocksdb_range_t range, bool reverse, const rocksdb_read_options_t *options, rocksdb_iterator_cb cb) {
   iterator->db = db;
+  iterator->options = options ? *options : rocksdb__default_read_options;
   iterator->range = range;
   iterator->reverse = reverse;
   iterator->cb = cb;
@@ -406,6 +456,8 @@ rocksdb_iterator_open (rocksdb_t *db, rocksdb_iterator_t *iterator, rocksdb_rang
   return uv_queue_work(iterator->db->loop, &iterator->worker, rocksdb__on_iterator_open, rocksdb__on_after_iterator);
 }
 
+namespace {
+
 static void
 rocksdb__on_iterator_close (uv_work_t *handle) {
   auto req = reinterpret_cast<rocksdb_iterator_t *>(handle->data);
@@ -415,6 +467,8 @@ rocksdb__on_iterator_close (uv_work_t *handle) {
   delete iterator;
 }
 
+} // namespace
+
 extern "C" int
 rocksdb_iterator_close (rocksdb_iterator_t *iterator, rocksdb_iterator_cb cb) {
   iterator->cb = cb;
@@ -422,6 +476,8 @@ rocksdb_iterator_close (rocksdb_iterator_t *iterator, rocksdb_iterator_cb cb) {
   return uv_queue_work(iterator->db->loop, &iterator->worker, rocksdb__on_iterator_close, rocksdb__on_after_iterator);
 }
 
+namespace {
+
 static void
 rocksdb__on_iterator_refresh (uv_work_t *handle) {
   auto req = reinterpret_cast<rocksdb_iterator_t *>(handle->data);
@@ -439,8 +495,11 @@ rocksdb__on_iterator_refresh (uv_work_t *handle) {
   }
 }
 
+} // namespace
+
 extern "C" int
-rocksdb_iterator_refresh (rocksdb_iterator_t *iterator, rocksdb_range_t range, bool reverse, rocksdb_iterator_cb cb) {
+rocksdb_iterator_refresh (rocksdb_iterator_t *iterator, rocksdb_range_t range, bool reverse, const rocksdb_read_options_t *options, rocksdb_iterator_cb cb) {
+  iterator->options = options ? *options : rocksdb__default_read_options;
   iterator->range = range;
   iterator->reverse = reverse;
   iterator->cb = cb;
@@ -448,6 +507,8 @@ rocksdb_iterator_refresh (rocksdb_iterator_t *iterator, rocksdb_range_t range, b
   return uv_queue_work(iterator->db->loop, &iterator->worker, rocksdb__on_iterator_close, rocksdb__on_after_iterator);
 }
 
+namespace {
+
 static void
 rocksdb__on_iterator_read (uv_work_t *handle) {
   auto req = reinterpret_cast<rocksdb_iterator_t *>(handle->data);
@@ -465,6 +526,8 @@ rocksdb__on_iterator_read (uv_work_t *handle) {
   }
 }
 
+} // namespace
+
 extern "C" int
 rocksdb_iterator_read (rocksdb_iterator_t *iterator, rocksdb_slice_t *keys, rocksdb_slice_t *values, size_t capacity, rocksdb_iterator_cb cb) {
   iterator->cb = cb;
@@ -476,6 +539,8 @@ rocksdb_iterator_read (rocksdb_iterator_t *iterator, rocksdb_slice_t *keys, rock
   return uv_queue_work(iterator->db->loop, &iterator->worker, rocksdb__on_iterator_read, rocksdb__on_after_iterator);
 }
 
+namespace {
+
 static void
 rocksdb__on_after_read (uv_work_t *handle, int status) {
   auto req = reinterpret_cast<rocksdb_read_batch_t *>(handle->data);
@@ -488,7 +553,7 @@ rocksdb__on_after_read (uv_work_t *handle, int status) {
 }
 
 static void
-rocksdb__on_batch_read (uv_work_t *handle) {
+rocksdb__on_read (uv_work_t *handle) {
   auto req = reinterpret_cast<rocksdb_read_batch_t *>(handle->data);
 
   auto db = reinterpret_cast<DB *>(req->db->handle);
@@ -510,7 +575,15 @@ rocksdb__on_batch_read (uv_work_t *handle) {
 
     std::vector<Status> statuses(req->len);
 
-    db->MultiGet(ReadOptions(), db->DefaultColumnFamily(), req->len, keys.data(), values.data(), statuses.data());
+    auto snapshot = rocksdb__option<&rocksdb_read_options_t::snapshot, rocksdb_snapshot_t *>(
+      &req->options, 0
+    );
+
+    ReadOptions options;
+
+    if (snapshot) options.snapshot = reinterpret_cast<const Snapshot *>(snapshot->handle);
+
+    db->MultiGet(options, db->DefaultColumnFamily(), req->len, keys.data(), values.data(), statuses.data());
 
     for (size_t i = 0, n = req->len; i < n; i++) {
       auto op = &req->reads[i];
@@ -539,9 +612,12 @@ rocksdb__on_batch_read (uv_work_t *handle) {
   }
 }
 
+} // namespace
+
 extern "C" int
-rocksdb_read (rocksdb_t *db, rocksdb_read_batch_t *batch, rocksdb_read_t *reads, char **errors, size_t len, rocksdb_read_batch_cb cb) {
+rocksdb_read (rocksdb_t *db, rocksdb_read_batch_t *batch, rocksdb_read_t *reads, char **errors, size_t len, const rocksdb_read_options_t *options, rocksdb_read_batch_cb cb) {
   batch->db = db;
+  batch->options = options ? *options : rocksdb__default_read_options;
   batch->reads = reads;
   batch->errors = errors;
   batch->len = len;
@@ -549,9 +625,11 @@ rocksdb_read (rocksdb_t *db, rocksdb_read_batch_t *batch, rocksdb_read_t *reads,
 
   batch->worker.data = static_cast<void *>(batch);
 
-  return uv_queue_work(batch->db->loop, &batch->worker, rocksdb__on_batch_read, rocksdb__on_after_read);
+  return uv_queue_work(batch->db->loop, &batch->worker, rocksdb__on_read, rocksdb__on_after_read);
 }
 
+namespace {
+
 static void
 rocksdb__on_after_write (uv_work_t *handle, int status) {
   auto req = reinterpret_cast<rocksdb_write_batch_t *>(handle->data);
@@ -564,7 +642,7 @@ rocksdb__on_after_write (uv_work_t *handle, int status) {
 }
 
 static void
-rocksdb__on_batch_write (uv_work_t *handle) {
+rocksdb__on_write (uv_work_t *handle) {
   auto req = reinterpret_cast<rocksdb_write_batch_t *>(handle->data);
 
   auto db = reinterpret_cast<DB *>(req->db->handle);
@@ -590,7 +668,9 @@ rocksdb__on_batch_write (uv_work_t *handle) {
       }
     }
 
-    auto status = db->Write(WriteOptions(), &batch);
+    WriteOptions options;
+
+    auto status = db->Write(options, &batch);
 
     if (status.ok()) {
       req->error = nullptr;
@@ -600,9 +680,12 @@ rocksdb__on_batch_write (uv_work_t *handle) {
   }
 }
 
+} // namespace
+
 extern "C" int
-rocksdb_write (rocksdb_t *db, rocksdb_write_batch_t *batch, rocksdb_write_t *writes, size_t len, rocksdb_write_batch_cb cb) {
+rocksdb_write (rocksdb_t *db, rocksdb_write_batch_t *batch, rocksdb_write_t *writes, size_t len, const rocksdb_write_options_t *options, rocksdb_write_batch_cb cb) {
   batch->db = db;
+  batch->options = options ? *options : rocksdb__default_write_options;
   batch->writes = writes;
   batch->error = nullptr;
   batch->len = len;
@@ -610,5 +693,22 @@ rocksdb_write (rocksdb_t *db, rocksdb_write_batch_t *batch, rocksdb_write_t *wri
 
   batch->worker.data = static_cast<void *>(batch);
 
-  return uv_queue_work(batch->db->loop, &batch->worker, rocksdb__on_batch_write, rocksdb__on_after_write);
+  return uv_queue_work(batch->db->loop, &batch->worker, rocksdb__on_write, rocksdb__on_after_write);
+}
+
+extern "C" int
+rocksdb_snapshot_create (rocksdb_t *db, rocksdb_snapshot_t *snapshot) {
+  auto handle = reinterpret_cast<DB *>(db->handle)->GetSnapshot();
+
+  if (handle == nullptr) return -1;
+
+  snapshot->db = db;
+  snapshot->handle = handle;
+
+  return 0;
+}
+
+extern "C" void
+rocksdb_snapshot_destroy (rocksdb_snapshot_t *snapshot) {
+  reinterpret_cast<DB *>(snapshot->db->handle)->ReleaseSnapshot(reinterpret_cast<const Snapshot *>(snapshot->handle));
 }

package/vendor/librocksdb/vendor/rocksdb/CMakeLists.txt

@@ -267,8 +267,8 @@ endif(CMAKE_SYSTEM_PROCESSOR MATCHES "s390x")
 if(CMAKE_SYSTEM_PROCESSOR MATCHES "loongarch64")
   CHECK_C_COMPILER_FLAG("-march=loongarch64" HAS_LOONGARCH64)
   if(HAS_LOONGARCH64)
-    set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mcpu=loongarch64 -mtune=loongarch64")
-    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -mcpu=loongarch64 -mtune=loongarch64")
+    set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -march=loongarch64 -mtune=loongarch64")
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -march=loongarch64 -mtune=loongarch64")
   endif(HAS_LOONGARCH64)
 endif(CMAKE_SYSTEM_PROCESSOR MATCHES "loongarch64")
 
@@ -488,6 +488,8 @@ if(CMAKE_SYSTEM_NAME MATCHES "Cygwin")
   add_definitions(-fno-builtin-memcmp -DCYGWIN)
 elseif(CMAKE_SYSTEM_NAME MATCHES "Darwin")
   add_definitions(-DOS_MACOSX)
+elseif(CMAKE_SYSTEM_NAME MATCHES "iOS")
+  add_definitions(-DOS_MACOSX -DIOS_CROSS_COMPILE)
 elseif(CMAKE_SYSTEM_NAME MATCHES "Linux")
   add_definitions(-DOS_LINUX)
 elseif(CMAKE_SYSTEM_NAME MATCHES "SunOS")
@@ -615,7 +617,7 @@ if(USE_FOLLY)
     FMT_INST_PATH)
     exec_program(ls ARGS -d ${FOLLY_INST_PATH}/../gflags* OUTPUT_VARIABLE
     GFLAGS_INST_PATH)
-    set(Boost_DIR ${BOOST_INST_PATH}/lib/cmake/Boost-1.78.0)
+    set(Boost_DIR ${BOOST_INST_PATH}/lib/cmake/Boost-1.83.0)
     if(EXISTS ${FMT_INST_PATH}/lib64)
       set(fmt_DIR ${FMT_INST_PATH}/lib64/cmake/fmt)
     else()
@@ -653,6 +655,7 @@ set(SOURCES
         cache/sharded_cache.cc
         cache/tiered_secondary_cache.cc
         db/arena_wrapped_db_iter.cc
+        db/attribute_group_iterator_impl.cc
         db/blob/blob_contents.cc
         db/blob/blob_fetcher.cc
         db/blob/blob_file_addition.cc
@@ -669,6 +672,7 @@ set(SOURCES
         db/blob/prefetch_buffer_collection.cc
         db/builder.cc
         db/c.cc
+        db/coalescing_iterator.cc
         db/column_family.cc
         db/compaction/compaction.cc
         db/compaction/compaction_iterator.cc
@@ -689,6 +693,7 @@ set(SOURCES
         db/db_impl/db_impl_write.cc
         db/db_impl/db_impl_compaction_flush.cc
         db/db_impl/db_impl_files.cc
+        db/db_impl/db_impl_follower.cc
         db/db_impl/db_impl_open.cc
         db/db_impl/db_impl_debug.cc
         db/db_impl/db_impl_experimental.cc
@@ -715,7 +720,6 @@ set(SOURCES
         db/memtable_list.cc
         db/merge_helper.cc
         db/merge_operator.cc
-        db/multi_cf_iterator.cc
         db/output_validator.cc
         db/periodic_task_scheduler.cc
         db/range_del_aggregator.cc
@@ -747,6 +751,7 @@ set(SOURCES
         env/env_encryption.cc
         env/file_system.cc
         env/file_system_tracer.cc
+        env/fs_on_demand.cc
         env/fs_remap.cc
         env/mock_env.cc
         env/unique_id_gen.cc
@@ -934,6 +939,7 @@ set(SOURCES
         utilities/persistent_cache/volatile_tier_impl.cc
         utilities/simulator_cache/cache_simulator.cc
         utilities/simulator_cache/sim_cache.cc
+        utilities/table_properties_collectors/compact_for_tiering_collector.cc
         utilities/table_properties_collectors/compact_on_deletion_collector.cc
         utilities/trace/file_trace_reader_writer.cc
         utilities/trace/replayer_impl.cc
@@ -954,6 +960,7 @@ set(SOURCES
         utilities/transactions/write_prepared_txn_db.cc
         utilities/transactions/write_unprepared_txn.cc
         utilities/transactions/write_unprepared_txn_db.cc
+        utilities/types_util.cc
         utilities/ttl/db_ttl_impl.cc
         utilities/wal_filter.cc
         utilities/write_batch_with_index/write_batch_with_index.cc
@@ -1045,6 +1052,7 @@ if(USE_FOLLY_LITE)
   list(APPEND SOURCES
     third-party/folly/folly/container/detail/F14Table.cpp
     third-party/folly/folly/detail/Futex.cpp
+    third-party/folly/folly/lang/Exception.cpp
     third-party/folly/folly/lang/SafeAssert.cpp
     third-party/folly/folly/lang/ToAscii.cpp
     third-party/folly/folly/ScopeGuard.cpp
@@ -1052,6 +1060,12 @@ if(USE_FOLLY_LITE)
     third-party/folly/folly/synchronization/DistributedMutex.cpp
     third-party/folly/folly/synchronization/ParkingLot.cpp)
   include_directories(${PROJECT_SOURCE_DIR}/third-party/folly)
+  exec_program(python3 ${PROJECT_SOURCE_DIR}/third-party/folly ARGS
+  build/fbcode_builder/getdeps.py show-source-dir boost OUTPUT_VARIABLE
+  BOOST_SOURCE_PATH)
+  exec_program(ls ARGS -d ${BOOST_SOURCE_PATH}/boost* OUTPUT_VARIABLE
+  BOOST_INCLUDE_DIR)
+  include_directories(${BOOST_INCLUDE_DIR})
   add_definitions(-DUSE_FOLLY -DFOLLY_NO_CONFIG)
   list(APPEND THIRDPARTY_LIBS glog)
 endif()
@@ -1361,6 +1375,7 @@ if(WITH_TESTS)
         db/file_indexer_test.cc
         db/filename_test.cc
         db/flush_job_test.cc
+        db/db_follower_test.cc
         db/import_column_family_test.cc
         db/listener_test.cc
         db/log_test.cc
@@ -1474,6 +1489,7 @@ if(WITH_TESTS)
         utilities/persistent_cache/persistent_cache_test.cc
         utilities/simulator_cache/cache_simulator_test.cc
         utilities/simulator_cache/sim_cache_test.cc
+        utilities/table_properties_collectors/compact_for_tiering_collector_test.cc
         utilities/table_properties_collectors/compact_on_deletion_collector_test.cc
         utilities/transactions/optimistic_transaction_test.cc
         utilities/transactions/transaction_test.cc
@@ -1484,6 +1500,7 @@ if(WITH_TESTS)
         utilities/transactions/lock/range/range_locking_test.cc
         utilities/transactions/timestamped_snapshot_test.cc
         utilities/ttl/ttl_test.cc
+        utilities/types_util_test.cc
         utilities/util_merge_operators_test.cc
         utilities/write_batch_with_index/write_batch_with_index_test.cc
 	${PLUGIN_TESTS}

package/vendor/librocksdb/vendor/rocksdb/cache/secondary_cache_adapter.cc

@@ -134,12 +134,14 @@ bool CacheWithSecondaryAdapter::EvictionHandler(const Slice& key,
     auto obj = target_->Value(handle);
     // Ignore dummy entry
     if (obj != kDummyObj) {
-      bool hit = false;
+      bool force = false;
       if (adm_policy_ == TieredAdmissionPolicy::kAdmPolicyAllowCacheHits) {
-        hit = was_hit;
+        force = was_hit;
+      } else if (adm_policy_ == TieredAdmissionPolicy::kAdmPolicyAllowAll) {
+        force = true;
       }
       // Spill into secondary cache.
-      secondary_cache_->Insert(key, obj, helper, hit).PermitUncheckedError();
+      secondary_cache_->Insert(key, obj, helper, force).PermitUncheckedError();
     }
   }
   // Never takes ownership of obj
@@ -661,6 +663,7 @@ std::shared_ptr<Cache> NewTieredCache(const TieredCacheOptions& _opts) {
         break;
       case TieredAdmissionPolicy::kAdmPolicyPlaceholder:
       case TieredAdmissionPolicy::kAdmPolicyAllowCacheHits:
+      case TieredAdmissionPolicy::kAdmPolicyAllowAll:
         if (opts.nvm_sec_cache) {
           valid_adm_policy = false;
         }

package/vendor/librocksdb/vendor/rocksdb/db/arena_wrapped_db_iter.cc

@@ -138,10 +138,10 @@ Status ArenaWrappedDBIter::Refresh(const Snapshot* snapshot) {
             reinit_internal_iter();
             break;
           } else {
-            delete *memtable_range_tombstone_iter_;
-            *memtable_range_tombstone_iter_ = new TruncatedRangeDelIterator(
-                std::unique_ptr<FragmentedRangeTombstoneIterator>(t),
-                &cfd->internal_comparator(), nullptr, nullptr);
+            *memtable_range_tombstone_iter_ =
+                std::make_unique<TruncatedRangeDelIterator>(
+                    std::unique_ptr<FragmentedRangeTombstoneIterator>(t),
+                    &cfd->internal_comparator(), nullptr, nullptr);
           }
         }
         db_impl->ReturnAndCleanupSuperVersion(cfd, sv);

package/vendor/librocksdb/vendor/rocksdb/db/arena_wrapped_db_iter.h

@@ -55,7 +55,8 @@ class ArenaWrappedDBIter : public Iterator {
     db_iter_->SetIter(iter);
   }
 
-  void SetMemtableRangetombstoneIter(TruncatedRangeDelIterator** iter) {
+  void SetMemtableRangetombstoneIter(
+      std::unique_ptr<TruncatedRangeDelIterator>* iter) {
     memtable_range_tombstone_iter_ = iter;
   }
 
@@ -110,7 +111,8 @@ class ArenaWrappedDBIter : public Iterator {
   bool allow_refresh_ = true;
   // If this is nullptr, it means the mutable memtable does not contain range
   // tombstone when added under this DBIter.
-  TruncatedRangeDelIterator** memtable_range_tombstone_iter_ = nullptr;
+  std::unique_ptr<TruncatedRangeDelIterator>* memtable_range_tombstone_iter_ =
+      nullptr;
 };
 
 // Generate the arena wrapped iterator class.

package/vendor/librocksdb/vendor/rocksdb/db/attribute_group_iterator_impl.cc

@@ -0,0 +1,20 @@
+//  Copyright (c) Meta Platforms, Inc. and affiliates.
+//  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/attribute_group_iterator_impl.h"
+
+namespace ROCKSDB_NAMESPACE {
+
+const AttributeGroups kNoAttributeGroups;
+const IteratorAttributeGroups kNoIteratorAttributeGroups;
+
+void AttributeGroupIteratorImpl::AddToAttributeGroups(
+    const autovector<MultiCfIteratorInfo>& items) {
+  for (const auto& item : items) {
+    attribute_groups_.emplace_back(item.cfh, &item.iterator->columns());
+  }
+}
+
+}  // namespace ROCKSDB_NAMESPACE

package/vendor/librocksdb/vendor/rocksdb/db/attribute_group_iterator_impl.h

@@ -0,0 +1,83 @@
+//  Copyright (c) Meta Platforms, Inc. and affiliates.
+//  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).
+
+#pragma once
+
+#include "db/multi_cf_iterator_impl.h"
+#include "rocksdb/attribute_groups.h"
+
+namespace ROCKSDB_NAMESPACE {
+
+class AttributeGroupIteratorImpl : public AttributeGroupIterator {
+ public:
+  AttributeGroupIteratorImpl(
+      const Comparator* comparator,
+      const std::vector<ColumnFamilyHandle*>& column_families,
+      const std::vector<Iterator*>& child_iterators)
+      : impl_(
+            comparator, column_families, child_iterators, [this]() { Reset(); },
+            [this](const autovector<MultiCfIteratorInfo>& items) {
+              AddToAttributeGroups(items);
+            }) {}
+  ~AttributeGroupIteratorImpl() override {}
+
+  // No copy allowed
+  AttributeGroupIteratorImpl(const AttributeGroupIteratorImpl&) = delete;
+  AttributeGroupIteratorImpl& operator=(const AttributeGroupIteratorImpl&) =
+      delete;
+
+  bool Valid() const override { return impl_.Valid(); }
+  void SeekToFirst() override { impl_.SeekToFirst(); }
+  void SeekToLast() override { impl_.SeekToLast(); }
+  void Seek(const Slice& target) override { impl_.Seek(target); }
+  void SeekForPrev(const Slice& target) override { impl_.SeekForPrev(target); }
+  void Next() override { impl_.Next(); }
+  void Prev() override { impl_.Prev(); }
+  Slice key() const override { return impl_.key(); }
+  Status status() const override { return impl_.status(); }
+
+  const IteratorAttributeGroups& attribute_groups() const override {
+    assert(Valid());
+    return attribute_groups_;
+  }
+
+  void Reset() { attribute_groups_.clear(); }
+
+ private:
+  MultiCfIteratorImpl impl_;
+  IteratorAttributeGroups attribute_groups_;
+  void AddToAttributeGroups(const autovector<MultiCfIteratorInfo>& items);
+};
+
+class EmptyAttributeGroupIterator : public AttributeGroupIterator {
+ public:
+  explicit EmptyAttributeGroupIterator(const Status& s) : status_(s) {}
+  bool Valid() const override { return false; }
+  void Seek(const Slice& /*target*/) override {}
+  void SeekForPrev(const Slice& /*target*/) override {}
+  void SeekToFirst() override {}
+  void SeekToLast() override {}
+  void Next() override { assert(false); }
+  void Prev() override { assert(false); }
+  Slice key() const override {
+    assert(false);
+    return Slice();
+  }
+  Status status() const override { return status_; }
+
+  const IteratorAttributeGroups& attribute_groups() const override {
+    return kNoIteratorAttributeGroups;
+  }
+
+ private:
+  Status status_;
+};
+
+inline std::unique_ptr<AttributeGroupIterator> NewAttributeGroupErrorIterator(
+    const Status& status) {
+  return std::make_unique<EmptyAttributeGroupIterator>(status);
+}
+
+}  // namespace ROCKSDB_NAMESPACE

package/vendor/librocksdb/vendor/rocksdb/db/builder.cc

@@ -64,7 +64,7 @@ Status BuildTable(
     std::vector<std::unique_ptr<FragmentedRangeTombstoneIterator>>
         range_del_iters,
     FileMetaData* meta, std::vector<BlobFileAddition>* blob_file_additions,
-    std::vector<SequenceNumber> snapshots,
+    std::vector<SequenceNumber> snapshots, SequenceNumber earliest_snapshot,
     SequenceNumber earliest_write_conflict_snapshot,
     SequenceNumber job_snapshot, SnapshotChecker* snapshot_checker,
     bool paranoid_file_checks, InternalStats* internal_stats,
@@ -195,7 +195,7 @@ Status BuildTable(
 
     const std::atomic<bool> kManualCompactionCanceledFalse{false};
     CompactionIterator c_iter(
-        iter, ucmp, &merge, kMaxSequenceNumber, &snapshots,
+        iter, ucmp, &merge, kMaxSequenceNumber, &snapshots, earliest_snapshot,
         earliest_write_conflict_snapshot, job_snapshot, snapshot_checker, env,
         ShouldReportDetailedTime(env, ioptions.stats),
         true /* internal key corruption is not ok */, range_del_agg.get(),
@@ -210,6 +210,7 @@ Status BuildTable(
     const bool logical_strip_timestamp =
         ts_sz > 0 && !ioptions.persist_user_defined_timestamps;
 
+    SequenceNumber smallest_preferred_seqno = kMaxSequenceNumber;
     std::string key_after_flush_buf;
     std::string value_buf;
     c_iter.SeekToFirst();
@@ -242,6 +243,8 @@ Status BuildTable(
         if (preferred_seqno < ikey.sequence) {
           value_after_flush =
               PackValueAndSeqno(unpacked_value, preferred_seqno, &value_buf);
+          smallest_preferred_seqno =
+              std::min(smallest_preferred_seqno, preferred_seqno);
         } else {
           // Cannot get a useful preferred seqno, convert it to a kTypeValue.
           UpdateInternalKey(&key_after_flush_buf, ikey.sequence, kTypeValue);
@@ -326,9 +329,10 @@ Status BuildTable(
     } else {
       SeqnoToTimeMapping relevant_mapping;
       if (seqno_to_time_mapping) {
-        relevant_mapping.CopyFromSeqnoRange(*seqno_to_time_mapping,
-                                            meta->fd.smallest_seqno,
-                                            meta->fd.largest_seqno);
+        relevant_mapping.CopyFromSeqnoRange(
+            *seqno_to_time_mapping,
+            std::min(meta->fd.smallest_seqno, smallest_preferred_seqno),
+            meta->fd.largest_seqno);
         relevant_mapping.SetCapacity(kMaxSeqnoTimePairsPerSST);
         relevant_mapping.Enforce(tboptions.file_creation_time);
       }

package/vendor/librocksdb/vendor/rocksdb/db/builder.h

@@ -57,7 +57,7 @@ Status BuildTable(
     std::vector<std::unique_ptr<FragmentedRangeTombstoneIterator>>
         range_del_iters,
     FileMetaData* meta, std::vector<BlobFileAddition>* blob_file_additions,
-    std::vector<SequenceNumber> snapshots,
+    std::vector<SequenceNumber> snapshots, SequenceNumber earliest_snapshot,
     SequenceNumber earliest_write_conflict_snapshot,
     SequenceNumber job_snapshot, SnapshotChecker* snapshot_checker,
     bool paranoid_file_checks, InternalStats* internal_stats,