|
|
|
// Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
|
|
|
|
// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
|
|
|
|
// Use of this source code is governed by a BSD-style license that can be
|
|
|
|
// found in the LICENSE file. See the AUTHORS file for names of contributors.
|
|
|
|
|
|
|
|
#pragma once
|
|
|
|
|
|
|
|
#ifndef ROCKSDB_LITE
|
|
|
|
#include <unordered_map>
|
|
|
|
#include <memory>
|
|
|
|
#include <vector>
|
|
|
|
#include <string>
|
|
|
|
#include <stdint.h>
|
|
|
|
|
|
|
|
#include "db/dbformat.h"
|
|
|
|
#include "memory/arena.h"
|
|
|
|
#include "rocksdb/env.h"
|
|
|
|
#include "rocksdb/iterator.h"
|
|
|
|
#include "rocksdb/slice_transform.h"
|
|
|
|
#include "rocksdb/table.h"
|
|
|
|
#include "rocksdb/table_properties.h"
|
|
|
|
#include "table/plain/plain_table_factory.h"
|
|
|
|
#include "table/plain/plain_table_index.h"
|
|
|
|
#include "table/table_reader.h"
|
|
|
|
#include "util/dynamic_bloom.h"
|
Move rate_limiter, write buffering, most perf context instrumentation and most random kill out of Env
Summary: We want to keep Env a think layer for better portability. Less platform dependent codes should be moved out of Env. In this patch, I create a wrapper of file readers and writers, and put rate limiting, write buffering, as well as most perf context instrumentation and random kill out of Env. It will make it easier to maintain multiple Env in the future.
Test Plan: Run all existing unit tests.
Reviewers: anthony, kradhakrishnan, IslamAbdelRahman, yhchiang, igor
Reviewed By: igor
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D42321
9 years ago
|
|
|
#include "util/file_reader_writer.h"
|
|
|
|
|
|
|
|
namespace rocksdb {
|
|
|
|
|
|
|
|
class Block;
|
|
|
|
struct BlockContents;
|
|
|
|
class BlockHandle;
|
|
|
|
class Footer;
|
|
|
|
struct Options;
|
|
|
|
class RandomAccessFile;
|
|
|
|
struct ReadOptions;
|
|
|
|
class TableCache;
|
|
|
|
class TableReader;
|
|
|
|
class InternalKeyComparator;
|
|
|
|
class PlainTableKeyDecoder;
|
|
|
|
class GetContext;
|
|
|
|
|
|
|
|
extern const uint32_t kPlainTableVariableLength;
|
|
|
|
|
|
|
|
struct PlainTableReaderFileInfo {
|
|
|
|
bool is_mmap_mode;
|
|
|
|
Slice file_data;
|
|
|
|
uint32_t data_end_offset;
|
|
|
|
std::unique_ptr<RandomAccessFileReader> file;
|
|
|
|
|
|
|
|
PlainTableReaderFileInfo(std::unique_ptr<RandomAccessFileReader>&& _file,
|
|
|
|
const EnvOptions& storage_options,
|
|
|
|
uint32_t _data_size_offset)
|
|
|
|
: is_mmap_mode(storage_options.use_mmap_reads),
|
|
|
|
data_end_offset(_data_size_offset),
|
|
|
|
file(std::move(_file)) {}
|
|
|
|
};
|
|
|
|
|
|
|
|
// The reader class of PlainTable. For description of PlainTable format
|
|
|
|
// See comments of class PlainTableFactory, where instances of
|
|
|
|
// PlainTableReader are created.
|
|
|
|
class PlainTableReader: public TableReader {
|
|
|
|
public:
|
|
|
|
// Based on following output file format shown in plain_table_factory.h
|
|
|
|
// When opening the output file, PlainTableReader creates a hash table
|
|
|
|
// from key prefixes to offset of the output file. PlainTable will decide
|
|
|
|
// whether it points to the data offset of the first key with the key prefix
|
|
|
|
// or the offset of it. If there are too many keys share this prefix, it will
|
|
|
|
// create a binary search-able index from the suffix to offset on disk.
|
|
|
|
static Status Open(const ImmutableCFOptions& ioptions,
|
|
|
|
const EnvOptions& env_options,
|
|
|
|
const InternalKeyComparator& internal_comparator,
|
|
|
|
std::unique_ptr<RandomAccessFileReader>&& file,
|
|
|
|
uint64_t file_size, std::unique_ptr<TableReader>* table,
|
|
|
|
const int bloom_bits_per_key, double hash_table_ratio,
|
|
|
|
size_t index_sparseness, size_t huge_page_tlb_size,
|
|
|
|
bool full_scan_mode, const bool immortal_table = false,
|
|
|
|
const SliceTransform* prefix_extractor = nullptr);
|
|
|
|
|
|
|
|
InternalIterator* NewIterator(const ReadOptions&,
|
|
|
|
const SliceTransform* prefix_extractor,
|
|
|
|
Arena* arena = nullptr,
|
|
|
|
bool skip_filters = false,
|
|
|
|
bool for_compaction = false) override;
|
|
|
|
|
|
|
|
void Prepare(const Slice& target) override;
|
|
|
|
|
|
|
|
Status Get(const ReadOptions& readOptions, const Slice& key,
|
|
|
|
GetContext* get_context, const SliceTransform* prefix_extractor,
|
Skip bottom-level filter block caching when hit-optimized
Summary:
When Get() or NewIterator() trigger file loads, skip caching the filter block if
(1) optimize_filters_for_hits is set and (2) the file is on the bottommost
level. Also skip checking filters under the same conditions, which means that
for a preloaded file or a file that was trivially-moved to the bottom level, its
filter block will eventually expire from the cache.
- added parameters/instance variables in various places in order to propagate the config ("skip_filters") from version_set to block_based_table_reader
- in BlockBasedTable::Rep, this optimization prevents filter from being loaded when the file is opened simply by setting filter_policy = nullptr
- in BlockBasedTable::Get/BlockBasedTable::NewIterator, this optimization prevents filter from being used (even if it was loaded already) by setting filter = nullptr
Test Plan:
updated unit test:
$ ./db_test --gtest_filter=DBTest.OptimizeFiltersForHits
will also run 'make check'
Reviewers: sdong, igor, paultuckfield, anthony, rven, kradhakrishnan, IslamAbdelRahman, yhchiang
Reviewed By: yhchiang
Subscribers: leveldb
Differential Revision: https://reviews.facebook.net/D51633
9 years ago
|
|
|
bool skip_filters = false) override;
|
|
|
|
|
|
|
|
uint64_t ApproximateOffsetOf(const Slice& key,
|
|
|
|
bool for_compaction = false) override;
|
|
|
|
|
|
|
|
uint32_t GetIndexSize() const { return index_.GetIndexSize(); }
|
|
|
|
void SetupForCompaction() override;
|
|
|
|
|
|
|
|
std::shared_ptr<const TableProperties> GetTableProperties() const override {
|
|
|
|
return table_properties_;
|
|
|
|
}
|
|
|
|
|
|
|
|
virtual size_t ApproximateMemoryUsage() const override {
|
|
|
|
return arena_.MemoryAllocatedBytes();
|
|
|
|
}
|
|
|
|
|
|
|
|
PlainTableReader(const ImmutableCFOptions& ioptions,
|
|
|
|
std::unique_ptr<RandomAccessFileReader>&& file,
|
|
|
|
const EnvOptions& env_options,
|
|
|
|
const InternalKeyComparator& internal_comparator,
|
|
|
|
EncodingType encoding_type, uint64_t file_size,
|
|
|
|
const TableProperties* table_properties,
|
|
|
|
const SliceTransform* prefix_extractor);
|
|
|
|
virtual ~PlainTableReader();
|
|
|
|
|
|
|
|
protected:
|
|
|
|
// Check bloom filter to see whether it might contain this prefix.
|
|
|
|
// The hash of the prefix is given, since it can be reused for index lookup
|
|
|
|
// too.
|
|
|
|
virtual bool MatchBloom(uint32_t hash) const;
|
|
|
|
|
|
|
|
// PopulateIndex() builds index of keys. It must be called before any query
|
|
|
|
// to the table.
|
|
|
|
//
|
|
|
|
// props: the table properties object that need to be stored. Ownership of
|
|
|
|
// the object will be passed.
|
|
|
|
//
|
|
|
|
|
|
|
|
Status PopulateIndex(TableProperties* props, int bloom_bits_per_key,
|
|
|
|
double hash_table_ratio, size_t index_sparseness,
|
|
|
|
size_t huge_page_tlb_size);
|
|
|
|
|
|
|
|
Status MmapDataIfNeeded();
|
|
|
|
|
|
|
|
private:
|
|
|
|
const InternalKeyComparator internal_comparator_;
|
|
|
|
EncodingType encoding_type_;
|
|
|
|
// represents plain table's current status.
|
|
|
|
Status status_;
|
|
|
|
|
|
|
|
PlainTableIndex index_;
|
|
|
|
bool full_scan_mode_;
|
|
|
|
|
|
|
|
// data_start_offset_ and data_end_offset_ defines the range of the
|
|
|
|
// sst file that stores data.
|
|
|
|
const uint32_t data_start_offset_ = 0;
|
|
|
|
const uint32_t user_key_len_;
|
|
|
|
const SliceTransform* prefix_extractor_;
|
|
|
|
|
|
|
|
static const size_t kNumInternalBytes = 8;
|
|
|
|
|
|
|
|
// Bloom filter is used to rule out non-existent key
|
|
|
|
bool enable_bloom_;
|
|
|
|
DynamicBloom bloom_;
|
|
|
|
PlainTableReaderFileInfo file_info_;
|
|
|
|
Arena arena_;
|
|
|
|
CacheAllocationPtr index_block_alloc_;
|
|
|
|
CacheAllocationPtr bloom_block_alloc_;
|
|
|
|
|
|
|
|
const ImmutableCFOptions& ioptions_;
|
|
|
|
std::unique_ptr<Cleanable> dummy_cleanable_;
|
|
|
|
uint64_t file_size_;
|
|
|
|
std::shared_ptr<const TableProperties> table_properties_;
|
|
|
|
|
|
|
|
bool IsFixedLength() const {
|
|
|
|
return user_key_len_ != kPlainTableVariableLength;
|
|
|
|
}
|
|
|
|
|
|
|
|
size_t GetFixedInternalKeyLength() const {
|
|
|
|
return user_key_len_ + kNumInternalBytes;
|
|
|
|
}
|
|
|
|
|
|
|
|
Slice GetPrefix(const Slice& target) const {
|
|
|
|
assert(target.size() >= 8); // target is internal key
|
|
|
|
return GetPrefixFromUserKey(GetUserKey(target));
|
|
|
|
}
|
|
|
|
|
|
|
|
Slice GetPrefix(const ParsedInternalKey& target) const {
|
|
|
|
return GetPrefixFromUserKey(target.user_key);
|
|
|
|
}
|
|
|
|
|
|
|
|
Slice GetUserKey(const Slice& key) const {
|
|
|
|
return Slice(key.data(), key.size() - 8);
|
|
|
|
}
|
|
|
|
|
|
|
|
Slice GetPrefixFromUserKey(const Slice& user_key) const {
|
|
|
|
if (!IsTotalOrderMode()) {
|
|
|
|
return prefix_extractor_->Transform(user_key);
|
|
|
|
} else {
|
|
|
|
// Use empty slice as prefix if prefix_extractor is not set.
|
|
|
|
// In that case,
|
|
|
|
// it falls back to pure binary search and
|
|
|
|
// total iterator seek is supported.
|
|
|
|
return Slice();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
friend class TableCache;
|
|
|
|
friend class PlainTableIterator;
|
|
|
|
|
|
|
|
// Internal helper function to generate an IndexRecordList object from all
|
|
|
|
// the rows, which contains index records as a list.
|
|
|
|
// If bloom_ is not null, all the keys' full-key hash will be added to the
|
|
|
|
// bloom filter.
|
|
|
|
Status PopulateIndexRecordList(PlainTableIndexBuilder* index_builder,
|
|
|
|
std::vector<uint32_t>* prefix_hashes);
|
|
|
|
|
|
|
|
// Internal helper function to allocate memory for bloom filter and fill it
|
|
|
|
void AllocateAndFillBloom(int bloom_bits_per_key, int num_prefixes,
|
|
|
|
size_t huge_page_tlb_size,
|
|
|
|
std::vector<uint32_t>* prefix_hashes);
|
|
|
|
|
|
|
|
void FillBloom(std::vector<uint32_t>* prefix_hashes);
|
|
|
|
|
|
|
|
// Read the key and value at `offset` to parameters for keys, the and
|
|
|
|
// `seekable`.
|
|
|
|
// On success, `offset` will be updated as the offset for the next key.
|
|
|
|
// `parsed_key` will be key in parsed format.
|
|
|
|
// if `internal_key` is not empty, it will be filled with key with slice
|
|
|
|
// format.
|
|
|
|
// if `seekable` is not null, it will return whether we can directly read
|
|
|
|
// data using this offset.
|
|
|
|
Status Next(PlainTableKeyDecoder* decoder, uint32_t* offset,
|
|
|
|
ParsedInternalKey* parsed_key, Slice* internal_key, Slice* value,
|
|
|
|
bool* seekable = nullptr) const;
|
|
|
|
// Get file offset for key target.
|
|
|
|
// return value prefix_matched is set to true if the offset is confirmed
|
|
|
|
// for a key with the same prefix as target.
|
|
|
|
Status GetOffset(PlainTableKeyDecoder* decoder, const Slice& target,
|
|
|
|
const Slice& prefix, uint32_t prefix_hash,
|
|
|
|
bool& prefix_matched, uint32_t* offset) const;
|
|
|
|
|
|
|
|
bool IsTotalOrderMode() const { return (prefix_extractor_ == nullptr); }
|
|
|
|
|
|
|
|
// No copying allowed
|
|
|
|
explicit PlainTableReader(const TableReader&) = delete;
|
|
|
|
void operator=(const TableReader&) = delete;
|
|
|
|
};
|
|
|
|
} // namespace rocksdb
|
|
|
|
#endif // ROCKSDB_LITE
|