db_test.cc 69.1 KB
Newer Older
J
jorlow@chromium.org 已提交
1 2 3 4
// 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.

5
#include <algorithm>
6
#include "leveldb/db.h"
S
Sanjay Ghemawat 已提交
7
#include "leveldb/filter_policy.h"
J
jorlow@chromium.org 已提交
8 9 10 11
#include "db/db_impl.h"
#include "db/filename.h"
#include "db/version_set.h"
#include "db/write_batch_internal.h"
S
Sanjay Ghemawat 已提交
12
#include "leveldb/cache.h"
13 14
#include "leveldb/env.h"
#include "leveldb/table.h"
S
Sanjay Ghemawat 已提交
15
#include "util/hash.h"
J
jorlow@chromium.org 已提交
16
#include "util/logging.h"
17
#include "util/mutexlock.h"
J
jorlow@chromium.org 已提交
18 19 20 21 22
#include "util/testharness.h"
#include "util/testutil.h"

namespace leveldb {

23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40
static bool SnappyCompressionSupported() {	
  std::string out;	
  Slice in = "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa";	
  return port::Snappy_Compress(in.data(), in.size(), &out);	
}	
	
static bool ZlibCompressionSupported() {	
  std::string out;	
  Slice in = "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa";	
  return port::Zlib_Compress(in.data(), in.size(), &out);	
}	
	
static bool BZip2CompressionSupported() {	
  std::string out;	
  Slice in = "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa";	
  return port::BZip2_Compress(in.data(), in.size(), &out);	
}	

J
jorlow@chromium.org 已提交
41 42 43 44 45 46
static std::string RandomString(Random* rnd, int len) {
  std::string r;
  test::RandomString(rnd, len, &r);
  return r;
}

S
Sanjay Ghemawat 已提交
47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68
namespace {
class AtomicCounter {
 private:
  port::Mutex mu_;
  int count_;
 public:
  AtomicCounter() : count_(0) { }
  void Increment() {
    MutexLock l(&mu_);
    count_++;
  }
  int Read() {
    MutexLock l(&mu_);
    return count_;
  }
  void Reset() {
    MutexLock l(&mu_);
    count_ = 0;
  }
};
}

69 70 71 72 73 74
// Special Env used to delay background operations
class SpecialEnv : public EnvWrapper {
 public:
  // sstable Sync() calls are blocked while this pointer is non-NULL.
  port::AtomicPointer delay_sstable_sync_;

75 76 77
  // Simulate no-space errors while this pointer is non-NULL.
  port::AtomicPointer no_space_;

78 79 80
  // Simulate non-writable file system while this pointer is non-NULL
  port::AtomicPointer non_writable_;

S
Sanjay Ghemawat 已提交
81 82 83
  bool count_random_reads_;
  AtomicCounter random_read_counter_;

84 85
  AtomicCounter sleep_counter_;

86 87
  explicit SpecialEnv(Env* base) : EnvWrapper(base) {
    delay_sstable_sync_.Release_Store(NULL);
88
    no_space_.Release_Store(NULL);
89
    non_writable_.Release_Store(NULL);
S
Sanjay Ghemawat 已提交
90
    count_random_reads_ = false;
91 92 93 94 95 96 97 98 99 100 101 102 103
  }

  Status NewWritableFile(const std::string& f, WritableFile** r) {
    class SSTableFile : public WritableFile {
     private:
      SpecialEnv* env_;
      WritableFile* base_;

     public:
      SSTableFile(SpecialEnv* env, WritableFile* base)
          : env_(env),
            base_(base) {
      }
104
      ~SSTableFile() { delete base_; }
105 106 107 108 109 110 111 112
      Status Append(const Slice& data) {
        if (env_->no_space_.Acquire_Load() != NULL) {
          // Drop writes on the floor
          return Status::OK();
        } else {
          return base_->Append(data);
        }
      }
113 114 115 116 117 118 119 120 121 122
      Status Close() { return base_->Close(); }
      Status Flush() { return base_->Flush(); }
      Status Sync() {
        while (env_->delay_sstable_sync_.Acquire_Load() != NULL) {
          env_->SleepForMicroseconds(100000);
        }
        return base_->Sync();
      }
    };

123 124 125 126
    if (non_writable_.Acquire_Load() != NULL) {
       return Status::IOError("simulated write error");
     }

127 128 129 130 131 132 133 134
    Status s = target()->NewWritableFile(f, r);
    if (s.ok()) {
      if (strstr(f.c_str(), ".sst") != NULL) {
        *r = new SSTableFile(this, *r);
      }
    }
    return s;
  }
S
Sanjay Ghemawat 已提交
135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158

  Status NewRandomAccessFile(const std::string& f, RandomAccessFile** r) {
    class CountingFile : public RandomAccessFile {
     private:
      RandomAccessFile* target_;
      AtomicCounter* counter_;
     public:
      CountingFile(RandomAccessFile* target, AtomicCounter* counter)
          : target_(target), counter_(counter) {
      }
      virtual ~CountingFile() { delete target_; }
      virtual Status Read(uint64_t offset, size_t n, Slice* result,
                          char* scratch) const {
        counter_->Increment();
        return target_->Read(offset, n, result, scratch);
      }
    };

    Status s = target()->NewRandomAccessFile(f, r);
    if (s.ok() && count_random_reads_) {
      *r = new CountingFile(*r, &random_read_counter_);
    }
    return s;
  }
159 160 161 162 163

  virtual void SleepForMicroseconds(int micros) {
    sleep_counter_.Increment();
    target()->SleepForMicroseconds(micros);
  }
164 165
};

J
jorlow@chromium.org 已提交
166
class DBTest {
S
Sanjay Ghemawat 已提交
167 168 169 170 171 172 173
 private:
  const FilterPolicy* filter_policy_;

  // Sequence of option configurations to try
  enum OptionConfig {
    kDefault,
    kFilter,
H
heyongqiang 已提交
174
    kUncompressed,
175
    kNumLevel_3,
H
heyongqiang 已提交
176
    kDBLogDir,
S
Sanjay Ghemawat 已提交
177 178 179 180
    kEnd
  };
  int option_config_;

J
jorlow@chromium.org 已提交
181 182
 public:
  std::string dbname_;
183
  SpecialEnv* env_;
J
jorlow@chromium.org 已提交
184 185 186 187
  DB* db_;

  Options last_options_;

S
Sanjay Ghemawat 已提交
188 189 190
  DBTest() : option_config_(kDefault),
             env_(new SpecialEnv(Env::Default())) {
    filter_policy_ = NewBloomFilterPolicy(10);
J
jorlow@chromium.org 已提交
191 192 193 194 195 196 197 198 199
    dbname_ = test::TmpDir() + "/db_test";
    DestroyDB(dbname_, Options());
    db_ = NULL;
    Reopen();
  }

  ~DBTest() {
    delete db_;
    DestroyDB(dbname_, Options());
200
    delete env_;
S
Sanjay Ghemawat 已提交
201 202 203 204 205 206
    delete filter_policy_;
  }

  // Switch to a fresh database with the next option configuration to
  // test.  Return false if there are no more configurations to test.
  bool ChangeOptions() {
H
heyongqiang 已提交
207 208
    option_config_++;
    if (option_config_ >= kEnd) {
S
Sanjay Ghemawat 已提交
209 210 211 212 213 214 215 216 217 218 219 220 221 222
      return false;
    } else {
      DestroyAndReopen();
      return true;
    }
  }

  // Return the current option configuration.
  Options CurrentOptions() {
    Options options;
    switch (option_config_) {
      case kFilter:
        options.filter_policy = filter_policy_;
        break;
H
heyongqiang 已提交
223
      case kUncompressed:
H
heyongqiang 已提交
224 225
        options.compression = kNoCompression;
        break;
226
      case kNumLevel_3:
H
heyongqiang 已提交
227 228 229 230 231
        options.num_levels = 3;
        break;
      case kDBLogDir:
        options.db_log_dir = test::TmpDir();
        break;
S
Sanjay Ghemawat 已提交
232 233 234 235
      default:
        break;
    }
    return options;
J
jorlow@chromium.org 已提交
236 237 238 239 240 241 242 243 244 245
  }

  DBImpl* dbfull() {
    return reinterpret_cast<DBImpl*>(db_);
  }

  void Reopen(Options* options = NULL) {
    ASSERT_OK(TryReopen(options));
  }

S
Sanjay Ghemawat 已提交
246 247 248 249 250
  void Close() {
    delete db_;
    db_ = NULL;
  }

J
jorlow@chromium.org 已提交
251 252 253 254 255 256 257
  void DestroyAndReopen(Options* options = NULL) {
    delete db_;
    db_ = NULL;
    DestroyDB(dbname_, Options());
    ASSERT_OK(TryReopen(options));
  }

258 259 260 261
  Status PureReopen(Options* options, DB** db) {
    return DB::Open(*options, dbname_, db);
  }

J
jorlow@chromium.org 已提交
262 263 264 265 266 267 268
  Status TryReopen(Options* options) {
    delete db_;
    db_ = NULL;
    Options opts;
    if (options != NULL) {
      opts = *options;
    } else {
S
Sanjay Ghemawat 已提交
269
      opts = CurrentOptions();
J
jorlow@chromium.org 已提交
270 271 272 273 274 275 276 277
      opts.create_if_missing = true;
    }
    last_options_ = opts;

    return DB::Open(opts, dbname_, &db_);
  }

  Status Put(const std::string& k, const std::string& v) {
278
    return db_->Put(WriteOptions(), k, v);
J
jorlow@chromium.org 已提交
279 280 281
  }

  Status Delete(const std::string& k) {
282
    return db_->Delete(WriteOptions(), k);
J
jorlow@chromium.org 已提交
283 284 285 286 287 288 289 290 291 292 293 294 295 296 297
  }

  std::string Get(const std::string& k, const Snapshot* snapshot = NULL) {
    ReadOptions options;
    options.snapshot = snapshot;
    std::string result;
    Status s = db_->Get(options, k, &result);
    if (s.IsNotFound()) {
      result = "NOT_FOUND";
    } else if (!s.ok()) {
      result = s.ToString();
    }
    return result;
  }

H
Hans Wennborg 已提交
298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324
  // Return a string that contains all key,value pairs in order,
  // formatted like "(k1->v1)(k2->v2)".
  std::string Contents() {
    std::vector<std::string> forward;
    std::string result;
    Iterator* iter = db_->NewIterator(ReadOptions());
    for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
      std::string s = IterStatus(iter);
      result.push_back('(');
      result.append(s);
      result.push_back(')');
      forward.push_back(s);
    }

    // Check reverse iteration results are the reverse of forward results
    int matched = 0;
    for (iter->SeekToLast(); iter->Valid(); iter->Prev()) {
      ASSERT_LT(matched, forward.size());
      ASSERT_EQ(IterStatus(iter), forward[forward.size() - matched - 1]);
      matched++;
    }
    ASSERT_EQ(matched, forward.size());

    delete iter;
    return result;
  }

J
jorlow@chromium.org 已提交
325 326 327 328 329 330 331 332 333 334 335 336 337 338 339
  std::string AllEntriesFor(const Slice& user_key) {
    Iterator* iter = dbfull()->TEST_NewInternalIterator();
    InternalKey target(user_key, kMaxSequenceNumber, kTypeValue);
    iter->Seek(target.Encode());
    std::string result;
    if (!iter->status().ok()) {
      result = iter->status().ToString();
    } else {
      result = "[ ";
      bool first = true;
      while (iter->Valid()) {
        ParsedInternalKey ikey;
        if (!ParseInternalKey(iter->key(), &ikey)) {
          result += "CORRUPTED";
        } else {
S
Sanjay Ghemawat 已提交
340
          if (last_options_.comparator->Compare(ikey.user_key, user_key) != 0) {
J
jorlow@chromium.org 已提交
341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367
            break;
          }
          if (!first) {
            result += ", ";
          }
          first = false;
          switch (ikey.type) {
            case kTypeValue:
              result += iter->value().ToString();
              break;
            case kTypeDeletion:
              result += "DEL";
              break;
          }
        }
        iter->Next();
      }
      if (!first) {
        result += " ";
      }
      result += "]";
    }
    delete iter;
    return result;
  }

  int NumTableFilesAtLevel(int level) {
368
    std::string property;
J
jorlow@chromium.org 已提交
369 370
    ASSERT_TRUE(
        db_->GetProperty("leveldb.num-files-at-level" + NumberToString(level),
371 372
                         &property));
    return atoi(property.c_str());
J
jorlow@chromium.org 已提交
373 374
  }

375 376
  int TotalTableFiles() {
    int result = 0;
377
    for (int level = 0; level < db_->NumberLevels(); level++) {
378 379 380 381 382
      result += NumTableFilesAtLevel(level);
    }
    return result;
  }

G
Gabor Cselle 已提交
383 384 385 386
  // Return spread of files per level
  std::string FilesPerLevel() {
    std::string result;
    int last_non_zero_offset = 0;
387
    for (int level = 0; level < db_->NumberLevels(); level++) {
G
Gabor Cselle 已提交
388 389 390 391 392 393 394 395 396 397 398 399
      int f = NumTableFilesAtLevel(level);
      char buf[100];
      snprintf(buf, sizeof(buf), "%s%d", (level ? "," : ""), f);
      result += buf;
      if (f > 0) {
        last_non_zero_offset = result.size();
      }
    }
    result.resize(last_non_zero_offset);
    return result;
  }

400 401 402 403 404 405
  int CountFiles() {
    std::vector<std::string> files;
    env_->GetChildren(dbname_, &files);
    return static_cast<int>(files.size());
  }

406 407 408 409 410 411 412
  int CountLiveFiles() {
    std::vector<std::string> files;
    uint64_t manifest_file_size;
    db_->GetLiveFiles(files, &manifest_file_size);
    return files.size();
  }

J
jorlow@chromium.org 已提交
413 414 415 416 417 418 419
  uint64_t Size(const Slice& start, const Slice& limit) {
    Range r(start, limit);
    uint64_t size;
    db_->GetApproximateSizes(&r, 1, &size);
    return size;
  }

420
  void Compact(const Slice& start, const Slice& limit) {
G
Gabor Cselle 已提交
421 422 423 424 425 426 427 428 429 430
    db_->CompactRange(&start, &limit);
  }

  // Do n memtable compactions, each of which produces an sstable
  // covering the range [small,large].
  void MakeTables(int n, const std::string& small, const std::string& large) {
    for (int i = 0; i < n; i++) {
      Put(small, "begin");
      Put(large, "end");
      dbfull()->TEST_CompactMemTable();
431 432 433
    }
  }

434 435 436
  // Prevent pushing of new sstables into deeper levels by adding
  // tables that cover a specified range to all levels.
  void FillLevels(const std::string& smallest, const std::string& largest) {
437
    MakeTables(db_->NumberLevels(), smallest, largest);
438 439
  }

440 441 442 443 444
  void DumpFileCounts(const char* label) {
    fprintf(stderr, "---\n%s:\n", label);
    fprintf(stderr, "maxoverlap: %lld\n",
            static_cast<long long>(
                dbfull()->TEST_MaxNextLevelOverlappingBytes()));
445
    for (int level = 0; level < db_->NumberLevels(); level++) {
446 447 448 449 450 451
      int num = NumTableFilesAtLevel(level);
      if (num > 0) {
        fprintf(stderr, "  level %3d : %d files\n", level, num);
      }
    }
  }
J
jorlow@chromium.org 已提交
452

G
Gabor Cselle 已提交
453 454 455 456 457 458
  std::string DumpSSTableList() {
    std::string property;
    db_->GetProperty("leveldb.sstables", &property);
    return property;
  }

J
jorlow@chromium.org 已提交
459 460 461 462 463 464 465 466 467
  std::string IterStatus(Iterator* iter) {
    std::string result;
    if (iter->Valid()) {
      result = iter->key().ToString() + "->" + iter->value().ToString();
    } else {
      result = "(invalid)";
    }
    return result;
  }
J
jorlow@chromium.org 已提交
468 469 470
};

TEST(DBTest, Empty) {
S
Sanjay Ghemawat 已提交
471 472 473 474
  do {
    ASSERT_TRUE(db_ != NULL);
    ASSERT_EQ("NOT_FOUND", Get("foo"));
  } while (ChangeOptions());
J
jorlow@chromium.org 已提交
475 476 477
}

TEST(DBTest, ReadWrite) {
S
Sanjay Ghemawat 已提交
478 479 480 481 482 483 484 485
  do {
    ASSERT_OK(Put("foo", "v1"));
    ASSERT_EQ("v1", Get("foo"));
    ASSERT_OK(Put("bar", "v2"));
    ASSERT_OK(Put("foo", "v3"));
    ASSERT_EQ("v3", Get("foo"));
    ASSERT_EQ("v2", Get("bar"));
  } while (ChangeOptions());
J
jorlow@chromium.org 已提交
486 487 488
}

TEST(DBTest, PutDeleteGet) {
S
Sanjay Ghemawat 已提交
489 490 491 492 493 494 495 496
  do {
    ASSERT_OK(db_->Put(WriteOptions(), "foo", "v1"));
    ASSERT_EQ("v1", Get("foo"));
    ASSERT_OK(db_->Put(WriteOptions(), "foo", "v2"));
    ASSERT_EQ("v2", Get("foo"));
    ASSERT_OK(db_->Delete(WriteOptions(), "foo"));
    ASSERT_EQ("NOT_FOUND", Get("foo"));
  } while (ChangeOptions());
J
jorlow@chromium.org 已提交
497 498
}

499
TEST(DBTest, GetFromImmutableLayer) {
S
Sanjay Ghemawat 已提交
500 501 502 503 504
  do {
    Options options = CurrentOptions();
    options.env = env_;
    options.write_buffer_size = 100000;  // Small write buffer
    Reopen(&options);
505

S
Sanjay Ghemawat 已提交
506 507
    ASSERT_OK(Put("foo", "v1"));
    ASSERT_EQ("v1", Get("foo"));
508

S
Sanjay Ghemawat 已提交
509 510 511 512 513 514
    env_->delay_sstable_sync_.Release_Store(env_);   // Block sync calls
    Put("k1", std::string(100000, 'x'));             // Fill memtable
    Put("k2", std::string(100000, 'y'));             // Trigger compaction
    ASSERT_EQ("v1", Get("foo"));
    env_->delay_sstable_sync_.Release_Store(NULL);   // Release sync calls
  } while (ChangeOptions());
515 516 517
}

TEST(DBTest, GetFromVersions) {
S
Sanjay Ghemawat 已提交
518 519 520 521 522
  do {
    ASSERT_OK(Put("foo", "v1"));
    dbfull()->TEST_CompactMemTable();
    ASSERT_EQ("v1", Get("foo"));
  } while (ChangeOptions());
523 524 525
}

TEST(DBTest, GetSnapshot) {
S
Sanjay Ghemawat 已提交
526 527 528 529 530 531 532 533 534 535 536 537 538 539 540
  do {
    // Try with both a short key and a long key
    for (int i = 0; i < 2; i++) {
      std::string key = (i == 0) ? std::string("foo") : std::string(200, 'x');
      ASSERT_OK(Put(key, "v1"));
      const Snapshot* s1 = db_->GetSnapshot();
      ASSERT_OK(Put(key, "v2"));
      ASSERT_EQ("v2", Get(key));
      ASSERT_EQ("v1", Get(key, s1));
      dbfull()->TEST_CompactMemTable();
      ASSERT_EQ("v2", Get(key));
      ASSERT_EQ("v1", Get(key, s1));
      db_->ReleaseSnapshot(s1);
    }
  } while (ChangeOptions());
541 542 543
}

TEST(DBTest, GetLevel0Ordering) {
S
Sanjay Ghemawat 已提交
544 545 546 547 548 549 550 551 552 553 554 555
  do {
    // Check that we process level-0 files in correct order.  The code
    // below generates two level-0 files where the earlier one comes
    // before the later one in the level-0 file list since the earlier
    // one has a smaller "smallest" key.
    ASSERT_OK(Put("bar", "b"));
    ASSERT_OK(Put("foo", "v1"));
    dbfull()->TEST_CompactMemTable();
    ASSERT_OK(Put("foo", "v2"));
    dbfull()->TEST_CompactMemTable();
    ASSERT_EQ("v2", Get("foo"));
  } while (ChangeOptions());
556 557 558
}

TEST(DBTest, GetOrderedByLevels) {
S
Sanjay Ghemawat 已提交
559 560 561 562 563 564 565 566 567
  do {
    ASSERT_OK(Put("foo", "v1"));
    Compact("a", "z");
    ASSERT_EQ("v1", Get("foo"));
    ASSERT_OK(Put("foo", "v2"));
    ASSERT_EQ("v2", Get("foo"));
    dbfull()->TEST_CompactMemTable();
    ASSERT_EQ("v2", Get("foo"));
  } while (ChangeOptions());
568 569 570
}

TEST(DBTest, GetPicksCorrectFile) {
S
Sanjay Ghemawat 已提交
571 572 573 574 575 576 577 578 579 580 581 582
  do {
    // Arrange to have multiple files in a non-level-0 level.
    ASSERT_OK(Put("a", "va"));
    Compact("a", "b");
    ASSERT_OK(Put("x", "vx"));
    Compact("x", "y");
    ASSERT_OK(Put("f", "vf"));
    Compact("f", "g");
    ASSERT_EQ("va", Get("a"));
    ASSERT_EQ("vf", Get("f"));
    ASSERT_EQ("vx", Get("x"));
  } while (ChangeOptions());
583 584
}

585
TEST(DBTest, GetEncountersEmptyLevel) {
S
Sanjay Ghemawat 已提交
586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604
  do {
    // Arrange for the following to happen:
    //   * sstable A in level 0
    //   * nothing in level 1
    //   * sstable B in level 2
    // Then do enough Get() calls to arrange for an automatic compaction
    // of sstable A.  A bug would cause the compaction to be marked as
    // occuring at level 1 (instead of the correct level 0).

    // Step 1: First place sstables in levels 0 and 2
    int compaction_count = 0;
    while (NumTableFilesAtLevel(0) == 0 ||
           NumTableFilesAtLevel(2) == 0) {
      ASSERT_LE(compaction_count, 100) << "could not fill levels 0 and 2";
      compaction_count++;
      Put("a", "begin");
      Put("z", "end");
      dbfull()->TEST_CompactMemTable();
    }
605

S
Sanjay Ghemawat 已提交
606 607 608 609 610 611
    // Step 2: clear level 1 if necessary.
    dbfull()->TEST_CompactRange(1, NULL, NULL);
    ASSERT_EQ(NumTableFilesAtLevel(0), 1);
    ASSERT_EQ(NumTableFilesAtLevel(1), 0);
    ASSERT_EQ(NumTableFilesAtLevel(2), 1);

H
heyongqiang 已提交
612 613
    // Step 3: read a bunch of times
    for (int i = 0; i < 1000; i++) {
S
Sanjay Ghemawat 已提交
614 615
      ASSERT_EQ("NOT_FOUND", Get("missing"));
    }
H
heyongqiang 已提交
616 617 618 619

    // Step 4: Wait for compaction to finish
    env_->SleepForMicroseconds(1000000);

620
    ASSERT_EQ(NumTableFilesAtLevel(0), 1); // XXX
S
Sanjay Ghemawat 已提交
621
  } while (ChangeOptions());
622 623
}

J
jorlow@chromium.org 已提交
624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797
TEST(DBTest, IterEmpty) {
  Iterator* iter = db_->NewIterator(ReadOptions());

  iter->SeekToFirst();
  ASSERT_EQ(IterStatus(iter), "(invalid)");

  iter->SeekToLast();
  ASSERT_EQ(IterStatus(iter), "(invalid)");

  iter->Seek("foo");
  ASSERT_EQ(IterStatus(iter), "(invalid)");

  delete iter;
}

TEST(DBTest, IterSingle) {
  ASSERT_OK(Put("a", "va"));
  Iterator* iter = db_->NewIterator(ReadOptions());

  iter->SeekToFirst();
  ASSERT_EQ(IterStatus(iter), "a->va");
  iter->Next();
  ASSERT_EQ(IterStatus(iter), "(invalid)");
  iter->SeekToFirst();
  ASSERT_EQ(IterStatus(iter), "a->va");
  iter->Prev();
  ASSERT_EQ(IterStatus(iter), "(invalid)");

  iter->SeekToLast();
  ASSERT_EQ(IterStatus(iter), "a->va");
  iter->Next();
  ASSERT_EQ(IterStatus(iter), "(invalid)");
  iter->SeekToLast();
  ASSERT_EQ(IterStatus(iter), "a->va");
  iter->Prev();
  ASSERT_EQ(IterStatus(iter), "(invalid)");

  iter->Seek("");
  ASSERT_EQ(IterStatus(iter), "a->va");
  iter->Next();
  ASSERT_EQ(IterStatus(iter), "(invalid)");

  iter->Seek("a");
  ASSERT_EQ(IterStatus(iter), "a->va");
  iter->Next();
  ASSERT_EQ(IterStatus(iter), "(invalid)");

  iter->Seek("b");
  ASSERT_EQ(IterStatus(iter), "(invalid)");

  delete iter;
}

TEST(DBTest, IterMulti) {
  ASSERT_OK(Put("a", "va"));
  ASSERT_OK(Put("b", "vb"));
  ASSERT_OK(Put("c", "vc"));
  Iterator* iter = db_->NewIterator(ReadOptions());

  iter->SeekToFirst();
  ASSERT_EQ(IterStatus(iter), "a->va");
  iter->Next();
  ASSERT_EQ(IterStatus(iter), "b->vb");
  iter->Next();
  ASSERT_EQ(IterStatus(iter), "c->vc");
  iter->Next();
  ASSERT_EQ(IterStatus(iter), "(invalid)");
  iter->SeekToFirst();
  ASSERT_EQ(IterStatus(iter), "a->va");
  iter->Prev();
  ASSERT_EQ(IterStatus(iter), "(invalid)");

  iter->SeekToLast();
  ASSERT_EQ(IterStatus(iter), "c->vc");
  iter->Prev();
  ASSERT_EQ(IterStatus(iter), "b->vb");
  iter->Prev();
  ASSERT_EQ(IterStatus(iter), "a->va");
  iter->Prev();
  ASSERT_EQ(IterStatus(iter), "(invalid)");
  iter->SeekToLast();
  ASSERT_EQ(IterStatus(iter), "c->vc");
  iter->Next();
  ASSERT_EQ(IterStatus(iter), "(invalid)");

  iter->Seek("");
  ASSERT_EQ(IterStatus(iter), "a->va");
  iter->Seek("a");
  ASSERT_EQ(IterStatus(iter), "a->va");
  iter->Seek("ax");
  ASSERT_EQ(IterStatus(iter), "b->vb");
  iter->Seek("b");
  ASSERT_EQ(IterStatus(iter), "b->vb");
  iter->Seek("z");
  ASSERT_EQ(IterStatus(iter), "(invalid)");

  // Switch from reverse to forward
  iter->SeekToLast();
  iter->Prev();
  iter->Prev();
  iter->Next();
  ASSERT_EQ(IterStatus(iter), "b->vb");

  // Switch from forward to reverse
  iter->SeekToFirst();
  iter->Next();
  iter->Next();
  iter->Prev();
  ASSERT_EQ(IterStatus(iter), "b->vb");

  // Make sure iter stays at snapshot
  ASSERT_OK(Put("a",  "va2"));
  ASSERT_OK(Put("a2", "va3"));
  ASSERT_OK(Put("b",  "vb2"));
  ASSERT_OK(Put("c",  "vc2"));
  ASSERT_OK(Delete("b"));
  iter->SeekToFirst();
  ASSERT_EQ(IterStatus(iter), "a->va");
  iter->Next();
  ASSERT_EQ(IterStatus(iter), "b->vb");
  iter->Next();
  ASSERT_EQ(IterStatus(iter), "c->vc");
  iter->Next();
  ASSERT_EQ(IterStatus(iter), "(invalid)");
  iter->SeekToLast();
  ASSERT_EQ(IterStatus(iter), "c->vc");
  iter->Prev();
  ASSERT_EQ(IterStatus(iter), "b->vb");
  iter->Prev();
  ASSERT_EQ(IterStatus(iter), "a->va");
  iter->Prev();
  ASSERT_EQ(IterStatus(iter), "(invalid)");

  delete iter;
}

TEST(DBTest, IterSmallAndLargeMix) {
  ASSERT_OK(Put("a", "va"));
  ASSERT_OK(Put("b", std::string(100000, 'b')));
  ASSERT_OK(Put("c", "vc"));
  ASSERT_OK(Put("d", std::string(100000, 'd')));
  ASSERT_OK(Put("e", std::string(100000, 'e')));

  Iterator* iter = db_->NewIterator(ReadOptions());

  iter->SeekToFirst();
  ASSERT_EQ(IterStatus(iter), "a->va");
  iter->Next();
  ASSERT_EQ(IterStatus(iter), "b->" + std::string(100000, 'b'));
  iter->Next();
  ASSERT_EQ(IterStatus(iter), "c->vc");
  iter->Next();
  ASSERT_EQ(IterStatus(iter), "d->" + std::string(100000, 'd'));
  iter->Next();
  ASSERT_EQ(IterStatus(iter), "e->" + std::string(100000, 'e'));
  iter->Next();
  ASSERT_EQ(IterStatus(iter), "(invalid)");

  iter->SeekToLast();
  ASSERT_EQ(IterStatus(iter), "e->" + std::string(100000, 'e'));
  iter->Prev();
  ASSERT_EQ(IterStatus(iter), "d->" + std::string(100000, 'd'));
  iter->Prev();
  ASSERT_EQ(IterStatus(iter), "c->vc");
  iter->Prev();
  ASSERT_EQ(IterStatus(iter), "b->" + std::string(100000, 'b'));
  iter->Prev();
  ASSERT_EQ(IterStatus(iter), "a->va");
  iter->Prev();
  ASSERT_EQ(IterStatus(iter), "(invalid)");

  delete iter;
}

798
TEST(DBTest, IterMultiWithDelete) {
S
Sanjay Ghemawat 已提交
799 800 801 802 803 804
  do {
    ASSERT_OK(Put("a", "va"));
    ASSERT_OK(Put("b", "vb"));
    ASSERT_OK(Put("c", "vc"));
    ASSERT_OK(Delete("b"));
    ASSERT_EQ("NOT_FOUND", Get("b"));
805

S
Sanjay Ghemawat 已提交
806 807 808 809 810 811 812
    Iterator* iter = db_->NewIterator(ReadOptions());
    iter->Seek("c");
    ASSERT_EQ(IterStatus(iter), "c->vc");
    iter->Prev();
    ASSERT_EQ(IterStatus(iter), "a->va");
    delete iter;
  } while (ChangeOptions());
813 814
}

J
jorlow@chromium.org 已提交
815
TEST(DBTest, Recover) {
S
Sanjay Ghemawat 已提交
816 817 818
  do {
    ASSERT_OK(Put("foo", "v1"));
    ASSERT_OK(Put("baz", "v5"));
J
jorlow@chromium.org 已提交
819

S
Sanjay Ghemawat 已提交
820 821
    Reopen();
    ASSERT_EQ("v1", Get("foo"));
J
jorlow@chromium.org 已提交
822

S
Sanjay Ghemawat 已提交
823 824 825 826
    ASSERT_EQ("v1", Get("foo"));
    ASSERT_EQ("v5", Get("baz"));
    ASSERT_OK(Put("bar", "v2"));
    ASSERT_OK(Put("foo", "v3"));
J
jorlow@chromium.org 已提交
827

S
Sanjay Ghemawat 已提交
828 829 830 831 832 833 834
    Reopen();
    ASSERT_EQ("v3", Get("foo"));
    ASSERT_OK(Put("foo", "v4"));
    ASSERT_EQ("v4", Get("foo"));
    ASSERT_EQ("v2", Get("bar"));
    ASSERT_EQ("v5", Get("baz"));
  } while (ChangeOptions());
J
jorlow@chromium.org 已提交
835 836
}

H
heyongqiang 已提交
837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852
TEST(DBTest, RollLog) {
  do {
    ASSERT_OK(Put("foo", "v1"));
    ASSERT_OK(Put("baz", "v5"));

    Reopen();
    for (int i = 0; i < 10; i++) {
      Reopen();
    }
    ASSERT_OK(Put("foo", "v4"));
    for (int i = 0; i < 10; i++) {
      Reopen();
    }
  } while (ChangeOptions());
}

H
heyongqiang 已提交
853 854 855 856 857
TEST(DBTest, WAL) {
  Options options = CurrentOptions();
  WriteOptions writeOpt = WriteOptions();
  writeOpt.disableWAL = true;
  ASSERT_OK(dbfull()->Put(writeOpt, "foo", "v1"));
H
heyongqiang 已提交
858
  ASSERT_OK(dbfull()->Put(writeOpt, "bar", "v1"));
H
heyongqiang 已提交
859 860 861

  Reopen();
  ASSERT_EQ("NOT_FOUND", Get("foo"));
H
heyongqiang 已提交
862
  ASSERT_EQ("NOT_FOUND", Get("bar"));
H
heyongqiang 已提交
863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885

  writeOpt.disableWAL = false;
  ASSERT_OK(dbfull()->Put(writeOpt, "bar", "v2"));
  writeOpt.disableWAL = true;
  ASSERT_OK(dbfull()->Put(writeOpt, "foo", "v2"));

  Reopen();
  // We garantee the 'bar' will be there
  // because its put has WAL enabled.
  // But 'foo' may or may not be there.
  ASSERT_EQ("v2", Get("bar"));

  writeOpt.disableWAL = true;
  ASSERT_OK(dbfull()->Put(writeOpt, "bar", "v3"));
  writeOpt.disableWAL = false;
  ASSERT_OK(dbfull()->Put(writeOpt, "foo", "v3"));

  Reopen();
  // 'foo' should be there because its put
  // has WAL enabled.
  ASSERT_EQ("v3", Get("foo"));
}

886 887 888 889 890 891 892
TEST(DBTest, CheckLock) {
  DB* localdb;
  Options options = CurrentOptions();
  ASSERT_TRUE(TryReopen(&options).ok());
  ASSERT_TRUE(!(PureReopen(&options, &localdb).ok())); // second open should fail
}

H
heyongqiang 已提交
893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926
TEST(DBTest, FLUSH) {
  Options options = CurrentOptions();
  WriteOptions writeOpt = WriteOptions();
  writeOpt.disableWAL = true;
  ASSERT_OK(dbfull()->Put(writeOpt, "foo", "v1"));
  // this will not flush the last 2 writes
  dbfull()->Flush(FlushOptions());
  ASSERT_OK(dbfull()->Put(writeOpt, "bar", "v1"));

  Reopen();
  ASSERT_EQ("v1", Get("foo"));
  ASSERT_EQ("NOT_FOUND", Get("bar"));

  writeOpt.disableWAL = true;
  ASSERT_OK(dbfull()->Put(writeOpt, "bar", "v2"));
  ASSERT_OK(dbfull()->Put(writeOpt, "foo", "v2"));
  dbfull()->Flush(FlushOptions());

  Reopen();
  ASSERT_EQ("v2", Get("bar"));
  ASSERT_EQ("v2", Get("foo"));

  writeOpt.disableWAL = false;
  ASSERT_OK(dbfull()->Put(writeOpt, "bar", "v3"));
  ASSERT_OK(dbfull()->Put(writeOpt, "foo", "v3"));
  dbfull()->Flush(FlushOptions());

  Reopen();
  // 'foo' should be there because its put
  // has WAL enabled.
  ASSERT_EQ("v3", Get("foo"));
  ASSERT_EQ("v3", Get("bar"));
}

J
jorlow@chromium.org 已提交
927
TEST(DBTest, RecoveryWithEmptyLog) {
S
Sanjay Ghemawat 已提交
928 929 930 931 932 933 934 935 936
  do {
    ASSERT_OK(Put("foo", "v1"));
    ASSERT_OK(Put("foo", "v2"));
    Reopen();
    Reopen();
    ASSERT_OK(Put("foo", "v3"));
    Reopen();
    ASSERT_EQ("v3", Get("foo"));
  } while (ChangeOptions());
J
jorlow@chromium.org 已提交
937 938
}

939 940 941
// Check that writes done during a memtable compaction are recovered
// if the database is shutdown during the memtable compaction.
TEST(DBTest, RecoverDuringMemtableCompaction) {
S
Sanjay Ghemawat 已提交
942 943 944 945 946
  do {
    Options options = CurrentOptions();
    options.env = env_;
    options.write_buffer_size = 1000000;
    Reopen(&options);
947

S
Sanjay Ghemawat 已提交
948 949 950 951 952
    // Trigger a long memtable compaction and reopen the database during it
    ASSERT_OK(Put("foo", "v1"));                         // Goes to 1st log file
    ASSERT_OK(Put("big1", std::string(10000000, 'x')));  // Fills memtable
    ASSERT_OK(Put("big2", std::string(1000, 'y')));      // Triggers compaction
    ASSERT_OK(Put("bar", "v2"));                         // Goes to new log file
953

S
Sanjay Ghemawat 已提交
954 955 956 957 958 959
    Reopen(&options);
    ASSERT_EQ("v1", Get("foo"));
    ASSERT_EQ("v2", Get("bar"));
    ASSERT_EQ(std::string(10000000, 'x'), Get("big1"));
    ASSERT_EQ(std::string(1000, 'y'), Get("big2"));
  } while (ChangeOptions());
960 961
}

J
jorlow@chromium.org 已提交
962 963 964 965 966 967 968
static std::string Key(int i) {
  char buf[100];
  snprintf(buf, sizeof(buf), "key%06d", i);
  return std::string(buf);
}

TEST(DBTest, MinorCompactionsHappen) {
S
Sanjay Ghemawat 已提交
969
  Options options = CurrentOptions();
J
jorlow@chromium.org 已提交
970 971 972
  options.write_buffer_size = 10000;
  Reopen(&options);

973
  const int N = 500;
J
jorlow@chromium.org 已提交
974

975
  int starting_num_tables = TotalTableFiles();
J
jorlow@chromium.org 已提交
976 977 978
  for (int i = 0; i < N; i++) {
    ASSERT_OK(Put(Key(i), Key(i) + std::string(1000, 'v')));
  }
979
  int ending_num_tables = TotalTableFiles();
J
jorlow@chromium.org 已提交
980 981 982 983 984 985 986 987 988 989 990 991 992 993 994
  ASSERT_GT(ending_num_tables, starting_num_tables);

  for (int i = 0; i < N; i++) {
    ASSERT_EQ(Key(i) + std::string(1000, 'v'), Get(Key(i)));
  }

  Reopen();

  for (int i = 0; i < N; i++) {
    ASSERT_EQ(Key(i) + std::string(1000, 'v'), Get(Key(i)));
  }
}

TEST(DBTest, RecoverWithLargeLog) {
  {
S
Sanjay Ghemawat 已提交
995
    Options options = CurrentOptions();
J
jorlow@chromium.org 已提交
996 997 998 999 1000 1001 1002 1003 1004 1005
    Reopen(&options);
    ASSERT_OK(Put("big1", std::string(200000, '1')));
    ASSERT_OK(Put("big2", std::string(200000, '2')));
    ASSERT_OK(Put("small3", std::string(10, '3')));
    ASSERT_OK(Put("small4", std::string(10, '4')));
    ASSERT_EQ(NumTableFilesAtLevel(0), 0);
  }

  // Make sure that if we re-open with a small write buffer size that
  // we flush table files in the middle of a large log file.
S
Sanjay Ghemawat 已提交
1006
  Options options = CurrentOptions();
J
jorlow@chromium.org 已提交
1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017
  options.write_buffer_size = 100000;
  Reopen(&options);
  ASSERT_EQ(NumTableFilesAtLevel(0), 3);
  ASSERT_EQ(std::string(200000, '1'), Get("big1"));
  ASSERT_EQ(std::string(200000, '2'), Get("big2"));
  ASSERT_EQ(std::string(10, '3'), Get("small3"));
  ASSERT_EQ(std::string(10, '4'), Get("small4"));
  ASSERT_GT(NumTableFilesAtLevel(0), 1);
}

TEST(DBTest, CompactionsGenerateMultipleFiles) {
S
Sanjay Ghemawat 已提交
1018
  Options options = CurrentOptions();
J
jorlow@chromium.org 已提交
1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033
  options.write_buffer_size = 100000000;        // Large write buffer
  Reopen(&options);

  Random rnd(301);

  // Write 8MB (80 values, each 100K)
  ASSERT_EQ(NumTableFilesAtLevel(0), 0);
  std::vector<std::string> values;
  for (int i = 0; i < 80; i++) {
    values.push_back(RandomString(&rnd, 100000));
    ASSERT_OK(Put(Key(i), values[i]));
  }

  // Reopening moves updates to level-0
  Reopen(&options);
G
Gabor Cselle 已提交
1034
  dbfull()->TEST_CompactRange(0, NULL, NULL);
J
jorlow@chromium.org 已提交
1035 1036 1037 1038 1039 1040 1041 1042

  ASSERT_EQ(NumTableFilesAtLevel(0), 0);
  ASSERT_GT(NumTableFilesAtLevel(1), 1);
  for (int i = 0; i < 80; i++) {
    ASSERT_EQ(Get(Key(i)), values[i]);
  }
}

1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078
TEST(DBTest, CompactionTrigger) {
  Options options = CurrentOptions();
  options.write_buffer_size = 100<<10; //100KB
  options.num_levels = 3;
  options.max_mem_compaction_level = 0;
  options.level0_file_num_compaction_trigger = 3;
  Reopen(&options);

  Random rnd(301);

  for (int num = 0;
		num < options.level0_file_num_compaction_trigger - 1;
		num++)
  {
    std::vector<std::string> values;
    // Write 120KB (12 values, each 10K)
    for (int i = 0; i < 12; i++) {
      values.push_back(RandomString(&rnd, 10000));
      ASSERT_OK(Put(Key(i), values[i]));
    }
    dbfull()->TEST_WaitForCompactMemTable();
    ASSERT_EQ(NumTableFilesAtLevel(0), num + 1);
  }

  //generate one more file in level-0, and should trigger level-0 compaction
  std::vector<std::string> values;
  for (int i = 0; i < 12; i++) {
    values.push_back(RandomString(&rnd, 10000));
    ASSERT_OK(Put(Key(i), values[i]));
  }
  dbfull()->TEST_WaitForCompact();

  ASSERT_EQ(NumTableFilesAtLevel(0), 0);
  ASSERT_EQ(NumTableFilesAtLevel(1), 1);
}

1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152
void MinLevelHelper(DBTest* self, Options& options) {	
  Random rnd(301);	
	
  for (int num = 0;	
    num < options.level0_file_num_compaction_trigger - 1;	
    num++)	
  {	
    std::vector<std::string> values;	
    // Write 120KB (12 values, each 10K)	
    for (int i = 0; i < 12; i++) {	
      values.push_back(RandomString(&rnd, 10000));	
      ASSERT_OK(self->Put(Key(i), values[i]));	
    }	
    self->dbfull()->TEST_WaitForCompactMemTable();	
    ASSERT_EQ(self->NumTableFilesAtLevel(0), num + 1);	
  }	
	
  //generate one more file in level-0, and should trigger level-0 compaction	
  std::vector<std::string> values;	
  for (int i = 0; i < 12; i++) {	
    values.push_back(RandomString(&rnd, 10000));	
    ASSERT_OK(self->Put(Key(i), values[i]));	
  }	
  self->dbfull()->TEST_WaitForCompact();	
	
  ASSERT_EQ(self->NumTableFilesAtLevel(0), 0);	
  ASSERT_EQ(self->NumTableFilesAtLevel(1), 1);	
}	
	
TEST(DBTest, MinLevelToCompress) {	
  Options options = CurrentOptions();	
  options.write_buffer_size = 100<<10; //100KB	
  options.num_levels = 3;	
  options.max_mem_compaction_level = 0;	
  options.level0_file_num_compaction_trigger = 3;	
  options.create_if_missing = true;
  CompressionType type;
	
  if (SnappyCompressionSupported()) {	
    type = kSnappyCompression;	
    fprintf(stderr, "using snappy\n");	
  } else if (ZlibCompressionSupported()) {	
    type = kZlibCompression;	
    fprintf(stderr, "using zlib\n");	
  } else if (BZip2CompressionSupported()) {	
    type = kBZip2Compression;	
    fprintf(stderr, "using bzip2\n");	
  } else {	
    fprintf(stderr, "skipping test, compression disabled\n");	
    return;	
  }	
  options.compression_per_level = new CompressionType[options.num_levels];

  // do not compress L0
  for (int i = 0; i < 1; i++) {
    options.compression_per_level[i] = kNoCompression;
  }
  for (int i = 1; i < options.num_levels; i++) {
    options.compression_per_level[i] = type;
  }
  Reopen(&options);	
  MinLevelHelper(this, options);	
	
  // do not compress L0 and L1
  for (int i = 0; i < 2; i++) {
    options.compression_per_level[i] = kNoCompression;
  }
  for (int i = 2; i < options.num_levels; i++) {
    options.compression_per_level[i] = type;
  }
  DestroyAndReopen(&options);	
  MinLevelHelper(this, options);	
}	

1153
TEST(DBTest, RepeatedWritesToSameKey) {
S
Sanjay Ghemawat 已提交
1154
  Options options = CurrentOptions();
1155 1156 1157 1158 1159 1160
  options.env = env_;
  options.write_buffer_size = 100000;  // Small write buffer
  Reopen(&options);

  // We must have at most one file per level except for level-0,
  // which may have up to kL0_StopWritesTrigger files.
1161 1162
  const int kMaxFiles = dbfull()->NumberLevels() +
		dbfull()->Level0StopWriteTrigger();
1163 1164 1165 1166 1167 1168 1169 1170 1171 1172

  Random rnd(301);
  std::string value = RandomString(&rnd, 2 * options.write_buffer_size);
  for (int i = 0; i < 5 * kMaxFiles; i++) {
    Put("key", value);
    ASSERT_LE(TotalTableFiles(), kMaxFiles);
    fprintf(stderr, "after %d: %d files\n", int(i+1), TotalTableFiles());
  }
}

1173
TEST(DBTest, SparseMerge) {
S
Sanjay Ghemawat 已提交
1174
  Options options = CurrentOptions();
1175 1176 1177
  options.compression = kNoCompression;
  Reopen(&options);

1178 1179
  FillLevels("A", "Z");

1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194
  // Suppose there is:
  //    small amount of data with prefix A
  //    large amount of data with prefix B
  //    small amount of data with prefix C
  // and that recent updates have made small changes to all three prefixes.
  // Check that we do not do a compaction that merges all of B in one shot.
  const std::string value(1000, 'x');
  Put("A", "va");
  // Write approximately 100MB of "B" values
  for (int i = 0; i < 100000; i++) {
    char key[100];
    snprintf(key, sizeof(key), "B%010d", i);
    Put(key, value);
  }
  Put("C", "vc");
1195
  dbfull()->TEST_CompactMemTable();
G
Gabor Cselle 已提交
1196
  dbfull()->TEST_CompactRange(0, NULL, NULL);
1197 1198 1199 1200 1201 1202 1203 1204 1205 1206

  // Make sparse update
  Put("A",    "va2");
  Put("B100", "bvalue2");
  Put("C",    "vc2");
  dbfull()->TEST_CompactMemTable();

  // Compactions should not cause us to create a situation where
  // a file overlaps too much data at the next level.
  ASSERT_LE(dbfull()->TEST_MaxNextLevelOverlappingBytes(), 20*1048576);
G
Gabor Cselle 已提交
1207
  dbfull()->TEST_CompactRange(0, NULL, NULL);
1208
  ASSERT_LE(dbfull()->TEST_MaxNextLevelOverlappingBytes(), 20*1048576);
G
Gabor Cselle 已提交
1209
  dbfull()->TEST_CompactRange(1, NULL, NULL);
1210 1211 1212
  ASSERT_LE(dbfull()->TEST_MaxNextLevelOverlappingBytes(), 20*1048576);
}

J
jorlow@chromium.org 已提交
1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224
static bool Between(uint64_t val, uint64_t low, uint64_t high) {
  bool result = (val >= low) && (val <= high);
  if (!result) {
    fprintf(stderr, "Value %llu is not in range [%llu, %llu]\n",
            (unsigned long long)(val),
            (unsigned long long)(low),
            (unsigned long long)(high));
  }
  return result;
}

TEST(DBTest, ApproximateSizes) {
S
Sanjay Ghemawat 已提交
1225 1226 1227 1228 1229
  do {
    Options options = CurrentOptions();
    options.write_buffer_size = 100000000;        // Large write buffer
    options.compression = kNoCompression;
    DestroyAndReopen();
J
jorlow@chromium.org 已提交
1230

S
Sanjay Ghemawat 已提交
1231 1232 1233
    ASSERT_TRUE(Between(Size("", "xyz"), 0, 0));
    Reopen(&options);
    ASSERT_TRUE(Between(Size("", "xyz"), 0, 0));
J
jorlow@chromium.org 已提交
1234

S
Sanjay Ghemawat 已提交
1235 1236 1237 1238 1239 1240 1241 1242 1243
    // Write 8MB (80 values, each 100K)
    ASSERT_EQ(NumTableFilesAtLevel(0), 0);
    const int N = 80;
    static const int S1 = 100000;
    static const int S2 = 105000;  // Allow some expansion from metadata
    Random rnd(301);
    for (int i = 0; i < N; i++) {
      ASSERT_OK(Put(Key(i), RandomString(&rnd, S1)));
    }
J
jorlow@chromium.org 已提交
1244

S
Sanjay Ghemawat 已提交
1245 1246
    // 0 because GetApproximateSizes() does not account for memtable space
    ASSERT_TRUE(Between(Size("", Key(50)), 0, 0));
D
 
dgrogan@chromium.org 已提交
1247

S
Sanjay Ghemawat 已提交
1248 1249 1250
    // Check sizes across recovery by reopening a few times
    for (int run = 0; run < 3; run++) {
      Reopen(&options);
D
dgrogan@chromium.org 已提交
1251

S
Sanjay Ghemawat 已提交
1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265
      for (int compact_start = 0; compact_start < N; compact_start += 10) {
        for (int i = 0; i < N; i += 10) {
          ASSERT_TRUE(Between(Size("", Key(i)), S1*i, S2*i));
          ASSERT_TRUE(Between(Size("", Key(i)+".suffix"), S1*(i+1), S2*(i+1)));
          ASSERT_TRUE(Between(Size(Key(i), Key(i+10)), S1*10, S2*10));
        }
        ASSERT_TRUE(Between(Size("", Key(50)), S1*50, S2*50));
        ASSERT_TRUE(Between(Size("", Key(50)+".suffix"), S1*50, S2*50));

        std::string cstart_str = Key(compact_start);
        std::string cend_str = Key(compact_start + 9);
        Slice cstart = cstart_str;
        Slice cend = cend_str;
        dbfull()->TEST_CompactRange(0, &cstart, &cend);
J
jorlow@chromium.org 已提交
1266
      }
D
dgrogan@chromium.org 已提交
1267

S
Sanjay Ghemawat 已提交
1268 1269 1270 1271
      ASSERT_EQ(NumTableFilesAtLevel(0), 0);
      ASSERT_GT(NumTableFilesAtLevel(1), 0);
    }
  } while (ChangeOptions());
J
jorlow@chromium.org 已提交
1272 1273 1274
}

TEST(DBTest, ApproximateSizes_MixOfSmallAndLarge) {
S
Sanjay Ghemawat 已提交
1275 1276 1277 1278
  do {
    Options options = CurrentOptions();
    options.compression = kNoCompression;
    Reopen();
J
jorlow@chromium.org 已提交
1279

S
Sanjay Ghemawat 已提交
1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309
    Random rnd(301);
    std::string big1 = RandomString(&rnd, 100000);
    ASSERT_OK(Put(Key(0), RandomString(&rnd, 10000)));
    ASSERT_OK(Put(Key(1), RandomString(&rnd, 10000)));
    ASSERT_OK(Put(Key(2), big1));
    ASSERT_OK(Put(Key(3), RandomString(&rnd, 10000)));
    ASSERT_OK(Put(Key(4), big1));
    ASSERT_OK(Put(Key(5), RandomString(&rnd, 10000)));
    ASSERT_OK(Put(Key(6), RandomString(&rnd, 300000)));
    ASSERT_OK(Put(Key(7), RandomString(&rnd, 10000)));

    // Check sizes across recovery by reopening a few times
    for (int run = 0; run < 3; run++) {
      Reopen(&options);

      ASSERT_TRUE(Between(Size("", Key(0)), 0, 0));
      ASSERT_TRUE(Between(Size("", Key(1)), 10000, 11000));
      ASSERT_TRUE(Between(Size("", Key(2)), 20000, 21000));
      ASSERT_TRUE(Between(Size("", Key(3)), 120000, 121000));
      ASSERT_TRUE(Between(Size("", Key(4)), 130000, 131000));
      ASSERT_TRUE(Between(Size("", Key(5)), 230000, 231000));
      ASSERT_TRUE(Between(Size("", Key(6)), 240000, 241000));
      ASSERT_TRUE(Between(Size("", Key(7)), 540000, 541000));
      ASSERT_TRUE(Between(Size("", Key(8)), 550000, 560000));

      ASSERT_TRUE(Between(Size(Key(3), Key(5)), 110000, 111000));

      dbfull()->TEST_CompactRange(0, NULL, NULL);
    }
  } while (ChangeOptions());
J
jorlow@chromium.org 已提交
1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334
}

TEST(DBTest, IteratorPinsRef) {
  Put("foo", "hello");

  // Get iterator that will yield the current contents of the DB.
  Iterator* iter = db_->NewIterator(ReadOptions());

  // Write to force compactions
  Put("foo", "newvalue1");
  for (int i = 0; i < 100; i++) {
    ASSERT_OK(Put(Key(i), Key(i) + std::string(100000, 'v'))); // 100K values
  }
  Put("foo", "newvalue2");

  iter->SeekToFirst();
  ASSERT_TRUE(iter->Valid());
  ASSERT_EQ("foo", iter->key().ToString());
  ASSERT_EQ("hello", iter->value().ToString());
  iter->Next();
  ASSERT_TRUE(!iter->Valid());
  delete iter;
}

TEST(DBTest, Snapshot) {
S
Sanjay Ghemawat 已提交
1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352
  do {
    Put("foo", "v1");
    const Snapshot* s1 = db_->GetSnapshot();
    Put("foo", "v2");
    const Snapshot* s2 = db_->GetSnapshot();
    Put("foo", "v3");
    const Snapshot* s3 = db_->GetSnapshot();

    Put("foo", "v4");
    ASSERT_EQ("v1", Get("foo", s1));
    ASSERT_EQ("v2", Get("foo", s2));
    ASSERT_EQ("v3", Get("foo", s3));
    ASSERT_EQ("v4", Get("foo"));

    db_->ReleaseSnapshot(s3);
    ASSERT_EQ("v1", Get("foo", s1));
    ASSERT_EQ("v2", Get("foo", s2));
    ASSERT_EQ("v4", Get("foo"));
J
jorlow@chromium.org 已提交
1353

S
Sanjay Ghemawat 已提交
1354 1355 1356
    db_->ReleaseSnapshot(s1);
    ASSERT_EQ("v2", Get("foo", s2));
    ASSERT_EQ("v4", Get("foo"));
1357

S
Sanjay Ghemawat 已提交
1358 1359 1360 1361
    db_->ReleaseSnapshot(s2);
    ASSERT_EQ("v4", Get("foo"));
  } while (ChangeOptions());
}
J
jorlow@chromium.org 已提交
1362

S
Sanjay Ghemawat 已提交
1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388
TEST(DBTest, HiddenValuesAreRemoved) {
  do {
    Random rnd(301);
    FillLevels("a", "z");

    std::string big = RandomString(&rnd, 50000);
    Put("foo", big);
    Put("pastfoo", "v");
    const Snapshot* snapshot = db_->GetSnapshot();
    Put("foo", "tiny");
    Put("pastfoo2", "v2");        // Advance sequence number one more

    ASSERT_OK(dbfull()->TEST_CompactMemTable());
    ASSERT_GT(NumTableFilesAtLevel(0), 0);

    ASSERT_EQ(big, Get("foo", snapshot));
    ASSERT_TRUE(Between(Size("", "pastfoo"), 50000, 60000));
    db_->ReleaseSnapshot(snapshot);
    ASSERT_EQ(AllEntriesFor("foo"), "[ tiny, " + big + " ]");
    Slice x("x");
    dbfull()->TEST_CompactRange(0, NULL, &x);
    ASSERT_EQ(AllEntriesFor("foo"), "[ tiny ]");
    ASSERT_EQ(NumTableFilesAtLevel(0), 0);
    ASSERT_GE(NumTableFilesAtLevel(1), 1);
    dbfull()->TEST_CompactRange(1, NULL, &x);
    ASSERT_EQ(AllEntriesFor("foo"), "[ tiny ]");
J
jorlow@chromium.org 已提交
1389

S
Sanjay Ghemawat 已提交
1390 1391
    ASSERT_TRUE(Between(Size("", "pastfoo"), 0, 1000));
  } while (ChangeOptions());
J
jorlow@chromium.org 已提交
1392 1393 1394 1395 1396
}

TEST(DBTest, DeletionMarkers1) {
  Put("foo", "v1");
  ASSERT_OK(dbfull()->TEST_CompactMemTable());
1397
  const int last = dbfull()->MaxMemCompactionLevel();
1398 1399 1400 1401 1402 1403 1404 1405 1406
  ASSERT_EQ(NumTableFilesAtLevel(last), 1);   // foo => v1 is now in last level

  // Place a table at level last-1 to prevent merging with preceding mutation
  Put("a", "begin");
  Put("z", "end");
  dbfull()->TEST_CompactMemTable();
  ASSERT_EQ(NumTableFilesAtLevel(last), 1);
  ASSERT_EQ(NumTableFilesAtLevel(last-1), 1);

J
jorlow@chromium.org 已提交
1407 1408 1409
  Delete("foo");
  Put("foo", "v2");
  ASSERT_EQ(AllEntriesFor("foo"), "[ v2, DEL, v1 ]");
1410
  ASSERT_OK(dbfull()->TEST_CompactMemTable());  // Moves to level last-2
J
jorlow@chromium.org 已提交
1411
  ASSERT_EQ(AllEntriesFor("foo"), "[ v2, DEL, v1 ]");
G
Gabor Cselle 已提交
1412 1413
  Slice z("z");
  dbfull()->TEST_CompactRange(last-2, NULL, &z);
J
jorlow@chromium.org 已提交
1414 1415 1416
  // DEL eliminated, but v1 remains because we aren't compacting that level
  // (DEL can be eliminated because v2 hides v1).
  ASSERT_EQ(AllEntriesFor("foo"), "[ v2, v1 ]");
G
Gabor Cselle 已提交
1417
  dbfull()->TEST_CompactRange(last-1, NULL, NULL);
1418 1419
  // Merging last-1 w/ last, so we are the base level for "foo", so
  // DEL is removed.  (as is v1).
J
jorlow@chromium.org 已提交
1420 1421 1422 1423 1424 1425
  ASSERT_EQ(AllEntriesFor("foo"), "[ v2 ]");
}

TEST(DBTest, DeletionMarkers2) {
  Put("foo", "v1");
  ASSERT_OK(dbfull()->TEST_CompactMemTable());
1426
  const int last = dbfull()->MaxMemCompactionLevel();
1427 1428 1429 1430 1431 1432 1433 1434 1435
  ASSERT_EQ(NumTableFilesAtLevel(last), 1);   // foo => v1 is now in last level

  // Place a table at level last-1 to prevent merging with preceding mutation
  Put("a", "begin");
  Put("z", "end");
  dbfull()->TEST_CompactMemTable();
  ASSERT_EQ(NumTableFilesAtLevel(last), 1);
  ASSERT_EQ(NumTableFilesAtLevel(last-1), 1);

J
jorlow@chromium.org 已提交
1436 1437
  Delete("foo");
  ASSERT_EQ(AllEntriesFor("foo"), "[ DEL, v1 ]");
1438
  ASSERT_OK(dbfull()->TEST_CompactMemTable());  // Moves to level last-2
J
jorlow@chromium.org 已提交
1439
  ASSERT_EQ(AllEntriesFor("foo"), "[ DEL, v1 ]");
G
Gabor Cselle 已提交
1440
  dbfull()->TEST_CompactRange(last-2, NULL, NULL);
1441
  // DEL kept: "last" file overlaps
J
jorlow@chromium.org 已提交
1442
  ASSERT_EQ(AllEntriesFor("foo"), "[ DEL, v1 ]");
G
Gabor Cselle 已提交
1443
  dbfull()->TEST_CompactRange(last-1, NULL, NULL);
1444 1445
  // Merging last-1 w/ last, so we are the base level for "foo", so
  // DEL is removed.  (as is v1).
J
jorlow@chromium.org 已提交
1446 1447 1448
  ASSERT_EQ(AllEntriesFor("foo"), "[ ]");
}

G
Gabor Cselle 已提交
1449
TEST(DBTest, OverlapInLevel0) {
S
Sanjay Ghemawat 已提交
1450
  do {
1451 1452
	int tmp = dbfull()->MaxMemCompactionLevel();
    ASSERT_EQ(tmp, 2) << "Fix test to match config";
G
Gabor Cselle 已提交
1453

S
Sanjay Ghemawat 已提交
1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474
    // Fill levels 1 and 2 to disable the pushing of new memtables to levels > 0.
    ASSERT_OK(Put("100", "v100"));
    ASSERT_OK(Put("999", "v999"));
    dbfull()->TEST_CompactMemTable();
    ASSERT_OK(Delete("100"));
    ASSERT_OK(Delete("999"));
    dbfull()->TEST_CompactMemTable();
    ASSERT_EQ("0,1,1", FilesPerLevel());

    // Make files spanning the following ranges in level-0:
    //  files[0]  200 .. 900
    //  files[1]  300 .. 500
    // Note that files are sorted by smallest key.
    ASSERT_OK(Put("300", "v300"));
    ASSERT_OK(Put("500", "v500"));
    dbfull()->TEST_CompactMemTable();
    ASSERT_OK(Put("200", "v200"));
    ASSERT_OK(Put("600", "v600"));
    ASSERT_OK(Put("900", "v900"));
    dbfull()->TEST_CompactMemTable();
    ASSERT_EQ("2,1,1", FilesPerLevel());
G
Gabor Cselle 已提交
1475

S
Sanjay Ghemawat 已提交
1476 1477 1478 1479
    // Compact away the placeholder files we created initially
    dbfull()->TEST_CompactRange(1, NULL, NULL);
    dbfull()->TEST_CompactRange(2, NULL, NULL);
    ASSERT_EQ("2", FilesPerLevel());
G
Gabor Cselle 已提交
1480

S
Sanjay Ghemawat 已提交
1481 1482 1483 1484 1485 1486 1487 1488
    // Do a memtable compaction.  Before bug-fix, the compaction would
    // not detect the overlap with level-0 files and would incorrectly place
    // the deletion in a deeper level.
    ASSERT_OK(Delete("600"));
    dbfull()->TEST_CompactMemTable();
    ASSERT_EQ("3", FilesPerLevel());
    ASSERT_EQ("NOT_FOUND", Get("600"));
  } while (ChangeOptions());
G
Gabor Cselle 已提交
1489 1490
}

H
Hans Wennborg 已提交
1491
TEST(DBTest, L0_CompactionBug_Issue44_a) {
S
Sanjay Ghemawat 已提交
1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505
  Reopen();
  ASSERT_OK(Put("b", "v"));
  Reopen();
  ASSERT_OK(Delete("b"));
  ASSERT_OK(Delete("a"));
  Reopen();
  ASSERT_OK(Delete("a"));
  Reopen();
  ASSERT_OK(Put("a", "v"));
  Reopen();
  Reopen();
  ASSERT_EQ("(a->v)", Contents());
  env_->SleepForMicroseconds(1000000);  // Wait for compaction to finish
  ASSERT_EQ("(a->v)", Contents());
H
Hans Wennborg 已提交
1506 1507 1508
}

TEST(DBTest, L0_CompactionBug_Issue44_b) {
S
Sanjay Ghemawat 已提交
1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531
  Reopen();
  Put("","");
  Reopen();
  Delete("e");
  Put("","");
  Reopen();
  Put("c", "cv");
  Reopen();
  Put("","");
  Reopen();
  Put("","");
  env_->SleepForMicroseconds(1000000);  // Wait for compaction to finish
  Reopen();
  Put("d","dv");
  Reopen();
  Put("","");
  Reopen();
  Delete("d");
  Delete("b");
  Reopen();
  ASSERT_EQ("(->)(c->cv)", Contents());
  env_->SleepForMicroseconds(1000000);  // Wait for compaction to finish
  ASSERT_EQ("(->)(c->cv)", Contents());
H
Hans Wennborg 已提交
1532 1533
}

J
jorlow@chromium.org 已提交
1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548
TEST(DBTest, ComparatorCheck) {
  class NewComparator : public Comparator {
   public:
    virtual const char* Name() const { return "leveldb.NewComparator"; }
    virtual int Compare(const Slice& a, const Slice& b) const {
      return BytewiseComparator()->Compare(a, b);
    }
    virtual void FindShortestSeparator(std::string* s, const Slice& l) const {
      BytewiseComparator()->FindShortestSeparator(s, l);
    }
    virtual void FindShortSuccessor(std::string* key) const {
      BytewiseComparator()->FindShortSuccessor(key);
    }
  };
  NewComparator cmp;
S
Sanjay Ghemawat 已提交
1549
  Options new_options = CurrentOptions();
J
jorlow@chromium.org 已提交
1550 1551 1552 1553 1554 1555 1556
  new_options.comparator = &cmp;
  Status s = TryReopen(&new_options);
  ASSERT_TRUE(!s.ok());
  ASSERT_TRUE(s.ToString().find("comparator") != std::string::npos)
      << s.ToString();
}

H
Hans Wennborg 已提交
1557 1558 1559 1560 1561
TEST(DBTest, CustomComparator) {
  class NumberComparator : public Comparator {
   public:
    virtual const char* Name() const { return "test.NumberComparator"; }
    virtual int Compare(const Slice& a, const Slice& b) const {
1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580
      return ToNumber(a) - ToNumber(b);
    }
    virtual void FindShortestSeparator(std::string* s, const Slice& l) const {
      ToNumber(*s);     // Check format
      ToNumber(l);      // Check format
    }
    virtual void FindShortSuccessor(std::string* key) const {
      ToNumber(*key);   // Check format
    }
   private:
    static int ToNumber(const Slice& x) {
      // Check that there are no extra characters.
      ASSERT_TRUE(x.size() >= 2 && x[0] == '[' && x[x.size()-1] == ']')
          << EscapeString(x);
      int val;
      char ignored;
      ASSERT_TRUE(sscanf(x.ToString().c_str(), "[%i]%c", &val, &ignored) == 1)
          << EscapeString(x);
      return val;
H
Hans Wennborg 已提交
1581 1582 1583
    }
  };
  NumberComparator cmp;
S
Sanjay Ghemawat 已提交
1584
  Options new_options = CurrentOptions();
H
Hans Wennborg 已提交
1585 1586
  new_options.create_if_missing = true;
  new_options.comparator = &cmp;
S
Sanjay Ghemawat 已提交
1587
  new_options.filter_policy = NULL;     // Cannot use bloom filters
1588
  new_options.write_buffer_size = 1000;  // Compact more often
H
Hans Wennborg 已提交
1589
  DestroyAndReopen(&new_options);
1590 1591
  ASSERT_OK(Put("[10]", "ten"));
  ASSERT_OK(Put("[0x14]", "twenty"));
H
Hans Wennborg 已提交
1592
  for (int i = 0; i < 2; i++) {
1593 1594 1595 1596
    ASSERT_EQ("ten", Get("[10]"));
    ASSERT_EQ("ten", Get("[0xa]"));
    ASSERT_EQ("twenty", Get("[20]"));
    ASSERT_EQ("twenty", Get("[0x14]"));
S
Sanjay Ghemawat 已提交
1597 1598
    ASSERT_EQ("NOT_FOUND", Get("[15]"));
    ASSERT_EQ("NOT_FOUND", Get("[0xf]"));
1599 1600 1601 1602 1603 1604 1605 1606 1607 1608
    Compact("[0]", "[9999]");
  }

  for (int run = 0; run < 2; run++) {
    for (int i = 0; i < 1000; i++) {
      char buf[100];
      snprintf(buf, sizeof(buf), "[%d]", i*10);
      ASSERT_OK(Put(buf, buf));
    }
    Compact("[0]", "[1000000]");
H
Hans Wennborg 已提交
1609 1610 1611
  }
}

G
Gabor Cselle 已提交
1612
TEST(DBTest, ManualCompaction) {
1613
  ASSERT_EQ(dbfull()->MaxMemCompactionLevel(), 2)
G
Gabor Cselle 已提交
1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645
      << "Need to update this test to match kMaxMemCompactLevel";

  MakeTables(3, "p", "q");
  ASSERT_EQ("1,1,1", FilesPerLevel());

  // Compaction range falls before files
  Compact("", "c");
  ASSERT_EQ("1,1,1", FilesPerLevel());

  // Compaction range falls after files
  Compact("r", "z");
  ASSERT_EQ("1,1,1", FilesPerLevel());

  // Compaction range overlaps files
  Compact("p1", "p9");
  ASSERT_EQ("0,0,1", FilesPerLevel());

  // Populate a different range
  MakeTables(3, "c", "e");
  ASSERT_EQ("1,1,2", FilesPerLevel());

  // Compact just the new range
  Compact("b", "f");
  ASSERT_EQ("0,0,2", FilesPerLevel());

  // Compact all
  MakeTables(1, "a", "z");
  ASSERT_EQ("0,1,2", FilesPerLevel());
  db_->CompactRange(NULL, NULL);
  ASSERT_EQ("0,0,1", FilesPerLevel());
}

J
jorlow@chromium.org 已提交
1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684
TEST(DBTest, DBOpen_Options) {
  std::string dbname = test::TmpDir() + "/db_options_test";
  DestroyDB(dbname, Options());

  // Does not exist, and create_if_missing == false: error
  DB* db = NULL;
  Options opts;
  opts.create_if_missing = false;
  Status s = DB::Open(opts, dbname, &db);
  ASSERT_TRUE(strstr(s.ToString().c_str(), "does not exist") != NULL);
  ASSERT_TRUE(db == NULL);

  // Does not exist, and create_if_missing == true: OK
  opts.create_if_missing = true;
  s = DB::Open(opts, dbname, &db);
  ASSERT_OK(s);
  ASSERT_TRUE(db != NULL);

  delete db;
  db = NULL;

  // Does exist, and error_if_exists == true: error
  opts.create_if_missing = false;
  opts.error_if_exists = true;
  s = DB::Open(opts, dbname, &db);
  ASSERT_TRUE(strstr(s.ToString().c_str(), "exists") != NULL);
  ASSERT_TRUE(db == NULL);

  // Does exist, and error_if_exists == false: OK
  opts.create_if_missing = true;
  opts.error_if_exists = false;
  s = DB::Open(opts, dbname, &db);
  ASSERT_OK(s);
  ASSERT_TRUE(db != NULL);

  delete db;
  db = NULL;
}

1685 1686 1687 1688 1689
TEST(DBTest, DBOpen_Change_NumLevels) {
  std::string dbname = test::TmpDir() + "/db_change_num_levels";
  DestroyDB(dbname, Options());
  Options opts;
  Status s;
H
heyongqiang 已提交
1690
  DB* db = NULL;
1691
  opts.create_if_missing = true;
H
heyongqiang 已提交
1692
  s = DB::Open(opts, dbname, &db);
1693
  ASSERT_OK(s);
H
heyongqiang 已提交
1694 1695 1696 1697 1698 1699
  ASSERT_TRUE(db != NULL);
  db->Put(WriteOptions(), "a", "123");
  db->Put(WriteOptions(), "b", "234");
  db->CompactRange(NULL, NULL);
  delete db;
  db = NULL;
1700 1701 1702

  opts.create_if_missing = false;
  opts.num_levels = 2;
H
heyongqiang 已提交
1703
  s = DB::Open(opts, dbname, &db);
1704
  ASSERT_TRUE(strstr(s.ToString().c_str(), "Corruption") != NULL);
H
heyongqiang 已提交
1705
  ASSERT_TRUE(db == NULL);
1706 1707
}

1708 1709
// Check that number of files does not grow when we are out of space
TEST(DBTest, NoSpace) {
S
Sanjay Ghemawat 已提交
1710
  Options options = CurrentOptions();
1711 1712 1713 1714 1715 1716 1717 1718
  options.env = env_;
  Reopen(&options);

  ASSERT_OK(Put("foo", "v1"));
  ASSERT_EQ("v1", Get("foo"));
  Compact("a", "z");
  const int num_files = CountFiles();
  env_->no_space_.Release_Store(env_);   // Force out-of-space errors
H
heyongqiang 已提交
1719
  env_->sleep_counter_.Reset();
1720
  for (int i = 0; i < 5; i++) {
1721
    for (int level = 0; level < dbfull()->NumberLevels()-1; level++) {
1722 1723 1724 1725
      dbfull()->TEST_CompactRange(level, NULL, NULL);
    }
  }
  env_->no_space_.Release_Store(NULL);
1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750
  ASSERT_LT(CountFiles(), num_files + 3);

  // Check that compaction attempts slept after errors
  ASSERT_GE(env_->sleep_counter_.Read(), 5);
}

TEST(DBTest, NonWritableFileSystem)
{
  Options options = CurrentOptions();
  options.write_buffer_size = 1000;
  options.env = env_;
  Reopen(&options);
  ASSERT_OK(Put("foo", "v1"));
  env_->non_writable_.Release_Store(env_); // Force errors for new files
  std::string big(100000, 'x');
  int errors = 0;
  for (int i = 0; i < 20; i++) {
    fprintf(stderr, "iter %d; errors %d\n", i, errors);
    if (!Put("foo", big).ok()) {
      errors++;
      env_->SleepForMicroseconds(100000);
    }
  }
  ASSERT_GT(errors, 0);
  env_->non_writable_.Release_Store(NULL);
1751 1752 1753 1754 1755
}

TEST(DBTest, FilesDeletedAfterCompaction) {
  ASSERT_OK(Put("foo", "v2"));
  Compact("a", "z");
1756
  const int num_files = CountLiveFiles();
1757 1758 1759 1760
  for (int i = 0; i < 10; i++) {
    ASSERT_OK(Put("foo", "v2"));
    Compact("a", "z");
  }
1761
  ASSERT_EQ(CountLiveFiles(), num_files);
1762 1763
}

S
Sanjay Ghemawat 已提交
1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810
TEST(DBTest, BloomFilter) {
  env_->count_random_reads_ = true;
  Options options = CurrentOptions();
  options.env = env_;
  options.block_cache = NewLRUCache(0);  // Prevent cache hits
  options.filter_policy = NewBloomFilterPolicy(10);
  Reopen(&options);

  // Populate multiple layers
  const int N = 10000;
  for (int i = 0; i < N; i++) {
    ASSERT_OK(Put(Key(i), Key(i)));
  }
  Compact("a", "z");
  for (int i = 0; i < N; i += 100) {
    ASSERT_OK(Put(Key(i), Key(i)));
  }
  dbfull()->TEST_CompactMemTable();

  // Prevent auto compactions triggered by seeks
  env_->delay_sstable_sync_.Release_Store(env_);

  // Lookup present keys.  Should rarely read from small sstable.
  env_->random_read_counter_.Reset();
  for (int i = 0; i < N; i++) {
    ASSERT_EQ(Key(i), Get(Key(i)));
  }
  int reads = env_->random_read_counter_.Read();
  fprintf(stderr, "%d present => %d reads\n", N, reads);
  ASSERT_GE(reads, N);
  ASSERT_LE(reads, N + 2*N/100);

  // Lookup present keys.  Should rarely read from either sstable.
  env_->random_read_counter_.Reset();
  for (int i = 0; i < N; i++) {
    ASSERT_EQ("NOT_FOUND", Get(Key(i) + ".missing"));
  }
  reads = env_->random_read_counter_.Read();
  fprintf(stderr, "%d missing => %d reads\n", N, reads);
  ASSERT_LE(reads, 3*N/100);

  env_->delay_sstable_sync_.Release_Store(NULL);
  Close();
  delete options.block_cache;
  delete options.filter_policy;
}

1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829
TEST(DBTest, SnapshotFiles) {
  Options options = CurrentOptions();
  options.write_buffer_size = 100000000;        // Large write buffer
  Reopen(&options);

  Random rnd(301);

  // Write 8MB (80 values, each 100K)
  ASSERT_EQ(NumTableFilesAtLevel(0), 0);
  std::vector<std::string> values;
  for (int i = 0; i < 80; i++) {
    values.push_back(RandomString(&rnd, 100000));
    ASSERT_OK(Put(Key(i), values[i]));
  }

  // assert that nothing makes it to disk yet.
  ASSERT_EQ(NumTableFilesAtLevel(0), 0);

  // get a file snapshot
1830 1831
  uint64_t manifest_number = 0;
  uint64_t manifest_size = 0;
1832 1833
  std::vector<std::string> files;
  dbfull()->DisableFileDeletions();
1834
  dbfull()->GetLiveFiles(files, &manifest_size);
1835 1836 1837 1838

  // CURRENT, MANIFEST, *.sst files
  ASSERT_EQ(files.size(), 3);

1839 1840 1841
  uint64_t number = 0;
  FileType type;

1842 1843 1844 1845
  // copy these files to a new snapshot directory
  std::string snapdir = dbname_ + ".snapdir/";
  std::string mkdir = "mkdir -p " + snapdir;
  ASSERT_EQ(system(mkdir.c_str()), 0);
1846 1847

  for (unsigned int i = 0; i < files.size(); i++) {
1848 1849
    std::string src = dbname_ + "/" + files[i];
    std::string dest = snapdir + "/" + files[i];
1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880

    uint64_t size;
    ASSERT_OK(env_->GetFileSize(src, &size));

    // record the number and the size of the 
    // latest manifest file
    if (ParseFileName(files[i].substr(1), &number, &type)) {
      if (type == kDescriptorFile) {
        if (number > manifest_number) {
          manifest_number = number;
          ASSERT_GE(size, manifest_size);
          size = manifest_size; // copy only valid MANIFEST data
        }
      }
    }
    SequentialFile* srcfile;
    ASSERT_OK(env_->NewSequentialFile(src, &srcfile));
    WritableFile* destfile;
    ASSERT_OK(env_->NewWritableFile(dest, &destfile));
    
    char buffer[4096];
    Slice slice;
    while (size > 0) {
      uint64_t one = std::min(sizeof(buffer), size);
      ASSERT_OK(srcfile->Read(one, &slice, buffer));
      ASSERT_OK(destfile->Append(slice));
      size -= slice.size();
    }
    ASSERT_OK(destfile->Close());
    delete destfile;
    delete srcfile;
1881
  }
1882

1883 1884 1885
  // release file snapshot
  dbfull()->DisableFileDeletions();

1886 1887 1888 1889 1890 1891 1892
  // overwrite one key, this key should not appear in the snapshot
  std::vector<std::string> extras;
  for (unsigned int i = 0; i < 1; i++) {
    extras.push_back(RandomString(&rnd, 100000));
    ASSERT_OK(Put(Key(i), extras[i]));
  }
  
1893 1894 1895 1896 1897 1898 1899 1900 1901
  // verify that data in the snapshot are correct
  Options opts;
  DB* snapdb;
  opts.create_if_missing = false;
  Status stat = DB::Open(opts, snapdir, &snapdb);
  ASSERT_TRUE(stat.ok());

  ReadOptions roptions;
  std::string val;
1902
  for (unsigned int i = 0; i < 80; i++) {
1903 1904 1905 1906
    stat = snapdb->Get(roptions, Key(i), &val);
    ASSERT_EQ(values[i].compare(val), 0);
  }
  delete snapdb;
1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939

  // look at the new live files after we added an 'extra' key 
  // and after we took the first snapshot.
  uint64_t new_manifest_number = 0;
  uint64_t new_manifest_size = 0;
  std::vector<std::string> newfiles;
  dbfull()->DisableFileDeletions();
  dbfull()->GetLiveFiles(newfiles, &new_manifest_size);

  // find the new manifest file. assert that this manifest file is
  // the same one as in the previous snapshot. But its size should be
  // larger because we added an extra key after taking the
  // previous shapshot.
  for (unsigned int i = 0; i < newfiles.size(); i++) {
    std::string src = dbname_ + "/" + newfiles[i];
    // record the lognumber and the size of the 
    // latest manifest file
    if (ParseFileName(newfiles[i].substr(1), &number, &type)) {
      if (type == kDescriptorFile) {
        if (number > new_manifest_number) {
          uint64_t size;
          new_manifest_number = number;
          ASSERT_OK(env_->GetFileSize(src, &size));
          ASSERT_GE(size, new_manifest_size);
        }
      }
    }
  }
  ASSERT_EQ(manifest_number, new_manifest_number);
  ASSERT_GT(new_manifest_size, manifest_size);
  
  // release file snapshot
  dbfull()->DisableFileDeletions();
1940 1941
}

1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005

TEST(DBTest, ReadCompaction) {
  std::string value(4096, '4'); // a string of size 4K
  {
    Options options = CurrentOptions();
    options.create_if_missing = true;
    options.max_open_files = 20; // only 10 file in file-cache
    options.target_file_size_base = 512;
    options.write_buffer_size = 64 * 1024;
    options.filter_policy = NULL;
    options.block_size = 4096;
    options.block_cache = NewLRUCache(0);  // Prevent cache hits

    Reopen(&options);

    // Write 8MB (2000 values, each 4K)
    ASSERT_EQ(NumTableFilesAtLevel(0), 0);
    std::vector<std::string> values;
    for (int i = 0; i < 2000; i++) {
      ASSERT_OK(Put(Key(i), value));
    }

    // clear level 0 and 1 if necessary.
    dbfull()->TEST_CompactMemTable();
    dbfull()->TEST_CompactRange(0, NULL, NULL);
    dbfull()->TEST_CompactRange(1, NULL, NULL);
    ASSERT_EQ(NumTableFilesAtLevel(0), 0);
    ASSERT_EQ(NumTableFilesAtLevel(1), 0);

    // write some new keys into level 0
    for (int i = 0; i < 2000; i = i + 16) {
      ASSERT_OK(Put(Key(i), value));
    }
    dbfull()->Flush(FlushOptions());

    // Wait for any write compaction to finish
    dbfull()->TEST_WaitForCompact();

    // remember number of files in each level
    int l1 = NumTableFilesAtLevel(0);
    int l2 = NumTableFilesAtLevel(1);
    int l3 = NumTableFilesAtLevel(3);
    ASSERT_NE(NumTableFilesAtLevel(0), 0);
    ASSERT_NE(NumTableFilesAtLevel(1), 0);
    ASSERT_NE(NumTableFilesAtLevel(2), 0);

    // read a bunch of times, trigger read compaction
    for (int j = 0; j < 100; j++) {
      for (int i = 0; i < 2000; i++) {
        Get(Key(i));
      }
    }
    // wait for read compaction to finish
    env_->SleepForMicroseconds(1000000);

    // verify that the number of files have decreased
    // in some level, indicating that there was a compaction
    ASSERT_TRUE(NumTableFilesAtLevel(0) < l1 ||
                NumTableFilesAtLevel(1) < l2 ||
                NumTableFilesAtLevel(2) < l3); 
    delete options.block_cache;
  }
}

2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026
// Multi-threaded test:
namespace {

static const int kNumThreads = 4;
static const int kTestSeconds = 10;
static const int kNumKeys = 1000;

struct MTState {
  DBTest* test;
  port::AtomicPointer stop;
  port::AtomicPointer counter[kNumThreads];
  port::AtomicPointer thread_done[kNumThreads];
};

struct MTThread {
  MTState* state;
  int id;
};

static void MTThreadBody(void* arg) {
  MTThread* t = reinterpret_cast<MTThread*>(arg);
2027
  int id = t->id;
2028 2029
  DB* db = t->state->test->db_;
  uintptr_t counter = 0;
2030 2031
  fprintf(stderr, "... starting thread %d\n", id);
  Random rnd(1000 + id);
2032 2033 2034
  std::string value;
  char valbuf[1500];
  while (t->state->stop.Acquire_Load() == NULL) {
2035
    t->state->counter[id].Release_Store(reinterpret_cast<void*>(counter));
2036 2037 2038 2039 2040 2041 2042 2043 2044

    int key = rnd.Uniform(kNumKeys);
    char keybuf[20];
    snprintf(keybuf, sizeof(keybuf), "%016d", key);

    if (rnd.OneIn(2)) {
      // Write values of the form <key, my id, counter>.
      // We add some padding for force compactions.
      snprintf(valbuf, sizeof(valbuf), "%d.%d.%-1000d",
2045
               key, id, static_cast<int>(counter));
2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065
      ASSERT_OK(db->Put(WriteOptions(), Slice(keybuf), Slice(valbuf)));
    } else {
      // Read a value and verify that it matches the pattern written above.
      Status s = db->Get(ReadOptions(), Slice(keybuf), &value);
      if (s.IsNotFound()) {
        // Key has not yet been written
      } else {
        // Check that the writer thread counter is >= the counter in the value
        ASSERT_OK(s);
        int k, w, c;
        ASSERT_EQ(3, sscanf(value.c_str(), "%d.%d.%d", &k, &w, &c)) << value;
        ASSERT_EQ(k, key);
        ASSERT_GE(w, 0);
        ASSERT_LT(w, kNumThreads);
        ASSERT_LE(c, reinterpret_cast<uintptr_t>(
            t->state->counter[w].Acquire_Load()));
      }
    }
    counter++;
  }
2066 2067
  t->state->thread_done[id].Release_Store(t);
  fprintf(stderr, "... stopping thread %d after %d ops\n", id, int(counter));
2068 2069
}

H
Hans Wennborg 已提交
2070
}  // namespace
2071 2072

TEST(DBTest, MultiThreaded) {
S
Sanjay Ghemawat 已提交
2073 2074 2075 2076 2077 2078 2079 2080 2081
  do {
    // Initialize state
    MTState mt;
    mt.test = this;
    mt.stop.Release_Store(0);
    for (int id = 0; id < kNumThreads; id++) {
      mt.counter[id].Release_Store(0);
      mt.thread_done[id].Release_Store(0);
    }
2082

S
Sanjay Ghemawat 已提交
2083 2084 2085 2086 2087 2088 2089
    // Start threads
    MTThread thread[kNumThreads];
    for (int id = 0; id < kNumThreads; id++) {
      thread[id].state = &mt;
      thread[id].id = id;
      env_->StartThread(MTThreadBody, &thread[id]);
    }
2090

S
Sanjay Ghemawat 已提交
2091 2092
    // Let them run for a while
    env_->SleepForMicroseconds(kTestSeconds * 1000000);
2093

S
Sanjay Ghemawat 已提交
2094 2095 2096 2097 2098 2099
    // Stop the threads and wait for them to finish
    mt.stop.Release_Store(&mt);
    for (int id = 0; id < kNumThreads; id++) {
      while (mt.thread_done[id].Acquire_Load() == NULL) {
        env_->SleepForMicroseconds(100000);
      }
2100
    }
S
Sanjay Ghemawat 已提交
2101
  } while (ChangeOptions());
2102 2103
}

2104 2105 2106 2107
namespace {
typedef std::map<std::string, std::string> KVMap;
}

J
jorlow@chromium.org 已提交
2108 2109
class ModelDB: public DB {
 public:
2110 2111 2112 2113 2114
  class ModelSnapshot : public Snapshot {
   public:
    KVMap map_;
  };

J
jorlow@chromium.org 已提交
2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134
  explicit ModelDB(const Options& options): options_(options) { }
  ~ModelDB() { }
  virtual Status Put(const WriteOptions& o, const Slice& k, const Slice& v) {
    return DB::Put(o, k, v);
  }
  virtual Status Delete(const WriteOptions& o, const Slice& key) {
    return DB::Delete(o, key);
  }
  virtual Status Get(const ReadOptions& options,
                     const Slice& key, std::string* value) {
    assert(false);      // Not implemented
    return Status::NotFound(key);
  }
  virtual Iterator* NewIterator(const ReadOptions& options) {
    if (options.snapshot == NULL) {
      KVMap* saved = new KVMap;
      *saved = map_;
      return new ModelIter(saved, true);
    } else {
      const KVMap* snapshot_state =
2135
          &(reinterpret_cast<const ModelSnapshot*>(options.snapshot)->map_);
J
jorlow@chromium.org 已提交
2136 2137 2138 2139
      return new ModelIter(snapshot_state, false);
    }
  }
  virtual const Snapshot* GetSnapshot() {
2140 2141 2142
    ModelSnapshot* snapshot = new ModelSnapshot;
    snapshot->map_ = map_;
    return snapshot;
J
jorlow@chromium.org 已提交
2143 2144 2145
  }

  virtual void ReleaseSnapshot(const Snapshot* snapshot) {
2146
    delete reinterpret_cast<const ModelSnapshot*>(snapshot);
J
jorlow@chromium.org 已提交
2147 2148
  }
  virtual Status Write(const WriteOptions& options, WriteBatch* batch) {
2149 2150 2151 2152 2153
    class Handler : public WriteBatch::Handler {
     public:
      KVMap* map_;
      virtual void Put(const Slice& key, const Slice& value) {
        (*map_)[key.ToString()] = value.ToString();
J
jorlow@chromium.org 已提交
2154
      }
2155 2156 2157 2158 2159 2160 2161
      virtual void Delete(const Slice& key) {
        map_->erase(key.ToString());
      }
    };
    Handler handler;
    handler.map_ = &map_;
    return batch->Iterate(&handler);
J
jorlow@chromium.org 已提交
2162 2163
  }

2164
  virtual bool GetProperty(const Slice& property, std::string* value) {
J
jorlow@chromium.org 已提交
2165 2166 2167 2168 2169 2170 2171
    return false;
  }
  virtual void GetApproximateSizes(const Range* r, int n, uint64_t* sizes) {
    for (int i = 0; i < n; i++) {
      sizes[i] = 0;
    }
  }
G
Gabor Cselle 已提交
2172 2173 2174
  virtual void CompactRange(const Slice* start, const Slice* end) {
  }

2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189
  virtual int NumberLevels()
  {
	return 1;
  }

  virtual int MaxMemCompactionLevel()
  {
	return 1;
  }

  virtual int Level0StopWriteTrigger()
  {
	return -1;
  }

H
heyongqiang 已提交
2190 2191 2192 2193 2194
  virtual Status Flush(const leveldb::FlushOptions& options) {
    Status ret;
    return ret;
  }

2195 2196 2197 2198 2199 2200
  virtual Status DisableFileDeletions() {
    return Status::OK();
  }
  virtual Status EnableFileDeletions() {
    return Status::OK();
  }
2201
  virtual Status GetLiveFiles(std::vector<std::string>&, uint64_t* size) {
2202 2203 2204
    return Status::OK();
  }

J
jorlow@chromium.org 已提交
2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296
 private:
  class ModelIter: public Iterator {
   public:
    ModelIter(const KVMap* map, bool owned)
        : map_(map), owned_(owned), iter_(map_->end()) {
    }
    ~ModelIter() {
      if (owned_) delete map_;
    }
    virtual bool Valid() const { return iter_ != map_->end(); }
    virtual void SeekToFirst() { iter_ = map_->begin(); }
    virtual void SeekToLast() {
      if (map_->empty()) {
        iter_ = map_->end();
      } else {
        iter_ = map_->find(map_->rbegin()->first);
      }
    }
    virtual void Seek(const Slice& k) {
      iter_ = map_->lower_bound(k.ToString());
    }
    virtual void Next() { ++iter_; }
    virtual void Prev() { --iter_; }
    virtual Slice key() const { return iter_->first; }
    virtual Slice value() const { return iter_->second; }
    virtual Status status() const { return Status::OK(); }
   private:
    const KVMap* const map_;
    const bool owned_;  // Do we own map_
    KVMap::const_iterator iter_;
  };
  const Options options_;
  KVMap map_;
};

static std::string RandomKey(Random* rnd) {
  int len = (rnd->OneIn(3)
             ? 1                // Short sometimes to encourage collisions
             : (rnd->OneIn(100) ? rnd->Skewed(10) : rnd->Uniform(10)));
  return test::RandomKey(rnd, len);
}

static bool CompareIterators(int step,
                             DB* model,
                             DB* db,
                             const Snapshot* model_snap,
                             const Snapshot* db_snap) {
  ReadOptions options;
  options.snapshot = model_snap;
  Iterator* miter = model->NewIterator(options);
  options.snapshot = db_snap;
  Iterator* dbiter = db->NewIterator(options);
  bool ok = true;
  int count = 0;
  for (miter->SeekToFirst(), dbiter->SeekToFirst();
       ok && miter->Valid() && dbiter->Valid();
       miter->Next(), dbiter->Next()) {
    count++;
    if (miter->key().compare(dbiter->key()) != 0) {
      fprintf(stderr, "step %d: Key mismatch: '%s' vs. '%s'\n",
              step,
              EscapeString(miter->key()).c_str(),
              EscapeString(dbiter->key()).c_str());
      ok = false;
      break;
    }

    if (miter->value().compare(dbiter->value()) != 0) {
      fprintf(stderr, "step %d: Value mismatch for key '%s': '%s' vs. '%s'\n",
              step,
              EscapeString(miter->key()).c_str(),
              EscapeString(miter->value()).c_str(),
              EscapeString(miter->value()).c_str());
      ok = false;
    }
  }

  if (ok) {
    if (miter->Valid() != dbiter->Valid()) {
      fprintf(stderr, "step %d: Mismatch at end of iterators: %d vs. %d\n",
              step, miter->Valid(), dbiter->Valid());
      ok = false;
    }
  }
  fprintf(stderr, "%d entries compared: ok=%d\n", count, ok);
  delete miter;
  delete dbiter;
  return ok;
}

TEST(DBTest, Randomized) {
  Random rnd(test::RandomSeed());
S
Sanjay Ghemawat 已提交
2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339
  do {
    ModelDB model(CurrentOptions());
    const int N = 10000;
    const Snapshot* model_snap = NULL;
    const Snapshot* db_snap = NULL;
    std::string k, v;
    for (int step = 0; step < N; step++) {
      if (step % 100 == 0) {
        fprintf(stderr, "Step %d of %d\n", step, N);
      }
      // TODO(sanjay): Test Get() works
      int p = rnd.Uniform(100);
      if (p < 45) {                               // Put
        k = RandomKey(&rnd);
        v = RandomString(&rnd,
                         rnd.OneIn(20)
                         ? 100 + rnd.Uniform(100)
                         : rnd.Uniform(8));
        ASSERT_OK(model.Put(WriteOptions(), k, v));
        ASSERT_OK(db_->Put(WriteOptions(), k, v));

      } else if (p < 90) {                        // Delete
        k = RandomKey(&rnd);
        ASSERT_OK(model.Delete(WriteOptions(), k));
        ASSERT_OK(db_->Delete(WriteOptions(), k));


      } else {                                    // Multi-element batch
        WriteBatch b;
        const int num = rnd.Uniform(8);
        for (int i = 0; i < num; i++) {
          if (i == 0 || !rnd.OneIn(10)) {
            k = RandomKey(&rnd);
          } else {
            // Periodically re-use the same key from the previous iter, so
            // we have multiple entries in the write batch for the same key
          }
          if (rnd.OneIn(2)) {
            v = RandomString(&rnd, rnd.Uniform(10));
            b.Put(k, v);
          } else {
            b.Delete(k);
          }
J
jorlow@chromium.org 已提交
2340
        }
S
Sanjay Ghemawat 已提交
2341 2342
        ASSERT_OK(model.Write(WriteOptions(), &b));
        ASSERT_OK(db_->Write(WriteOptions(), &b));
J
jorlow@chromium.org 已提交
2343 2344
      }

S
Sanjay Ghemawat 已提交
2345 2346 2347 2348 2349 2350 2351 2352
      if ((step % 100) == 0) {
        ASSERT_TRUE(CompareIterators(step, &model, db_, NULL, NULL));
        ASSERT_TRUE(CompareIterators(step, &model, db_, model_snap, db_snap));
        // Save a snapshot from each DB this time that we'll use next
        // time we compare things, to make sure the current state is
        // preserved with the snapshot
        if (model_snap != NULL) model.ReleaseSnapshot(model_snap);
        if (db_snap != NULL) db_->ReleaseSnapshot(db_snap);
J
jorlow@chromium.org 已提交
2353

S
Sanjay Ghemawat 已提交
2354 2355
        Reopen();
        ASSERT_TRUE(CompareIterators(step, &model, db_, NULL, NULL));
J
jorlow@chromium.org 已提交
2356

S
Sanjay Ghemawat 已提交
2357 2358 2359
        model_snap = model.GetSnapshot();
        db_snap = db_->GetSnapshot();
      }
J
jorlow@chromium.org 已提交
2360
    }
S
Sanjay Ghemawat 已提交
2361 2362 2363
    if (model_snap != NULL) model.ReleaseSnapshot(model_snap);
    if (db_snap != NULL) db_->ReleaseSnapshot(db_snap);
  } while (ChangeOptions());
J
jorlow@chromium.org 已提交
2364 2365
}

2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387
std::string MakeKey(unsigned int num) {
  char buf[30];
  snprintf(buf, sizeof(buf), "%016u", num);
  return std::string(buf);
}

void BM_LogAndApply(int iters, int num_base_files) {
  std::string dbname = test::TmpDir() + "/leveldb_test_benchmark";
  DestroyDB(dbname, Options());

  DB* db = NULL;
  Options opts;
  opts.create_if_missing = true;
  Status s = DB::Open(opts, dbname, &db);
  ASSERT_OK(s);
  ASSERT_TRUE(db != NULL);

  delete db;
  db = NULL;

  Env* env = Env::Default();

2388 2389 2390
  port::Mutex mu;
  MutexLock l(&mu);

2391 2392 2393 2394
  InternalKeyComparator cmp(BytewiseComparator());
  Options options;
  VersionSet vset(dbname, &options, NULL, &cmp);
  ASSERT_OK(vset.Recover());
2395
  VersionEdit vbase(vset.NumberLevels());
2396 2397 2398 2399 2400 2401
  uint64_t fnum = 1;
  for (int i = 0; i < num_base_files; i++) {
    InternalKey start(MakeKey(2*fnum), 1, kTypeValue);
    InternalKey limit(MakeKey(2*fnum+1), 1, kTypeDeletion);
    vbase.AddFile(2, fnum++, 1 /* file size */, start, limit);
  }
2402
  ASSERT_OK(vset.LogAndApply(&vbase, &mu));
2403 2404 2405 2406

  uint64_t start_micros = env->NowMicros();

  for (int i = 0; i < iters; i++) {
2407
    VersionEdit vedit(vset.NumberLevels());
2408 2409 2410 2411
    vedit.DeleteFile(2, fnum);
    InternalKey start(MakeKey(2*fnum), 1, kTypeValue);
    InternalKey limit(MakeKey(2*fnum+1), 1, kTypeDeletion);
    vedit.AddFile(2, fnum++, 1 /* file size */, start, limit);
2412
    vset.LogAndApply(&vedit, &mu);
2413 2414 2415 2416 2417 2418 2419 2420 2421 2422
  }
  uint64_t stop_micros = env->NowMicros();
  unsigned int us = stop_micros - start_micros;
  char buf[16];
  snprintf(buf, sizeof(buf), "%d", num_base_files);
  fprintf(stderr,
          "BM_LogAndApply/%-6s   %8d iters : %9u us (%7.0f us / iter)\n",
          buf, iters, us, ((float)us) / iters);
}

H
Hans Wennborg 已提交
2423
}  // namespace leveldb
J
jorlow@chromium.org 已提交
2424 2425

int main(int argc, char** argv) {
2426 2427 2428 2429 2430 2431 2432 2433
  if (argc > 1 && std::string(argv[1]) == "--benchmark") {
    leveldb::BM_LogAndApply(1000, 1);
    leveldb::BM_LogAndApply(1000, 100);
    leveldb::BM_LogAndApply(1000, 10000);
    leveldb::BM_LogAndApply(100, 100000);
    return 0;
  }

J
jorlow@chromium.org 已提交
2434 2435
  return leveldb::test::RunAllTests();
}