db_test.cc 59.2 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 "leveldb/db.h"
S
Sanjay Ghemawat 已提交
6
#include "leveldb/filter_policy.h"
J
jorlow@chromium.org 已提交
7 8 9 10
#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
11
#include "leveldb/cache.h"
12 13
#include "leveldb/env.h"
#include "leveldb/table.h"
S
Sanjay Ghemawat 已提交
14
#include "util/hash.h"
J
jorlow@chromium.org 已提交
15
#include "util/logging.h"
16
#include "util/mutexlock.h"
J
jorlow@chromium.org 已提交
17 18 19 20 21 22 23 24 25 26 27
#include "util/testharness.h"
#include "util/testutil.h"

namespace leveldb {

static std::string RandomString(Random* rnd, int len) {
  std::string r;
  test::RandomString(rnd, len, &r);
  return r;
}

S
Sanjay Ghemawat 已提交
28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49
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;
  }
};
}

50 51 52 53 54 55
// 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_;

56 57 58
  // Simulate no-space errors while this pointer is non-NULL.
  port::AtomicPointer no_space_;

59 60 61
  // Simulate non-writable file system while this pointer is non-NULL
  port::AtomicPointer non_writable_;

S
Sanjay Ghemawat 已提交
62 63 64
  bool count_random_reads_;
  AtomicCounter random_read_counter_;

65 66
  AtomicCounter sleep_counter_;

67 68
  explicit SpecialEnv(Env* base) : EnvWrapper(base) {
    delay_sstable_sync_.Release_Store(NULL);
69
    no_space_.Release_Store(NULL);
70
    non_writable_.Release_Store(NULL);
S
Sanjay Ghemawat 已提交
71
    count_random_reads_ = false;
72 73 74 75 76 77 78 79 80 81 82 83 84
  }

  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) {
      }
85
      ~SSTableFile() { delete base_; }
86 87 88 89 90 91 92 93
      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);
        }
      }
94 95 96 97 98 99 100 101 102 103
      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();
      }
    };

104 105 106 107
    if (non_writable_.Acquire_Load() != NULL) {
       return Status::IOError("simulated write error");
     }

108 109 110 111 112 113 114 115
    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
116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139

  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;
  }
140 141 142 143 144

  virtual void SleepForMicroseconds(int micros) {
    sleep_counter_.Increment();
    target()->SleepForMicroseconds(micros);
  }
145 146
};

J
jorlow@chromium.org 已提交
147
class DBTest {
S
Sanjay Ghemawat 已提交
148 149 150 151 152 153 154
 private:
  const FilterPolicy* filter_policy_;

  // Sequence of option configurations to try
  enum OptionConfig {
    kDefault,
    kFilter,
155
    kNumLevel_3,
S
Sanjay Ghemawat 已提交
156 157 158 159
    kEnd
  };
  int option_config_;

J
jorlow@chromium.org 已提交
160 161
 public:
  std::string dbname_;
162
  SpecialEnv* env_;
J
jorlow@chromium.org 已提交
163 164 165 166
  DB* db_;

  Options last_options_;

S
Sanjay Ghemawat 已提交
167 168 169
  DBTest() : option_config_(kDefault),
             env_(new SpecialEnv(Env::Default())) {
    filter_policy_ = NewBloomFilterPolicy(10);
J
jorlow@chromium.org 已提交
170 171 172 173 174 175 176 177 178
    dbname_ = test::TmpDir() + "/db_test";
    DestroyDB(dbname_, Options());
    db_ = NULL;
    Reopen();
  }

  ~DBTest() {
    delete db_;
    DestroyDB(dbname_, Options());
179
    delete env_;
S
Sanjay Ghemawat 已提交
180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201
    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() {
    if (option_config_ == kEnd) {
      return false;
    } else {
      option_config_++;
      DestroyAndReopen();
      return true;
    }
  }

  // Return the current option configuration.
  Options CurrentOptions() {
    Options options;
    switch (option_config_) {
      case kFilter:
        options.filter_policy = filter_policy_;
        break;
202 203 204
      case kNumLevel_3:
	options.num_levels = 3;
	break;
S
Sanjay Ghemawat 已提交
205 206 207 208
      default:
        break;
    }
    return options;
J
jorlow@chromium.org 已提交
209 210 211 212 213 214 215 216 217 218
  }

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

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

S
Sanjay Ghemawat 已提交
219 220 221 222 223
  void Close() {
    delete db_;
    db_ = NULL;
  }

J
jorlow@chromium.org 已提交
224 225 226 227 228 229 230
  void DestroyAndReopen(Options* options = NULL) {
    delete db_;
    db_ = NULL;
    DestroyDB(dbname_, Options());
    ASSERT_OK(TryReopen(options));
  }

231 232 233 234
  Status PureReopen(Options* options, DB** db) {
    return DB::Open(*options, dbname_, db);
  }

J
jorlow@chromium.org 已提交
235 236 237 238 239 240 241
  Status TryReopen(Options* options) {
    delete db_;
    db_ = NULL;
    Options opts;
    if (options != NULL) {
      opts = *options;
    } else {
S
Sanjay Ghemawat 已提交
242
      opts = CurrentOptions();
J
jorlow@chromium.org 已提交
243 244 245 246 247 248 249 250
      opts.create_if_missing = true;
    }
    last_options_ = opts;

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

  Status Put(const std::string& k, const std::string& v) {
251
    return db_->Put(WriteOptions(), k, v);
J
jorlow@chromium.org 已提交
252 253 254
  }

  Status Delete(const std::string& k) {
255
    return db_->Delete(WriteOptions(), k);
J
jorlow@chromium.org 已提交
256 257 258 259 260 261 262 263 264 265 266 267 268 269 270
  }

  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
271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297
  // 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
298 299 300 301 302 303 304 305 306 307 308 309 310 311 312
  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
313
          if (last_options_.comparator->Compare(ikey.user_key, user_key) != 0) {
J
jorlow@chromium.org 已提交
314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340
            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) {
341
    std::string property;
J
jorlow@chromium.org 已提交
342 343
    ASSERT_TRUE(
        db_->GetProperty("leveldb.num-files-at-level" + NumberToString(level),
344 345
                         &property));
    return atoi(property.c_str());
J
jorlow@chromium.org 已提交
346 347
  }

348 349
  int TotalTableFiles() {
    int result = 0;
350
    for (int level = 0; level < db_->NumberLevels(); level++) {
351 352 353 354 355
      result += NumTableFilesAtLevel(level);
    }
    return result;
  }

G
Gabor Cselle 已提交
356 357 358 359
  // Return spread of files per level
  std::string FilesPerLevel() {
    std::string result;
    int last_non_zero_offset = 0;
360
    for (int level = 0; level < db_->NumberLevels(); level++) {
G
Gabor Cselle 已提交
361 362 363 364 365 366 367 368 369 370 371 372
      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;
  }

373 374 375 376 377 378
  int CountFiles() {
    std::vector<std::string> files;
    env_->GetChildren(dbname_, &files);
    return static_cast<int>(files.size());
  }

J
jorlow@chromium.org 已提交
379 380 381 382 383 384 385
  uint64_t Size(const Slice& start, const Slice& limit) {
    Range r(start, limit);
    uint64_t size;
    db_->GetApproximateSizes(&r, 1, &size);
    return size;
  }

386
  void Compact(const Slice& start, const Slice& limit) {
G
Gabor Cselle 已提交
387 388 389 390 391 392 393 394 395 396
    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();
397 398 399
    }
  }

400 401 402
  // 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) {
403
    MakeTables(db_->NumberLevels(), smallest, largest);
404 405
  }

406 407 408 409 410
  void DumpFileCounts(const char* label) {
    fprintf(stderr, "---\n%s:\n", label);
    fprintf(stderr, "maxoverlap: %lld\n",
            static_cast<long long>(
                dbfull()->TEST_MaxNextLevelOverlappingBytes()));
411
    for (int level = 0; level < db_->NumberLevels(); level++) {
412 413 414 415 416 417
      int num = NumTableFilesAtLevel(level);
      if (num > 0) {
        fprintf(stderr, "  level %3d : %d files\n", level, num);
      }
    }
  }
J
jorlow@chromium.org 已提交
418

G
Gabor Cselle 已提交
419 420 421 422 423 424
  std::string DumpSSTableList() {
    std::string property;
    db_->GetProperty("leveldb.sstables", &property);
    return property;
  }

J
jorlow@chromium.org 已提交
425 426 427 428 429 430 431 432 433
  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
434 435 436
};

TEST(DBTest, Empty) {
S
Sanjay Ghemawat 已提交
437 438 439 440
  do {
    ASSERT_TRUE(db_ != NULL);
    ASSERT_EQ("NOT_FOUND", Get("foo"));
  } while (ChangeOptions());
J
jorlow@chromium.org 已提交
441 442 443
}

TEST(DBTest, ReadWrite) {
S
Sanjay Ghemawat 已提交
444 445 446 447 448 449 450 451
  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
452 453 454
}

TEST(DBTest, PutDeleteGet) {
S
Sanjay Ghemawat 已提交
455 456 457 458 459 460 461 462
  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
463 464
}

465
TEST(DBTest, GetFromImmutableLayer) {
S
Sanjay Ghemawat 已提交
466 467 468 469 470
  do {
    Options options = CurrentOptions();
    options.env = env_;
    options.write_buffer_size = 100000;  // Small write buffer
    Reopen(&options);
471

S
Sanjay Ghemawat 已提交
472 473
    ASSERT_OK(Put("foo", "v1"));
    ASSERT_EQ("v1", Get("foo"));
474

S
Sanjay Ghemawat 已提交
475 476 477 478 479 480
    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());
481 482 483
}

TEST(DBTest, GetFromVersions) {
S
Sanjay Ghemawat 已提交
484 485 486 487 488
  do {
    ASSERT_OK(Put("foo", "v1"));
    dbfull()->TEST_CompactMemTable();
    ASSERT_EQ("v1", Get("foo"));
  } while (ChangeOptions());
489 490 491
}

TEST(DBTest, GetSnapshot) {
S
Sanjay Ghemawat 已提交
492 493 494 495 496 497 498 499 500 501 502 503 504 505 506
  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());
507 508 509
}

TEST(DBTest, GetLevel0Ordering) {
S
Sanjay Ghemawat 已提交
510 511 512 513 514 515 516 517 518 519 520 521
  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());
522 523 524
}

TEST(DBTest, GetOrderedByLevels) {
S
Sanjay Ghemawat 已提交
525 526 527 528 529 530 531 532 533
  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());
534 535 536
}

TEST(DBTest, GetPicksCorrectFile) {
S
Sanjay Ghemawat 已提交
537 538 539 540 541 542 543 544 545 546 547 548
  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());
549 550
}

551
TEST(DBTest, GetEncountersEmptyLevel) {
S
Sanjay Ghemawat 已提交
552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570
  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();
    }
571

S
Sanjay Ghemawat 已提交
572 573 574 575 576 577 578 579 580 581 582 583 584 585
    // 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);

    // Step 3: read until level 0 compaction disappears.
    int read_count = 0;
    while (NumTableFilesAtLevel(0) > 0) {
      ASSERT_LE(read_count, 10000) << "did not trigger level 0 compaction";
      read_count++;
      ASSERT_EQ("NOT_FOUND", Get("missing"));
    }
  } while (ChangeOptions());
586 587
}

J
jorlow@chromium.org 已提交
588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761
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;
}

762
TEST(DBTest, IterMultiWithDelete) {
S
Sanjay Ghemawat 已提交
763 764 765 766 767 768
  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"));
769

S
Sanjay Ghemawat 已提交
770 771 772 773 774 775 776
    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());
777 778
}

J
jorlow@chromium.org 已提交
779
TEST(DBTest, Recover) {
S
Sanjay Ghemawat 已提交
780 781 782
  do {
    ASSERT_OK(Put("foo", "v1"));
    ASSERT_OK(Put("baz", "v5"));
J
jorlow@chromium.org 已提交
783

S
Sanjay Ghemawat 已提交
784 785
    Reopen();
    ASSERT_EQ("v1", Get("foo"));
J
jorlow@chromium.org 已提交
786

S
Sanjay Ghemawat 已提交
787 788 789 790
    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
791

S
Sanjay Ghemawat 已提交
792 793 794 795 796 797 798
    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
799 800
}

H
heyongqiang 已提交
801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816
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
817 818 819 820 821
TEST(DBTest, WAL) {
  Options options = CurrentOptions();
  WriteOptions writeOpt = WriteOptions();
  writeOpt.disableWAL = true;
  ASSERT_OK(dbfull()->Put(writeOpt, "foo", "v1"));
H
heyongqiang 已提交
822
  ASSERT_OK(dbfull()->Put(writeOpt, "bar", "v1"));
H
heyongqiang 已提交
823 824 825

  Reopen();
  ASSERT_EQ("NOT_FOUND", Get("foo"));
H
heyongqiang 已提交
826
  ASSERT_EQ("NOT_FOUND", Get("bar"));
H
heyongqiang 已提交
827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849

  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"));
}

850 851 852 853 854 855 856
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
857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890
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
891
TEST(DBTest, RecoveryWithEmptyLog) {
S
Sanjay Ghemawat 已提交
892 893 894 895 896 897 898 899 900
  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
901 902
}

903 904 905
// 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
906 907 908 909 910
  do {
    Options options = CurrentOptions();
    options.env = env_;
    options.write_buffer_size = 1000000;
    Reopen(&options);
911

S
Sanjay Ghemawat 已提交
912 913 914 915 916
    // 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
917

S
Sanjay Ghemawat 已提交
918 919 920 921 922 923
    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());
924 925
}

J
jorlow@chromium.org 已提交
926 927 928 929 930 931 932
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
933
  Options options = CurrentOptions();
J
jorlow@chromium.org 已提交
934 935 936
  options.write_buffer_size = 10000;
  Reopen(&options);

937
  const int N = 500;
J
jorlow@chromium.org 已提交
938

939
  int starting_num_tables = TotalTableFiles();
J
jorlow@chromium.org 已提交
940 941 942
  for (int i = 0; i < N; i++) {
    ASSERT_OK(Put(Key(i), Key(i) + std::string(1000, 'v')));
  }
943
  int ending_num_tables = TotalTableFiles();
J
jorlow@chromium.org 已提交
944 945 946 947 948 949 950 951 952 953 954 955 956 957 958
  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
959
    Options options = CurrentOptions();
J
jorlow@chromium.org 已提交
960 961 962 963 964 965 966 967 968 969
    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
970
  Options options = CurrentOptions();
J
jorlow@chromium.org 已提交
971 972 973 974 975 976 977 978 979 980 981
  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
982
  Options options = CurrentOptions();
J
jorlow@chromium.org 已提交
983 984 985 986 987 988 989 990 991 992 993 994 995 996 997
  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
998
  dbfull()->TEST_CompactRange(0, NULL, NULL);
J
jorlow@chromium.org 已提交
999 1000 1001 1002 1003 1004 1005 1006

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

1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042
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);
}

1043
TEST(DBTest, RepeatedWritesToSameKey) {
S
Sanjay Ghemawat 已提交
1044
  Options options = CurrentOptions();
1045 1046 1047 1048 1049 1050
  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.
1051 1052
  const int kMaxFiles = dbfull()->NumberLevels() +
		dbfull()->Level0StopWriteTrigger();
1053 1054 1055 1056 1057 1058 1059 1060 1061 1062

  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());
  }
}

1063
TEST(DBTest, SparseMerge) {
S
Sanjay Ghemawat 已提交
1064
  Options options = CurrentOptions();
1065 1066 1067
  options.compression = kNoCompression;
  Reopen(&options);

1068 1069
  FillLevels("A", "Z");

1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084
  // 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");
1085
  dbfull()->TEST_CompactMemTable();
G
Gabor Cselle 已提交
1086
  dbfull()->TEST_CompactRange(0, NULL, NULL);
1087 1088 1089 1090 1091 1092 1093 1094 1095 1096

  // 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
1097
  dbfull()->TEST_CompactRange(0, NULL, NULL);
1098
  ASSERT_LE(dbfull()->TEST_MaxNextLevelOverlappingBytes(), 20*1048576);
G
Gabor Cselle 已提交
1099
  dbfull()->TEST_CompactRange(1, NULL, NULL);
1100 1101 1102
  ASSERT_LE(dbfull()->TEST_MaxNextLevelOverlappingBytes(), 20*1048576);
}

J
jorlow@chromium.org 已提交
1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114
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
1115 1116 1117 1118 1119
  do {
    Options options = CurrentOptions();
    options.write_buffer_size = 100000000;        // Large write buffer
    options.compression = kNoCompression;
    DestroyAndReopen();
J
jorlow@chromium.org 已提交
1120

S
Sanjay Ghemawat 已提交
1121 1122 1123
    ASSERT_TRUE(Between(Size("", "xyz"), 0, 0));
    Reopen(&options);
    ASSERT_TRUE(Between(Size("", "xyz"), 0, 0));
J
jorlow@chromium.org 已提交
1124

S
Sanjay Ghemawat 已提交
1125 1126 1127 1128 1129 1130 1131 1132 1133
    // 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
1134

S
Sanjay Ghemawat 已提交
1135 1136
    // 0 because GetApproximateSizes() does not account for memtable space
    ASSERT_TRUE(Between(Size("", Key(50)), 0, 0));
D
 
dgrogan@chromium.org 已提交
1137

S
Sanjay Ghemawat 已提交
1138 1139 1140
    // 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
1141

S
Sanjay Ghemawat 已提交
1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155
      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
1156
      }
D
dgrogan@chromium.org 已提交
1157

S
Sanjay Ghemawat 已提交
1158 1159 1160 1161
      ASSERT_EQ(NumTableFilesAtLevel(0), 0);
      ASSERT_GT(NumTableFilesAtLevel(1), 0);
    }
  } while (ChangeOptions());
J
jorlow@chromium.org 已提交
1162 1163 1164
}

TEST(DBTest, ApproximateSizes_MixOfSmallAndLarge) {
S
Sanjay Ghemawat 已提交
1165 1166 1167 1168
  do {
    Options options = CurrentOptions();
    options.compression = kNoCompression;
    Reopen();
J
jorlow@chromium.org 已提交
1169

S
Sanjay Ghemawat 已提交
1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199
    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
1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224
}

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
1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242
  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
1243

S
Sanjay Ghemawat 已提交
1244 1245 1246
    db_->ReleaseSnapshot(s1);
    ASSERT_EQ("v2", Get("foo", s2));
    ASSERT_EQ("v4", Get("foo"));
1247

S
Sanjay Ghemawat 已提交
1248 1249 1250 1251
    db_->ReleaseSnapshot(s2);
    ASSERT_EQ("v4", Get("foo"));
  } while (ChangeOptions());
}
J
jorlow@chromium.org 已提交
1252

S
Sanjay Ghemawat 已提交
1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278
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
1279

S
Sanjay Ghemawat 已提交
1280 1281
    ASSERT_TRUE(Between(Size("", "pastfoo"), 0, 1000));
  } while (ChangeOptions());
J
jorlow@chromium.org 已提交
1282 1283 1284 1285 1286
}

TEST(DBTest, DeletionMarkers1) {
  Put("foo", "v1");
  ASSERT_OK(dbfull()->TEST_CompactMemTable());
1287
  const int last = dbfull()->MaxMemCompactionLevel();
1288 1289 1290 1291 1292 1293 1294 1295 1296
  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
1297 1298 1299
  Delete("foo");
  Put("foo", "v2");
  ASSERT_EQ(AllEntriesFor("foo"), "[ v2, DEL, v1 ]");
1300
  ASSERT_OK(dbfull()->TEST_CompactMemTable());  // Moves to level last-2
J
jorlow@chromium.org 已提交
1301
  ASSERT_EQ(AllEntriesFor("foo"), "[ v2, DEL, v1 ]");
G
Gabor Cselle 已提交
1302 1303
  Slice z("z");
  dbfull()->TEST_CompactRange(last-2, NULL, &z);
J
jorlow@chromium.org 已提交
1304 1305 1306
  // 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
1307
  dbfull()->TEST_CompactRange(last-1, NULL, NULL);
1308 1309
  // 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
1310 1311 1312 1313 1314 1315
  ASSERT_EQ(AllEntriesFor("foo"), "[ v2 ]");
}

TEST(DBTest, DeletionMarkers2) {
  Put("foo", "v1");
  ASSERT_OK(dbfull()->TEST_CompactMemTable());
1316
  const int last = dbfull()->MaxMemCompactionLevel();
1317 1318 1319 1320 1321 1322 1323 1324 1325
  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
1326 1327
  Delete("foo");
  ASSERT_EQ(AllEntriesFor("foo"), "[ DEL, v1 ]");
1328
  ASSERT_OK(dbfull()->TEST_CompactMemTable());  // Moves to level last-2
J
jorlow@chromium.org 已提交
1329
  ASSERT_EQ(AllEntriesFor("foo"), "[ DEL, v1 ]");
G
Gabor Cselle 已提交
1330
  dbfull()->TEST_CompactRange(last-2, NULL, NULL);
1331
  // DEL kept: "last" file overlaps
J
jorlow@chromium.org 已提交
1332
  ASSERT_EQ(AllEntriesFor("foo"), "[ DEL, v1 ]");
G
Gabor Cselle 已提交
1333
  dbfull()->TEST_CompactRange(last-1, NULL, NULL);
1334 1335
  // 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
1336 1337 1338
  ASSERT_EQ(AllEntriesFor("foo"), "[ ]");
}

G
Gabor Cselle 已提交
1339
TEST(DBTest, OverlapInLevel0) {
S
Sanjay Ghemawat 已提交
1340
  do {
1341 1342
	int tmp = dbfull()->MaxMemCompactionLevel();
    ASSERT_EQ(tmp, 2) << "Fix test to match config";
G
Gabor Cselle 已提交
1343

S
Sanjay Ghemawat 已提交
1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364
    // 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
1365

S
Sanjay Ghemawat 已提交
1366 1367 1368 1369
    // 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
1370

S
Sanjay Ghemawat 已提交
1371 1372 1373 1374 1375 1376 1377 1378
    // 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
1379 1380
}

H
Hans Wennborg 已提交
1381
TEST(DBTest, L0_CompactionBug_Issue44_a) {
S
Sanjay Ghemawat 已提交
1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395
  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
1396 1397 1398
}

TEST(DBTest, L0_CompactionBug_Issue44_b) {
S
Sanjay Ghemawat 已提交
1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421
  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
1422 1423
}

J
jorlow@chromium.org 已提交
1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438
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
1439
  Options new_options = CurrentOptions();
J
jorlow@chromium.org 已提交
1440 1441 1442 1443 1444 1445 1446
  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
1447 1448 1449 1450 1451
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 {
1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470
      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
1471 1472 1473
    }
  };
  NumberComparator cmp;
S
Sanjay Ghemawat 已提交
1474
  Options new_options = CurrentOptions();
H
Hans Wennborg 已提交
1475 1476
  new_options.create_if_missing = true;
  new_options.comparator = &cmp;
S
Sanjay Ghemawat 已提交
1477
  new_options.filter_policy = NULL;     // Cannot use bloom filters
1478
  new_options.write_buffer_size = 1000;  // Compact more often
H
Hans Wennborg 已提交
1479
  DestroyAndReopen(&new_options);
1480 1481
  ASSERT_OK(Put("[10]", "ten"));
  ASSERT_OK(Put("[0x14]", "twenty"));
H
Hans Wennborg 已提交
1482
  for (int i = 0; i < 2; i++) {
1483 1484 1485 1486
    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
1487 1488
    ASSERT_EQ("NOT_FOUND", Get("[15]"));
    ASSERT_EQ("NOT_FOUND", Get("[0xf]"));
1489 1490 1491 1492 1493 1494 1495 1496 1497 1498
    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
1499 1500 1501
  }
}

G
Gabor Cselle 已提交
1502
TEST(DBTest, ManualCompaction) {
1503
  ASSERT_EQ(dbfull()->MaxMemCompactionLevel(), 2)
G
Gabor Cselle 已提交
1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535
      << "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
1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574
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;
}

1575 1576
// Check that number of files does not grow when we are out of space
TEST(DBTest, NoSpace) {
S
Sanjay Ghemawat 已提交
1577
  Options options = CurrentOptions();
1578 1579 1580 1581 1582 1583 1584 1585
  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
1586
  for (int i = 0; i < 5; i++) {
1587
    for (int level = 0; level < dbfull()->NumberLevels()-1; level++) {
1588 1589 1590 1591
      dbfull()->TEST_CompactRange(level, NULL, NULL);
    }
  }
  env_->no_space_.Release_Store(NULL);
1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616
  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);
1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629
}

TEST(DBTest, FilesDeletedAfterCompaction) {
  ASSERT_OK(Put("foo", "v2"));
  Compact("a", "z");
  const int num_files = CountFiles();
  for (int i = 0; i < 10; i++) {
    ASSERT_OK(Put("foo", "v2"));
    Compact("a", "z");
  }
  ASSERT_EQ(CountFiles(), num_files);
}

S
Sanjay Ghemawat 已提交
1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 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
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;
}

1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697
// 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);
1698
  int id = t->id;
1699 1700
  DB* db = t->state->test->db_;
  uintptr_t counter = 0;
1701 1702
  fprintf(stderr, "... starting thread %d\n", id);
  Random rnd(1000 + id);
1703 1704 1705
  std::string value;
  char valbuf[1500];
  while (t->state->stop.Acquire_Load() == NULL) {
1706
    t->state->counter[id].Release_Store(reinterpret_cast<void*>(counter));
1707 1708 1709 1710 1711 1712 1713 1714 1715

    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",
1716
               key, id, static_cast<int>(counter));
1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736
      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++;
  }
1737 1738
  t->state->thread_done[id].Release_Store(t);
  fprintf(stderr, "... stopping thread %d after %d ops\n", id, int(counter));
1739 1740
}

H
Hans Wennborg 已提交
1741
}  // namespace
1742 1743

TEST(DBTest, MultiThreaded) {
S
Sanjay Ghemawat 已提交
1744 1745 1746 1747 1748 1749 1750 1751 1752
  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);
    }
1753

S
Sanjay Ghemawat 已提交
1754 1755 1756 1757 1758 1759 1760
    // 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]);
    }
1761

S
Sanjay Ghemawat 已提交
1762 1763
    // Let them run for a while
    env_->SleepForMicroseconds(kTestSeconds * 1000000);
1764

S
Sanjay Ghemawat 已提交
1765 1766 1767 1768 1769 1770
    // 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);
      }
1771
    }
S
Sanjay Ghemawat 已提交
1772
  } while (ChangeOptions());
1773 1774
}

1775 1776 1777 1778
namespace {
typedef std::map<std::string, std::string> KVMap;
}

J
jorlow@chromium.org 已提交
1779 1780
class ModelDB: public DB {
 public:
1781 1782 1783 1784 1785
  class ModelSnapshot : public Snapshot {
   public:
    KVMap map_;
  };

J
jorlow@chromium.org 已提交
1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805
  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 =
1806
          &(reinterpret_cast<const ModelSnapshot*>(options.snapshot)->map_);
J
jorlow@chromium.org 已提交
1807 1808 1809 1810
      return new ModelIter(snapshot_state, false);
    }
  }
  virtual const Snapshot* GetSnapshot() {
1811 1812 1813
    ModelSnapshot* snapshot = new ModelSnapshot;
    snapshot->map_ = map_;
    return snapshot;
J
jorlow@chromium.org 已提交
1814 1815 1816
  }

  virtual void ReleaseSnapshot(const Snapshot* snapshot) {
1817
    delete reinterpret_cast<const ModelSnapshot*>(snapshot);
J
jorlow@chromium.org 已提交
1818 1819
  }
  virtual Status Write(const WriteOptions& options, WriteBatch* batch) {
1820 1821 1822 1823 1824
    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
1825
      }
1826 1827 1828 1829 1830 1831 1832
      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
1833 1834
  }

1835
  virtual bool GetProperty(const Slice& property, std::string* value) {
J
jorlow@chromium.org 已提交
1836 1837 1838 1839 1840 1841 1842
    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
1843 1844 1845
  virtual void CompactRange(const Slice* start, const Slice* end) {
  }

1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860
  virtual int NumberLevels()
  {
	return 1;
  }

  virtual int MaxMemCompactionLevel()
  {
	return 1;
  }

  virtual int Level0StopWriteTrigger()
  {
	return -1;
  }

H
heyongqiang 已提交
1861 1862 1863 1864 1865
  virtual Status Flush(const leveldb::FlushOptions& options) {
    Status ret;
    return ret;
  }

J
jorlow@chromium.org 已提交
1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 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 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957
 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
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
  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
2001
        }
S
Sanjay Ghemawat 已提交
2002 2003
        ASSERT_OK(model.Write(WriteOptions(), &b));
        ASSERT_OK(db_->Write(WriteOptions(), &b));
J
jorlow@chromium.org 已提交
2004 2005
      }

S
Sanjay Ghemawat 已提交
2006 2007 2008 2009 2010 2011 2012 2013
      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
2014

S
Sanjay Ghemawat 已提交
2015 2016
        Reopen();
        ASSERT_TRUE(CompareIterators(step, &model, db_, NULL, NULL));
J
jorlow@chromium.org 已提交
2017

S
Sanjay Ghemawat 已提交
2018 2019 2020
        model_snap = model.GetSnapshot();
        db_snap = db_->GetSnapshot();
      }
J
jorlow@chromium.org 已提交
2021
    }
S
Sanjay Ghemawat 已提交
2022 2023 2024
    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
2025 2026
}

2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048
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();

2049 2050 2051
  port::Mutex mu;
  MutexLock l(&mu);

2052 2053 2054 2055
  InternalKeyComparator cmp(BytewiseComparator());
  Options options;
  VersionSet vset(dbname, &options, NULL, &cmp);
  ASSERT_OK(vset.Recover());
2056
  VersionEdit vbase(vset.NumberLevels());
2057 2058 2059 2060 2061 2062
  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);
  }
2063
  ASSERT_OK(vset.LogAndApply(&vbase, &mu));
2064 2065 2066 2067

  uint64_t start_micros = env->NowMicros();

  for (int i = 0; i < iters; i++) {
2068
    VersionEdit vedit(vset.NumberLevels());
2069 2070 2071 2072
    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);
2073
    vset.LogAndApply(&vedit, &mu);
2074 2075 2076 2077 2078 2079 2080 2081 2082 2083
  }
  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
2084
}  // namespace leveldb
J
jorlow@chromium.org 已提交
2085 2086

int main(int argc, char** argv) {
2087 2088 2089 2090 2091 2092 2093 2094
  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
2095 2096
  return leveldb::test::RunAllTests();
}