ha_tokudb.cc 90.8 KB
Newer Older
1
#ifdef USE_PRAGMA_IMPLEMENTATION
2
#pragma implementation          // gcc: Class implementation
3 4 5 6 7
#endif

#define MYSQL_SERVER 1
#include "mysql_priv.h"

8 9 10 11
#if !defined(HA_END_SPACE_KEY) || HA_END_SPACE_KEY != 0
#error
#endif

12 13 14 15
unsigned long my_getphyspages() {
    return sysconf(_SC_PHYS_PAGES);
}

16 17 18 19 20 21
#include <syscall.h>

unsigned int my_tid() {
    return syscall(__NR_gettid);
}

22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39
#undef PACKAGE
#undef VERSION
#undef HAVE_DTRACE
#undef _DTRACE_VERSION

#include "tokudb_config.h"

/* We define DTRACE after mysql_priv.h in case it disabled dtrace in the main server */
#ifdef HAVE_DTRACE
#define _DTRACE_VERSION 1
#else
#endif

#include "tokudb_probes.h"

#include "ha_tokudb.h"
#include <mysql/plugin.h>

40
static handler *tokudb_create_handler(handlerton * hton, TABLE_SHARE * table, MEM_ROOT * mem_root);
41

42
handlerton *tokudb_hton;
43 44

typedef struct st_tokudb_trx_data {
45 46 47 48
    DB_TXN *all;
    DB_TXN *stmt;
    DB_TXN *sp_level;
    uint tokudb_lock_count;
49 50
} tokudb_trx_data;

51 52
// QQQ how to tune these?
#define HA_TOKUDB_ROWS_IN_TABLE 10000   /* to get optimization right */
53
#define HA_TOKUDB_RANGE_COUNT   100
54
#define HA_TOKUDB_MAX_ROWS      10000000        /* Max rows in table */
55 56 57 58 59 60
/* extra rows for estimate_rows_upper_bound() */
#define HA_TOKUDB_EXTRA_ROWS    100

/* Bits for share->status */
#define STATUS_PRIMARY_KEY_INIT 1
#define STATUS_ROW_COUNT_INIT   2
61 62 63 64
#define STATUS_TOKUDB_ANALYZE      4

const char *ha_tokudb_ext = ".tokudb";

Rich Prohaska's avatar
Rich Prohaska committed
65
//static my_bool tokudb_shared_data = FALSE;
66 67
static u_int32_t tokudb_init_flags = DB_PRIVATE | DB_RECOVER;
static u_int32_t tokudb_env_flags = DB_LOG_AUTOREMOVE;
Rich Prohaska's avatar
Rich Prohaska committed
68
//static u_int32_t tokudb_lock_type = DB_LOCK_DEFAULT;
69 70
//static ulong tokudb_log_buffer_size = 0;
//static ulong tokudb_log_file_size = 0;
71
static ulonglong tokudb_cache_size = 0;
72
static char *tokudb_home;
73 74 75
//static char *tokudb_tmpdir;
static char *tokudb_data_dir;
static char *tokudb_log_dir;
Rich Prohaska's avatar
Rich Prohaska committed
76 77
//static long tokudb_lock_scan_time = 0;
//static ulong tokudb_region_size = 0;
78
//static ulong tokudb_cache_parts = 1;
79
static ulong tokudb_trans_retry = 1;
80 81 82 83
static ulong tokudb_max_lock;

static DB_ENV *db_env;

84 85 86
static const char tokudb_hton_name[] = "TokuDB";
static const int tokudb_hton_name_length = sizeof(tokudb_hton_name) - 1;

87
#if 0 // QQQ do we need this
88 89 90 91 92 93 94 95
const char *tokudb_lock_names[] = { "DEFAULT", "OLDEST", "RANDOM", "YOUNGEST", "EXPIRE", "MAXLOCKS",
    "MAXWRITE", "MINLOCKS", "MINWRITE", 0
};

u_int32_t tokudb_lock_types[] = { DB_LOCK_DEFAULT, DB_LOCK_OLDEST, DB_LOCK_RANDOM, DB_LOCK_YOUNGEST,
    DB_LOCK_EXPIRE, DB_LOCK_MAXLOCKS, DB_LOCK_MAXWRITE, DB_LOCK_MINLOCKS,
    DB_LOCK_MINWRITE
};
96

97 98 99
TYPELIB tokudb_lock_typelib = { array_elements(tokudb_lock_names) - 1, "",
    tokudb_lock_names, NULL
};
100
#endif
101 102

static void tokudb_print_error(const DB_ENV * db_env, const char *db_errpfx, const char *buffer);
103
static void tokudb_cleanup_log_files(void);
104 105 106 107 108 109 110 111 112 113
static TOKUDB_SHARE *get_share(const char *table_name, TABLE * table);
static int free_share(TOKUDB_SHARE * share, TABLE * table, uint hidden_primary_key, bool mutex_is_locked);
static int write_status(DB * status_block, char *buff, uint length);
static void update_status(TOKUDB_SHARE * share, TABLE * table);
static int tokudb_end(handlerton * hton, ha_panic_function type);
static bool tokudb_flush_logs(handlerton * hton);
static bool tokudb_show_status(handlerton * hton, THD * thd, stat_print_fn * print, enum ha_stat_type);
static int tokudb_close_connection(handlerton * hton, THD * thd);
static int tokudb_commit(handlerton * hton, THD * thd, bool all);
static int tokudb_rollback(handlerton * hton, THD * thd, bool all);
114
#if 0
115 116 117
static int tokudb_rollback_to_savepoint(handlerton * hton, THD * thd, void *savepoint);
static int tokudb_savepoint(handlerton * hton, THD * thd, void *savepoint);
static int tokudb_release_savepoint(handlerton * hton, THD * thd, void *savepoint);
118
#endif
119
static bool tokudb_show_logs(THD * thd, stat_print_fn * stat_print);
120 121 122 123

static HASH tokudb_open_tables;
pthread_mutex_t tokudb_mutex;

124 125 126
static uchar *tokudb_get_key(TOKUDB_SHARE * share, size_t * length, my_bool not_used __attribute__ ((unused))) {
    *length = share->table_name_length;
    return (uchar *) share->table_name;
127 128 129 130 131
}

// declare a thread specific variable
// stole this from my_thr_init.h with MontyW's advice
#ifdef USE_TLS
132
pthread_key(void **, tokudb_thr_private);
133
#else
134
pthread_key(void *, tokudb_thr_private);
135 136
#endif /* USE_TLS */

137 138 139 140
int tokudb_thr_private_create(void) {
    if ((pthread_key_create(&tokudb_thr_private, NULL)) != 0)
        return 1;
    return 0;
141 142
}

143 144 145 146 147 148
void *tokudb_thr_private_get(void) {
    void **tmp;
    tmp = my_pthread_getspecific(void **, tokudb_thr_private);
    if (tmp == NULL)
        return NULL;
    return *tmp;
149 150
}

151 152 153 154 155 156
void tokudb_thr_private_set(void *v) {
    // if i alloc'ed v to store here,
    //  make SURE to set a thread cleanup function to de-alloc it
    //  because the thread may terminate unexpectedly early
    pthread_setspecific(tokudb_thr_private, v);
    return;
157 158
}

159 160
static int tokudb_init_func(void *p) {
    DBUG_ENTER("tokudb_init_func");
161

162
    tokudb_hton = (handlerton *) p;
163

164 165 166
    if (tokudb_thr_private_create()) {
        DBUG_RETURN(TRUE);
    }
167

168 169
    VOID(pthread_mutex_init(&tokudb_mutex, MY_MUTEX_INIT_FAST));
    (void) hash_init(&tokudb_open_tables, system_charset_info, 32, 0, 0, (hash_get_key) tokudb_get_key, 0, 0);
170

171
    tokudb_hton->state = SHOW_OPTION_YES;
172
    // tokudb_hton->flags= HTON_CAN_RECREATE;  // QQQ this came from skeleton
173
    tokudb_hton->flags = HTON_CLOSE_CURSORS_AT_COMMIT | HTON_FLUSH_AFTER_RENAME;
174
#ifdef DB_TYPE_TOKUDB
175 176 177
    tokudb_hton->db_type = DB_TYPE_TOKUDB;
#else
    tokudb_hton->db_type = DB_TYPE_UNKNOWN;
178 179
#endif

180 181
    tokudb_hton->create = tokudb_create_handler;
    tokudb_hton->close_connection = tokudb_close_connection;
182 183
#if 0
    tokudb_hton->savepoint_offset = sizeof(DB_TXN *);
184 185 186
    tokudb_hton->savepoint_set = tokudb_savepoint;
    tokudb_hton->savepoint_rollback = tokudb_rollback_to_savepoint;
    tokudb_hton->savepoint_release = tokudb_release_savepoint;
187
#endif
188 189 190 191 192 193
    tokudb_hton->commit = tokudb_commit;
    tokudb_hton->rollback = tokudb_rollback;
    tokudb_hton->panic = tokudb_end;
    tokudb_hton->flush_logs = tokudb_flush_logs;
    tokudb_hton->show_status = tokudb_show_status;

194
#if 0
195 196 197
    if (!tokudb_tmpdir)
        tokudb_tmpdir = mysql_tmpdir;
    DBUG_PRINT("info", ("tokudb_tmpdir: %s", tokudb_tmpdir));
198
#endif
199 200 201
    if (!tokudb_home)
        tokudb_home = mysql_real_data_home;
    DBUG_PRINT("info", ("tokudb_home: %s", tokudb_home));
202
#if 0
203
    if (!tokudb_log_buffer_size) { // QQQ
204 205 206 207 208 209 210
        tokudb_log_buffer_size = max(table_cache_size * 512, 32 * 1024);
        DBUG_PRINT("info", ("computing tokudb_log_buffer_size %ld\n", tokudb_log_buffer_size));
    }
    tokudb_log_file_size = tokudb_log_buffer_size * 4;
    tokudb_log_file_size = MY_ALIGN(tokudb_log_file_size, 1024 * 1024L);
    tokudb_log_file_size = max(tokudb_log_file_size, 10 * 1024 * 1024L);
    DBUG_PRINT("info", ("computing tokudb_log_file_size: %ld\n", tokudb_log_file_size));
211
#endif
212 213 214
    int r;
    if ((r = db_env_create(&db_env, 0))) {
        DBUG_PRINT("info", ("db_env_create %d\n", r));
215
        goto error;
216 217
    }

Rich Prohaska's avatar
Rich Prohaska committed
218
    DBUG_PRINT("info", ("tokudb_env_flags: 0x%x\n", tokudb_env_flags));
219 220 221 222 223
    r = db_env->set_flags(db_env, tokudb_env_flags, 1);
    if (r) // QQQ
        printf("WARNING: flags %x r %d\n", tokudb_env_flags, r); // goto error;

    // config error handling
224
    db_env->set_errcall(db_env, tokudb_print_error);
225 226 227
    db_env->set_errpfx(db_env, "TokuDB");

    // config directories
228
#if 0
229 230
    DBUG_PRINT("info", ("tokudb_tmpdir: %s\n", tokudb_tmpdir));
    db_env->set_tmp_dir(db_env, tokudb_tmpdir);
231 232 233 234 235 236 237 238 239 240 241 242 243
#endif

    {
    char *data_dir = tokudb_data_dir;
    if (data_dir == 0) 
        data_dir = mysql_data_home;
    DBUG_PRINT("info", ("tokudb_data_dir: %s\n", data_dir));
    db_env->set_data_dir(db_env, data_dir);
    }

    if (tokudb_log_dir) {
        DBUG_PRINT("info", ("tokudb_log_dir: %s\n", tokudb_log_dir));
        db_env->set_lg_dir(db_env, tokudb_log_dir);
244 245
    }

246
    // config the cache table
247 248 249 250 251 252
    if (tokudb_cache_size == 0) {
        unsigned long pagesize = my_getpagesize();
        unsigned long long npages = my_getphyspages();
        unsigned long long physmem = npages * pagesize;
        tokudb_cache_size = physmem / 8;
    }
253 254
    if (tokudb_cache_size) {
        DBUG_PRINT("info", ("tokudb_cache_size: %lld\n", tokudb_cache_size));
255
        r = db_env->set_cachesize(db_env, tokudb_cache_size / (1024 * 1024L * 1024L), tokudb_cache_size % (1024L * 1024L * 1024L), 1);
256 257 258 259
        if (r) {
            DBUG_PRINT("info", ("set_cachesize %d\n", r));
            goto error; 
        }
260 261 262 263
    }
    u_int32_t gbytes, bytes; int parts;
    r = db_env->get_cachesize(db_env, &gbytes, &bytes, &parts);
    if (r == 0) 
264
        printf("tokudb_cache_size %lld\n", ((unsigned long long) gbytes << 30) + bytes);
265

266 267
#if 0
    // QQQ config the logs
268 269 270 271 272 273
    DBUG_PRINT("info", ("tokudb_log_file_size: %ld\n", tokudb_log_file_size));
    db_env->set_lg_max(db_env, tokudb_log_file_size);
    DBUG_PRINT("info", ("tokudb_log_buffer_size: %ld\n", tokudb_log_buffer_size));
    db_env->set_lg_bsize(db_env, tokudb_log_buffer_size);
    // DBUG_PRINT("info",("tokudb_region_size: %ld\n", tokudb_region_size));
    // db_env->set_lg_regionmax(db_env, tokudb_region_size);
274
#endif
275

276 277
    // config the locks
#if 0 // QQQ no lock types yet
278 279
    DBUG_PRINT("info", ("tokudb_lock_type: 0x%lx\n", tokudb_lock_type));
    db_env->set_lk_detect(db_env, tokudb_lock_type);
280 281 282 283 284 285 286 287 288
#endif
    if (tokudb_max_lock) {
        DBUG_PRINT("info",("tokudb_max_lock: %ld\n", tokudb_max_lock));
        r = db_env->set_lk_max_locks(db_env, tokudb_max_lock);
        if (r) {
            DBUG_PRINT("info", ("tokudb_set_max_locks %d\n", r));
            goto error;
        }
    }
289

290 291 292 293
    if ((r = db_env->open(db_env, tokudb_home, 
                          tokudb_init_flags | DB_INIT_LOCK | DB_INIT_LOG | DB_INIT_MPOOL | DB_INIT_TXN | DB_CREATE | DB_THREAD, 
                          0666))) {
        DBUG_PRINT("info", ("env->open %d\n", r));
294 295 296
        goto error;
    }
    DBUG_RETURN(FALSE);
297 298 299 300 301 302

error:
    if (db_env) {
        db_env->close(db_env, 0);
        db_env = 0;
    }
303 304 305 306 307 308 309 310 311 312 313 314 315 316
    DBUG_RETURN(TRUE);
}

static int tokudb_done_func(void *p) {
    DBUG_ENTER("tokudb_done_func");
    int error = 0;

    if (tokudb_open_tables.records)
        error = 1;
    hash_free(&tokudb_open_tables);
    pthread_mutex_destroy(&tokudb_mutex);
    pthread_key_delete(tokudb_thr_private);

    DBUG_RETURN(0);
317 318 319 320 321 322 323
}

/** @brief
    Simple lock controls. The "share" it creates is a structure we will
    pass to each tokudb handler. Do you have to have one of these? Well, you have
    pieces that are used for locking, and they are needed to function.
*/
324 325 326
static TOKUDB_SHARE *get_share(const char *table_name, TABLE * table) {
    TOKUDB_SHARE *share;
    uint length;
327

328 329 330 331 332 333 334 335 336 337
    pthread_mutex_lock(&tokudb_mutex);
    length = (uint) strlen(table_name);

    if (!(share = (TOKUDB_SHARE *) hash_search(&tokudb_open_tables, (uchar *) table_name, length))) {
        ulong *rec_per_key;
        char *tmp_name;
        DB **key_file;
        u_int32_t *key_type;
        uint keys = table->s->keys;

338 339 340 341 342 343 344 345
        if (!(share = (TOKUDB_SHARE *) 
            my_multi_malloc(MYF(MY_WME | MY_ZEROFILL), 
                            &share, sizeof(*share),
                            &tmp_name, length + 1, 
                            &rec_per_key, keys * sizeof(ha_rows), 
                            &key_file, (keys + 1) * sizeof(*key_file), 
                            &key_type, (keys + 1) * sizeof(u_int32_t), 
                            NullS))) {
346 347 348 349 350 351 352 353
            pthread_mutex_unlock(&tokudb_mutex);
            return NULL;
        }

        share->use_count = 0;
        share->table_name_length = length;
        share->table_name = tmp_name;
        strmov(share->table_name, table_name);
354

355 356 357 358 359 360 361 362
        share->rec_per_key = rec_per_key;
        share->key_file = key_file;
        share->key_type = key_type;

        if (my_hash_insert(&tokudb_open_tables, (uchar *) share))
            goto error;
        thr_lock_init(&share->lock);
        pthread_mutex_init(&share->mutex, MY_MUTEX_INIT_FAST);
363
    }
364
    pthread_mutex_unlock(&tokudb_mutex);
365

366
    return share;
367

368
  error:
369 370 371
    pthread_mutex_destroy(&share->mutex);
    my_free((uchar *) share, MYF(0));

372
    return NULL;
373 374
}

375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405
static int free_share(TOKUDB_SHARE * share, TABLE * table, uint hidden_primary_key, bool mutex_is_locked) {
    int error, result = 0;
    uint keys = table->s->keys + test(hidden_primary_key);

    pthread_mutex_lock(&tokudb_mutex);

    if (mutex_is_locked)
        pthread_mutex_unlock(&share->mutex);
    if (!--share->use_count) {
        DBUG_PRINT("info", ("share->use_count %u", share->use_count));
        DB **key_file = share->key_file;

        /* this does share->file->close() implicitly */
        update_status(share, table);

        for (uint i = 0; i < keys; i++) {
            if (key_file[i] && (error = key_file[i]->close(key_file[i], 0)))
                result = error;
        }

        if (share->status_block && (error = share->status_block->close(share->status_block, 0)))
            result = error;

        hash_delete(&tokudb_open_tables, (uchar *) share);
        thr_lock_delete(&share->lock);
        pthread_mutex_destroy(&share->mutex);
        my_free((uchar *) share, MYF(0));
    }
    pthread_mutex_unlock(&tokudb_mutex);

    return result;
406 407
}

408 409 410 411 412 413 414 415 416 417 418 419 420 421
static handler *tokudb_create_handler(handlerton * hton, TABLE_SHARE * table, MEM_ROOT * mem_root) {
    return new(mem_root) ha_tokudb(hton, table);
}

int tokudb_end(handlerton * hton, ha_panic_function type) {
    DBUG_ENTER("tokudb_end");
    int error = 0;
    if (db_env) {
        tokudb_cleanup_log_files();
        error = db_env->close(db_env, 0);       // Error is logged
        db_env = 0;
#if 0   // avoid a double mutex destroy during shutdown
        hash_free(&tokudb_open_tables);
        pthread_mutex_destroy(&tokudb_mutex);
422
#endif
423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451
    }
    DBUG_RETURN(error);
}

static int tokudb_close_connection(handlerton * hton, THD * thd) {
    my_free((void *) thd->ha_data[hton->slot], MYF(0));
    return 0;
}

bool tokudb_flush_logs(handlerton * hton) {
    DBUG_ENTER("tokudb_flush_logs");
    int error;
    bool result = 0;
    if ((error = db_env->log_flush(db_env, 0))) {
        my_error(ER_ERROR_DURING_FLUSH_LOGS, MYF(0), error);
        result = 1;
    }
    if ((error = db_env->txn_checkpoint(db_env, 0, 0, 0))) {
        my_error(ER_ERROR_DURING_CHECKPOINT, MYF(0), error);
        result = 1;
    }
    DBUG_RETURN(result);
}

static int tokudb_commit(handlerton * hton, THD * thd, bool all) {
    DBUG_ENTER("tokudb_commit");
    DBUG_PRINT("trans", ("ending transaction %s", all ? "all" : "stmt"));
    tokudb_trx_data *trx = (tokudb_trx_data *) thd->ha_data[hton->slot];
    DB_TXN **txn = all ? &trx->all : &trx->stmt;
452 453 454 455 456 457 458 459
    int error = 0;
    if (*txn) {
        error = (*txn)->commit(*txn, 0);
        if (*txn == trx->sp_level)
            trx->sp_level = 0;
        *txn = 0;
    } else
        printf("%s:%d:warning\n", __FILE__, __LINE__);
460 461 462 463 464 465 466 467
    DBUG_RETURN(error);
}

static int tokudb_rollback(handlerton * hton, THD * thd, bool all) {
    DBUG_ENTER("tokudb_rollback");
    DBUG_PRINT("trans", ("aborting transaction %s", all ? "all" : "stmt"));
    tokudb_trx_data *trx = (tokudb_trx_data *) thd->ha_data[hton->slot];
    DB_TXN **txn = all ? &trx->all : &trx->stmt;
468 469 470 471 472 473 474 475
    int error = 0;
    if (*txn) {
        error = (*txn)->abort(*txn);
	if (*txn == trx->sp_level)
	    trx->sp_level = 0;
	*txn = 0;
    } else
        printf("%s:%d:warning\n", __FILE__, __LINE__);
476 477 478
    DBUG_RETURN(error);
}

479 480
#if 0

481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516
static int tokudb_savepoint(handlerton * hton, THD * thd, void *savepoint) {
    DBUG_ENTER("tokudb_savepoint");
    int error;
    DB_TXN **save_txn = (DB_TXN **) savepoint;
    tokudb_trx_data *trx = (tokudb_trx_data *) thd->ha_data[hton->slot];
    if (!(error = db_env->txn_begin(db_env, trx->sp_level, save_txn, 0))) {
        trx->sp_level = *save_txn;
    }
    DBUG_RETURN(error);
}

static int tokudb_rollback_to_savepoint(handlerton * hton, THD * thd, void *savepoint) {
    DBUG_ENTER("tokudb_rollback_to_savepoint");
    int error;
    DB_TXN *parent, **save_txn = (DB_TXN **) savepoint;
    tokudb_trx_data *trx = (tokudb_trx_data *) thd->ha_data[hton->slot];
    parent = (*save_txn)->parent;
    if (!(error = (*save_txn)->abort(*save_txn))) {
        trx->sp_level = parent;
        error = tokudb_savepoint(hton, thd, savepoint);
    }
    DBUG_RETURN(error);
}

static int tokudb_release_savepoint(handlerton * hton, THD * thd, void *savepoint) {
    DBUG_ENTER("tokudb_release_savepoint");
    int error;
    DB_TXN *parent, **save_txn = (DB_TXN **) savepoint;
    tokudb_trx_data *trx = (tokudb_trx_data *) thd->ha_data[hton->slot];
    parent = (*save_txn)->parent;
    if (!(error = (*save_txn)->commit(*save_txn, 0))) {
        trx->sp_level = parent;
        *save_txn = 0;
    }
    DBUG_RETURN(error);
}
517

518
#endif
519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579

static bool tokudb_show_logs(THD * thd, stat_print_fn * stat_print) {
    DBUG_ENTER("tokudb_show_logs");
    char **all_logs, **free_logs, **a, **f;
    int error = 1;
    MEM_ROOT **root_ptr = my_pthread_getspecific_ptr(MEM_ROOT **, THR_MALLOC);
    MEM_ROOT show_logs_root, *old_mem_root = *root_ptr;

    init_sql_alloc(&show_logs_root, BDB_LOG_ALLOC_BLOCK_SIZE, BDB_LOG_ALLOC_BLOCK_SIZE);
    *root_ptr = &show_logs_root;
    all_logs = free_logs = 0;

    if ((error = db_env->log_archive(db_env, &all_logs, DB_ARCH_ABS | DB_ARCH_LOG)) || (error = db_env->log_archive(db_env, &free_logs, DB_ARCH_ABS))) {
        DBUG_PRINT("error", ("log_archive failed (error %d)", error));
        db_env->err(db_env, error, "log_archive: DB_ARCH_ABS");
        if (error == DB_NOTFOUND)
            error = 0;          // No log files
        goto err;
    }
    /* Error is 0 here */
    if (all_logs) {
        for (a = all_logs, f = free_logs; *a; ++a) {
            if (f && *f && strcmp(*a, *f) == 0) {
                f++;
                if ((error = stat_print(thd, tokudb_hton_name, tokudb_hton_name_length, *a, strlen(*a), STRING_WITH_LEN(SHOW_LOG_STATUS_FREE))))
                    break;
            } else {
                if ((error = stat_print(thd, tokudb_hton_name, tokudb_hton_name_length, *a, strlen(*a), STRING_WITH_LEN(SHOW_LOG_STATUS_INUSE))))
                    break;
            }
        }
    }
  err:
    if (all_logs)
        free(all_logs);
    if (free_logs)
        free(free_logs);
    free_root(&show_logs_root, MYF(0));
    *root_ptr = old_mem_root;
    DBUG_RETURN(error);
}

bool tokudb_show_status(handlerton * hton, THD * thd, stat_print_fn * stat_print, enum ha_stat_type stat_type) {
    switch (stat_type) {
    case HA_ENGINE_LOGS:
        return tokudb_show_logs(thd, stat_print);
    default:
        return FALSE;
    }
}

static void tokudb_print_error(const DB_ENV * db_env, const char *db_errpfx, const char *buffer) {
    sql_print_error("%s:  %s", db_errpfx, buffer);
}

void tokudb_cleanup_log_files(void) {
    DBUG_ENTER("tokudb_cleanup_log_files");
    char **names;
    int error;

    // by HF. Sometimes it crashes. TODO - find out why
580
#ifndef EMBEDDED_LIBRARY
581 582 583 584
    /* XXX: Probably this should be done somewhere else, and
     * should be tunable by the user. */
    if ((error = db_env->txn_checkpoint(db_env, 0, 0, 0)))
        my_error(ER_ERROR_DURING_CHECKPOINT, MYF(0), error);
585
#endif
586 587 588 589 590
    if ((error = db_env->log_archive(db_env, &names, DB_ARCH_ABS)) != 0) {
        DBUG_PRINT("error", ("log_archive failed (error %d)", error));
        db_env->err(db_env, error, "log_archive: DB_ARCH_ABS");
        DBUG_VOID_RETURN;
    }
591

592 593 594 595
    if (names) {
        char **np;
        for (np = names; *np; ++np)
            my_delete(*np, MYF(MY_WME));
596

597 598
        free(names);
    }
599

600
    DBUG_VOID_RETURN;
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
static int get_name_length(const char *name) {
    int n = 0;
    const char *newname = name;
    if (tokudb_data_dir) {
        n += strlen(tokudb_data_dir) + 1;
        if (strncmp("./", name, 2) == 0) 
            newname = name + 2;
    }
    n += strlen(newname);
    n += strlen(ha_tokudb_ext);
    return n;
}

static void make_name(char *newname, const char *tablename, const char *dictname) {
    const char *newtablename = tablename;
    char *nn = newname;
    if (tokudb_data_dir) {
        nn += sprintf(nn, "%s/", tokudb_data_dir);
        if (strncmp("./", tablename, 2) == 0)
            newtablename = tablename + 2;
    }
    nn += sprintf(nn, "%s%s", newtablename, ha_tokudb_ext);
    if (dictname)
        nn += sprintf(nn, "/%s%s", dictname, ha_tokudb_ext);
}

629 630 631
ha_tokudb::ha_tokudb(handlerton * hton, TABLE_SHARE * table_arg)
:  
    handler(hton, table_arg), alloc_ptr(0), rec_buff(0), file(0),
632
    // QQQ what do these flags do?
633 634 635 636
    int_table_flags(HA_REC_NOT_IN_SEQ | HA_FAST_KEY_READ | HA_NULL_IN_KEY | HA_CAN_INDEX_BLOBS | HA_PRIMARY_KEY_IN_READ_INDEX | 
                    HA_FILE_BASED | HA_CAN_GEOMETRY | HA_AUTO_PART_KEY | HA_TABLE_SCAN_ON_INDEX), 
    changed_rows(0), last_dup_key((uint) - 1), version(0), using_ignore(0) {
}
637

638 639 640
static const char *ha_tokudb_exts[] = {
    ha_tokudb_ext,
    NullS
641 642
};

643 644 645 646
const char **ha_tokudb::bas_ext() const {
    return ha_tokudb_exts;
}

647
// QQQ what do these flags do?
648 649 650 651 652 653 654 655 656 657 658 659 660
ulong ha_tokudb::index_flags(uint idx, uint part, bool all_parts) const {
    ulong flags = (HA_READ_NEXT | HA_READ_PREV | HA_READ_ORDER | HA_KEYREAD_ONLY | HA_READ_RANGE);
    for (uint i = all_parts ? 0 : part; i <= part; i++) {
        KEY_PART_INFO *key_part = table_share->key_info[idx].key_part + i;
        if (key_part->field->type() == FIELD_TYPE_BLOB) {
            /* We can't use BLOBS to shortcut sorts */
            flags &= ~(HA_READ_ORDER | HA_KEYREAD_ONLY | HA_READ_RANGE);
            break;
        }
        switch (key_part->field->key_type()) {
        case HA_KEYTYPE_TEXT:
        case HA_KEYTYPE_VARTEXT1:
        case HA_KEYTYPE_VARTEXT2:
661
            /* QQQ
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
               As BDB stores only one copy of equal strings, we can't use key read
               on these. Binary collations do support key read though.
             */
            if (!(key_part->field->charset()->state & MY_CS_BINSORT))
                flags &= ~HA_KEYREAD_ONLY;
            break;
        default:               // Keep compiler happy
            break;
        }
    }
    return flags;
}

static int tokudb_cmp_hidden_key(DB * file, const DBT * new_key, const DBT * saved_key) {
    ulonglong a = uint5korr((char *) new_key->data);
    ulonglong b = uint5korr((char *) saved_key->data);
    return a < b ? -1 : (a > b ? 1 : 0);
}

static int tokudb_cmp_packed_key(DB * file, const DBT * new_key, const DBT * saved_key) {
    assert(file->app_private != 0);
    KEY *key = (KEY *) file->app_private;
    uchar *new_key_ptr = (uchar *) new_key->data;
    uchar *saved_key_ptr = (uchar *) saved_key->data;
    KEY_PART_INFO *key_part = key->key_part, *end = key_part + key->key_parts;
    uint key_length = new_key->size;

    //DBUG_DUMP("key_in_index", saved_key_ptr, saved_key->size);
    for (; key_part != end && (int) key_length > 0; key_part++) {
        int cmp;
        uint length;
        if (key_part->null_bit) {
            if (*new_key_ptr != *saved_key_ptr++)
                return ((int) *new_key_ptr - (int) saved_key_ptr[-1]);
            key_length--;
            if (!*new_key_ptr++)
                continue;
        }
Rich Prohaska's avatar
Rich Prohaska committed
700
        if ((cmp = key_part->field->pack_cmp(new_key_ptr, saved_key_ptr, key_part->length, 0)))
701 702 703 704 705 706
            return cmp;
        length = key_part->field->packed_col_length(new_key_ptr, key_part->length);
        new_key_ptr += length;
        key_length -= length;
        saved_key_ptr += key_part->field->packed_col_length(saved_key_ptr, key_part->length);
    }
707
    return 0;
708 709 710
}

/* Compare key against row */
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
static bool tokudb_key_cmp(TABLE * table, KEY * key_info, const uchar * key, uint key_length) {
    KEY_PART_INFO *key_part = key_info->key_part, *end = key_part + key_info->key_parts;

    for (; key_part != end && (int) key_length > 0; key_part++) {
        int cmp;
        uint length;
        if (key_part->null_bit) {
            key_length--;
            /*
               With the current usage, the following case will always be FALSE,
               because NULL keys are sorted before any other key
             */
            if (*key != (table->record[0][key_part->null_offset] & key_part->null_bit) ? 0 : 1)
                return 1;
            if (!*key++)        // Null value
                continue;
        }
        /*
           Last argument has to be 0 as we are also using this to function to see
           if a key like 'a  ' matched a row with 'a'
         */
        if ((cmp = key_part->field->pack_cmp(key, key_part->length, 0)))
            return cmp;
        length = key_part->field->packed_col_length(key, key_part->length);
        key += length;
        key_length -= length;
737
    }
738 739 740 741 742 743 744 745
    return 0;                   // Identical keys
}

int ha_tokudb::open(const char *name, int mode, uint test_if_locked) {
    DBUG_ENTER("ha_tokudb::open");
    TOKUDB_OPEN();

    char name_buff[FN_REFLEN];
746
    uint open_flags = (mode == O_RDONLY ? DB_RDONLY : 0) | DB_THREAD;
747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763
    uint max_key_length;
    int error;

    /* Open primary key */
    hidden_primary_key = 0;
    if ((primary_key = table_share->primary_key) >= MAX_KEY) {
        // No primary key
        primary_key = table_share->keys;
        key_used_on_scan = MAX_KEY;
        ref_length = hidden_primary_key = TOKUDB_HIDDEN_PRIMARY_KEY_LENGTH;
    } else
        key_used_on_scan = primary_key;

    /* Need some extra memory in case of packed keys */
    max_key_length = table_share->max_key_length + MAX_REF_PARTS * 3;
    if (!(alloc_ptr =
          my_multi_malloc(MYF(MY_WME),
764 765 766 767
                          &key_buff, max_key_length, 
                          &key_buff2, max_key_length, 
                          &primary_key_buff, (hidden_primary_key ? 0 : table_share->key_info[table_share->primary_key].key_length), 
                          NullS)))
768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787
        DBUG_RETURN(1);
    if (!(rec_buff = (uchar *) my_malloc((alloced_rec_buff_length = table_share->rec_buff_length), MYF(MY_WME)))) {
        my_free(alloc_ptr, MYF(0));
        DBUG_RETURN(1);
    }

    /* Init shared structure */
    if (!(share = get_share(name, table))) {
        my_free((char *) rec_buff, MYF(0));
        my_free(alloc_ptr, MYF(0));
        DBUG_RETURN(1);
    }
    thr_lock_data_init(&share->lock, &lock, NULL);
    key_file = share->key_file;
    key_type = share->key_type;
    bzero((void *) &current_row, sizeof(current_row));

    /* Fill in shared structure, if needed */
    pthread_mutex_lock(&share->mutex);
    file = share->file;
788
    printf("%s:%d:tokudbopen:%p:share=%p:file=%p:table=%p:table->s=%p:%d:tid=%u\n", __FILE__, __LINE__, this, share, share->file, table, table->s, share->use_count, my_tid());
789 790 791 792 793 794 795 796 797 798 799 800 801 802
    if (!share->use_count++) {
        DBUG_PRINT("info", ("share->use_count %u", share->use_count));

        if ((error = db_create(&file, db_env, 0))) {
            free_share(share, table, hidden_primary_key, 1);
            my_free((char *) rec_buff, MYF(0));
            my_free(alloc_ptr, MYF(0));
            my_errno = error;
            DBUG_RETURN(1);
        }
        share->file = file;

        file->set_bt_compare(file, (hidden_primary_key ? tokudb_cmp_hidden_key : tokudb_cmp_packed_key));
        if (!hidden_primary_key)
803
            file->app_private = (void *) (table_share->key_info + table_share->primary_key);
804
        char newname[strlen(name) + 32];
805 806
        make_name(newname, name, "main");
        fn_format(name_buff, newname, "", 0, MY_UNPACK_FILENAME);
807
        if ((error = file->open(file, 0, name_buff, "main", DB_BTREE, open_flags + DB_AUTO_COMMIT, 0))) {
808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828
            free_share(share, table, hidden_primary_key, 1);
            my_free((char *) rec_buff, MYF(0));
            my_free(alloc_ptr, MYF(0));
            my_errno = error;
            DBUG_RETURN(1);
        }

        /* Open other keys;  These are part of the share structure */
        key_file[primary_key] = file;
        key_type[primary_key] = hidden_primary_key ? 0 : DB_NOOVERWRITE;

        DB **ptr = key_file;
        for (uint i = 0, used_keys = 0; i < table_share->keys; i++, ptr++) {
            char part[16];
            if (i != primary_key) {
                if ((error = db_create(ptr, db_env, 0))) {
                    __close(1);
                    my_errno = error;
                    DBUG_RETURN(1);
                }
                sprintf(part, "key%d", ++used_keys);
829 830
                make_name(newname, name, part);
                fn_format(name_buff, newname, "", 0, MY_UNPACK_FILENAME);
831 832
                key_type[i] = table->key_info[i].flags & HA_NOSAME ? DB_NOOVERWRITE : DB_YESOVERWRITE;
                (*ptr)->set_bt_compare(*ptr, tokudb_cmp_packed_key);
833
                (*ptr)->app_private = (void *) (table_share->key_info + i);
834 835 836 837
                if (!(table->key_info[i].flags & HA_NOSAME)) {
                    DBUG_PRINT("info", ("Setting DB_DUP+DB_DUPSORT for key %u", i));
                    (*ptr)->set_flags(*ptr, DB_DUP + DB_DUPSORT);
                }
838
                if ((error = (*ptr)->open(*ptr, 0, name_buff, part, DB_BTREE, open_flags + DB_AUTO_COMMIT, 0))) {
839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863
                    __close(1);
                    my_errno = error;
                    DBUG_RETURN(1);
                }
            }
        }
        /* Calculate pack_length of primary key */
        share->fixed_length_primary_key = 1;
        if (!hidden_primary_key) {
            ref_length = 0;
            KEY_PART_INFO *key_part = table->key_info[primary_key].key_part;
            KEY_PART_INFO *end = key_part + table->key_info[primary_key].key_parts;
            for (; key_part != end; key_part++)
                ref_length += key_part->field->max_packed_col_length(key_part->length);
            share->fixed_length_primary_key = (ref_length == table->key_info[primary_key].key_length);
            share->status |= STATUS_PRIMARY_KEY_INIT;
        }
        share->ref_length = ref_length;
    }
    ref_length = share->ref_length;     // If second open
    pthread_mutex_unlock(&share->mutex);

    transaction = 0;
    cursor = 0;
    key_read = 0;
864
    stats.block_size = 1<<20;    // QQQ Tokudb DB block size
865 866
    share->fixed_length_row = !(table_share->db_create_options & HA_OPTION_PACK_RECORD);

867
    // QQQ what happens if get_status fails
868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886
    get_status();
    info(HA_STATUS_NO_LOCK | HA_STATUS_VARIABLE | HA_STATUS_CONST);

    DBUG_RETURN(0);
}

int ha_tokudb::close(void) {
    DBUG_ENTER("ha_tokudb::close");
    TOKUDB_CLOSE();
    DBUG_RETURN(__close(0));
}

int ha_tokudb::__close(int mutex_is_locked) {
    DBUG_ENTER("ha_tokudb::__close");
    printf("%s:%d:close:%p\n", __FILE__, __LINE__, this);
    my_free(rec_buff, MYF(MY_ALLOW_ZERO_PTR));
    my_free(alloc_ptr, MYF(MY_ALLOW_ZERO_PTR));
    ha_tokudb::reset();         // current_row buffer
    DBUG_RETURN(free_share(share, table, hidden_primary_key, mutex_is_locked));
887 888 889
}

/* Reallocate buffer if needed */
890 891 892 893 894 895 896 897 898
bool ha_tokudb::fix_rec_buff_for_blob(ulong length) {
    if (!rec_buff || length > alloced_rec_buff_length) {
        uchar *newptr;
        if (!(newptr = (uchar *) my_realloc((void *) rec_buff, length, MYF(MY_ALLOW_ZERO_PTR))))
            return 1;
        rec_buff = newptr;
        alloced_rec_buff_length = length;
    }
    return 0;
899 900 901
}

/* Calculate max length needed for row */
902 903 904 905 906 907 908 909
ulong ha_tokudb::max_row_length(const uchar * buf) {
    ulong length = table_share->reclength + table_share->fields * 2;
    uint *ptr, *end;
    for (ptr = table_share->blob_field, end = ptr + table_share->blob_fields; ptr != end; ptr++) {
        Field_blob *blob = ((Field_blob *) table->field[*ptr]);
        length += blob->get_length((uchar *) (buf + blob->offset((uchar *) buf))) + 2;
    }
    return length;
910 911 912 913 914 915 916 917 918 919 920
}

/*
  Pack a row for storage.
  If the row is of fixed length, just store the  row 'as is'.
  If not, we will generate a packed row suitable for storage.
  This will only fail if we don't have enough memory to pack the row,
  which may only happen in rows with blobs, as the default row length is
  pre-allocated.
*/

921 922 923 924 925 926 927 928 929 930 931 932
int ha_tokudb::pack_row(DBT * row, const uchar * record, bool new_row) {
    uchar *ptr;
    bzero((void *) row, sizeof(*row));
    if (share->fixed_length_row) {
        row->data = (void *) record;
        row->size = table_share->reclength + hidden_primary_key;
        if (hidden_primary_key) {
            if (new_row)
                get_auto_primary_key(current_ident);
            memcpy_fixed((char *) record + table_share->reclength, (char *) current_ident, TOKUDB_HIDDEN_PRIMARY_KEY_LENGTH);
        }
        return 0;
933
    }
934 935 936 937 938
    if (table_share->blob_fields) {
        if (fix_rec_buff_for_blob(max_row_length(record)))
            return HA_ERR_OUT_OF_MEM;
    }

939
    /* Copy null bits */
940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970
    memcpy(rec_buff, record, table_share->null_bytes);
    ptr = rec_buff + table_share->null_bytes;

    for (Field ** field = table->field; *field; field++)
        ptr = (*field)->pack(ptr, (const uchar *)
                             (record + (*field)->offset((uchar *) record)));

    if (hidden_primary_key) {
        if (new_row)
            get_auto_primary_key(current_ident);
        memcpy_fixed((char *) ptr, (char *) current_ident, TOKUDB_HIDDEN_PRIMARY_KEY_LENGTH);
        ptr += TOKUDB_HIDDEN_PRIMARY_KEY_LENGTH;
    }
    row->data = rec_buff;
    row->size = (size_t) (ptr - rec_buff);
    return 0;
}

void ha_tokudb::unpack_row(uchar * record, DBT * row) {
    if (share->fixed_length_row)
        memcpy(record, (void *) row->data, table_share->reclength + hidden_primary_key);
    else {
        /* Copy null bits */
        my_bitmap_map *old_map = dbug_tmp_use_all_columns(table, table->write_set);
        const uchar *ptr = (const uchar *) row->data;
        memcpy(record, ptr, table_share->null_bytes);
        ptr += table_share->null_bytes;
        for (Field ** field = table->field; *field; field++)
            ptr = (*field)->unpack(record + (*field)->offset(record), ptr);
        dbug_tmp_restore_column_map(table->write_set, old_map);
    }
971 972 973
}

/* Store the key and the primary key into the row */
974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991
void ha_tokudb::unpack_key(uchar * record, DBT * key, uint index) {
    KEY *key_info = table->key_info + index;
    KEY_PART_INFO *key_part = key_info->key_part, *end = key_part + key_info->key_parts;
    uchar *pos = (uchar *) key->data;

    for (; key_part != end; key_part++) {
        if (key_part->null_bit) {
            if (!*pos++) {        // Null value
                /*
                   We don't need to reset the record data as we will not access it
                   if the null data is set
                 */
                record[key_part->null_offset] |= key_part->null_bit;
                continue;
            }
            record[key_part->null_offset] &= ~key_part->null_bit;
        }
        pos = (uchar *) key_part->field->unpack_key(record + key_part->field->offset(record), pos,
992
#if MYSQL_VERSION_ID < 50123
993
                                                    key_part->length);
994
#else
995
                                                    key_part->length, table->s->db_low_byte_first);
996
#endif
997
    }
998 999 1000 1001 1002 1003 1004 1005 1006
}

/*
  Create a packed key from a row. This key will be written as such
  to the index tree.

  This will never fail as the key buffer is pre-allocated.
*/

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
DBT *ha_tokudb::create_key(DBT * key, uint keynr, uchar * buff, const uchar * record, int key_length) {
    DBUG_ENTER("ha_tokudb::create_key");
    bzero((void *) key, sizeof(*key));
    if (hidden_primary_key && keynr == primary_key) {
        key->data = current_ident;
        key->size = TOKUDB_HIDDEN_PRIMARY_KEY_LENGTH;
        DBUG_RETURN(key);
    }

    KEY *key_info = table->key_info + keynr;
    KEY_PART_INFO *key_part = key_info->key_part;
    KEY_PART_INFO *end = key_part + key_info->key_parts;
    my_bitmap_map *old_map = dbug_tmp_use_all_columns(table, table->write_set);

    key->data = buff;
    for (; key_part != end && key_length > 0; key_part++) {
        if (key_part->null_bit) {
            /* Store 0 if the key part is a NULL part */
            if (record[key_part->null_offset] & key_part->null_bit) {
                *buff++ = 0;
                key->flags |= DB_DBT_DUPOK;
                continue;
            }
            *buff++ = 1;        // Store NOT NULL marker
        }
        buff = key_part->field->pack_key(buff, (uchar *) (record + key_part->offset),
1033
#if MYSQL_VERSION_ID < 50123
1034
                                         key_part->length);
1035
#else
1036
                                         key_part->length, table->s->db_low_byte_first);
1037
#endif
1038 1039 1040 1041 1042 1043
        key_length -= key_part->length;
    }
    key->size = (buff - (uchar *) key->data);
    DBUG_DUMP("key", (uchar *) key->data, key->size);
    dbug_tmp_restore_column_map(table->write_set, old_map);
    DBUG_RETURN(key);
1044 1045 1046 1047 1048 1049 1050 1051 1052 1053
}


/*
  Create a packed key from from a MySQL unpacked key (like the one that is
  sent from the index_read()

  This key is to be used to read a row
*/

1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076
DBT *ha_tokudb::pack_key(DBT * key, uint keynr, uchar * buff, const uchar * key_ptr, uint key_length) {
    DBUG_ENTER("ha_tokudb::pack_key");
    KEY *key_info = table->key_info + keynr;
    KEY_PART_INFO *key_part = key_info->key_part;
    KEY_PART_INFO *end = key_part + key_info->key_parts;
    my_bitmap_map *old_map = dbug_tmp_use_all_columns(table, table->write_set);

    bzero((void *) key, sizeof(*key));
    key->data = buff;

    for (; key_part != end && (int) key_length > 0; key_part++) {
        uint offset = 0;
        if (key_part->null_bit) {
            if (!(*buff++ = (*key_ptr == 0)))   // Store 0 if NULL
            {
                key_length -= key_part->store_length;
                key_ptr += key_part->store_length;
                key->flags |= DB_DBT_DUPOK;
                continue;
            }
            offset = 1;         // Data is at key_ptr+1
        }
        buff = key_part->field->pack_key_from_key_image(buff, (uchar *) key_ptr + offset,
1077
#if MYSQL_VERSION_ID < 50123
1078
                                                        key_part->length);
1079
#else
1080
                                                        key_part->length, table->s->db_low_byte_first);
1081
#endif
1082 1083 1084 1085 1086 1087 1088
        key_ptr += key_part->store_length;
        key_length -= key_part->store_length;
    }
    key->size = (buff - (uchar *) key->data);
    DBUG_DUMP("key", (uchar *) key->data, key->size);
    dbug_tmp_restore_column_map(table->write_set, old_map);
    DBUG_RETURN(key);
1089 1090 1091 1092 1093 1094
}

/** @brief
    Get status information that is stored in the 'status' sub database
    and the max used value for the hidden primary key.
*/
1095 1096 1097 1098 1099
void ha_tokudb::get_status() {
    if (!test_all_bits(share->status, (STATUS_PRIMARY_KEY_INIT | STATUS_ROW_COUNT_INIT))) {
        pthread_mutex_lock(&share->mutex);
        if (!(share->status & STATUS_PRIMARY_KEY_INIT)) {
            (void) extra(HA_EXTRA_KEYREAD);
1100 1101 1102 1103 1104 1105
            int do_commit = 0;
            if (transaction == 0) {
                int r = db_env->txn_begin(db_env, 0, &transaction, 0);
                assert(r == 0);
                do_commit = 1;
            }
1106 1107 1108 1109
            index_init(primary_key, 0);
            if (!index_last(table->record[1]))
                share->auto_ident = uint5korr(current_ident);
            index_end();
1110 1111 1112 1113 1114
            if (do_commit) {
                int r = transaction->commit(transaction, 0);
                assert(r == 0);
                transaction = 0;
            }
1115 1116 1117 1118
            (void) extra(HA_EXTRA_NO_KEYREAD);
        }
        if (!share->status_block) {
            char name_buff[FN_REFLEN];
1119 1120 1121
            char newname[get_name_length(share->table_name) + 32];
            make_name(newname, share->table_name, "status");
            fn_format(name_buff, newname, "", 0, MY_UNPACK_FILENAME);
1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132
            uint open_mode = (((table->db_stat & HA_READ_ONLY) ? DB_RDONLY : 0)
                              | DB_THREAD);
            if (!db_create(&share->status_block, db_env, 0)) {
                if (share->status_block->open(share->status_block, NULL, name_buff, "status", DB_BTREE, open_mode, 0)) {
                    share->status_block->close(share->status_block, 0);
                    share->status_block = 0;
                }
            }
        }
        if (!(share->status & STATUS_ROW_COUNT_INIT) && share->status_block) {
            share->org_rows = share->rows = table_share->max_rows ? table_share->max_rows : HA_TOKUDB_MAX_ROWS;
1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148
            DB_TXN *transaction = 0;
            int r = db_env->txn_begin(db_env, 0, &transaction, 0);
            if (r == 0) {
                r = share->status_block->cursor(share->status_block, transaction, &cursor, 0);
                if (r == 0) {
                    DBT row;
                    char rec_buff[64];
                    bzero((void *) &row, sizeof(row));
                    bzero((void *) &last_key, sizeof(last_key));
                    row.data = rec_buff;
                    row.ulen = sizeof(rec_buff);
                    row.flags = DB_DBT_USERMEM;
                    if (!cursor->c_get(cursor, &last_key, &row, DB_FIRST)) {
                        uint i;
                        uchar *pos = (uchar *) row.data;
                        share->org_rows = share->rows = uint4korr(pos);
1149
                        pos += 4;
1150 1151 1152 1153
                        for (i = 0; i < table_share->keys; i++) {
                            share->rec_per_key[i] = uint4korr(pos);
                            pos += 4;
                        }
1154
                    }
1155
                    cursor->c_close(cursor);
1156
                }
1157 1158
                r = transaction->commit(transaction, 0);
                transaction = 0;
1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193
            }
            cursor = 0;
        }
        share->status |= STATUS_PRIMARY_KEY_INIT | STATUS_ROW_COUNT_INIT;
        pthread_mutex_unlock(&share->mutex);
    }
}

static int write_status(DB * status_block, char *buff, uint length) {
    DBUG_ENTER("write_status");
    DBT row, key;
    int error;
    const char *key_buff = "status";

    bzero((void *) &row, sizeof(row));
    bzero((void *) &key, sizeof(key));
    row.data = buff;
    key.data = (void *) key_buff;
    key.size = sizeof(key_buff);
    row.size = length;
    error = status_block->put(status_block, 0, &key, &row, 0);
    DBUG_RETURN(error);
}

static void update_status(TOKUDB_SHARE * share, TABLE * table) {
    DBUG_ENTER("update_status");
    if (share->rows != share->org_rows || (share->status & STATUS_TOKUDB_ANALYZE)) {
        pthread_mutex_lock(&share->mutex);
        if (!share->status_block) {
            /*
               Create sub database 'status' if it doesn't exist from before
               (This '*should*' always exist for table created with MySQL)
             */

            char name_buff[FN_REFLEN];
1194 1195 1196
            char newname[get_name_length(share->table_name) + 32];
            make_name(newname, share->table_name, "status");
            fn_format(name_buff, newname, "", 0, MY_UNPACK_FILENAME);
1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219
            if (db_create(&share->status_block, db_env, 0))
                goto end;
            share->status_block->set_flags(share->status_block, 0);
            if (share->status_block->open(share->status_block, NULL, name_buff, "status", DB_BTREE, DB_THREAD | DB_CREATE, my_umask))
                goto end;
        }
        {
            char rec_buff[4 + MAX_KEY * 4], *pos = rec_buff;
            int4store(pos, share->rows);
            pos += 4;
            for (uint i = 0; i < table->s->keys; i++) {
                int4store(pos, share->rec_per_key[i]);
                pos += 4;
            }
            DBUG_PRINT("info", ("updating status for %s", share->table_name));
            (void) write_status(share->status_block, rec_buff, (uint) (pos - rec_buff));
            share->status &= ~STATUS_TOKUDB_ANALYZE;
            share->org_rows = share->rows;
        }
      end:
        pthread_mutex_unlock(&share->mutex);
    }
    DBUG_VOID_RETURN;
1220 1221 1222 1223 1224 1225
}

/** @brief
    Return an estimated of the number of rows in the table.
    Used when sorting to allocate buffers and by the optimizer.
*/
1226 1227 1228
ha_rows ha_tokudb::estimate_rows_upper_bound() {
    return share->rows + HA_TOKUDB_EXTRA_ROWS;
}
1229

1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249
int ha_tokudb::cmp_ref(const uchar * ref1, const uchar * ref2) {
    if (hidden_primary_key)
        return memcmp(ref1, ref2, TOKUDB_HIDDEN_PRIMARY_KEY_LENGTH);

    int result;
    Field *field;
    KEY *key_info = table->key_info + table_share->primary_key;
    KEY_PART_INFO *key_part = key_info->key_part;
    KEY_PART_INFO *end = key_part + key_info->key_parts;

    for (; key_part != end; key_part++) {
        field = key_part->field;
        result = field->pack_cmp((const uchar *) ref1, (const uchar *) ref2, key_part->length, 0);
        if (result)
            return result;
        ref1 += field->packed_col_length((const uchar *) ref1, key_part->length);
        ref2 += field->packed_col_length((const uchar *) ref2, key_part->length);
    }

    return 0;
1250 1251
}

1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270
bool ha_tokudb::check_if_incompatible_data(HA_CREATE_INFO * info, uint table_changes) {
    if (table_changes < IS_EQUAL_YES)
        return COMPATIBLE_DATA_NO;
    return COMPATIBLE_DATA_YES;
}

int ha_tokudb::write_row(uchar * record) {
    DBUG_ENTER("write_row");
    DBT row, prim_key, key;
    int error;

    statistic_increment(table->in_use->status_var.ha_write_count, &LOCK_status);
    if (table->timestamp_field_type & TIMESTAMP_AUTO_SET_ON_INSERT)
        table->timestamp_field->set_time();
    if (table->next_number_field && record == table->record[0])
        update_auto_increment();
    if ((error = pack_row(&row, (const uchar *) record, 1)))
        DBUG_RETURN(error);

1271 1272
    u_int32_t put_flags = key_type[primary_key];
    THD *thd = ha_thd();
1273
    if (thd_test_options(thd, OPTION_RELAXED_UNIQUE_CHECKS)) {
1274
        put_flags = DB_YESOVERWRITE;
1275
    }
1276

1277
    if (table_share->keys + test(hidden_primary_key) == 1) {
1278
        error = file->put(file, transaction, create_key(&prim_key, primary_key, key_buff, record), &row, put_flags);
1279 1280 1281
        last_dup_key = primary_key;
    } else {
        DB_TXN *sub_trans = transaction;
1282
        /* QQQ Don't use sub transactions in temporary tables */
1283 1284
        for (uint retry = 0; retry < tokudb_trans_retry; retry++) {
            key_map changed_keys(0);
1285
            if (!(error = file->put(file, sub_trans, create_key(&prim_key, primary_key, key_buff, record), &row, put_flags))) {
1286 1287 1288 1289
                changed_keys.set_bit(primary_key);
                for (uint keynr = 0; keynr < table_share->keys; keynr++) {
                    if (keynr == primary_key)
                        continue;
1290 1291 1292 1293
                    put_flags = key_type[keynr];
                    if (put_flags == DB_NOOVERWRITE && thd_test_options(thd, OPTION_RELAXED_UNIQUE_CHECKS))
                        put_flags = DB_YESOVERWRITE;
                    if ((error = key_file[keynr]->put(key_file[keynr], sub_trans, create_key(&key, keynr, key_buff2, record), &prim_key, put_flags))) {
1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323
                        last_dup_key = keynr;
                        break;
                    }
                    changed_keys.set_bit(keynr);
                }
            } else
                last_dup_key = primary_key;
            if (error) {
                /* Remove inserted row */
                DBUG_PRINT("error", ("Got error %d", error));
                if (using_ignore) {
                    int new_error = 0;
                    if (!changed_keys.is_clear_all()) {
                        new_error = 0;
                        for (uint keynr = 0; keynr < table_share->keys + test(hidden_primary_key); keynr++) {
                            if (changed_keys.is_set(keynr)) {
                                if ((new_error = remove_key(sub_trans, keynr, record, &prim_key)))
                                    break;
                            }
                        }
                    }
                    if (new_error) {
                        error = new_error;      // This shouldn't happen
                        break;
                    }
                }
            }
            if (error != DB_LOCK_DEADLOCK)
                break;
        }
1324
    }
1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349
    if (error == DB_KEYEXIST)
        error = HA_ERR_FOUND_DUPP_KEY;
    else if (!error)
        changed_rows++;
    DBUG_RETURN(error);
}

/* Compare if a key in a row has changed */
int ha_tokudb::key_cmp(uint keynr, const uchar * old_row, const uchar * new_row) {
    KEY_PART_INFO *key_part = table->key_info[keynr].key_part;
    KEY_PART_INFO *end = key_part + table->key_info[keynr].key_parts;

    for (; key_part != end; key_part++) {
        if (key_part->null_bit) {
            if ((old_row[key_part->null_offset] & key_part->null_bit) != (new_row[key_part->null_offset] & key_part->null_bit))
                return 1;
        }
        if (key_part->key_part_flag & (HA_BLOB_PART | HA_VAR_LENGTH_PART)) {

            if (key_part->field->cmp_binary((uchar *) (old_row + key_part->offset), (uchar *) (new_row + key_part->offset), (ulong) key_part->length))
                return 1;
        } else {
            if (memcmp(old_row + key_part->offset, new_row + key_part->offset, key_part->length))
                return 1;
        }
1350
    }
1351
    return 0;
1352 1353 1354 1355 1356 1357 1358
}


/*
  Update a row from one value to another.
  Clobbers key_buff2
*/
1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385
int ha_tokudb::update_primary_key(DB_TXN * trans, bool primary_key_changed, const uchar * old_row, DBT * old_key, const uchar * new_row, DBT * new_key, bool local_using_ignore) {
    DBUG_ENTER("update_primary_key");
    DBT row;
    int error;

    if (primary_key_changed) {
        // Primary key changed or we are updating a key that can have duplicates.
        // Delete the old row and add a new one
        if (!(error = remove_key(trans, primary_key, old_row, old_key))) {
            if (!(error = pack_row(&row, new_row, 0))) {
                if ((error = file->put(file, trans, new_key, &row, key_type[primary_key]))) {
                    // Probably a duplicated key; restore old key and row if needed
                    last_dup_key = primary_key;
                    if (local_using_ignore) {
                        int new_error;
                        if ((new_error = pack_row(&row, old_row, 0)) || (new_error = file->put(file, trans, old_key, &row, key_type[primary_key])))
                            error = new_error;  // fatal error
                    }
                }
            }
        }
    } else {
        // Primary key didn't change;  just update the row data
        if (!(error = pack_row(&row, new_row, 0)))
            error = file->put(file, trans, new_key, &row, 0);
    }
    DBUG_RETURN(error);
1386 1387 1388 1389 1390 1391 1392
}

/*
  Restore changed keys, when a non-fatal error aborts the insert/update
  of one row.
  Clobbers keybuff2
*/
1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417
int ha_tokudb::restore_keys(DB_TXN * trans, key_map * changed_keys, uint primary_key, const uchar * old_row, DBT * old_key, const uchar * new_row, DBT * new_key) {
    DBUG_ENTER("restore_keys");
    int error;
    DBT tmp_key;
    uint keynr;

    /* Restore the old primary key, and the old row, but don't ignore
       duplicate key failure */
    if ((error = update_primary_key(trans, TRUE, new_row, new_key, old_row, old_key, FALSE)))
        goto err;

    /* Remove the new key, and put back the old key
       changed_keys is a map of all non-primary keys that need to be
       rolled back.  The last key set in changed_keys is the one that
       triggered the duplicate key error (it wasn't inserted), so for
       that one just put back the old value. */
    if (!changed_keys->is_clear_all()) {
        for (keynr = 0; keynr < table_share->keys + test(hidden_primary_key); keynr++) {
            if (changed_keys->is_set(keynr)) {
                if (changed_keys->is_prefix(1) && (error = remove_key(trans, keynr, new_row, new_key)))
                    break;
                if ((error = key_file[keynr]->put(key_file[keynr], trans, create_key(&tmp_key, keynr, key_buff2, old_row), old_key, key_type[keynr])))
                    break;
            }
        }
1418
    }
1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492

  err:
    DBUG_ASSERT(error != DB_KEYEXIST);
    DBUG_RETURN(error);
}

int ha_tokudb::update_row(const uchar * old_row, uchar * new_row) {
    DBUG_ENTER("update_row");
    DBT prim_key, key, old_prim_key;
    int error;
    DB_TXN *sub_trans;
    bool primary_key_changed;

    LINT_INIT(error);
    statistic_increment(table->in_use->status_var.ha_update_count, &LOCK_status);
    if (table->timestamp_field_type & TIMESTAMP_AUTO_SET_ON_UPDATE)
        table->timestamp_field->set_time();

    if (hidden_primary_key) {
        primary_key_changed = 0;
        bzero((void *) &prim_key, sizeof(prim_key));
        prim_key.data = (void *) current_ident;
        prim_key.size = TOKUDB_HIDDEN_PRIMARY_KEY_LENGTH;
        old_prim_key = prim_key;
    } else {
        create_key(&prim_key, primary_key, key_buff, new_row);

        if ((primary_key_changed = key_cmp(primary_key, old_row, new_row)))
            create_key(&old_prim_key, primary_key, primary_key_buff, old_row);
        else
            old_prim_key = prim_key;
    }

    sub_trans = transaction;
    for (uint retry = 0; retry < tokudb_trans_retry; retry++) {
        key_map changed_keys(0);
        /* Start by updating the primary key */
        if (!(error = update_primary_key(sub_trans, primary_key_changed, old_row, &old_prim_key, new_row, &prim_key, using_ignore))) {
            // Update all other keys
            for (uint keynr = 0; keynr < table_share->keys; keynr++) {
                if (keynr == primary_key)
                    continue;
                if (key_cmp(keynr, old_row, new_row) || primary_key_changed) {
                    if ((error = remove_key(sub_trans, keynr, old_row, &old_prim_key))) {
                        DBUG_RETURN(error);     // Fatal error
                    }
                    changed_keys.set_bit(keynr);
                    if ((error = key_file[keynr]->put(key_file[keynr], sub_trans, create_key(&key, keynr, key_buff2, new_row), &prim_key, key_type[keynr]))) {
                        last_dup_key = keynr;
                        break;
                    }
                }
            }
        }
        if (error) {
            /* Remove inserted row */
            DBUG_PRINT("error", ("Got error %d", error));
            if (using_ignore) {
                int new_error = 0;
                if (!changed_keys.is_clear_all())
                    new_error = restore_keys(transaction, &changed_keys, primary_key, old_row, &old_prim_key, new_row, &prim_key);
                if (new_error) {
                    /* This shouldn't happen */
                    error = new_error;
                    break;
                }
            }
        }
        if (error != DB_LOCK_DEADLOCK)
            break;
    }
    if (error == DB_KEYEXIST)
        error = HA_ERR_FOUND_DUPP_KEY;
    DBUG_RETURN(error);
1493 1494 1495 1496 1497 1498 1499 1500
}

/*
  Delete one key
  This uses key_buff2, when keynr != primary key, so it's important that
  a function that calls this doesn't use this buffer for anything else.
*/

1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512
int ha_tokudb::remove_key(DB_TXN * trans, uint keynr, const uchar * record, DBT * prim_key) {
    DBUG_ENTER("remove_key");
    int error;
    DBT key;
    DBUG_PRINT("enter", ("index: %d", keynr));

    if (keynr == active_index && cursor)
        error = cursor->c_del(cursor, 0);
    else if (keynr == primary_key || ((table->key_info[keynr].flags & (HA_NOSAME | HA_NULL_PART_KEY)) == HA_NOSAME)) {  // Unique key
        DBUG_ASSERT(keynr == primary_key || prim_key->data != key_buff2);
        error = key_file[keynr]->del(key_file[keynr], trans, keynr == primary_key ? prim_key : create_key(&key, keynr, key_buff2, record), 0);
    } else {
1513
        /* QQQ use toku_db_delboth(key_file[keynr], key, val, trans);
1514 1515 1516 1517 1518 1519 1520
           To delete the not duplicated key, we need to open an cursor on the
           row to find the key to be delete and delete it.
           We will never come here with keynr = primary_key
         */
        DBUG_ASSERT(keynr != primary_key && prim_key->data != key_buff2);
        DBC *tmp_cursor;
        if (!(error = key_file[keynr]->cursor(key_file[keynr], trans, &tmp_cursor, 0))) {
1521
            if (!(error = tmp_cursor->c_get(tmp_cursor, create_key(&key, keynr, key_buff2, record), prim_key, DB_GET_BOTH))) { 
1522 1523 1524 1525 1526 1527 1528 1529
                error = tmp_cursor->c_del(tmp_cursor, 0);
            }
            int result = tmp_cursor->c_close(tmp_cursor);
            if (!error)
                error = result;
        }
    }
    DBUG_RETURN(error);
1530 1531 1532
}

/* Delete all keys for new_record */
1533 1534 1535 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
int ha_tokudb::remove_keys(DB_TXN * trans, const uchar * record, DBT * new_record, DBT * prim_key, key_map * keys) {
    int result = 0;
    for (uint keynr = 0; keynr < table_share->keys + test(hidden_primary_key); keynr++) {
        if (keys->is_set(keynr)) {
            int new_error = remove_key(trans, keynr, record, prim_key);
            if (new_error) {
                result = new_error;     // Return last error
                break;          // Let rollback correct things
            }
        }
    }
    return result;
}

int ha_tokudb::delete_row(const uchar * record) {
    DBUG_ENTER("delete_row");
    int error;
    DBT row, prim_key;
    key_map keys = table_share->keys_in_use;
    statistic_increment(table->in_use->status_var.ha_delete_count, &LOCK_status);

    if ((error = pack_row(&row, record, 0)))
        DBUG_RETURN((error));
    create_key(&prim_key, primary_key, key_buff, record);
    if (hidden_primary_key)
        keys.set_bit(primary_key);

    /* Subtransactions may be used in order to retry the delete in
       case we get a DB_LOCK_DEADLOCK error. */
    DB_TXN *sub_trans = transaction;
    for (uint retry = 0; retry < tokudb_trans_retry; retry++) {
        error = remove_keys(sub_trans, record, &row, &prim_key, &keys);
        if (error) {
            DBUG_PRINT("error", ("Got error %d", error));
            break;              // No retry - return error
        }
        if (error != DB_LOCK_DEADLOCK)
            break;
1571 1572
    }
#ifdef CANT_COUNT_DELETED_ROWS
1573 1574
    if (!error)
        changed_rows--;
1575
#endif
1576
    DBUG_RETURN(error);
1577 1578
}

1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597
int ha_tokudb::index_init(uint keynr, bool sorted) {
    DBUG_ENTER("ha_tokudb::index_init");
    int error;
    DBUG_PRINT("enter", ("table: '%s'  key: %d", table_share->table_name.str, keynr));

    /*
       Under some very rare conditions (like full joins) we may already have
       an active cursor at this point
     */
    if (cursor) {
        DBUG_PRINT("note", ("Closing active cursor"));
        cursor->c_close(cursor);
    }
    active_index = keynr;
    DBUG_ASSERT(keynr <= table->s->keys);
    DBUG_ASSERT(key_file[keynr]);
    if ((error = key_file[keynr]->cursor(key_file[keynr], transaction, &cursor, table->reginfo.lock_type > TL_WRITE_ALLOW_READ ? 0 : 0)))
        cursor = 0;             // Safety
    bzero((void *) &last_key, sizeof(last_key));
1598
    DBUG_RETURN(error);
1599 1600 1601 1602 1603 1604 1605 1606 1607
}

int ha_tokudb::index_end() {
    DBUG_ENTER("ha_tokudb::index_end");
    int error = 0;
    if (cursor) {
        DBUG_PRINT("enter", ("table: '%s'", table_share->table_name.str));
        error = cursor->c_close(cursor);
        cursor = 0;
1608
    }
1609 1610 1611 1612 1613 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 1646 1647
    active_index = MAX_KEY;
    DBUG_RETURN(error);
}

/* What to do after we have read a row based on an index */
int ha_tokudb::read_row(int error, uchar * buf, uint keynr, DBT * row, DBT * found_key, bool read_next) {
    DBUG_ENTER("ha_tokudb::read_row");
    if (error) {
        if (error == DB_NOTFOUND || error == DB_KEYEMPTY)
            error = read_next ? HA_ERR_END_OF_FILE : HA_ERR_KEY_NOT_FOUND;
        table->status = STATUS_NOT_FOUND;
        DBUG_RETURN(error);
    }
    if (hidden_primary_key)
        memcpy_fixed(current_ident, (char *) row->data + row->size - TOKUDB_HIDDEN_PRIMARY_KEY_LENGTH, TOKUDB_HIDDEN_PRIMARY_KEY_LENGTH);
    table->status = 0;
    if (keynr != primary_key) {
        /* We only found the primary key.  Now we have to use this to find the row data */
        if (key_read && found_key) {
            unpack_key(buf, found_key, keynr);
            if (!hidden_primary_key)
                unpack_key(buf, row, primary_key);
            DBUG_RETURN(0);
        }
        DBT key;
        bzero((void *) &key, sizeof(key));
        key.data = key_buff;
        key.size = row->size;
        memcpy(key_buff, row->data, row->size);
        /* Read the data into current_row */
        current_row.flags = DB_DBT_REALLOC;
        if ((error = file->get(file, transaction, &key, &current_row, 0))) {
            table->status = STATUS_NOT_FOUND;
            DBUG_RETURN(error == DB_NOTFOUND ? HA_ERR_CRASHED : error);
        }
        row = &current_row;
    }
    unpack_row(buf, row);
    DBUG_RETURN(0);
1648 1649 1650 1651 1652
}

/*
  This is only used to read whole keys
*/
1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667
int ha_tokudb::index_read_idx(uchar * buf, uint keynr, const uchar * key, uint key_len, enum ha_rkey_function find_flag) {
    DBUG_ENTER("ha_tokudb::index_read_idx");
    table->in_use->status_var.ha_read_key_count++;
    current_row.flags = DB_DBT_REALLOC;
    active_index = MAX_KEY;
    DBUG_RETURN(read_row(key_file[keynr]->get(key_file[keynr], transaction, pack_key(&last_key, keynr, key_buff, key, key_len), &current_row, 0), buf, keynr, &current_row, &last_key, 0));
}

int ha_tokudb::index_read(uchar * buf, const uchar * key, uint key_len, enum ha_rkey_function find_flag) {
    DBUG_ENTER("ha_tokudb::index_read");
    DBT row;
    int error;

    table->in_use->status_var.ha_read_key_count++;
    bzero((void *) &row, sizeof(row));
1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679
    pack_key(&last_key, active_index, key_buff, key, key_len);

    switch (find_flag) {
    case HA_READ_KEY_EXACT:
        error = cursor->c_get(cursor, &last_key, &row, DB_SET);
        error = read_row(error, buf, active_index, &row, 0, 0);
        break;
    case HA_READ_AFTER_KEY:
        error = cursor->c_get(cursor, &last_key, &row, DB_SET_RANGE);
        if (error == 0) {
            DBT orig_key;
            pack_key(&orig_key, active_index, key_buff2, key, key_len);
1680
            if (tokudb_cmp_packed_key(share->key_file[active_index], &orig_key, &last_key) == 0)
1681
                error = cursor->c_get(cursor, &last_key, &row, DB_NEXT_NODUP);
1682
        }
1683 1684
        error = read_row(error, buf, active_index, &row, 0, 0);
        break;
1685 1686 1687 1688 1689 1690 1691 1692
    case HA_READ_BEFORE_KEY:
        error = cursor->c_get(cursor, &last_key, &row, DB_SET_RANGE);
        if (error == 0)
            error = cursor->c_get(cursor, &last_key, &row, DB_PREV);
        else
            error = cursor->c_get(cursor, &last_key, &row, DB_LAST);
        assert(0);
        break;
1693 1694 1695 1696
    case HA_READ_KEY_OR_NEXT:
        error = cursor->c_get(cursor, &last_key, &row, DB_SET_RANGE);
        error = read_row(error, buf, active_index, &row, 0, 0);
        break;
1697
    case HA_READ_KEY_OR_PREV:
1698
        error = cursor->c_get(cursor, &last_key, &row, DB_SET_RANGE);
1699 1700 1701 1702 1703 1704 1705 1706
        if (error == 0) {
            DBT orig_key; 
            pack_key(&orig_key, active_index, key_buff2, key, key_len);
            if (tokudb_cmp_packed_key(share->key_file[active_index], &orig_key, &last_key) != 0)
                error = cursor->c_get(cursor, &last_key, &row, DB_PREV_NODUP);
        } else 
            error = cursor->c_get(cursor, &last_key, &row, DB_LAST);
        assert(0);
1707 1708
        break;
    default:
1709
        assert(0); // QQQ others?
1710
        break;
1711
    }
1712

1713
    DBUG_RETURN(error);
1714 1715 1716 1717 1718 1719
}

/*
  Read last key is solved by reading the next key and then reading
  the previous key
*/
1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732
int ha_tokudb::index_read_last(uchar * buf, const uchar * key, uint key_len) {
    DBUG_ENTER("ha_tokudb::index_read_last");
    DBT row;
    int error;
    KEY *key_info = &table->key_info[active_index];

    statistic_increment(table->in_use->status_var.ha_read_key_count, &LOCK_status);
    bzero((void *) &row, sizeof(row));

    /* read of partial key */
    pack_key(&last_key, active_index, key_buff, key, key_len);
    /* Store for compare */
    memcpy(key_buff2, key_buff, (key_len = last_key.size));
1733
    assert(0);
1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 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 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835
    key_info->handler.bdb_return_if_eq = 1;
    error = read_row(cursor->c_get(cursor, &last_key, &row, DB_SET_RANGE), buf, active_index, &row, (DBT *) 0, 0);
    key_info->handler.bdb_return_if_eq = 0;
    bzero((void *) &row, sizeof(row));
    if (read_row(cursor->c_get(cursor, &last_key, &row, DB_PREV), buf, active_index, &row, &last_key, 1) || tokudb_key_cmp(table, key_info, key_buff2, key_len))
        error = HA_ERR_KEY_NOT_FOUND;
    DBUG_RETURN(error);
}


int ha_tokudb::index_next(uchar * buf) {
    DBUG_ENTER("ha_tokudb::index_next");
    DBT row;
    statistic_increment(table->in_use->status_var.ha_read_next_count, &LOCK_status);
    bzero((void *) &row, sizeof(row));
    DBUG_RETURN(read_row(cursor->c_get(cursor, &last_key, &row, DB_NEXT), buf, active_index, &row, &last_key, 1));
}

int ha_tokudb::index_next_same(uchar * buf, const uchar * key, uint keylen) {
    DBUG_ENTER("ha_tokudb::index_next_same");
    DBT row;
    int error;
    statistic_increment(table->in_use->status_var.ha_read_next_count, &LOCK_status);
    bzero((void *) &row, sizeof(row));
    if (keylen == table->key_info[active_index].key_length && !(table->key_info[active_index].flags & HA_END_SPACE_KEY))
        error = read_row(cursor->c_get(cursor, &last_key, &row, DB_NEXT_DUP), buf, active_index, &row, &last_key, 1);
    else {
        error = read_row(cursor->c_get(cursor, &last_key, &row, DB_NEXT), buf, active_index, &row, &last_key, 1);
        if (!error &&::key_cmp_if_same(table, key, active_index, keylen))
            error = HA_ERR_END_OF_FILE;
    }
    DBUG_RETURN(error);
}

int ha_tokudb::index_prev(uchar * buf) {
    DBUG_ENTER("ha_tokudb::index_prev");
    DBT row;
    statistic_increment(table->in_use->status_var.ha_read_prev_count, &LOCK_status);
    bzero((void *) &row, sizeof(row));
    DBUG_RETURN(read_row(cursor->c_get(cursor, &last_key, &row, DB_PREV), buf, active_index, &row, &last_key, 1));
}

int ha_tokudb::index_first(uchar * buf) {
    DBUG_ENTER("ha_tokudb::index_first");
    DBT row;
    statistic_increment(table->in_use->status_var.ha_read_first_count, &LOCK_status);
    bzero((void *) &row, sizeof(row));
    DBUG_RETURN(read_row(cursor->c_get(cursor, &last_key, &row, DB_FIRST), buf, active_index, &row, &last_key, 1));
}

int ha_tokudb::index_last(uchar * buf) {
    DBUG_ENTER("ha_tokudb::index_last");
    DBT row;
    statistic_increment(table->in_use->status_var.ha_read_last_count, &LOCK_status);
    bzero((void *) &row, sizeof(row));
    DBUG_RETURN(read_row(cursor->c_get(cursor, &last_key, &row, DB_LAST), buf, active_index, &row, &last_key, 0));
}

int ha_tokudb::rnd_init(bool scan) {
    DBUG_ENTER("ha_tokudb::rnd_init");
    current_row.flags = DB_DBT_REALLOC;
    DBUG_RETURN(index_init(primary_key, 0));
}

int ha_tokudb::rnd_end() {
    return index_end();
}

int ha_tokudb::rnd_next(uchar * buf) {
    DBUG_ENTER("ha_tokudb::ha_tokudb::rnd_next");
    DBT row;
    statistic_increment(table->in_use->status_var.ha_read_rnd_next_count, &LOCK_status);
    bzero((void *) &row, sizeof(row));
    DBUG_RETURN(read_row(cursor->c_get(cursor, &last_key, &row, DB_NEXT), buf, primary_key, &row, &last_key, 1));
}


DBT *ha_tokudb::get_pos(DBT * to, uchar * pos) {
    /* We don't need to set app_data here */
    bzero((void *) to, sizeof(*to));

    to->data = pos;
    if (share->fixed_length_primary_key)
        to->size = ref_length;
    else {
        KEY_PART_INFO *key_part = table->key_info[primary_key].key_part;
        KEY_PART_INFO *end = key_part + table->key_info[primary_key].key_parts;

        for (; key_part != end; key_part++)
            pos += key_part->field->packed_col_length(pos, key_part->length);
        to->size = (uint) (pos - (uchar *) to->data);
    }
    DBUG_DUMP("key", (const uchar *) to->data, to->size);
    return to;
}

int ha_tokudb::rnd_pos(uchar * buf, uchar * pos) {
    DBUG_ENTER("ha_tokudb::rnd_pos");
    DBT db_pos;
    statistic_increment(table->in_use->status_var.ha_read_rnd_count, &LOCK_status);
    active_index = MAX_KEY;
    DBUG_RETURN(read_row(file->get(file, transaction, get_pos(&db_pos, pos), &current_row, 0), buf, primary_key, &current_row, (DBT *) 0, 0));
1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859
}

/*
  Set a reference to the current record in (ref,ref_length).

  SYNOPSIS
  ha_tokudb::position()
  record                      The current record buffer

  DESCRIPTION
  The BDB handler stores the primary key in (ref,ref_length).
  There is either an explicit primary key, or an implicit (hidden)
  primary key.
  During open(), 'ref_length' is calculated as the maximum primary
  key length. When an actual key is shorter than that, the rest of
  the buffer must be cleared out. The row cannot be identified, if
  garbage follows behind the end of the key. There is no length
  field for the current key, so that the whole ref_length is used
  for comparison.

  RETURN
  nothing
*/

1860 1861 1862 1863 1864 1865 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
void ha_tokudb::position(const uchar * record) {
    DBUG_ENTER("ha_tokudb::position");
    DBT key;
    if (hidden_primary_key) {
        DBUG_ASSERT(ref_length == TOKUDB_HIDDEN_PRIMARY_KEY_LENGTH);
        memcpy_fixed(ref, (char *) current_ident, TOKUDB_HIDDEN_PRIMARY_KEY_LENGTH);
    } else {
        create_key(&key, primary_key, ref, record);
        if (key.size < ref_length)
            bzero(ref + key.size, ref_length - key.size);
    }
    DBUG_VOID_RETURN;
}

int ha_tokudb::info(uint flag) {
    DBUG_ENTER("ha_tokudb::info");
    if (flag & HA_STATUS_VARIABLE) {
        // Just to get optimizations right
        stats.records = share->rows + changed_rows;
        stats.deleted = 0;
    }
    if ((flag & HA_STATUS_CONST) || version != share->version) {
        version = share->version;
        for (uint i = 0; i < table_share->keys; i++) {
            table->key_info[i].rec_per_key[table->key_info[i].key_parts - 1] = share->rec_per_key[i];
        }
    }
    /* Don't return key if we got an error for the internal primary key */
    if (flag & HA_STATUS_ERRKEY && last_dup_key < table_share->keys)
        errkey = last_dup_key;
    DBUG_RETURN(0);
}


int ha_tokudb::extra(enum ha_extra_function operation) {
    switch (operation) {
    case HA_EXTRA_RESET_STATE:
        reset();
        break;
    case HA_EXTRA_KEYREAD:
        key_read = 1;           // Query satisfied with key
        break;
    case HA_EXTRA_NO_KEYREAD:
        key_read = 0;
        break;
    case HA_EXTRA_IGNORE_DUP_KEY:
        using_ignore = 1;
        break;
    case HA_EXTRA_NO_IGNORE_DUP_KEY:
        using_ignore = 0;
        break;
    default:
        break;
    }
    return 0;
}

int ha_tokudb::reset(void) {
    key_read = 0;
    using_ignore = 0;
    if (current_row.flags & (DB_DBT_MALLOC | DB_DBT_REALLOC)) {
        current_row.flags = 0;
        if (current_row.data) {
            free(current_row.data);
            current_row.data = 0;
        }
    }
    return 0;
1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938
}

/*
  As MySQL will execute an external lock for every new table it uses
  we can use this to start the transactions.
  If we are in auto_commit mode we just need to start a transaction
  for the statement to be able to rollback the statement.
  If not, we have to start a master transaction if there doesn't exist
  one from before.
*/

1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956
int ha_tokudb::external_lock(THD * thd, int lock_type) {
    DBUG_ENTER("ha_tokudb::external_lock");
    int error = 0;
    tokudb_trx_data *trx = (tokudb_trx_data *) thd->ha_data[tokudb_hton->slot];
    if (!trx) {
        thd->ha_data[tokudb_hton->slot] = trx = (tokudb_trx_data *)
            my_malloc(sizeof(*trx), MYF(MY_ZEROFILL));
        if (!trx)
            DBUG_RETURN(1);
    }
    if (trx->all == 0)
        trx->sp_level = 0;
    if (lock_type != F_UNLCK) {
        if (!trx->tokudb_lock_count++) {
            DBUG_ASSERT(trx->stmt == 0);
            transaction = 0;    // Safety
            /* First table lock, start transaction */
            if ((thd->options & (OPTION_NOT_AUTOCOMMIT | OPTION_BEGIN | OPTION_TABLE_LOCK)) && !trx->all) {
1957
                /* QQQ We have to start a master transaction */
1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968
                DBUG_PRINT("trans", ("starting transaction all:  options: 0x%lx", (ulong) thd->options));
                if ((error = db_env->txn_begin(db_env, NULL, &trx->all, 0))) {
                    trx->tokudb_lock_count--;      // We didn't get the lock
                    DBUG_RETURN(error);
                }
                trx->sp_level = trx->all;
                trans_register_ha(thd, TRUE, tokudb_hton);
                if (thd->in_lock_tables)
                    DBUG_RETURN(0);     // Don't create stmt trans
            }
            DBUG_PRINT("trans", ("starting transaction stmt"));
1969
	    printf("%s:%d:warning:stmt=%p\n", __FILE__, __LINE__, trx->stmt);
1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991
            if ((error = db_env->txn_begin(db_env, trx->sp_level, &trx->stmt, 0))) {
                /* We leave the possible master transaction open */
                trx->tokudb_lock_count--;  // We didn't get the lock
                DBUG_RETURN(error);
            }
            trans_register_ha(thd, FALSE, tokudb_hton);
        }
        transaction = trx->stmt;
    } else {
        lock.type = TL_UNLOCK;  // Unlocked
        thread_safe_add(share->rows, changed_rows, &share->mutex);
        changed_rows = 0;
        if (!--trx->tokudb_lock_count) {
            if (trx->stmt) {
                /*
                   F_UNLCK is done without a transaction commit / rollback.
                   This happens if the thread didn't update any rows
                   We must in this case commit the work to keep the row locks
                 */
                DBUG_PRINT("trans", ("commiting non-updating transaction"));
                error = trx->stmt->commit(trx->stmt, 0);
                trx->stmt = transaction = 0;
1992 1993
            }
        }
1994 1995
    }
    DBUG_RETURN(error);
1996 1997 1998 1999 2000 2001 2002 2003 2004
}


/*
  When using LOCK TABLE's external_lock is only called when the actual
  TABLE LOCK is done.
  Under LOCK TABLES, each used tables will force a call to start_stmt.
*/

2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021
int ha_tokudb::start_stmt(THD * thd, thr_lock_type lock_type) {
    DBUG_ENTER("ha_tokudb::start_stmt");
    int error = 0;
    tokudb_trx_data *trx = (tokudb_trx_data *) thd->ha_data[tokudb_hton->slot];
    DBUG_ASSERT(trx);
    /*
       note that trx->stmt may have been already initialized as start_stmt()
       is called for *each table* not for each storage engine,
       and there could be many bdb tables referenced in the query
     */
    if (!trx->stmt) {
        DBUG_PRINT("trans", ("starting transaction stmt"));
        error = db_env->txn_begin(db_env, trx->sp_level, &trx->stmt, 0);
        trans_register_ha(thd, FALSE, tokudb_hton);
    }
    transaction = trx->stmt;
    DBUG_RETURN(error);
2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051
}

/*
  The idea with handler::store_lock() is the following:

  The statement decided which locks we should need for the table
  for updates/deletes/inserts we get WRITE locks, for SELECT... we get
  read locks.

  Before adding the lock into the table lock handler (see thr_lock.c)
  mysqld calls store lock with the requested locks.  Store lock can now
  modify a write lock to a read lock (or some other lock), ignore the
  lock (if we don't want to use MySQL table locks at all) or add locks
  for many tables (like we do when we are using a MERGE handler).

  Tokudb DB changes all WRITE locks to TL_WRITE_ALLOW_WRITE (which
  signals that we are doing WRITES, but we are still allowing other
  reader's and writer's.

  When releasing locks, store_lock() are also called. In this case one
  usually doesn't have to do anything.

  In some exceptional cases MySQL may send a request for a TL_IGNORE;
  This means that we are requesting the same lock as last time and this
  should also be ignored. (This may happen when someone does a flush
  table when we have opened a part of the tables, in which case mysqld
  closes and reopens the tables and tries to get the same locks at last
  time).  In the future we will probably try to remove this.
*/

2052 2053 2054 2055 2056 2057
THR_LOCK_DATA **ha_tokudb::store_lock(THD * thd, THR_LOCK_DATA ** to, enum thr_lock_type lock_type) {
    if (lock_type != TL_IGNORE && lock.type == TL_UNLOCK) {
        /* If we are not doing a LOCK TABLE, then allow multiple writers */
        if ((lock_type >= TL_WRITE_CONCURRENT_INSERT && lock_type <= TL_WRITE) && !thd->in_lock_tables)
            lock_type = TL_WRITE_ALLOW_WRITE;
        lock.type = lock_type;
2058
    }
2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082
    *to++ = &lock;
    return to;
}


static int create_sub_table(const char *table_name, const char *sub_name, DBTYPE type, int flags) {
    DBUG_ENTER("create_sub_table");
    int error;
    DB *file;
    DBUG_PRINT("enter", ("sub_name: %s  flags: %d", sub_name, flags));

    if (!(error = db_create(&file, db_env, 0))) {
        file->set_flags(file, flags);
        error = (file->open(file, NULL, table_name, sub_name, type, DB_THREAD | DB_CREATE, my_umask));
        if (error) {
            DBUG_PRINT("error", ("Got error: %d when opening table '%s'", error, table_name));
            (void) file->remove(file, table_name, NULL, 0);
        } else
            (void) file->close(file, 0);
    } else {
        DBUG_PRINT("error", ("Got error: %d when creating table", error));
    }
    if (error)
        my_errno = error;
2083
    DBUG_RETURN(error);
2084 2085
}

2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102

static int mkdirpath(char *name, mode_t mode) {
    int r = mkdir(name, mode);
    if (r == -1 && errno == ENOENT) {
        char parent[strlen(name)+1];
        strcpy(parent, name);
        char *cp = strrchr(parent, '/');
        if (cp) {
            *cp = 0;
            r = mkdir(parent, 0755);
            if (r == 0)
                r = mkdir(name, mode);
        }
    }
    return r;
}

2103 2104 2105 2106
int ha_tokudb::create(const char *name, TABLE * form, HA_CREATE_INFO * create_info) {
    DBUG_ENTER("ha_tokudb::create");
    char name_buff[FN_REFLEN];
    int error;
2107
    char newname[get_name_length(name) + 32];
2108

2109
    // a table is a directory of dictionaries
2110 2111
    make_name(newname, name, 0);
    error = mkdirpath(newname, 0777);
2112 2113 2114
    if (error != 0) {
        DBUG_RETURN(errno);
    }
2115 2116 2117
    
    make_name(newname, name, "main");
    fn_format(name_buff, newname, "", 0, MY_UNPACK_FILENAME);
2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130

    /* Create the main table that will hold the real rows */
    if ((error = create_sub_table(name_buff, "main", DB_BTREE, 0)))
        DBUG_RETURN(error);

    primary_key = form->s->primary_key;

    /* Create the keys */
    char part[16];
    uint index = 1;
    for (uint i = 0; i < form->s->keys; i++) {
        if (i != primary_key) {
            sprintf(part, "key%d", index++);
2131 2132
            make_name(newname, name, part);
            fn_format(name_buff, newname, "", 0, MY_UNPACK_FILENAME);
2133 2134 2135 2136
            if ((error = create_sub_table(name_buff, part, DB_BTREE, (form->key_info[i].flags & HA_NOSAME) ? 0 : DB_DUP + DB_DUPSORT)))
                DBUG_RETURN(error);
        }
    }
2137

2138
    /* Create the status block to save information from last status command */
2139 2140
    /* QQQ Is DB_BTREE the best option here ? (QUEUE can't be used in sub tables) */
    // QQQ what is the status DB used for?
2141 2142 2143

    DB *status_block;
    if (!(error = (db_create(&status_block, db_env, 0)))) {
2144 2145
        make_name(newname, name, "status");
        fn_format(name_buff, newname, "", 0, MY_UNPACK_FILENAME);
2146 2147 2148 2149 2150 2151 2152 2153 2154 2155

        if (!(error = (status_block->open(status_block, NULL, name_buff, "status", DB_BTREE, DB_CREATE, 0)))) {
            char rec_buff[4 + MAX_KEY * 4];
            uint length = 4 + form->s->keys * 4;
            bzero(rec_buff, length);
            error = write_status(status_block, rec_buff, length);
            status_block->close(status_block, 0);
        }
    }
    DBUG_RETURN(error);
2156 2157
}

2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171
#include <dirent.h>

static int rmall(const char *dname) {
    int error = 0;
    DIR *d = opendir(dname);
    if (d) {
        struct dirent *dirent;
        while ((dirent = readdir(d)) != 0) {
            if (0 == strcmp(dirent->d_name, ".") || 0 == strcmp(dirent->d_name, ".."))
                continue;
            char fname[strlen(dname) + 1 + strlen(dirent->d_name) + 1];
            sprintf(fname, "%s/%s", dname, dirent->d_name);
            if (dirent->d_type == DT_DIR) {
                error = rmall(fname);
2172
            } else
2173
                error = unlink(fname);
2174 2175 2176 2177 2178 2179 2180
            if (error != 0) {
                error = errno;
                break;
            }
        }
        closedir(d);
        if (error == 0) {
2181
            error = rmdir(dname);
2182 2183 2184 2185 2186 2187 2188
            if (error != 0)
                error = errno;
        }
    } else
        error = errno;
    return error;
}
2189

2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213
int ha_tokudb::delete_table(const char *name) {
    DBUG_ENTER("ha_tokudb::delete_table");
    int error;
#if 0 // QQQ single file per table
    char name_buff[FN_REFLEN];
    char newname[strlen(name) + 32];

    sprintf(newname, "%s/main", name);
    fn_format(name_buff, newname, "", ha_tokudb_ext, MY_UNPACK_FILENAME | MY_APPEND_EXT);
    error = db_create(&file, db_env, 0);
    if (error != 0)
        goto exit;
    error = file->remove(file, name_buff, NULL, 0);

    sprintf(newname, "%s/status", name);
    fn_format(name_buff, newname, "", ha_tokudb_ext, MY_UNPACK_FILENAME | MY_APPEND_EXT);
    error = db_create(&file, db_env, 0);
    if (error != 0)
        goto exit;
    error = file->remove(file, name_buff, NULL, 0);

  exit:
    file = 0;                   // Safety
    my_errno = error;
2214
#else
2215
    // remove all of the dictionaries in the table directory 
2216 2217
    char newname[(tokudb_data_dir ? strlen(tokudb_data_dir) : 0) + strlen(name) + 32];
    make_name(newname, name, 0);
2218 2219
    error = rmall(newname);
    my_errno = error;
2220
#endif
2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237
    DBUG_RETURN(error);
}


int ha_tokudb::rename_table(const char *from, const char *to) {
    int error;
#if 0 // QQQ single file per table
    char from_buff[FN_REFLEN];
    char to_buff[FN_REFLEN];

    if ((error = db_create(&file, db_env, 0)))
        my_errno = error;
    else {
        /* On should not do a file->close() after rename returns */
        error = file->rename(file,
                             fn_format(from_buff, from, "", ha_tokudb_ext, MY_UNPACK_FILENAME | MY_APPEND_EXT), NULL, fn_format(to_buff, to, "", ha_tokudb_ext, MY_UNPACK_FILENAME | MY_APPEND_EXT), 0);
    }
2238
#else
2239 2240 2241 2242 2243 2244
    int n = get_name_length(from) + 32;
    char newfrom[n];
    make_name(newfrom, from, 0);
    n = get_name_length(to) + 32;
    char newto[n];
    make_name(newto, to, 0);
2245 2246 2247
    error = rename(newfrom, newto);
    if (error != 0)
        error = my_errno = errno;
2248
#endif
2249
    return error;
2250 2251 2252 2253 2254 2255 2256 2257
}


/*
  How many seeks it will take to read through the table
  This is to be comparable to the number returned by records_in_range so
  that we can decide if we should scan the table or use keys.
*/
2258
/// QQQ why divide by 3
2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273
double ha_tokudb::scan_time() {
    return rows2double(stats.records / 3);
}

ha_rows ha_tokudb::records_in_range(uint keynr, key_range * start_key, key_range * end_key) {
    DBUG_ENTER("ha_tokudb::records_in_range");
#if 0 // QQQ need key_range
    DBT key;
    DB_KEY_RANGE start_range, end_range;
    DB *kfile = key_file[keynr];
    double start_pos, end_pos, rows;
    bool error;
    KEY *key_info = &table->key_info[keynr];

    /* Ensure we get maximum range, even for varchar keys with different space */
2274
    assert(0);
2275 2276 2277 2278 2279 2280 2281
    key_info->handler.bdb_return_if_eq = -1;
    error = ((start_key && kfile->key_range(kfile, transaction, pack_key(&key, keynr, key_buff, start_key->key, start_key->length), &start_range, 0)));
    if (error) {
        key_info->handler.bdb_return_if_eq = 0;
        // Better than returning an error
        DBUG_RETURN(HA_TOKUDB_RANGE_COUNT);
    }
2282
    assert(0);
2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306
    key_info->handler.bdb_return_if_eq = 1;
    error = (end_key && kfile->key_range(kfile, transaction, pack_key(&key, keynr, key_buff, end_key->key, end_key->length), &end_range, 0));
    key_info->handler.bdb_return_if_eq = 0;
    if (error) {
        // Better than returning an error
        DBUG_RETURN(HA_TOKUDB_RANGE_COUNT);
    }

    if (!start_key)
        start_pos = 0.0;
    else if (start_key->flag == HA_READ_KEY_EXACT)
        start_pos = start_range.less;
    else
        start_pos = start_range.less + start_range.equal;

    if (!end_key)
        end_pos = 1.0;
    else if (end_key->flag == HA_READ_BEFORE_KEY)
        end_pos = end_range.less;
    else
        end_pos = end_range.less + end_range.equal;
    rows = (end_pos - start_pos) * stats.records;
    DBUG_PRINT("exit", ("rows: %g", rows));
    DBUG_RETURN((ha_rows) (rows <= 1.0 ? 1 : rows));
2307
#else
2308
    DBUG_RETURN(HA_TOKUDB_RANGE_COUNT);
2309
#endif
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 2340 2341 2342 2343 2344
void ha_tokudb::get_auto_increment(ulonglong offset, ulonglong increment, ulonglong nb_desired_values, ulonglong * first_value, ulonglong * nb_reserved_values) {
    /* Ideally in case of real error (not "empty table") nr should be ~ULL(0) */
    ulonglong nr = 1;           // Default if error or new key
    int error;
    (void) ha_tokudb::extra(HA_EXTRA_KEYREAD);

    /* Set 'active_index' */
    ha_tokudb::index_init(table_share->next_number_index, 0);

    if (!table_share->next_number_key_offset) { // Autoincrement at key-start
        error = ha_tokudb::index_last(table->record[1]);
        /* has taken read lock on page of max key so reserves to infinite  */
        *nb_reserved_values = ULONGLONG_MAX;
    } else {
        /*
           MySQL needs to call us for next row: assume we are inserting ("a",null)
           here, we return 3, and next this statement will want to insert ("b",null):
           there is no reason why ("b",3+1) would be the good row to insert: maybe it
           already exists, maybe 3+1 is too large...
         */
        *nb_reserved_values = 1;
        DBT row, old_key;
        bzero((void *) &row, sizeof(row));
        KEY *key_info = &table->key_info[active_index];

        /* Reading next available number for a sub key */
        ha_tokudb::create_key(&last_key, active_index, key_buff, table->record[0], table_share->next_number_key_offset);
        /* Store for compare */
        memcpy(old_key.data = key_buff2, key_buff, (old_key.size = last_key.size));
        error = 1;
        {
            /* Modify the compare so that we will find the next key */
2345
            assert(0);
2346
            key_info->handler.bdb_return_if_eq = 1;
2347
            /* QQQ We lock the next key as the new key will probl. be on the same page */
2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363
            error = cursor->c_get(cursor, &last_key, &row, DB_SET_RANGE | DB_RMW);
            key_info->handler.bdb_return_if_eq = 0;
            if (!error || error == DB_NOTFOUND) {
                /*
                   Now search go one step back and then we should have found the
                   biggest key with the given prefix
                 */
                error = 1;
                if (!cursor->c_get(cursor, &last_key, &row, DB_PREV | DB_RMW) && !tokudb_cmp_packed_key(key_file[active_index], &old_key, &last_key)) {
                    error = 0;  // Found value
                    unpack_key(table->record[1], &last_key, active_index);
                }
            }
        }
    }
    if (!error)
2364
        nr = (ulonglong) table->next_number_field->val_int_offset(0) + 1;
2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375
    ha_tokudb::index_end();
    (void) ha_tokudb::extra(HA_EXTRA_NO_KEYREAD);
    *first_value = nr;
}

void ha_tokudb::print_error(int error, myf errflag) {
    if (error == DB_LOCK_DEADLOCK)
        error = HA_ERR_LOCK_DEADLOCK;
    handler::print_error(error, errflag);
}

2376
#if 0 // QQQ use default
2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407
int ha_tokudb::analyze(THD * thd, HA_CHECK_OPT * check_opt) {
    uint i;
    DB_BTREE_STAT *stat = 0;
    DB_TXN_STAT *txn_stat_ptr = 0;
    tokudb_trx_data *trx = (tokudb_trx_data *) thd->ha_data[tokudb_hton->slot];
    DBUG_ASSERT(trx);

    for (i = 0; i < table_share->keys; i++) {
        if (stat) {
            free(stat);
            stat = 0;
        }
        if ((key_file[i]->stat) (key_file[i], trx->all, (void *) &stat, 0))
            goto err;
        share->rec_per_key[i] = (stat->bt_ndata / (stat->bt_nkeys ? stat->bt_nkeys : 1));
    }
    /* A hidden primary key is not in key_file[] */
    if (hidden_primary_key) {
        if (stat) {
            free(stat);
            stat = 0;
        }
        if ((file->stat) (file, trx->all, (void *) &stat, 0))
            goto err;
    }
    pthread_mutex_lock(&share->mutex);
    share->rows = stat->bt_ndata;
    share->status |= STATUS_TOKUDB_ANALYZE;        // Save status on close
    share->version++;           // Update stat in table
    pthread_mutex_unlock(&share->mutex);
    update_status(share, table);        // Write status to file
2408
    if (stat)
2409 2410 2411 2412
        free(stat);
    return ((share->status & STATUS_TOKUDB_ANALYZE) ? HA_ADMIN_FAILED : HA_ADMIN_OK);

  err:
2413
    if (stat)
2414 2415
        free(stat);
    return HA_ADMIN_FAILED;
2416
}
2417
#endif
2418

2419
#if 0 // QQQ use default
2420 2421
int ha_tokudb::optimize(THD * thd, HA_CHECK_OPT * check_opt) {
    return ha_tokudb::analyze(thd, check_opt);
2422
}
2423
#endif
2424

2425
#if 0 // QQQ use default
2426 2427 2428 2429
int ha_tokudb::check(THD * thd, HA_CHECK_OPT * check_opt) {
    DBUG_ENTER("ha_tokudb::check");
    DBUG_RETURN(HA_ADMIN_NOT_IMPLEMENTED);
    // look in old_ha_tokudb.cc for a start of an implementation
2430
}
2431
#endif
2432

2433
struct st_mysql_storage_engine storage_engine_structure = { MYSQL_HANDLERTON_INTERFACE_VERSION };
2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444

// options flags
//   PLUGIN_VAR_THDLOCAL  Variable is per-connection
//   PLUGIN_VAR_READONLY  Server variable is read only
//   PLUGIN_VAR_NOSYSVAR  Not a server variable
//   PLUGIN_VAR_NOCMDOPT  Not a command line option
//   PLUGIN_VAR_NOCMDARG  No argument for cmd line
//   PLUGIN_VAR_RQCMDARG  Argument required for cmd line
//   PLUGIN_VAR_OPCMDARG  Argument optional for cmd line
//   PLUGIN_VAR_MEMALLOC  String needs memory allocated

Rich Prohaska's avatar
Rich Prohaska committed
2445 2446 2447 2448
static MYSQL_SYSVAR_ULONGLONG(cache_size, tokudb_cache_size, PLUGIN_VAR_READONLY, "TokuDB cache table size", NULL, NULL, 0, 0, ~0LL, 0);

static MYSQL_SYSVAR_ULONG(max_lock, tokudb_max_lock, PLUGIN_VAR_READONLY, "TokuDB Max Locks", NULL, NULL, 8 * 1024, 0, ~0L, 0);

2449
static MYSQL_SYSVAR_STR(log_dir, tokudb_log_dir, PLUGIN_VAR_READONLY, "TokuDB Log Directory", NULL, NULL, NULL);
2450

2451
static MYSQL_SYSVAR_STR(data_dir, tokudb_data_dir, PLUGIN_VAR_READONLY, "TokuDB Data Directory", NULL, NULL, NULL);
2452 2453
#if 0

2454
static MYSQL_SYSVAR_ULONG(cache_parts, tokudb_cache_parts, PLUGIN_VAR_READONLY, "Sets TokuDB set_cache_parts", NULL, NULL, 0, 0, ~0L, 0);
2455 2456 2457

// this is really a u_int32_t
// ? use MYSQL_SYSVAR_SET
2458
static MYSQL_SYSVAR_UINT(env_flags, tokudb_env_flags, PLUGIN_VAR_READONLY, "Sets TokuDB env_flags", NULL, NULL, DB_LOG_AUTOREMOVE, 0, ~0, 0);
2459

2460
static MYSQL_SYSVAR_STR(home, tokudb_home, PLUGIN_VAR_READONLY, "Sets TokuDB env->open home", NULL, NULL, NULL);
2461 2462 2463

// this is really a u_int32_t
//? use MYSQL_SYSVAR_SET
2464
static MYSQL_SYSVAR_UINT(init_flags, tokudb_init_flags, PLUGIN_VAR_READONLY, "Sets TokuDB DB_ENV->open flags", NULL, NULL, DB_PRIVATE | DB_RECOVER, 0, ~0, 0);
2465 2466

// this looks to be unused
2467
static MYSQL_SYSVAR_LONG(lock_scan_time, tokudb_lock_scan_time, PLUGIN_VAR_READONLY, "Tokudb Lock Scan Time (UNUSED)", NULL, NULL, 0, 0, ~0L, 0);
2468 2469 2470

// this is really a u_int32_t
//? use MYSQL_SYSVAR_ENUM
2471 2472 2473 2474 2475 2476 2477 2478 2479
static MYSQL_SYSVAR_UINT(lock_type, tokudb_lock_type, PLUGIN_VAR_READONLY, "Sets set_lk_detect", NULL, NULL, DB_LOCK_DEFAULT, 0, ~0, 0);

static MYSQL_SYSVAR_ULONG(log_buffer_size, tokudb_log_buffer_size, PLUGIN_VAR_READONLY, "Tokudb Log Buffer Size", NULL, NULL, 0, 0, ~0L, 0);

static MYSQL_SYSVAR_ULONG(region_size, tokudb_region_size, PLUGIN_VAR_READONLY, "Tokudb Region Size", NULL, NULL, 128 * 1024, 0, ~0L, 0);

static MYSQL_SYSVAR_BOOL(shared_data, tokudb_shared_data, PLUGIN_VAR_READONLY, "Tokudb Shared Data", NULL, NULL, FALSE);

static MYSQL_SYSVAR_STR(tmpdir, tokudb_tmpdir, PLUGIN_VAR_READONLY, "Tokudb Tmp Dir", NULL, NULL, NULL);
2480
#endif
2481

2482 2483
static struct st_mysql_sys_var *tokudb_system_variables[] = {
    MYSQL_SYSVAR(cache_size),
2484
    MYSQL_SYSVAR(max_lock),
2485 2486
    MYSQL_SYSVAR(data_dir),
    MYSQL_SYSVAR(log_dir),
2487
#if 0
2488
    MYSQL_SYSVAR(cache_parts),
2489 2490 2491 2492 2493 2494 2495 2496 2497
    MYSQL_SYSVAR(env_flags),
    MYSQL_SYSVAR(home),
    MYSQL_SYSVAR(init_flags),
    MYSQL_SYSVAR(lock_scan_time),
    MYSQL_SYSVAR(lock_type),
    MYSQL_SYSVAR(log_buffer_size),
    MYSQL_SYSVAR(region_size),
    MYSQL_SYSVAR(shared_data),
    MYSQL_SYSVAR(tmpdir),
2498
#endif
2499
    NULL
2500 2501
};

2502
mysql_declare_plugin(tokudb) {
2503 2504 2505 2506
    MYSQL_STORAGE_ENGINE_PLUGIN, 
    &storage_engine_structure, 
    "TokuDB", 
    "Tokutek Inc", 
2507
    "Fractal trees with transactions, row level locks",
2508
    PLUGIN_LICENSE_PROPRIETARY,        /* QQQ license? */
2509
    tokudb_init_func,          /* plugin init */
2510
    tokudb_done_func,          /* plugin deinit */
2511 2512
    0x0200,                    /* QQQ 2.0 */
    NULL,                      /* status variables */
2513
    tokudb_system_variables,   /* system variables */
2514
    NULL                       /* config options */
2515 2516
}
mysql_declare_plugin_end;