2 * @brief memory-mapped database library
4 * @mainpage MDB Memory-Mapped Database Manager
5 * MDB is a Btree-based database management library modeled loosely on the
6 * BerkeleyDB API, but much simplified. The entire database is exposed
7 * in a memory map, and all data fetches return data directly
8 * from the mapped memory, so no malloc's or memcpy's occur during
9 * data fetches. As such, the library is extremely simple because it
10 * requires no page caching layer of its own, and it is extremely high
11 * performance and memory-efficient. It is also fully transactional with
12 * full ACID semantics, and when the memory map is read-only, the
13 * database integrity cannot be corrupted by stray pointer writes from
16 * The library is fully thread-aware and supports concurrent read/write
17 * access from multiple processes and threads. Data pages use a copy-on-
18 * write strategy so no active data pages are ever overwritten, which
19 * also provides resistance to corruption and eliminates the need of any
20 * special recovery procedures after a system crash. Writes are fully
21 * serialized; only one write transaction may be active at a time, which
22 * guarantees that writers can never deadlock. The database structure is
23 * multi-versioned so readers run with no locks; writers cannot block
24 * readers, and readers don't block writers.
26 * Unlike other well-known database mechanisms which use either write-ahead
27 * transaction logs or append-only data writes, MDB requires no maintenance
28 * during operation. Both write-ahead loggers and append-only databases
29 * require periodic checkpointing and/or compaction of their log or database
30 * files otherwise they grow without bound. MDB tracks free pages within
31 * the database and re-uses them for new write operations, so the database
32 * size does not grow without bound in normal use.
34 * The memory map can be used as a read-only or read-write map. It is
35 * read-only by default as this provides total immunity to corruption.
36 * Using read-write mode offers much higher write performance, but adds
37 * the possibility for stray application writes thru pointers to silently
38 * corrupt the database. Of course if your application code is known to
39 * be bug-free (...) then this is not an issue.
41 * Troubleshooting the lock file, plus semaphores on BSD systems:
43 * - A broken lockfile can cause sync issues.
44 * Stale reader transactions left behind by an aborted program
45 * cause further writes to grow the database quickly, and
46 * stale locks can block further operation.
48 * Fix: Terminate all programs using the database, or make
49 * them close it. Next database user will reset the lockfile.
51 * - On BSD systems or others configured with MDB_USE_POSIX_SEM,
52 * startup can fail due to semaphores owned by another userid.
54 * Fix: Open and close the database as the user which owns the
55 * semaphores (likely last user) or as root, while no other
56 * process is using the database.
58 * Restrictions/caveats (in addition to those listed for some functions):
60 * - Only the database owner should normally use the database on
61 * BSD systems or when otherwise configured with MDB_USE_POSIX_SEM.
62 * Multiple users can cause startup to fail later, as noted above.
64 * - A thread can only use one transaction at a time, plus any child
65 * transactions. Each transaction belongs to one thread. See below.
67 * - Use an MDB_env* in the process which opened it, without fork()ing.
69 * - Do not have open an MDB database twice in the same process at
70 * the same time. Not even from a plain open() call - close()ing it
71 * breaks flock() advisory locking.
73 * - Avoid long-lived transactions. Read transactions prevent
74 * reuse of pages freed by newer write transactions, thus the
75 * database can grow quickly. Write transactions prevent
76 * other write transactions, since writes are serialized.
78 * ...when several processes can use a database concurrently:
80 * - Avoid suspending a process with active transactions. These
81 * would then be "long-lived" as above.
83 * - Avoid aborting a process with an active transaction.
84 * The transaction becomes "long-lived" as above until the lockfile
85 * is reset, since the process may not remove it from the lockfile.
87 * - If you do that anyway, close the environment once in a while,
88 * so the lockfile can get reset.
90 * - Do not use MDB databases on remote filesystems, even between
91 * processes on the same host. This breaks flock() on some OSes,
92 * possibly memory map sync, and certainly sync between programs
95 * - Opening a database can fail if another process is opening or
96 * closing it at exactly the same time.
98 * @author Howard Chu, Symas Corporation.
100 * @copyright Copyright 2011-2012 Howard Chu, Symas Corp. All rights reserved.
102 * Redistribution and use in source and binary forms, with or without
103 * modification, are permitted only as authorized by the OpenLDAP
106 * A copy of this license is available in the file LICENSE in the
107 * top-level directory of the distribution or, alternatively, at
108 * <http://www.OpenLDAP.org/license.html>.
111 * This code is derived from btree.c written by Martin Hedenfalk.
113 * Copyright (c) 2009, 2010 Martin Hedenfalk <martin@bzero.se>
115 * Permission to use, copy, modify, and distribute this software for any
116 * purpose with or without fee is hereby granted, provided that the above
117 * copyright notice and this permission notice appear in all copies.
119 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
120 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
121 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
122 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
123 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
124 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
125 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
130 #include <sys/types.h>
136 /** @defgroup public Public API
139 /** @defgroup Version Version Macros
142 /** Library major version */
143 #define MDB_VERSION_MAJOR 0
144 /** Library minor version */
145 #define MDB_VERSION_MINOR 9
146 /** Library patch version */
147 #define MDB_VERSION_PATCH 4
149 /** Combine args a,b,c into a single integer for easy version comparisons */
150 #define MDB_VERINT(a,b,c) (((a) << 24) | ((b) << 16) | (c))
152 /** The full library version as a single integer */
153 #define MDB_VERSION_FULL \
154 MDB_VERINT(MDB_VERSION_MAJOR,MDB_VERSION_MINOR,MDB_VERSION_PATCH)
156 /** The release date of this library version */
157 #define MDB_VERSION_DATE "September 14, 2012"
159 /** A stringifier for the version info */
160 #define MDB_VERSTR(a,b,c,d) "MDB " #a "." #b "." #c ": (" d ")"
162 /** A helper for the stringifier macro */
163 #define MDB_VERFOO(a,b,c,d) MDB_VERSTR(a,b,c,d)
165 /** The full library version as a C string */
166 #define MDB_VERSION_STRING \
167 MDB_VERFOO(MDB_VERSION_MAJOR,MDB_VERSION_MINOR,MDB_VERSION_PATCH,MDB_VERSION_DATE)
170 /** @brief Opaque structure for a database environment.
172 * A DB environment supports multiple databases, all residing in the same
175 typedef struct MDB_env MDB_env;
177 /** @brief Opaque structure for a transaction handle.
179 * All database operations require a transaction handle. Transactions may be
180 * read-only or read-write.
182 typedef struct MDB_txn MDB_txn;
184 /** @brief A handle for an individual database in the DB environment. */
185 typedef unsigned int MDB_dbi;
187 /** @brief Opaque structure for navigating through a database */
188 typedef struct MDB_cursor MDB_cursor;
190 /** @brief Generic structure used for passing keys and data in and out of the database. */
191 typedef struct MDB_val {
192 size_t mv_size; /**< size of the data item */
193 void *mv_data; /**< address of the data item */
196 /** @brief A callback function used to compare two keys in a database */
197 typedef int (MDB_cmp_func)(const MDB_val *a, const MDB_val *b);
199 /** @brief A callback function used to relocate a position-dependent data item
200 * in a fixed-address database.
202 * The \b newptr gives the item's desired address in
203 * the memory map, and \b oldptr gives its previous address. The item's actual
204 * data resides at the address in \b item. This callback is expected to walk
205 * through the fields of the record in \b item and modify any
206 * values based at the \b oldptr address to be relative to the \b newptr address.
207 * @param[in,out] item The item that is to be relocated.
208 * @param[in] oldptr The previous address.
209 * @param[in] newptr The new address to relocate to.
210 * @param[in] relctx An application-provided context, set by #mdb_set_relctx().
211 * @todo This feature is currently unimplemented.
213 typedef void (MDB_rel_func)(MDB_val *item, void *oldptr, void *newptr, void *relctx);
215 /** @defgroup mdb_env Environment Flags
218 /** mmap at a fixed address */
219 #define MDB_FIXEDMAP 0x01
220 /** no environment directory */
221 #define MDB_NOSUBDIR 0x02
222 /** don't fsync after commit */
223 #define MDB_NOSYNC 0x10000
225 #define MDB_RDONLY 0x20000
226 /** don't fsync metapage after commit */
227 #define MDB_NOMETASYNC 0x40000
228 /** use writable mmap */
229 #define MDB_WRITEMAP 0x80000
230 /** use asynchronous msync */
231 #define MDB_MAPASYNC 0x100000
234 /** @defgroup mdb_open Database Flags
237 /** use reverse string keys */
238 #define MDB_REVERSEKEY 0x02
239 /** use sorted duplicates */
240 #define MDB_DUPSORT 0x04
241 /** numeric keys in native byte order.
242 * The keys must all be of the same size. */
243 #define MDB_INTEGERKEY 0x08
244 /** with #MDB_DUPSORT, sorted dup items have fixed size */
245 #define MDB_DUPFIXED 0x10
246 /** with #MDB_DUPSORT, dups are numeric in native byte order */
247 #define MDB_INTEGERDUP 0x20
248 /** with #MDB_DUPSORT, use reverse string dups */
249 #define MDB_REVERSEDUP 0x40
250 /** create DB if not already existing */
251 #define MDB_CREATE 0x40000
254 /** @defgroup mdb_put Write Flags
257 /** For put: Don't write if the key already exists. */
258 #define MDB_NOOVERWRITE 0x10
259 /** Only for #MDB_DUPSORT<br>
260 * For put: don't write if the key and data pair already exist.<br>
261 * For mdb_cursor_del: remove all duplicate data items.
263 #define MDB_NODUPDATA 0x20
264 /** For mdb_cursor_put: overwrite the current key/data pair */
265 #define MDB_CURRENT 0x40
266 /** For put: Just reserve space for data, don't copy it. Return a
267 * pointer to the reserved space.
269 #define MDB_RESERVE 0x10000
270 /** Data is being appended, don't split full pages. */
271 #define MDB_APPEND 0x20000
272 /** Duplicate data is being appended, don't split full pages. */
273 #define MDB_APPENDDUP 0x40000
274 /** Store multiple data items in one call. */
275 #define MDB_MULTIPLE 0x80000
278 /** @brief Cursor Get operations.
280 * This is the set of all operations for retrieving data
283 typedef enum MDB_cursor_op {
284 MDB_FIRST, /**< Position at first key/data item */
285 MDB_FIRST_DUP, /**< Position at first data item of current key.
286 Only for #MDB_DUPSORT */
287 MDB_GET_BOTH, /**< Position at key/data pair. Only for #MDB_DUPSORT */
288 MDB_GET_BOTH_RANGE, /**< position at key, nearest data. Only for #MDB_DUPSORT */
289 MDB_GET_CURRENT, /**< Return key/data at current cursor position */
290 MDB_GET_MULTIPLE, /**< Return all the duplicate data items at the current
291 cursor position. Only for #MDB_DUPFIXED */
292 MDB_LAST, /**< Position at last key/data item */
293 MDB_LAST_DUP, /**< Position at last data item of current key.
294 Only for #MDB_DUPSORT */
295 MDB_NEXT, /**< Position at next data item */
296 MDB_NEXT_DUP, /**< Position at next data item of current key.
297 Only for #MDB_DUPSORT */
298 MDB_NEXT_MULTIPLE, /**< Return all duplicate data items at the next
299 cursor position. Only for #MDB_DUPFIXED */
300 MDB_NEXT_NODUP, /**< Position at first data item of next key.
301 Only for #MDB_DUPSORT */
302 MDB_PREV, /**< Position at previous data item */
303 MDB_PREV_DUP, /**< Position at previous data item of current key.
304 Only for #MDB_DUPSORT */
305 MDB_PREV_NODUP, /**< Position at last data item of previous key.
306 Only for #MDB_DUPSORT */
307 MDB_SET, /**< Position at specified key */
308 MDB_SET_KEY, /**< Position at specified key, return key + data */
309 MDB_SET_RANGE /**< Position at first key greater than or equal to specified key. */
312 /** @defgroup errors Return Codes
314 * BerkeleyDB uses -30800 to -30999, we'll go under them
317 /** Successful result */
318 #define MDB_SUCCESS 0
319 /** key/data pair already exists */
320 #define MDB_KEYEXIST (-30799)
321 /** key/data pair not found (EOF) */
322 #define MDB_NOTFOUND (-30798)
323 /** Requested page not found - this usually indicates corruption */
324 #define MDB_PAGE_NOTFOUND (-30797)
325 /** Located page was wrong type */
326 #define MDB_CORRUPTED (-30796)
327 /** Update of meta page failed, probably I/O error */
328 #define MDB_PANIC (-30795)
329 /** Environment version mismatch */
330 #define MDB_VERSION_MISMATCH (-30794)
331 /** File is not a valid MDB file */
332 #define MDB_INVALID (-30793)
333 /** Environment mapsize reached */
334 #define MDB_MAP_FULL (-30792)
335 /** Environment maxdbs reached */
336 #define MDB_DBS_FULL (-30791)
337 /** Environment maxreaders reached */
338 #define MDB_READERS_FULL (-30790)
339 /** Too many TLS keys in use - Windows only */
340 #define MDB_TLS_FULL (-30789)
341 /** Nested txn has too many dirty pages */
342 #define MDB_TXN_FULL (-30788)
343 /** Cursor stack too deep - internal error */
344 #define MDB_CURSOR_FULL (-30787)
345 /** Page has not enough space - internal error */
346 #define MDB_PAGE_FULL (-30786)
347 #define MDB_LAST_ERRCODE MDB_PAGE_FULL
350 /** @brief Statistics for a database in the environment */
351 typedef struct MDB_stat {
352 unsigned int ms_psize; /**< Size of a database page.
353 This is currently the same for all databases. */
354 unsigned int ms_depth; /**< Depth (height) of the B-tree */
355 size_t ms_branch_pages; /**< Number of internal (non-leaf) pages */
356 size_t ms_leaf_pages; /**< Number of leaf pages */
357 size_t ms_overflow_pages; /**< Number of overflow pages */
358 size_t ms_entries; /**< Number of data items */
361 /** @brief Return the mdb library version information.
363 * @param[out] major if non-NULL, the library major version number is copied here
364 * @param[out] minor if non-NULL, the library minor version number is copied here
365 * @param[out] patch if non-NULL, the library patch version number is copied here
366 * @retval "version string" The library version as a string
368 char *mdb_version(int *major, int *minor, int *patch);
370 /** @brief Return a string describing a given error code.
372 * This function is a superset of the ANSI C X3.159-1989 (ANSI C) strerror(3)
373 * function. If the error code is greater than or equal to 0, then the string
374 * returned by the system function strerror(3) is returned. If the error code
375 * is less than 0, an error string corresponding to the MDB library error is
376 * returned. See @ref errors for a list of MDB-specific error codes.
377 * @param[in] err The error code
378 * @retval "error message" The description of the error
380 char *mdb_strerror(int err);
382 /** @brief Create an MDB environment handle.
384 * This function allocates memory for a #MDB_env structure. To release
385 * the allocated memory and discard the handle, call #mdb_env_close().
386 * Before the handle may be used, it must be opened using #mdb_env_open().
387 * Various other options may also need to be set before opening the handle,
388 * e.g. #mdb_env_set_mapsize(), #mdb_env_set_maxreaders(), #mdb_env_set_maxdbs(),
389 * depending on usage requirements.
390 * @param[out] env The address where the new handle will be stored
391 * @return A non-zero error value on failure and 0 on success.
393 int mdb_env_create(MDB_env **env);
395 /** @brief Open an environment handle.
397 * If this function fails, #mdb_env_close() must be called to discard the #MDB_env handle.
398 * @param[in] env An environment handle returned by #mdb_env_create()
399 * @param[in] path The directory in which the database files reside. This
400 * directory must already exist and be writable.
401 * @param[in] flags Special options for this environment. This parameter
402 * must be set to 0 or by bitwise OR'ing together one or more of the
403 * values described here.
406 * use a fixed address for the mmap region. This flag must be specified
407 * when creating the environment, and is stored persistently in the environment.
408 * If successful, the memory map will always reside at the same virtual address
409 * and pointers used to reference data items in the database will be constant
410 * across multiple invocations. This option may not always work, depending on
411 * how the operating system has allocated memory to shared libraries and other uses.
412 * The feature is highly experimental.
414 * By default, MDB creates its environment in a directory whose
415 * pathname is given in \b path, and creates its data and lock files
416 * under that directory. With this option, \b path is used as-is for
417 * the database main data file. The database lock file is the \b path
418 * with "-lock" appended.
420 * Don't perform a synchronous flush after committing a transaction. This means
421 * transactions will exhibit the ACI (atomicity, consistency, and isolation)
422 * properties, but not D (durability); that is database integrity will be
423 * maintained but it is possible some number of the most recently committed
424 * transactions may be undone after a system crash. The number of transactions
425 * at risk is governed by how often the system flushes dirty buffers to disk
426 * and how often #mdb_env_sync() is called. This flag may be changed
427 * at any time using #mdb_env_set_flags().
428 * <li>#MDB_NOMETASYNC
429 * Don't perform a synchronous flush of the meta page after committing
430 * a transaction. This is similar to the #MDB_NOSYNC case, but safer
431 * because the transaction data is still flushed. The meta page for any
432 * transaction N will be flushed by the data flush of transaction N+1.
433 * In case of a system crash, the last committed transaction may be
434 * lost. This flag may be changed at any time using #mdb_env_set_flags().
436 * Open the environment in read-only mode. No write operations will be allowed.
438 * @param[in] mode The UNIX permissions to set on created files. This parameter
439 * is ignored on Windows.
440 * @return A non-zero error value on failure and 0 on success. Some possible
443 * <li>#MDB_VERSION_MISMATCH - the version of the MDB library doesn't match the
444 * version that created the database environment.
445 * <li>EINVAL - the environment file headers are corrupted.
446 * <li>ENOENT - the directory specified by the path parameter doesn't exist.
447 * <li>EACCES - the user didn't have permission to access the environment files.
448 * <li>EAGAIN - the environment was locked by another process.
451 int mdb_env_open(MDB_env *env, const char *path, unsigned int flags, mode_t mode);
453 /** @brief Return statistics about the MDB environment.
455 * @param[in] env An environment handle returned by #mdb_env_create()
456 * @param[out] stat The address of an #MDB_stat structure
457 * where the statistics will be copied
459 int mdb_env_stat(MDB_env *env, MDB_stat *stat);
461 /** @brief Flush the data buffers to disk.
463 * Data is always written to disk when #mdb_txn_commit() is called,
464 * but the operating system may keep it buffered. MDB always flushes
465 * the OS buffers upon commit as well, unless the environment was
466 * opened with #MDB_NOSYNC.
467 * @param[in] env An environment handle returned by #mdb_env_create()
468 * @param[in] force If non-zero, force the flush to occur. Otherwise
469 * if the environment has the #MDB_NOSYNC flag set the flushes
471 * @return A non-zero error value on failure and 0 on success. Some possible
474 * <li>EINVAL - an invalid parameter was specified.
475 * <li>EIO - an error occurred during synchronization.
478 int mdb_env_sync(MDB_env *env, int force);
480 /** @brief Close the environment and release the memory map.
482 * Only a single thread may call this function. All transactions, databases,
483 * and cursors must already be closed before calling this function. Attempts to
484 * use any such handles after calling this function will cause a SIGSEGV.
485 * The environment handle will be freed and must not be used again after this call.
486 * @param[in] env An environment handle returned by #mdb_env_create()
488 void mdb_env_close(MDB_env *env);
490 /** @brief Set environment flags.
492 * This may be used to set some flags that weren't already set during
493 * #mdb_env_open(), or to unset these flags.
494 * @param[in] env An environment handle returned by #mdb_env_create()
495 * @param[in] flags The flags to change, bitwise OR'ed together
496 * @param[in] onoff A non-zero value sets the flags, zero clears them.
497 * @return A non-zero error value on failure and 0 on success. Some possible
500 * <li>EINVAL - an invalid parameter was specified.
503 int mdb_env_set_flags(MDB_env *env, unsigned int flags, int onoff);
505 /** @brief Get environment flags.
507 * @param[in] env An environment handle returned by #mdb_env_create()
508 * @param[out] flags The address of an integer to store the flags
509 * @return A non-zero error value on failure and 0 on success. Some possible
512 * <li>EINVAL - an invalid parameter was specified.
515 int mdb_env_get_flags(MDB_env *env, unsigned int *flags);
517 /** @brief Return the path that was used in #mdb_env_open().
519 * @param[in] env An environment handle returned by #mdb_env_create()
520 * @param[out] path Address of a string pointer to contain the path. This
521 * is the actual string in the environment, not a copy. It should not be
522 * altered in any way.
523 * @return A non-zero error value on failure and 0 on success. Some possible
526 * <li>EINVAL - an invalid parameter was specified.
529 int mdb_env_get_path(MDB_env *env, const char **path);
531 /** @brief Set the size of the memory map to use for this environment.
533 * The size should be a multiple of the OS page size. The default is
534 * 10485760 bytes. The size of the memory map is also the maximum size
535 * of the database. The value should be chosen as large as possible,
536 * to accommodate future growth of the database.
537 * This function may only be called after #mdb_env_create() and before #mdb_env_open().
538 * @param[in] env An environment handle returned by #mdb_env_create()
539 * @param[in] size The size in bytes
540 * @return A non-zero error value on failure and 0 on success. Some possible
543 * <li>EINVAL - an invalid parameter was specified, or the environment is already open.
546 int mdb_env_set_mapsize(MDB_env *env, size_t size);
548 /** @brief Set the maximum number of threads for the environment.
550 * This defines the number of slots in the lock table that is used to track readers in the
551 * the environment. The default is 126.
552 * This function may only be called after #mdb_env_create() and before #mdb_env_open().
553 * @param[in] env An environment handle returned by #mdb_env_create()
554 * @param[in] readers The maximum number of threads
555 * @return A non-zero error value on failure and 0 on success. Some possible
558 * <li>EINVAL - an invalid parameter was specified, or the environment is already open.
561 int mdb_env_set_maxreaders(MDB_env *env, unsigned int readers);
563 /** @brief Get the maximum number of threads for the environment.
565 * @param[in] env An environment handle returned by #mdb_env_create()
566 * @param[out] readers Address of an integer to store the number of readers
567 * @return A non-zero error value on failure and 0 on success. Some possible
570 * <li>EINVAL - an invalid parameter was specified.
573 int mdb_env_get_maxreaders(MDB_env *env, unsigned int *readers);
575 /** @brief Set the maximum number of databases for the environment.
577 * This function is only needed if multiple databases will be used in the
578 * environment. Simpler applications that only use a single database can ignore
580 * This function may only be called after #mdb_env_create() and before #mdb_env_open().
581 * @param[in] env An environment handle returned by #mdb_env_create()
582 * @param[in] dbs The maximum number of databases
583 * @return A non-zero error value on failure and 0 on success. Some possible
586 * <li>EINVAL - an invalid parameter was specified, or the environment is already open.
589 int mdb_env_set_maxdbs(MDB_env *env, MDB_dbi dbs);
591 /** @brief Create a transaction for use with the environment.
593 * The transaction handle may be discarded using #mdb_txn_abort() or #mdb_txn_commit().
594 * @note Transactions may not span threads; a transaction must only be used by a
595 * single thread. Also, a thread may only have a single transaction.
596 * @note Cursors may not span transactions; each cursor must be opened and closed
597 * within a single transaction.
598 * @param[in] env An environment handle returned by #mdb_env_create()
599 * @param[in] parent If this parameter is non-NULL, the new transaction
600 * will be a nested transaction, with the transaction indicated by \b parent
601 * as its parent. Transactions may be nested to any level. A parent
602 * transaction may not issue any other operations besides mdb_txn_begin,
603 * mdb_txn_abort, or mdb_txn_commit while it has active child transactions.
604 * @param[in] flags Special options for this transaction. This parameter
605 * must be set to 0 or by bitwise OR'ing together one or more of the
606 * values described here.
609 * This transaction will not perform any write operations.
611 * @param[out] txn Address where the new #MDB_txn handle will be stored
612 * @return A non-zero error value on failure and 0 on success. Some possible
615 * <li>#MDB_PANIC - a fatal error occurred earlier and the environment
617 * <li>ENOMEM - out of memory, or a read-only transaction was requested and
618 * the reader lock table is full. See #mdb_env_set_maxreaders().
621 int mdb_txn_begin(MDB_env *env, MDB_txn *parent, unsigned int flags, MDB_txn **txn);
623 /** @brief Commit all the operations of a transaction into the database.
625 * All cursors opened within the transaction will be closed by this call. The cursors
626 * and transaction handle will be freed and must not be used again after this call.
627 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
628 * @return A non-zero error value on failure and 0 on success. Some possible
631 * <li>EINVAL - an invalid parameter was specified.
632 * <li>ENOSPC - no more disk space.
633 * <li>EIO - a low-level I/O error occurred while writing.
634 * <li>ENOMEM - the transaction is nested and could not be merged into its parent.
637 int mdb_txn_commit(MDB_txn *txn);
639 /** @brief Abandon all the operations of the transaction instead of saving them.
641 * All cursors opened within the transaction will be closed by this call. The cursors
642 * and transaction handle will be freed and must not be used again after this call.
643 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
645 void mdb_txn_abort(MDB_txn *txn);
647 /** @brief Reset a read-only transaction.
649 * This releases the current reader lock but doesn't free the
650 * transaction handle, allowing it to be used again later by #mdb_txn_renew().
651 * It otherwise has the same effect as #mdb_txn_abort() but saves some memory
652 * allocation/deallocation overhead if a thread is going to start a new
653 * read-only transaction again soon.
654 * All cursors opened within the transaction must be closed before the transaction
656 * Reader locks generally don't interfere with writers, but they keep old
657 * versions of database pages allocated. Thus they prevent the old pages
658 * from being reused when writers commit new data, and so under heavy load
659 * the database size may grow much more rapidly than otherwise.
660 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
662 void mdb_txn_reset(MDB_txn *txn);
664 /** @brief Renew a read-only transaction.
666 * This acquires a new reader lock for a transaction handle that had been
667 * released by #mdb_txn_reset(). It must be called before a reset transaction
669 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
670 * @return A non-zero error value on failure and 0 on success. Some possible
673 * <li>#MDB_PANIC - a fatal error occurred earlier and the environment
675 * <li>EINVAL - an invalid parameter was specified.
678 int mdb_txn_renew(MDB_txn *txn);
680 /** @brief Open a database in the environment.
682 * The database handle may be discarded by calling #mdb_close(). The
683 * database handle resides in the shared environment, it is not owned
684 * by the given transaction. Only one thread should call this function;
685 * it is not mutex-protected in a read-only transaction.
686 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
687 * @param[in] name The name of the database to open. If only a single
688 * database is needed in the environment, this value may be NULL.
689 * @param[in] flags Special options for this database. This parameter
690 * must be set to 0 or by bitwise OR'ing together one or more of the
691 * values described here.
693 * <li>#MDB_REVERSEKEY
694 * Keys are strings to be compared in reverse order, from the end
695 * of the strings to the beginning. By default, Keys are treated as strings and
696 * compared from beginning to end.
698 * Duplicate keys may be used in the database. (Or, from another perspective,
699 * keys may have multiple data items, stored in sorted order.) By default
700 * keys must be unique and may have only a single data item.
701 * <li>#MDB_INTEGERKEY
702 * Keys are binary integers in native byte order. Setting this option
703 * requires all keys to be the same size, typically sizeof(int)
706 * This flag may only be used in combination with #MDB_DUPSORT. This option
707 * tells the library that the data items for this database are all the same
708 * size, which allows further optimizations in storage and retrieval. When
709 * all data items are the same size, the #MDB_GET_MULTIPLE and #MDB_NEXT_MULTIPLE
710 * cursor operations may be used to retrieve multiple items at once.
711 * <li>#MDB_INTEGERDUP
712 * This option specifies that duplicate data items are also integers, and
713 * should be sorted as such.
714 * <li>#MDB_REVERSEDUP
715 * This option specifies that duplicate data items should be compared as
716 * strings in reverse order.
718 * Create the named database if it doesn't exist. This option is not
719 * allowed in a read-only transaction or a read-only environment.
721 * @param[out] dbi Address where the new #MDB_dbi handle will be stored
722 * @return A non-zero error value on failure and 0 on success. Some possible
725 * <li>#MDB_NOTFOUND - the specified database doesn't exist in the environment
726 * and #MDB_CREATE was not specified.
727 * <li>ENFILE - too many databases have been opened. See #mdb_env_set_maxdbs().
730 int mdb_open(MDB_txn *txn, const char *name, unsigned int flags, MDB_dbi *dbi);
732 /** @brief Retrieve statistics for a database.
734 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
735 * @param[in] dbi A database handle returned by #mdb_open()
736 * @param[out] stat The address of an #MDB_stat structure
737 * where the statistics will be copied
738 * @return A non-zero error value on failure and 0 on success. Some possible
741 * <li>EINVAL - an invalid parameter was specified.
744 int mdb_stat(MDB_txn *txn, MDB_dbi dbi, MDB_stat *stat);
746 /** @brief Close a database handle.
748 * This call is not mutex protected. Handles should only be closed by
749 * a single thread, and only if no other threads are going to reference
750 * the database handle any further.
751 * @param[in] env An environment handle returned by #mdb_env_create()
752 * @param[in] dbi A database handle returned by #mdb_open()
754 void mdb_close(MDB_env *env, MDB_dbi dbi);
756 /** @brief Delete a database and/or free all its pages.
758 * If the \b del parameter is non-zero the DB handle will be closed
759 * and the DB will be deleted.
760 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
761 * @param[in] dbi A database handle returned by #mdb_open()
762 * @param[in] del non-zero to delete the DB from the environment,
763 * otherwise just free its pages.
764 * @return A non-zero error value on failure and 0 on success.
766 int mdb_drop(MDB_txn *txn, MDB_dbi dbi, int del);
768 /** @brief Set a custom key comparison function for a database.
770 * The comparison function is called whenever it is necessary to compare a
771 * key specified by the application with a key currently stored in the database.
772 * If no comparison function is specified, and no special key flags were specified
773 * with #mdb_open(), the keys are compared lexically, with shorter keys collating
774 * before longer keys.
775 * @warning This function must be called before any data access functions are used,
776 * otherwise data corruption may occur. The same comparison function must be used by every
777 * program accessing the database, every time the database is used.
778 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
779 * @param[in] dbi A database handle returned by #mdb_open()
780 * @param[in] cmp A #MDB_cmp_func function
781 * @return A non-zero error value on failure and 0 on success. Some possible
784 * <li>EINVAL - an invalid parameter was specified.
787 int mdb_set_compare(MDB_txn *txn, MDB_dbi dbi, MDB_cmp_func *cmp);
789 /** @brief Set a custom data comparison function for a #MDB_DUPSORT database.
791 * This comparison function is called whenever it is necessary to compare a data
792 * item specified by the application with a data item currently stored in the database.
793 * This function only takes effect if the database was opened with the #MDB_DUPSORT
795 * If no comparison function is specified, and no special key flags were specified
796 * with #mdb_open(), the data items are compared lexically, with shorter items collating
797 * before longer items.
798 * @warning This function must be called before any data access functions are used,
799 * otherwise data corruption may occur. The same comparison function must be used by every
800 * program accessing the database, every time the database is used.
801 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
802 * @param[in] dbi A database handle returned by #mdb_open()
803 * @param[in] cmp A #MDB_cmp_func function
804 * @return A non-zero error value on failure and 0 on success. Some possible
807 * <li>EINVAL - an invalid parameter was specified.
810 int mdb_set_dupsort(MDB_txn *txn, MDB_dbi dbi, MDB_cmp_func *cmp);
812 /** @brief Set a relocation function for a #MDB_FIXEDMAP database.
814 * @todo The relocation function is called whenever it is necessary to move the data
815 * of an item to a different position in the database (e.g. through tree
816 * balancing operations, shifts as a result of adds or deletes, etc.). It is
817 * intended to allow address/position-dependent data items to be stored in
818 * a database in an environment opened with the #MDB_FIXEDMAP option.
819 * Currently the relocation feature is unimplemented and setting
820 * this function has no effect.
821 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
822 * @param[in] dbi A database handle returned by #mdb_open()
823 * @param[in] rel A #MDB_rel_func function
824 * @return A non-zero error value on failure and 0 on success. Some possible
827 * <li>EINVAL - an invalid parameter was specified.
830 int mdb_set_relfunc(MDB_txn *txn, MDB_dbi dbi, MDB_rel_func *rel);
832 /** @brief Set a context pointer for a #MDB_FIXEDMAP database's relocation function.
834 * See #mdb_set_relfunc and #MDB_rel_func for more details.
835 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
836 * @param[in] dbi A database handle returned by #mdb_open()
837 * @param[in] ctx An arbitrary pointer for whatever the application needs.
838 * It will be passed to the callback function set by #mdb_set_relfunc
839 * as its \b relctx parameter whenever the callback is invoked.
840 * @return A non-zero error value on failure and 0 on success. Some possible
843 * <li>EINVAL - an invalid parameter was specified.
846 int mdb_set_relctx(MDB_txn *txn, MDB_dbi dbi, void *ctx);
848 /** @brief Get items from a database.
850 * This function retrieves key/data pairs from the database. The address
851 * and length of the data associated with the specified \b key are returned
852 * in the structure to which \b data refers.
853 * If the database supports duplicate keys (#MDB_DUPSORT) then the
854 * first data item for the key will be returned. Retrieval of other
855 * items requires the use of #mdb_cursor_get().
857 * @note The memory pointed to by the returned values is owned by the
858 * database. The caller need not dispose of the memory, and may not
859 * modify it in any way. For values returned in a read-only transaction
860 * any modification attempts will cause a SIGSEGV.
861 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
862 * @param[in] dbi A database handle returned by #mdb_open()
863 * @param[in] key The key to search for in the database
864 * @param[out] data The data corresponding to the key
865 * @return A non-zero error value on failure and 0 on success. Some possible
868 * <li>#MDB_NOTFOUND - the key was not in the database.
869 * <li>EINVAL - an invalid parameter was specified.
872 int mdb_get(MDB_txn *txn, MDB_dbi dbi, MDB_val *key, MDB_val *data);
874 /** @brief Store items into a database.
876 * This function stores key/data pairs in the database. The default behavior
877 * is to enter the new key/data pair, replacing any previously existing key
878 * if duplicates are disallowed, or adding a duplicate data item if
879 * duplicates are allowed (#MDB_DUPSORT).
880 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
881 * @param[in] dbi A database handle returned by #mdb_open()
882 * @param[in] key The key to store in the database
883 * @param[in,out] data The data to store
884 * @param[in] flags Special options for this operation. This parameter
885 * must be set to 0 or by bitwise OR'ing together one or more of the
886 * values described here.
888 * <li>#MDB_NODUPDATA - enter the new key/data pair only if it does not
889 * already appear in the database. This flag may only be specified
890 * if the database was opened with #MDB_DUPSORT. The function will
891 * return #MDB_KEYEXIST if the key/data pair already appears in the
893 * <li>#MDB_NOOVERWRITE - enter the new key/data pair only if the key
894 * does not already appear in the database. The function will return
895 * #MDB_KEYEXIST if the key already appears in the database, even if
896 * the database supports duplicates (#MDB_DUPSORT). The \b data
897 * parameter will be set to point to the existing item.
898 * <li>#MDB_RESERVE - reserve space for data of the given size, but
899 * don't copy the given data. Instead, return a pointer to the
900 * reserved space, which the caller can fill in later. This saves
901 * an extra memcpy if the data is being generated later.
902 * <li>#MDB_APPEND - append the given key/data pair to the end of the
903 * database. No key comparisons are performed. This option allows
904 * fast bulk loading when keys are already known to be in the
905 * correct order. Loading unsorted keys with this flag will cause
907 * <li>#MDB_APPENDDUP - as above, but for sorted dup data.
909 * @return A non-zero error value on failure and 0 on success. Some possible
912 * <li>EACCES - an attempt was made to write in a read-only transaction.
913 * <li>EINVAL - an invalid parameter was specified.
914 * <li>ENOMEM - the database is full, see #mdb_env_set_mapsize().
917 int mdb_put(MDB_txn *txn, MDB_dbi dbi, MDB_val *key, MDB_val *data,
920 /** @brief Delete items from a database.
922 * This function removes key/data pairs from the database.
923 * If the database does not support sorted duplicate data items
924 * (#MDB_DUPSORT) the data parameter is ignored.
925 * If the database supports sorted duplicates and the data parameter
926 * is NULL, all of the duplicate data items for the key will be
927 * deleted. Otherwise, if the data parameter is non-NULL
928 * only the matching data item will be deleted.
929 * This function will return #MDB_NOTFOUND if the specified key/data
930 * pair is not in the database.
931 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
932 * @param[in] dbi A database handle returned by #mdb_open()
933 * @param[in] key The key to delete from the database
934 * @param[in] data The data to delete
935 * @return A non-zero error value on failure and 0 on success. Some possible
938 * <li>EACCES - an attempt was made to write in a read-only transaction.
939 * <li>EINVAL - an invalid parameter was specified.
942 int mdb_del(MDB_txn *txn, MDB_dbi dbi, MDB_val *key, MDB_val *data);
944 /** @brief Create a cursor handle.
946 * Cursors are associated with a specific transaction and database and
947 * may not span threads.
948 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
949 * @param[in] dbi A database handle returned by #mdb_open()
950 * @param[out] cursor Address where the new #MDB_cursor handle will be stored
951 * @return A non-zero error value on failure and 0 on success. Some possible
954 * <li>EINVAL - an invalid parameter was specified.
957 int mdb_cursor_open(MDB_txn *txn, MDB_dbi dbi, MDB_cursor **cursor);
959 /** @brief Close a cursor handle.
961 * The cursor handle will be freed and must not be used again after this call.
962 * @param[in] cursor A cursor handle returned by #mdb_cursor_open()
964 void mdb_cursor_close(MDB_cursor *cursor);
966 /** @brief Renew a cursor handle.
968 * Cursors are associated with a specific transaction and database and
969 * may not span threads. Cursors that are only used in read-only
970 * transactions may be re-used, to avoid unnecessary malloc/free overhead.
971 * The cursor may be associated with a new read-only transaction, and
972 * referencing the same database handle as it was created with.
973 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
974 * @param[in] cursor A cursor handle returned by #mdb_cursor_open()
975 * @return A non-zero error value on failure and 0 on success. Some possible
978 * <li>EINVAL - an invalid parameter was specified.
981 int mdb_cursor_renew(MDB_txn *txn, MDB_cursor *cursor);
983 /** @brief Return the cursor's transaction handle.
985 * @param[in] cursor A cursor handle returned by #mdb_cursor_open()
987 MDB_txn *mdb_cursor_txn(MDB_cursor *cursor);
989 /** @brief Return the cursor's database handle.
991 * @param[in] cursor A cursor handle returned by #mdb_cursor_open()
993 MDB_dbi mdb_cursor_dbi(MDB_cursor *cursor);
995 /** @brief Retrieve by cursor.
997 * This function retrieves key/data pairs from the database. The address and length
998 * of the key are returned in the object to which \b key refers (except for the
999 * case of the #MDB_SET option, in which the \b key object is unchanged), and
1000 * the address and length of the data are returned in the object to which \b data
1002 * @param[in] cursor A cursor handle returned by #mdb_cursor_open()
1003 * @param[in,out] key The key for a retrieved item
1004 * @param[in,out] data The data of a retrieved item
1005 * @param[in] op A cursor operation #MDB_cursor_op
1006 * @return A non-zero error value on failure and 0 on success. Some possible
1009 * <li>#MDB_NOTFOUND - no matching key found.
1010 * <li>EINVAL - an invalid parameter was specified.
1013 int mdb_cursor_get(MDB_cursor *cursor, MDB_val *key, MDB_val *data,
1016 /** @brief Store by cursor.
1018 * This function stores key/data pairs into the database.
1019 * If the function fails for any reason, the state of the cursor will be
1020 * unchanged. If the function succeeds and an item is inserted into the
1021 * database, the cursor is always positioned to refer to the newly inserted item.
1022 * @param[in] cursor A cursor handle returned by #mdb_cursor_open()
1023 * @param[in] key The key operated on.
1024 * @param[in] data The data operated on.
1025 * @param[in] flags Options for this operation. This parameter
1026 * must be set to 0 or one of the values described here.
1028 * <li>#MDB_CURRENT - overwrite the data of the key/data pair to which
1029 * the cursor refers with the specified data item. The \b key
1030 * parameter is ignored.
1031 * <li>#MDB_NODUPDATA - enter the new key/data pair only if it does not
1032 * already appear in the database. This flag may only be specified
1033 * if the database was opened with #MDB_DUPSORT. The function will
1034 * return #MDB_KEYEXIST if the key/data pair already appears in the
1036 * <li>#MDB_NOOVERWRITE - enter the new key/data pair only if the key
1037 * does not already appear in the database. The function will return
1038 * #MDB_KEYEXIST if the key already appears in the database, even if
1039 * the database supports duplicates (#MDB_DUPSORT).
1040 * <li>#MDB_RESERVE - reserve space for data of the given size, but
1041 * don't copy the given data. Instead, return a pointer to the
1042 * reserved space, which the caller can fill in later. This saves
1043 * an extra memcpy if the data is being generated later.
1044 * <li>#MDB_APPEND - append the given key/data pair to the end of the
1045 * database. No key comparisons are performed. This option allows
1046 * fast bulk loading when keys are already known to be in the
1047 * correct order. Loading unsorted keys with this flag will cause
1049 * <li>#MDB_APPENDDUP - as above, but for sorted dup data.
1051 * @return A non-zero error value on failure and 0 on success. Some possible
1054 * <li>EACCES - an attempt was made to modify a read-only database.
1055 * <li>EINVAL - an invalid parameter was specified.
1058 int mdb_cursor_put(MDB_cursor *cursor, MDB_val *key, MDB_val *data,
1059 unsigned int flags);
1061 /** @brief Delete current key/data pair
1063 * This function deletes the key/data pair to which the cursor refers.
1064 * @param[in] cursor A cursor handle returned by #mdb_cursor_open()
1065 * @param[in] flags Options for this operation. This parameter
1066 * must be set to 0 or one of the values described here.
1068 * <li>#MDB_NODUPDATA - delete all of the data items for the current key.
1069 * This flag may only be specified if the database was opened with #MDB_DUPSORT.
1071 * @return A non-zero error value on failure and 0 on success. Some possible
1074 * <li>EACCES - an attempt was made to modify a read-only database.
1075 * <li>EINVAL - an invalid parameter was specified.
1078 int mdb_cursor_del(MDB_cursor *cursor, unsigned int flags);
1080 /** @brief Return count of duplicates for current key.
1082 * This call is only valid on databases that support sorted duplicate
1083 * data items #MDB_DUPSORT.
1084 * @param[in] cursor A cursor handle returned by #mdb_cursor_open()
1085 * @param[out] countp Address where the count will be stored
1086 * @return A non-zero error value on failure and 0 on success. Some possible
1089 * <li>EINVAL - cursor is not initialized, or an invalid parameter was specified.
1092 int mdb_cursor_count(MDB_cursor *cursor, size_t *countp);
1094 /** @brief Compare two data items according to a particular database.
1096 * This returns a comparison as if the two data items were keys in the
1097 * specified database.
1098 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
1099 * @param[in] dbi A database handle returned by #mdb_open()
1100 * @param[in] a The first item to compare
1101 * @param[in] b The second item to compare
1102 * @return < 0 if a < b, 0 if a == b, > 0 if a > b
1104 int mdb_cmp(MDB_txn *txn, MDB_dbi dbi, const MDB_val *a, const MDB_val *b);
1106 /** @brief Compare two data items according to a particular database.
1108 * This returns a comparison as if the two items were data items of
1109 * a sorted duplicates #MDB_DUPSORT database.
1110 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
1111 * @param[in] dbi A database handle returned by #mdb_open()
1112 * @param[in] a The first item to compare
1113 * @param[in] b The second item to compare
1114 * @return < 0 if a < b, 0 if a == b, > 0 if a > b
1116 int mdb_dcmp(MDB_txn *txn, MDB_dbi dbi, const MDB_val *a, const MDB_val *b);
1122 #endif /* _MDB_H_ */