2 * @brief Lightning memory-mapped database library
4 * @mainpage Lightning Memory-Mapped Database Manager (LMDB)
6 * @section intro_sec Introduction
7 * LMDB is a Btree-based database management library modeled loosely on the
8 * BerkeleyDB API, but much simplified. The entire database is exposed
9 * in a memory map, and all data fetches return data directly
10 * from the mapped memory, so no malloc's or memcpy's occur during
11 * data fetches. As such, the library is extremely simple because it
12 * requires no page caching layer of its own, and it is extremely high
13 * performance and memory-efficient. It is also fully transactional with
14 * full ACID semantics, and when the memory map is read-only, the
15 * database integrity cannot be corrupted by stray pointer writes from
18 * The library is fully thread-aware and supports concurrent read/write
19 * access from multiple processes and threads. Data pages use a copy-on-
20 * write strategy so no active data pages are ever overwritten, which
21 * also provides resistance to corruption and eliminates the need of any
22 * special recovery procedures after a system crash. Writes are fully
23 * serialized; only one write transaction may be active at a time, which
24 * guarantees that writers can never deadlock. The database structure is
25 * multi-versioned so readers run with no locks; writers cannot block
26 * readers, and readers don't block writers.
28 * Unlike other well-known database mechanisms which use either write-ahead
29 * transaction logs or append-only data writes, LMDB requires no maintenance
30 * during operation. Both write-ahead loggers and append-only databases
31 * require periodic checkpointing and/or compaction of their log or database
32 * files otherwise they grow without bound. LMDB tracks free pages within
33 * the database and re-uses them for new write operations, so the database
34 * size does not grow without bound in normal use.
36 * The memory map can be used as a read-only or read-write map. It is
37 * read-only by default as this provides total immunity to corruption.
38 * Using read-write mode offers much higher write performance, but adds
39 * the possibility for stray application writes thru pointers to silently
40 * corrupt the database. Of course if your application code is known to
41 * be bug-free (...) then this is not an issue.
43 * If this is your first time using a transactional embedded key/value
44 * store, you may find the \ref starting page to be helpful.
46 * @section caveats_sec Caveats
47 * Troubleshooting the lock file, plus semaphores on BSD systems:
49 * - A broken lockfile can cause sync issues.
50 * Stale reader transactions left behind by an aborted program
51 * cause further writes to grow the database quickly, and
52 * stale locks can block further operation.
54 * Fix: Check for stale readers periodically, using the
55 * #mdb_reader_check function or the \ref mdb_stat_1 "mdb_stat" tool.
56 * Stale writers will be cleared automatically on most systems:
57 * - Windows - automatic
58 * - BSD, systems using SysV semaphores - automatic
59 * - Linux, systems using POSIX mutexes with Robust option - automatic
60 * Otherwise just make all programs using the database close it;
61 * the lockfile is always reset on first open of the environment.
63 * - On BSD systems or others configured with MDB_USE_SYSV_SEM or
65 * startup can fail due to semaphores owned by another userid.
67 * Fix: Open and close the database as the user which owns the
68 * semaphores (likely last user) or as root, while no other
69 * process is using the database.
71 * Restrictions/caveats (in addition to those listed for some functions):
73 * - Only the database owner should normally use the database on
74 * BSD systems or when otherwise configured with MDB_USE_POSIX_SEM.
75 * Multiple users can cause startup to fail later, as noted above.
77 * - There is normally no pure read-only mode, since readers need write
78 * access to locks and lock file. Exceptions: On read-only filesystems
79 * or with the #MDB_NOLOCK flag described under #mdb_env_open().
81 * - An LMDB configuration will often reserve considerable \b unused
82 * memory address space and maybe file size for future growth.
83 * This does not use actual memory or disk space, but users may need
84 * to understand the difference so they won't be scared off.
86 * - By default, in versions before 0.9.10, unused portions of the data
87 * file might receive garbage data from memory freed by other code.
88 * (This does not happen when using the #MDB_WRITEMAP flag.) As of
89 * 0.9.10 the default behavior is to initialize such memory before
90 * writing to the data file. Since there may be a slight performance
91 * cost due to this initialization, applications may disable it using
92 * the #MDB_NOMEMINIT flag. Applications handling sensitive data
93 * which must not be written should not use this flag. This flag is
94 * irrelevant when using #MDB_WRITEMAP.
96 * - A thread can only use one transaction at a time, plus any child
97 * transactions. Each transaction belongs to one thread. See below.
98 * The #MDB_NOTLS flag changes this for read-only transactions.
100 * - Use an MDB_env* in the process which opened it, not after fork().
102 * - Do not have open an LMDB database twice in the same process at
103 * the same time. Not even from a plain open() call - close()ing it
104 * breaks fcntl() advisory locking. (It is OK to reopen it after
105 * fork() - exec*(), since the lockfile has FD_CLOEXEC set.)
107 * - Avoid long-lived transactions. Read transactions prevent
108 * reuse of pages freed by newer write transactions, thus the
109 * database can grow quickly. Write transactions prevent
110 * other write transactions, since writes are serialized.
112 * - Avoid suspending a process with active transactions. These
113 * would then be "long-lived" as above. Also read transactions
114 * suspended when writers commit could sometimes see wrong data.
116 * ...when several processes can use a database concurrently:
118 * - Avoid aborting a process with an active transaction.
119 * The transaction becomes "long-lived" as above until a check
120 * for stale readers is performed or the lockfile is reset,
121 * since the process may not remove it from the lockfile.
123 * This does not apply to write transactions if the system clears
124 * stale writers, see above.
126 * - If you do that anyway, do a periodic check for stale readers. Or
127 * close the environment once in a while, so the lockfile can get reset.
129 * - Do not use LMDB databases on remote filesystems, even between
130 * processes on the same host. This breaks flock() on some OSes,
131 * possibly memory map sync, and certainly sync between programs
132 * on different hosts.
134 * - Opening a database can fail if another process is opening or
135 * closing it at exactly the same time.
137 * @author Howard Chu, Symas Corporation.
139 * @copyright Copyright 2011-2017 Howard Chu, Symas Corp. All rights reserved.
141 * Redistribution and use in source and binary forms, with or without
142 * modification, are permitted only as authorized by the OpenLDAP
145 * A copy of this license is available in the file LICENSE in the
146 * top-level directory of the distribution or, alternatively, at
147 * <http://www.OpenLDAP.org/license.html>.
150 * This code is derived from btree.c written by Martin Hedenfalk.
152 * Copyright (c) 2009, 2010 Martin Hedenfalk <martin@bzero.se>
154 * Permission to use, copy, modify, and distribute this software for any
155 * purpose with or without fee is hereby granted, provided that the above
156 * copyright notice and this permission notice appear in all copies.
158 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
159 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
160 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
161 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
162 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
163 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
164 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
169 #include <sys/types.h>
170 #include <inttypes.h>
177 /** Unix permissions for creating files, or dummy definition for Windows */
179 typedef int mdb_mode_t;
181 typedef mode_t mdb_mode_t;
185 # define MDB_FMT_Z "I"
187 # define MDB_FMT_Z "z" /**< printf/scanf format modifier for size_t */
191 /** Unsigned type used for mapsize, entry counts and page/transaction IDs.
193 * It is normally size_t, hence the name. Defining MDB_VL32 makes it
194 * uint64_t, but do not try this unless you know what you are doing.
196 typedef size_t mdb_size_t;
197 # define MDB_SIZE_MAX SIZE_MAX /**< max #mdb_size_t */
198 /** #mdb_size_t printf formats, \b t = one of [diouxX] without quotes */
199 # define MDB_PRIy(t) MDB_FMT_Z #t
200 /** #mdb_size_t scanf formats, \b t = one of [dioux] without quotes */
201 # define MDB_SCNy(t) MDB_FMT_Z #t
203 typedef uint64_t mdb_size_t;
204 # define MDB_SIZE_MAX UINT64_MAX
205 # define MDB_PRIy(t) PRI##t##64
206 # define MDB_SCNy(t) SCN##t##64
207 # define mdb_env_create mdb_env_create_vl32 /**< Prevent mixing with non-VL32 builds */
210 /** An abstraction for a file handle.
211 * On POSIX systems file handles are small integers. On Windows
212 * they're opaque pointers.
215 typedef void *mdb_filehandle_t;
217 typedef int mdb_filehandle_t;
220 /** @defgroup mdb LMDB API
222 * @brief OpenLDAP Lightning Memory-Mapped Database Manager
224 /** @defgroup Version Version Macros
227 /** Library major version */
228 #define MDB_VERSION_MAJOR 0
229 /** Library minor version */
230 #define MDB_VERSION_MINOR 9
231 /** Library patch version */
232 #define MDB_VERSION_PATCH 70
234 /** Combine args a,b,c into a single integer for easy version comparisons */
235 #define MDB_VERINT(a,b,c) (((a) << 24) | ((b) << 16) | (c))
237 /** The full library version as a single integer */
238 #define MDB_VERSION_FULL \
239 MDB_VERINT(MDB_VERSION_MAJOR,MDB_VERSION_MINOR,MDB_VERSION_PATCH)
241 /** The release date of this library version */
242 #define MDB_VERSION_DATE "December 19, 2015"
244 /** A stringifier for the version info */
245 #define MDB_VERSTR(a,b,c,d) "LMDB " #a "." #b "." #c ": (" d ")"
247 /** A helper for the stringifier macro */
248 #define MDB_VERFOO(a,b,c,d) MDB_VERSTR(a,b,c,d)
250 /** The full library version as a C string */
251 #define MDB_VERSION_STRING \
252 MDB_VERFOO(MDB_VERSION_MAJOR,MDB_VERSION_MINOR,MDB_VERSION_PATCH,MDB_VERSION_DATE)
255 /** @brief Opaque structure for a database environment.
257 * A DB environment supports multiple databases, all residing in the same
260 typedef struct MDB_env MDB_env;
262 /** @brief Opaque structure for a transaction handle.
264 * All database operations require a transaction handle. Transactions may be
265 * read-only or read-write.
267 typedef struct MDB_txn MDB_txn;
269 /** @brief A handle for an individual database in the DB environment. */
270 typedef unsigned int MDB_dbi;
272 /** @brief Opaque structure for navigating through a database */
273 typedef struct MDB_cursor MDB_cursor;
275 /** @brief Generic structure used for passing keys and data in and out
278 * Values returned from the database are valid only until a subsequent
279 * update operation, or the end of the transaction. Do not modify or
280 * free them, they commonly point into the database itself.
282 * Key sizes must be between 1 and #mdb_env_get_maxkeysize() inclusive.
283 * The same applies to data sizes in databases with the #MDB_DUPSORT flag.
284 * Other data items can in theory be from 0 to 0xffffffff bytes long.
286 typedef struct MDB_val {
287 size_t mv_size; /**< size of the data item */
288 void *mv_data; /**< address of the data item */
291 /** @brief A callback function used to compare two keys in a database */
292 typedef int (MDB_cmp_func)(const MDB_val *a, const MDB_val *b);
294 /** @brief A callback function used to relocate a position-dependent data item
295 * in a fixed-address database.
297 * The \b newptr gives the item's desired address in
298 * the memory map, and \b oldptr gives its previous address. The item's actual
299 * data resides at the address in \b item. This callback is expected to walk
300 * through the fields of the record in \b item and modify any
301 * values based at the \b oldptr address to be relative to the \b newptr address.
302 * @param[in,out] item The item that is to be relocated.
303 * @param[in] oldptr The previous address.
304 * @param[in] newptr The new address to relocate to.
305 * @param[in] relctx An application-provided context, set by #mdb_set_relctx().
306 * @todo This feature is currently unimplemented.
308 typedef void (MDB_rel_func)(MDB_val *item, void *oldptr, void *newptr, void *relctx);
310 /** @defgroup mdb_env Environment Flags
313 /** mmap at a fixed address (experimental) */
314 #define MDB_FIXEDMAP 0x01
315 /** no environment directory */
316 #define MDB_NOSUBDIR 0x4000
317 /** don't fsync after commit */
318 #define MDB_NOSYNC 0x10000
320 #define MDB_RDONLY 0x20000
321 /** don't fsync metapage after commit */
322 #define MDB_NOMETASYNC 0x40000
323 /** use writable mmap */
324 #define MDB_WRITEMAP 0x80000
325 /** use asynchronous msync when #MDB_WRITEMAP is used */
326 #define MDB_MAPASYNC 0x100000
327 /** tie reader locktable slots to #MDB_txn objects instead of to threads */
328 #define MDB_NOTLS 0x200000
329 /** don't do any locking, caller must manage their own locks */
330 #define MDB_NOLOCK 0x400000
331 /** don't do readahead (no effect on Windows) */
332 #define MDB_NORDAHEAD 0x800000
333 /** don't initialize malloc'd memory before writing to datafile */
334 #define MDB_NOMEMINIT 0x1000000
335 /** use the previous meta page rather than the latest one */
336 #define MDB_PREVMETA 0x2000000
339 /** @defgroup mdb_dbi_open Database Flags
342 /** use reverse string keys */
343 #define MDB_REVERSEKEY 0x02
344 /** use sorted duplicates */
345 #define MDB_DUPSORT 0x04
346 /** numeric keys in native byte order, either unsigned int or #mdb_size_t.
347 * (lmdb expects 32-bit int <= size_t <= 32/64-bit mdb_size_t.)
348 * The keys must all be of the same size. */
349 #define MDB_INTEGERKEY 0x08
350 /** with #MDB_DUPSORT, sorted dup items have fixed size */
351 #define MDB_DUPFIXED 0x10
352 /** with #MDB_DUPSORT, dups are #MDB_INTEGERKEY-style integers */
353 #define MDB_INTEGERDUP 0x20
354 /** with #MDB_DUPSORT, use reverse string dups */
355 #define MDB_REVERSEDUP 0x40
356 /** create DB if not already existing */
357 #define MDB_CREATE 0x40000
360 /** @defgroup mdb_put Write Flags
363 /** For put: Don't write if the key already exists. */
364 #define MDB_NOOVERWRITE 0x10
365 /** Only for #MDB_DUPSORT<br>
366 * For put: don't write if the key and data pair already exist.<br>
367 * For mdb_cursor_del: remove all duplicate data items.
369 #define MDB_NODUPDATA 0x20
370 /** For mdb_cursor_put: overwrite the current key/data pair */
371 #define MDB_CURRENT 0x40
372 /** For put: Just reserve space for data, don't copy it. Return a
373 * pointer to the reserved space.
375 #define MDB_RESERVE 0x10000
376 /** Data is being appended, don't split full pages. */
377 #define MDB_APPEND 0x20000
378 /** Duplicate data is being appended, don't split full pages. */
379 #define MDB_APPENDDUP 0x40000
380 /** Store multiple data items in one call. Only for #MDB_DUPFIXED. */
381 #define MDB_MULTIPLE 0x80000
384 /** @defgroup mdb_copy Copy Flags
387 /** Compacting copy: Omit free space from copy, and renumber all
388 * pages sequentially.
390 #define MDB_CP_COMPACT 0x01
393 /** @brief Cursor Get operations.
395 * This is the set of all operations for retrieving data
398 typedef enum MDB_cursor_op {
399 MDB_FIRST, /**< Position at first key/data item */
400 MDB_FIRST_DUP, /**< Position at first data item of current key.
401 Only for #MDB_DUPSORT */
402 MDB_GET_BOTH, /**< Position at key/data pair. Only for #MDB_DUPSORT */
403 MDB_GET_BOTH_RANGE, /**< position at key, nearest data. Only for #MDB_DUPSORT */
404 MDB_GET_CURRENT, /**< Return key/data at current cursor position */
405 MDB_GET_MULTIPLE, /**< Return key and up to a page of duplicate data items
406 from current cursor position. Move cursor to prepare
407 for #MDB_NEXT_MULTIPLE. Only for #MDB_DUPFIXED */
408 MDB_LAST, /**< Position at last key/data item */
409 MDB_LAST_DUP, /**< Position at last data item of current key.
410 Only for #MDB_DUPSORT */
411 MDB_NEXT, /**< Position at next data item */
412 MDB_NEXT_DUP, /**< Position at next data item of current key.
413 Only for #MDB_DUPSORT */
414 MDB_NEXT_MULTIPLE, /**< Return key and up to a page of duplicate data items
415 from next cursor position. Move cursor to prepare
416 for #MDB_NEXT_MULTIPLE. Only for #MDB_DUPFIXED */
417 MDB_NEXT_NODUP, /**< Position at first data item of next key */
418 MDB_PREV, /**< Position at previous data item */
419 MDB_PREV_DUP, /**< Position at previous data item of current key.
420 Only for #MDB_DUPSORT */
421 MDB_PREV_NODUP, /**< Position at last data item of previous key */
422 MDB_SET, /**< Position at specified key */
423 MDB_SET_KEY, /**< Position at specified key, return key + data */
424 MDB_SET_RANGE, /**< Position at first key greater than or equal to specified key. */
425 MDB_PREV_MULTIPLE /**< Position at previous page and return key and up to
426 a page of duplicate data items. Only for #MDB_DUPFIXED */
429 /** @defgroup errors Return Codes
431 * BerkeleyDB uses -30800 to -30999, we'll go under them
434 /** Successful result */
435 #define MDB_SUCCESS 0
436 /** key/data pair already exists */
437 #define MDB_KEYEXIST (-30799)
438 /** key/data pair not found (EOF) */
439 #define MDB_NOTFOUND (-30798)
440 /** Requested page not found - this usually indicates corruption */
441 #define MDB_PAGE_NOTFOUND (-30797)
442 /** Located page was wrong type */
443 #define MDB_CORRUPTED (-30796)
444 /** Update of meta page failed or environment had fatal error */
445 #define MDB_PANIC (-30795)
446 /** Environment version mismatch */
447 #define MDB_VERSION_MISMATCH (-30794)
448 /** File is not a valid LMDB file */
449 #define MDB_INVALID (-30793)
450 /** Environment mapsize reached */
451 #define MDB_MAP_FULL (-30792)
452 /** Environment maxdbs reached */
453 #define MDB_DBS_FULL (-30791)
454 /** Environment maxreaders reached */
455 #define MDB_READERS_FULL (-30790)
456 /** Too many TLS keys in use - Windows only */
457 #define MDB_TLS_FULL (-30789)
458 /** Txn has too many dirty pages */
459 #define MDB_TXN_FULL (-30788)
460 /** Cursor stack too deep - internal error */
461 #define MDB_CURSOR_FULL (-30787)
462 /** Page has not enough space - internal error */
463 #define MDB_PAGE_FULL (-30786)
464 /** Database contents grew beyond environment mapsize */
465 #define MDB_MAP_RESIZED (-30785)
466 /** Operation and DB incompatible, or DB type changed. This can mean:
468 * <li>The operation expects an #MDB_DUPSORT / #MDB_DUPFIXED database.
469 * <li>Opening a named DB when the unnamed DB has #MDB_DUPSORT / #MDB_INTEGERKEY.
470 * <li>Accessing a data record as a database, or vice versa.
471 * <li>The database was dropped and recreated with different flags.
474 #define MDB_INCOMPATIBLE (-30784)
475 /** Invalid reuse of reader locktable slot */
476 #define MDB_BAD_RSLOT (-30783)
477 /** Transaction must abort, has a child, or is invalid */
478 #define MDB_BAD_TXN (-30782)
479 /** Unsupported size of key/DB name/data, or wrong DUPFIXED size */
480 #define MDB_BAD_VALSIZE (-30781)
481 /** The specified DBI was changed unexpectedly */
482 #define MDB_BAD_DBI (-30780)
483 /** Unexpected problem - txn should abort */
484 #define MDB_PROBLEM (-30779)
485 /** The last defined error code */
486 #define MDB_LAST_ERRCODE MDB_PROBLEM
489 /** @brief Statistics for a database in the environment */
490 typedef struct MDB_stat {
491 unsigned int ms_psize; /**< Size of a database page.
492 This is currently the same for all databases. */
493 unsigned int ms_depth; /**< Depth (height) of the B-tree */
494 mdb_size_t ms_branch_pages; /**< Number of internal (non-leaf) pages */
495 mdb_size_t ms_leaf_pages; /**< Number of leaf pages */
496 mdb_size_t ms_overflow_pages; /**< Number of overflow pages */
497 mdb_size_t ms_entries; /**< Number of data items */
500 /** @brief Information about the environment */
501 typedef struct MDB_envinfo {
502 void *me_mapaddr; /**< Address of map, if fixed */
503 mdb_size_t me_mapsize; /**< Size of the data memory map */
504 mdb_size_t me_last_pgno; /**< ID of the last used page */
505 mdb_size_t me_last_txnid; /**< ID of the last committed transaction */
506 unsigned int me_maxreaders; /**< max reader slots in the environment */
507 unsigned int me_numreaders; /**< max reader slots used in the environment */
510 /** @brief Return the LMDB library version information.
512 * @param[out] major if non-NULL, the library major version number is copied here
513 * @param[out] minor if non-NULL, the library minor version number is copied here
514 * @param[out] patch if non-NULL, the library patch version number is copied here
515 * @retval "version string" The library version as a string
517 char *mdb_version(int *major, int *minor, int *patch);
519 /** @brief Return a string describing a given error code.
521 * This function is a superset of the ANSI C X3.159-1989 (ANSI C) strerror(3)
522 * function. If the error code is greater than or equal to 0, then the string
523 * returned by the system function strerror(3) is returned. If the error code
524 * is less than 0, an error string corresponding to the LMDB library error is
525 * returned. See @ref errors for a list of LMDB-specific error codes.
526 * @param[in] err The error code
527 * @retval "error message" The description of the error
529 char *mdb_strerror(int err);
531 /** @brief Create an LMDB environment handle.
533 * This function allocates memory for a #MDB_env structure. To release
534 * the allocated memory and discard the handle, call #mdb_env_close().
535 * Before the handle may be used, it must be opened using #mdb_env_open().
536 * Various other options may also need to be set before opening the handle,
537 * e.g. #mdb_env_set_mapsize(), #mdb_env_set_maxreaders(), #mdb_env_set_maxdbs(),
538 * depending on usage requirements.
539 * @param[out] env The address where the new handle will be stored
540 * @return A non-zero error value on failure and 0 on success.
542 int mdb_env_create(MDB_env **env);
544 /** @brief Open an environment handle.
546 * If this function fails, #mdb_env_close() must be called to discard the #MDB_env handle.
547 * @param[in] env An environment handle returned by #mdb_env_create()
548 * @param[in] path The directory in which the database files reside. This
549 * directory must already exist and be writable.
550 * @param[in] flags Special options for this environment. This parameter
551 * must be set to 0 or by bitwise OR'ing together one or more of the
552 * values described here.
553 * Flags set by mdb_env_set_flags() are also used.
556 * use a fixed address for the mmap region. This flag must be specified
557 * when creating the environment, and is stored persistently in the environment.
558 * If successful, the memory map will always reside at the same virtual address
559 * and pointers used to reference data items in the database will be constant
560 * across multiple invocations. This option may not always work, depending on
561 * how the operating system has allocated memory to shared libraries and other uses.
562 * The feature is highly experimental.
564 * By default, LMDB creates its environment in a directory whose
565 * pathname is given in \b path, and creates its data and lock files
566 * under that directory. With this option, \b path is used as-is for
567 * the database main data file. The database lock file is the \b path
568 * with "-lock" appended.
570 * Open the environment in read-only mode. No write operations will be
571 * allowed. LMDB will still modify the lock file - except on read-only
572 * filesystems, where LMDB does not use locks.
574 * Use a writeable memory map unless MDB_RDONLY is set. This uses
575 * fewer mallocs but loses protection from application bugs
576 * like wild pointer writes and other bad updates into the database.
577 * This may be slightly faster for DBs that fit entirely in RAM, but
578 * is slower for DBs larger than RAM.
579 * Incompatible with nested transactions.
580 * Do not mix processes with and without MDB_WRITEMAP on the same
581 * environment. This can defeat durability (#mdb_env_sync etc).
582 * <li>#MDB_NOMETASYNC
583 * Flush system buffers to disk only once per transaction, omit the
584 * metadata flush. Defer that until the system flushes files to disk,
585 * or next non-MDB_RDONLY commit or #mdb_env_sync(). This optimization
586 * maintains database integrity, but a system crash may undo the last
587 * committed transaction. I.e. it preserves the ACI (atomicity,
588 * consistency, isolation) but not D (durability) database property.
589 * This flag may be changed at any time using #mdb_env_set_flags().
591 * Don't flush system buffers to disk when committing a transaction.
592 * This optimization means a system crash can corrupt the database or
593 * lose the last transactions if buffers are not yet flushed to disk.
594 * The risk is governed by how often the system flushes dirty buffers
595 * to disk and how often #mdb_env_sync() is called. However, if the
596 * filesystem preserves write order and the #MDB_WRITEMAP flag is not
597 * used, transactions exhibit ACI (atomicity, consistency, isolation)
598 * properties and only lose D (durability). I.e. database integrity
599 * is maintained, but a system crash may undo the final transactions.
600 * Note that (#MDB_NOSYNC | #MDB_WRITEMAP) leaves the system with no
601 * hint for when to write transactions to disk, unless #mdb_env_sync()
602 * is called. (#MDB_MAPASYNC | #MDB_WRITEMAP) may be preferable.
603 * This flag may be changed at any time using #mdb_env_set_flags().
605 * When using #MDB_WRITEMAP, use asynchronous flushes to disk.
606 * As with #MDB_NOSYNC, a system crash can then corrupt the
607 * database or lose the last transactions. Calling #mdb_env_sync()
608 * ensures on-disk database integrity until next commit.
609 * This flag may be changed at any time using #mdb_env_set_flags().
611 * Don't use Thread-Local Storage. Tie reader locktable slots to
612 * #MDB_txn objects instead of to threads. I.e. #mdb_txn_reset() keeps
613 * the slot reseved for the #MDB_txn object. A thread may use parallel
614 * read-only transactions. A read-only transaction may span threads if
615 * the user synchronizes its use. Applications that multiplex many
616 * user threads over individual OS threads need this option. Such an
617 * application must also serialize the write transactions in an OS
618 * thread, since LMDB's write locking is unaware of the user threads.
620 * Don't do any locking. If concurrent access is anticipated, the
621 * caller must manage all concurrency itself. For proper operation
622 * the caller must enforce single-writer semantics, and must ensure
623 * that no readers are using old transactions while a writer is
624 * active. The simplest approach is to use an exclusive lock so that
625 * no readers may be active at all when a writer begins.
627 * Turn off readahead. Most operating systems perform readahead on
628 * read requests by default. This option turns it off if the OS
629 * supports it. Turning it off may help random read performance
630 * when the DB is larger than RAM and system RAM is full.
631 * The option is not implemented on Windows.
633 * Don't initialize malloc'd memory before writing to unused spaces
634 * in the data file. By default, memory for pages written to the data
635 * file is obtained using malloc. While these pages may be reused in
636 * subsequent transactions, freshly malloc'd pages will be initialized
637 * to zeroes before use. This avoids persisting leftover data from other
638 * code (that used the heap and subsequently freed the memory) into the
639 * data file. Note that many other system libraries may allocate
640 * and free memory from the heap for arbitrary uses. E.g., stdio may
641 * use the heap for file I/O buffers. This initialization step has a
642 * modest performance cost so some applications may want to disable
643 * it using this flag. This option can be a problem for applications
644 * which handle sensitive data like passwords, and it makes memory
645 * checkers like Valgrind noisy. This flag is not needed with #MDB_WRITEMAP,
646 * which writes directly to the mmap instead of using malloc for pages. The
647 * initialization is also skipped if #MDB_RESERVE is used; the
648 * caller is expected to overwrite all of the memory that was
649 * reserved in that case.
650 * This flag may be changed at any time using #mdb_env_set_flags().
652 * Open the environment with the previous meta page rather than the latest
653 * one. This loses the latest transaction, but may help work around some
654 * types of corruption.
656 * @param[in] mode The UNIX permissions to set on created files and semaphores.
657 * This parameter is ignored on Windows.
658 * @return A non-zero error value on failure and 0 on success. Some possible
661 * <li>#MDB_VERSION_MISMATCH - the version of the LMDB library doesn't match the
662 * version that created the database environment.
663 * <li>#MDB_INVALID - the environment file headers are corrupted.
664 * <li>ENOENT - the directory specified by the path parameter doesn't exist.
665 * <li>EACCES - the user didn't have permission to access the environment files.
666 * <li>EAGAIN - the environment was locked by another process.
669 int mdb_env_open(MDB_env *env, const char *path, unsigned int flags, mdb_mode_t mode);
671 /** @brief Copy an LMDB environment to the specified path.
673 * This function may be used to make a backup of an existing environment.
674 * No lockfile is created, since it gets recreated at need.
675 * @note This call can trigger significant file size growth if run in
676 * parallel with write transactions, because it employs a read-only
677 * transaction. See long-lived transactions under @ref caveats_sec.
678 * @param[in] env An environment handle returned by #mdb_env_create(). It
679 * must have already been opened successfully.
680 * @param[in] path The directory in which the copy will reside. This
681 * directory must already exist and be writable but must otherwise be
683 * @return A non-zero error value on failure and 0 on success.
685 int mdb_env_copy(MDB_env *env, const char *path);
687 /** @brief Copy an LMDB environment to the specified file descriptor.
689 * This function may be used to make a backup of an existing environment.
690 * No lockfile is created, since it gets recreated at need.
691 * @note This call can trigger significant file size growth if run in
692 * parallel with write transactions, because it employs a read-only
693 * transaction. See long-lived transactions under @ref caveats_sec.
694 * @param[in] env An environment handle returned by #mdb_env_create(). It
695 * must have already been opened successfully.
696 * @param[in] fd The filedescriptor to write the copy to. It must
697 * have already been opened for Write access.
698 * @return A non-zero error value on failure and 0 on success.
700 int mdb_env_copyfd(MDB_env *env, mdb_filehandle_t fd);
702 /** @brief Copy an LMDB environment to the specified path, with options.
704 * This function may be used to make a backup of an existing environment.
705 * No lockfile is created, since it gets recreated at need.
706 * @note This call can trigger significant file size growth if run in
707 * parallel with write transactions, because it employs a read-only
708 * transaction. See long-lived transactions under @ref caveats_sec.
709 * @param[in] env An environment handle returned by #mdb_env_create(). It
710 * must have already been opened successfully.
711 * @param[in] path The directory in which the copy will reside. This
712 * directory must already exist and be writable but must otherwise be
714 * @param[in] flags Special options for this operation. This parameter
715 * must be set to 0 or by bitwise OR'ing together one or more of the
716 * values described here.
718 * <li>#MDB_CP_COMPACT - Perform compaction while copying: omit free
719 * pages and sequentially renumber all pages in output. This option
720 * consumes more CPU and runs more slowly than the default.
721 * Currently it fails if the environment has suffered a page leak.
723 * @return A non-zero error value on failure and 0 on success.
725 int mdb_env_copy2(MDB_env *env, const char *path, unsigned int flags);
727 /** @brief Copy an LMDB environment to the specified file descriptor,
730 * This function may be used to make a backup of an existing environment.
731 * No lockfile is created, since it gets recreated at need. See
732 * #mdb_env_copy2() for further details.
733 * @note This call can trigger significant file size growth if run in
734 * parallel with write transactions, because it employs a read-only
735 * transaction. See long-lived transactions under @ref caveats_sec.
736 * @param[in] env An environment handle returned by #mdb_env_create(). It
737 * must have already been opened successfully.
738 * @param[in] fd The filedescriptor to write the copy to. It must
739 * have already been opened for Write access.
740 * @param[in] flags Special options for this operation.
741 * See #mdb_env_copy2() for options.
742 * @return A non-zero error value on failure and 0 on success.
744 int mdb_env_copyfd2(MDB_env *env, mdb_filehandle_t fd, unsigned int flags);
746 /** @brief Return statistics about the LMDB environment.
748 * @param[in] env An environment handle returned by #mdb_env_create()
749 * @param[out] stat The address of an #MDB_stat structure
750 * where the statistics will be copied
752 int mdb_env_stat(MDB_env *env, MDB_stat *stat);
754 /** @brief Return information about the LMDB environment.
756 * @param[in] env An environment handle returned by #mdb_env_create()
757 * @param[out] stat The address of an #MDB_envinfo structure
758 * where the information will be copied
760 int mdb_env_info(MDB_env *env, MDB_envinfo *stat);
762 /** @brief Flush the data buffers to disk.
764 * Data is always written to disk when #mdb_txn_commit() is called,
765 * but the operating system may keep it buffered. LMDB always flushes
766 * the OS buffers upon commit as well, unless the environment was
767 * opened with #MDB_NOSYNC or in part #MDB_NOMETASYNC. This call is
768 * not valid if the environment was opened with #MDB_RDONLY.
769 * @param[in] env An environment handle returned by #mdb_env_create()
770 * @param[in] force If non-zero, force a synchronous flush. Otherwise
771 * if the environment has the #MDB_NOSYNC flag set the flushes
772 * will be omitted, and with #MDB_MAPASYNC they will be asynchronous.
773 * @return A non-zero error value on failure and 0 on success. Some possible
776 * <li>EACCES - the environment is read-only.
777 * <li>EINVAL - an invalid parameter was specified.
778 * <li>EIO - an error occurred during synchronization.
781 int mdb_env_sync(MDB_env *env, int force);
783 /** @brief Close the environment and release the memory map.
785 * Only a single thread may call this function. All transactions, databases,
786 * and cursors must already be closed before calling this function. Attempts to
787 * use any such handles after calling this function will cause a SIGSEGV.
788 * The environment handle will be freed and must not be used again after this call.
789 * @param[in] env An environment handle returned by #mdb_env_create()
791 void mdb_env_close(MDB_env *env);
793 /** @brief Set environment flags.
795 * This may be used to set some flags in addition to those from
796 * #mdb_env_open(), or to unset these flags. If several threads
797 * change the flags at the same time, the result is undefined.
798 * @param[in] env An environment handle returned by #mdb_env_create()
799 * @param[in] flags The flags to change, bitwise OR'ed together
800 * @param[in] onoff A non-zero value sets the flags, zero clears them.
801 * @return A non-zero error value on failure and 0 on success. Some possible
804 * <li>EINVAL - an invalid parameter was specified.
807 int mdb_env_set_flags(MDB_env *env, unsigned int flags, int onoff);
809 /** @brief Get environment flags.
811 * @param[in] env An environment handle returned by #mdb_env_create()
812 * @param[out] flags The address of an integer to store the flags
813 * @return A non-zero error value on failure and 0 on success. Some possible
816 * <li>EINVAL - an invalid parameter was specified.
819 int mdb_env_get_flags(MDB_env *env, unsigned int *flags);
821 /** @brief Return the path that was used in #mdb_env_open().
823 * @param[in] env An environment handle returned by #mdb_env_create()
824 * @param[out] path Address of a string pointer to contain the path. This
825 * is the actual string in the environment, not a copy. It should not be
826 * altered in any way.
827 * @return A non-zero error value on failure and 0 on success. Some possible
830 * <li>EINVAL - an invalid parameter was specified.
833 int mdb_env_get_path(MDB_env *env, const char **path);
835 /** @brief Return the filedescriptor for the given environment.
837 * This function may be called after fork(), so the descriptor can be
838 * closed before exec*(). Other LMDB file descriptors have FD_CLOEXEC.
839 * (Until LMDB 0.9.18, only the lockfile had that.)
841 * @param[in] env An environment handle returned by #mdb_env_create()
842 * @param[out] fd Address of a mdb_filehandle_t to contain the descriptor.
843 * @return A non-zero error value on failure and 0 on success. Some possible
846 * <li>EINVAL - an invalid parameter was specified.
849 int mdb_env_get_fd(MDB_env *env, mdb_filehandle_t *fd);
851 /** @brief Set the size of the memory map to use for this environment.
853 * The size should be a multiple of the OS page size. The default is
854 * 10485760 bytes. The size of the memory map is also the maximum size
855 * of the database. The value should be chosen as large as possible,
856 * to accommodate future growth of the database.
857 * This function should be called after #mdb_env_create() and before #mdb_env_open().
858 * It may be called at later times if no transactions are active in
859 * this process. Note that the library does not check for this condition,
860 * the caller must ensure it explicitly.
862 * The new size takes effect immediately for the current process but
863 * will not be persisted to any others until a write transaction has been
864 * committed by the current process. Also, only mapsize increases are
865 * persisted into the environment.
867 * If the mapsize is increased by another process, and data has grown
868 * beyond the range of the current mapsize, #mdb_txn_begin() will
869 * return #MDB_MAP_RESIZED. This function may be called with a size
870 * of zero to adopt the new size.
872 * Any attempt to set a size smaller than the space already consumed
873 * by the environment will be silently changed to the current size of the used space.
874 * @param[in] env An environment handle returned by #mdb_env_create()
875 * @param[in] size The size in bytes
876 * @return A non-zero error value on failure and 0 on success. Some possible
879 * <li>EINVAL - an invalid parameter was specified, or the environment has
880 * an active write transaction.
883 int mdb_env_set_mapsize(MDB_env *env, mdb_size_t size);
885 /** @brief Set the maximum number of threads/reader slots for the environment.
887 * This defines the number of slots in the lock table that is used to track readers in the
888 * the environment. The default is 126.
889 * Starting a read-only transaction normally ties a lock table slot to the
890 * current thread until the environment closes or the thread exits. If
891 * MDB_NOTLS is in use, #mdb_txn_begin() instead ties the slot to the
892 * MDB_txn object until it or the #MDB_env object is destroyed.
893 * This function may only be called after #mdb_env_create() and before #mdb_env_open().
894 * @param[in] env An environment handle returned by #mdb_env_create()
895 * @param[in] readers The maximum number of reader lock table slots
896 * @return A non-zero error value on failure and 0 on success. Some possible
899 * <li>EINVAL - an invalid parameter was specified, or the environment is already open.
902 int mdb_env_set_maxreaders(MDB_env *env, unsigned int readers);
904 /** @brief Get the maximum number of threads/reader slots for the environment.
906 * @param[in] env An environment handle returned by #mdb_env_create()
907 * @param[out] readers Address of an integer to store the number of readers
908 * @return A non-zero error value on failure and 0 on success. Some possible
911 * <li>EINVAL - an invalid parameter was specified.
914 int mdb_env_get_maxreaders(MDB_env *env, unsigned int *readers);
916 /** @brief Set the maximum number of named databases for the environment.
918 * This function is only needed if multiple databases will be used in the
919 * environment. Simpler applications that use the environment as a single
920 * unnamed database can ignore this option.
921 * This function may only be called after #mdb_env_create() and before #mdb_env_open().
923 * Currently a moderate number of slots are cheap but a huge number gets
924 * expensive: 7-120 words per transaction, and every #mdb_dbi_open()
925 * does a linear search of the opened slots.
926 * @param[in] env An environment handle returned by #mdb_env_create()
927 * @param[in] dbs The maximum number of databases
928 * @return A non-zero error value on failure and 0 on success. Some possible
931 * <li>EINVAL - an invalid parameter was specified, or the environment is already open.
934 int mdb_env_set_maxdbs(MDB_env *env, MDB_dbi dbs);
936 /** @brief Get the maximum size of keys and #MDB_DUPSORT data we can write.
938 * Depends on the compile-time constant #MDB_MAXKEYSIZE. Default 511.
940 * @param[in] env An environment handle returned by #mdb_env_create()
941 * @return The maximum size of a key we can write
943 int mdb_env_get_maxkeysize(MDB_env *env);
945 /** @brief Set application information associated with the #MDB_env.
947 * @param[in] env An environment handle returned by #mdb_env_create()
948 * @param[in] ctx An arbitrary pointer for whatever the application needs.
949 * @return A non-zero error value on failure and 0 on success.
951 int mdb_env_set_userctx(MDB_env *env, void *ctx);
953 /** @brief Get the application information associated with the #MDB_env.
955 * @param[in] env An environment handle returned by #mdb_env_create()
956 * @return The pointer set by #mdb_env_set_userctx().
958 void *mdb_env_get_userctx(MDB_env *env);
960 /** @brief A callback function for most LMDB assert() failures,
961 * called before printing the message and aborting.
963 * @param[in] env An environment handle returned by #mdb_env_create().
964 * @param[in] msg The assertion message, not including newline.
966 typedef void MDB_assert_func(MDB_env *env, const char *msg);
968 /** Set or reset the assert() callback of the environment.
969 * Disabled if liblmdb is buillt with NDEBUG.
970 * @note This hack should become obsolete as lmdb's error handling matures.
971 * @param[in] env An environment handle returned by #mdb_env_create().
972 * @param[in] func An #MDB_assert_func function, or 0.
973 * @return A non-zero error value on failure and 0 on success.
975 int mdb_env_set_assert(MDB_env *env, MDB_assert_func *func);
977 /** @brief Create a transaction for use with the environment.
979 * The transaction handle may be discarded using #mdb_txn_abort() or #mdb_txn_commit().
980 * @note A transaction and its cursors must only be used by a single
981 * thread, and a thread may only have a single transaction at a time.
982 * If #MDB_NOTLS is in use, this does not apply to read-only transactions.
983 * @note Cursors may not span transactions.
984 * @param[in] env An environment handle returned by #mdb_env_create()
985 * @param[in] parent If this parameter is non-NULL, the new transaction
986 * will be a nested transaction, with the transaction indicated by \b parent
987 * as its parent. Transactions may be nested to any level. A parent
988 * transaction and its cursors may not issue any other operations than
989 * mdb_txn_commit and mdb_txn_abort while it has active child transactions.
990 * @param[in] flags Special options for this transaction. This parameter
991 * must be set to 0 or by bitwise OR'ing together one or more of the
992 * values described here.
995 * This transaction will not perform any write operations.
997 * Don't flush system buffers to disk when committing this transaction.
998 * <li>#MDB_NOMETASYNC
999 * Flush system buffers but omit metadata flush when committing this transaction.
1001 * @param[out] txn Address where the new #MDB_txn handle will be stored
1002 * @return A non-zero error value on failure and 0 on success. Some possible
1005 * <li>#MDB_PANIC - a fatal error occurred earlier and the environment
1006 * must be shut down.
1007 * <li>#MDB_MAP_RESIZED - another process wrote data beyond this MDB_env's
1008 * mapsize and this environment's map must be resized as well.
1009 * See #mdb_env_set_mapsize().
1010 * <li>#MDB_READERS_FULL - a read-only transaction was requested and
1011 * the reader lock table is full. See #mdb_env_set_maxreaders().
1012 * <li>ENOMEM - out of memory.
1015 int mdb_txn_begin(MDB_env *env, MDB_txn *parent, unsigned int flags, MDB_txn **txn);
1017 /** @brief Returns the transaction's #MDB_env
1019 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
1021 MDB_env *mdb_txn_env(MDB_txn *txn);
1023 /** @brief Return the transaction's ID.
1025 * This returns the identifier associated with this transaction. For a
1026 * read-only transaction, this corresponds to the snapshot being read;
1027 * concurrent readers will frequently have the same transaction ID.
1029 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
1030 * @return A transaction ID, valid if input is an active transaction.
1032 mdb_size_t mdb_txn_id(MDB_txn *txn);
1034 /** @brief Commit all the operations of a transaction into the database.
1036 * The transaction handle is freed. It and its cursors must not be used
1037 * again after this call, except with #mdb_cursor_renew().
1038 * @note Earlier documentation incorrectly said all cursors would be freed.
1039 * Only write-transactions free cursors.
1040 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
1041 * @return A non-zero error value on failure and 0 on success. Some possible
1044 * <li>EINVAL - an invalid parameter was specified.
1045 * <li>ENOSPC - no more disk space.
1046 * <li>EIO - a low-level I/O error occurred while writing.
1047 * <li>ENOMEM - out of memory.
1050 int mdb_txn_commit(MDB_txn *txn);
1052 /** @brief Abandon all the operations of the transaction instead of saving them.
1054 * The transaction handle is freed. It and its cursors must not be used
1055 * again after this call, except with #mdb_cursor_renew().
1056 * @note Earlier documentation incorrectly said all cursors would be freed.
1057 * Only write-transactions free cursors.
1058 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
1060 void mdb_txn_abort(MDB_txn *txn);
1062 /** @brief Reset a read-only transaction.
1064 * Abort the transaction like #mdb_txn_abort(), but keep the transaction
1065 * handle. #mdb_txn_renew() may reuse the handle. This saves allocation
1066 * overhead if the process will start a new read-only transaction soon,
1067 * and also locking overhead if #MDB_NOTLS is in use. The reader table
1068 * lock is released, but the table slot stays tied to its thread or
1069 * #MDB_txn. Use mdb_txn_abort() to discard a reset handle, and to free
1070 * its lock table slot if MDB_NOTLS is in use.
1071 * Cursors opened within the transaction must not be used
1072 * again after this call, except with #mdb_cursor_renew().
1073 * Reader locks generally don't interfere with writers, but they keep old
1074 * versions of database pages allocated. Thus they prevent the old pages
1075 * from being reused when writers commit new data, and so under heavy load
1076 * the database size may grow much more rapidly than otherwise.
1077 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
1079 void mdb_txn_reset(MDB_txn *txn);
1081 /** @brief Renew a read-only transaction.
1083 * This acquires a new reader lock for a transaction handle that had been
1084 * released by #mdb_txn_reset(). It must be called before a reset transaction
1085 * may be used again.
1086 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
1087 * @return A non-zero error value on failure and 0 on success. Some possible
1090 * <li>#MDB_PANIC - a fatal error occurred earlier and the environment
1091 * must be shut down.
1092 * <li>EINVAL - an invalid parameter was specified.
1095 int mdb_txn_renew(MDB_txn *txn);
1097 /** Compat with version <= 0.9.4, avoid clash with libmdb from MDB Tools project */
1098 #define mdb_open(txn,name,flags,dbi) mdb_dbi_open(txn,name,flags,dbi)
1099 /** Compat with version <= 0.9.4, avoid clash with libmdb from MDB Tools project */
1100 #define mdb_close(env,dbi) mdb_dbi_close(env,dbi)
1102 /** @brief Open a database in the environment.
1104 * A database handle denotes the name and parameters of a database,
1105 * independently of whether such a database exists.
1106 * The database handle may be discarded by calling #mdb_dbi_close().
1107 * The old database handle is returned if the database was already open.
1108 * The handle may only be closed once.
1110 * The database handle will be private to the current transaction until
1111 * the transaction is successfully committed. If the transaction is
1112 * aborted the handle will be closed automatically.
1113 * After a successful commit the handle will reside in the shared
1114 * environment, and may be used by other transactions.
1116 * This function must not be called from multiple concurrent
1117 * transactions in the same process. A transaction that uses
1118 * this function must finish (either commit or abort) before
1119 * any other transaction in the process may use this function.
1121 * To use named databases (with name != NULL), #mdb_env_set_maxdbs()
1122 * must be called before opening the environment. Database names are
1123 * keys in the unnamed database, and may be read but not written.
1125 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
1126 * @param[in] name The name of the database to open. If only a single
1127 * database is needed in the environment, this value may be NULL.
1128 * @param[in] flags Special options for this database. This parameter
1129 * must be set to 0 or by bitwise OR'ing together one or more of the
1130 * values described here.
1132 * <li>#MDB_REVERSEKEY
1133 * Keys are strings to be compared in reverse order, from the end
1134 * of the strings to the beginning. By default, Keys are treated as strings and
1135 * compared from beginning to end.
1137 * Duplicate keys may be used in the database. (Or, from another perspective,
1138 * keys may have multiple data items, stored in sorted order.) By default
1139 * keys must be unique and may have only a single data item.
1140 * <li>#MDB_INTEGERKEY
1141 * Keys are binary integers in native byte order, either unsigned int
1142 * or #mdb_size_t, and will be sorted as such.
1143 * (lmdb expects 32-bit int <= size_t <= 32/64-bit mdb_size_t.)
1144 * The keys must all be of the same size.
1146 * This flag may only be used in combination with #MDB_DUPSORT. This option
1147 * tells the library that the data items for this database are all the same
1148 * size, which allows further optimizations in storage and retrieval. When
1149 * all data items are the same size, the #MDB_GET_MULTIPLE, #MDB_NEXT_MULTIPLE
1150 * and #MDB_PREV_MULTIPLE cursor operations may be used to retrieve multiple
1152 * <li>#MDB_INTEGERDUP
1153 * This option specifies that duplicate data items are binary integers,
1154 * similar to #MDB_INTEGERKEY keys.
1155 * <li>#MDB_REVERSEDUP
1156 * This option specifies that duplicate data items should be compared as
1157 * strings in reverse order.
1159 * Create the named database if it doesn't exist. This option is not
1160 * allowed in a read-only transaction or a read-only environment.
1162 * @param[out] dbi Address where the new #MDB_dbi handle will be stored
1163 * @return A non-zero error value on failure and 0 on success. Some possible
1166 * <li>#MDB_NOTFOUND - the specified database doesn't exist in the environment
1167 * and #MDB_CREATE was not specified.
1168 * <li>#MDB_DBS_FULL - too many databases have been opened. See #mdb_env_set_maxdbs().
1171 int mdb_dbi_open(MDB_txn *txn, const char *name, unsigned int flags, MDB_dbi *dbi);
1173 /** @brief Retrieve statistics for a database.
1175 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
1176 * @param[in] dbi A database handle returned by #mdb_dbi_open()
1177 * @param[out] stat The address of an #MDB_stat structure
1178 * where the statistics will be copied
1179 * @return A non-zero error value on failure and 0 on success. Some possible
1182 * <li>EINVAL - an invalid parameter was specified.
1185 int mdb_stat(MDB_txn *txn, MDB_dbi dbi, MDB_stat *stat);
1187 /** @brief Retrieve the DB flags for a database handle.
1189 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
1190 * @param[in] dbi A database handle returned by #mdb_dbi_open()
1191 * @param[out] flags Address where the flags will be returned.
1192 * @return A non-zero error value on failure and 0 on success.
1194 int mdb_dbi_flags(MDB_txn *txn, MDB_dbi dbi, unsigned int *flags);
1196 /** @brief Close a database handle. Normally unnecessary. Use with care:
1198 * This call is not mutex protected. Handles should only be closed by
1199 * a single thread, and only if no other threads are going to reference
1200 * the database handle or one of its cursors any further. Do not close
1201 * a handle if an existing transaction has modified its database.
1202 * Doing so can cause misbehavior from database corruption to errors
1203 * like MDB_BAD_VALSIZE (since the DB name is gone).
1205 * Closing a database handle is not necessary, but lets #mdb_dbi_open()
1206 * reuse the handle value. Usually it's better to set a bigger
1207 * #mdb_env_set_maxdbs(), unless that value would be large.
1209 * @param[in] env An environment handle returned by #mdb_env_create()
1210 * @param[in] dbi A database handle returned by #mdb_dbi_open()
1212 void mdb_dbi_close(MDB_env *env, MDB_dbi dbi);
1214 /** @brief Empty or delete+close a database.
1216 * See #mdb_dbi_close() for restrictions about closing the DB handle.
1217 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
1218 * @param[in] dbi A database handle returned by #mdb_dbi_open()
1219 * @param[in] del 0 to empty the DB, 1 to delete it from the
1220 * environment and close the DB handle.
1221 * @return A non-zero error value on failure and 0 on success.
1223 int mdb_drop(MDB_txn *txn, MDB_dbi dbi, int del);
1225 /** @brief Set a custom key comparison function for a database.
1227 * The comparison function is called whenever it is necessary to compare a
1228 * key specified by the application with a key currently stored in the database.
1229 * If no comparison function is specified, and no special key flags were specified
1230 * with #mdb_dbi_open(), the keys are compared lexically, with shorter keys collating
1231 * before longer keys.
1232 * @warning This function must be called before any data access functions are used,
1233 * otherwise data corruption may occur. The same comparison function must be used by every
1234 * program accessing the database, every time the database is used.
1235 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
1236 * @param[in] dbi A database handle returned by #mdb_dbi_open()
1237 * @param[in] cmp A #MDB_cmp_func function
1238 * @return A non-zero error value on failure and 0 on success. Some possible
1241 * <li>EINVAL - an invalid parameter was specified.
1244 int mdb_set_compare(MDB_txn *txn, MDB_dbi dbi, MDB_cmp_func *cmp);
1246 /** @brief Set a custom data comparison function for a #MDB_DUPSORT database.
1248 * This comparison function is called whenever it is necessary to compare a data
1249 * item specified by the application with a data item currently stored in the database.
1250 * This function only takes effect if the database was opened with the #MDB_DUPSORT
1252 * If no comparison function is specified, and no special key flags were specified
1253 * with #mdb_dbi_open(), the data items are compared lexically, with shorter items collating
1254 * before longer items.
1255 * @warning This function must be called before any data access functions are used,
1256 * otherwise data corruption may occur. The same comparison function must be used by every
1257 * program accessing the database, every time the database is used.
1258 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
1259 * @param[in] dbi A database handle returned by #mdb_dbi_open()
1260 * @param[in] cmp A #MDB_cmp_func function
1261 * @return A non-zero error value on failure and 0 on success. Some possible
1264 * <li>EINVAL - an invalid parameter was specified.
1267 int mdb_set_dupsort(MDB_txn *txn, MDB_dbi dbi, MDB_cmp_func *cmp);
1269 /** @brief Set a relocation function for a #MDB_FIXEDMAP database.
1271 * @todo The relocation function is called whenever it is necessary to move the data
1272 * of an item to a different position in the database (e.g. through tree
1273 * balancing operations, shifts as a result of adds or deletes, etc.). It is
1274 * intended to allow address/position-dependent data items to be stored in
1275 * a database in an environment opened with the #MDB_FIXEDMAP option.
1276 * Currently the relocation feature is unimplemented and setting
1277 * this function has no effect.
1278 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
1279 * @param[in] dbi A database handle returned by #mdb_dbi_open()
1280 * @param[in] rel A #MDB_rel_func function
1281 * @return A non-zero error value on failure and 0 on success. Some possible
1284 * <li>EINVAL - an invalid parameter was specified.
1287 int mdb_set_relfunc(MDB_txn *txn, MDB_dbi dbi, MDB_rel_func *rel);
1289 /** @brief Set a context pointer for a #MDB_FIXEDMAP database's relocation function.
1291 * See #mdb_set_relfunc and #MDB_rel_func for more details.
1292 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
1293 * @param[in] dbi A database handle returned by #mdb_dbi_open()
1294 * @param[in] ctx An arbitrary pointer for whatever the application needs.
1295 * It will be passed to the callback function set by #mdb_set_relfunc
1296 * as its \b relctx parameter whenever the callback is invoked.
1297 * @return A non-zero error value on failure and 0 on success. Some possible
1300 * <li>EINVAL - an invalid parameter was specified.
1303 int mdb_set_relctx(MDB_txn *txn, MDB_dbi dbi, void *ctx);
1305 /** @brief Get items from a database.
1307 * This function retrieves key/data pairs from the database. The address
1308 * and length of the data associated with the specified \b key are returned
1309 * in the structure to which \b data refers.
1310 * If the database supports duplicate keys (#MDB_DUPSORT) then the
1311 * first data item for the key will be returned. Retrieval of other
1312 * items requires the use of #mdb_cursor_get().
1314 * @note The memory pointed to by the returned values is owned by the
1315 * database. The caller need not dispose of the memory, and may not
1316 * modify it in any way. For values returned in a read-only transaction
1317 * any modification attempts will cause a SIGSEGV.
1318 * @note Values returned from the database are valid only until a
1319 * subsequent update operation, or the end of the transaction.
1320 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
1321 * @param[in] dbi A database handle returned by #mdb_dbi_open()
1322 * @param[in] key The key to search for in the database
1323 * @param[out] data The data corresponding to the key
1324 * @return A non-zero error value on failure and 0 on success. Some possible
1327 * <li>#MDB_NOTFOUND - the key was not in the database.
1328 * <li>EINVAL - an invalid parameter was specified.
1331 int mdb_get(MDB_txn *txn, MDB_dbi dbi, MDB_val *key, MDB_val *data);
1333 /** @brief Store items into a database.
1335 * This function stores key/data pairs in the database. The default behavior
1336 * is to enter the new key/data pair, replacing any previously existing key
1337 * if duplicates are disallowed, or adding a duplicate data item if
1338 * duplicates are allowed (#MDB_DUPSORT).
1339 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
1340 * @param[in] dbi A database handle returned by #mdb_dbi_open()
1341 * @param[in] key The key to store in the database
1342 * @param[in,out] data The data to store
1343 * @param[in] flags Special options for this operation. This parameter
1344 * must be set to 0 or by bitwise OR'ing together one or more of the
1345 * values described here.
1347 * <li>#MDB_NODUPDATA - enter the new key/data pair only if it does not
1348 * already appear in the database. This flag may only be specified
1349 * if the database was opened with #MDB_DUPSORT. The function will
1350 * return #MDB_KEYEXIST if the key/data pair already appears in the
1352 * <li>#MDB_NOOVERWRITE - enter the new key/data pair only if the key
1353 * does not already appear in the database. The function will return
1354 * #MDB_KEYEXIST if the key already appears in the database, even if
1355 * the database supports duplicates (#MDB_DUPSORT). The \b data
1356 * parameter will be set to point to the existing item.
1357 * <li>#MDB_RESERVE - reserve space for data of the given size, but
1358 * don't copy the given data. Instead, return a pointer to the
1359 * reserved space, which the caller can fill in later - before
1360 * the next update operation or the transaction ends. This saves
1361 * an extra memcpy if the data is being generated later.
1362 * LMDB does nothing else with this memory, the caller is expected
1363 * to modify all of the space requested. This flag must not be
1364 * specified if the database was opened with #MDB_DUPSORT.
1365 * <li>#MDB_APPEND - append the given key/data pair to the end of the
1366 * database. This option allows fast bulk loading when keys are
1367 * already known to be in the correct order. Loading unsorted keys
1368 * with this flag will cause a #MDB_KEYEXIST error.
1369 * <li>#MDB_APPENDDUP - as above, but for sorted dup data.
1371 * @return A non-zero error value on failure and 0 on success. Some possible
1374 * <li>#MDB_MAP_FULL - the database is full, see #mdb_env_set_mapsize().
1375 * <li>#MDB_TXN_FULL - the transaction has too many dirty pages.
1376 * <li>EACCES - an attempt was made to write in a read-only transaction.
1377 * <li>EINVAL - an invalid parameter was specified.
1380 int mdb_put(MDB_txn *txn, MDB_dbi dbi, MDB_val *key, MDB_val *data,
1381 unsigned int flags);
1383 /** @brief Delete items from a database.
1385 * This function removes key/data pairs from the database.
1386 * If the database does not support sorted duplicate data items
1387 * (#MDB_DUPSORT) the data parameter is ignored.
1388 * If the database supports sorted duplicates and the data parameter
1389 * is NULL, all of the duplicate data items for the key will be
1390 * deleted. Otherwise, if the data parameter is non-NULL
1391 * only the matching data item will be deleted.
1392 * This function will return #MDB_NOTFOUND if the specified key/data
1393 * pair is not in the database.
1394 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
1395 * @param[in] dbi A database handle returned by #mdb_dbi_open()
1396 * @param[in] key The key to delete from the database
1397 * @param[in] data The data to delete
1398 * @return A non-zero error value on failure and 0 on success. Some possible
1401 * <li>EACCES - an attempt was made to write in a read-only transaction.
1402 * <li>EINVAL - an invalid parameter was specified.
1405 int mdb_del(MDB_txn *txn, MDB_dbi dbi, MDB_val *key, MDB_val *data);
1407 /** @brief Create a cursor handle.
1409 * A cursor is associated with a specific transaction and database.
1410 * A cursor cannot be used when its database handle is closed. Nor
1411 * when its transaction has ended, except with #mdb_cursor_renew().
1412 * It can be discarded with #mdb_cursor_close().
1413 * A cursor in a write-transaction can be closed before its transaction
1414 * ends, and will otherwise be closed when its transaction ends.
1415 * A cursor in a read-only transaction must be closed explicitly, before
1416 * or after its transaction ends. It can be reused with
1417 * #mdb_cursor_renew() before finally closing it.
1418 * @note Earlier documentation said that cursors in every transaction
1419 * were closed when the transaction committed or aborted.
1420 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
1421 * @param[in] dbi A database handle returned by #mdb_dbi_open()
1422 * @param[out] cursor Address where the new #MDB_cursor handle will be stored
1423 * @return A non-zero error value on failure and 0 on success. Some possible
1426 * <li>EINVAL - an invalid parameter was specified.
1429 int mdb_cursor_open(MDB_txn *txn, MDB_dbi dbi, MDB_cursor **cursor);
1431 /** @brief Close a cursor handle.
1433 * The cursor handle will be freed and must not be used again after this call.
1434 * Its transaction must still be live if it is a write-transaction.
1435 * @param[in] cursor A cursor handle returned by #mdb_cursor_open()
1437 void mdb_cursor_close(MDB_cursor *cursor);
1439 /** @brief Renew a cursor handle.
1441 * A cursor is associated with a specific transaction and database.
1442 * Cursors that are only used in read-only
1443 * transactions may be re-used, to avoid unnecessary malloc/free overhead.
1444 * The cursor may be associated with a new read-only transaction, and
1445 * referencing the same database handle as it was created with.
1446 * This may be done whether the previous transaction is live or dead.
1447 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
1448 * @param[in] cursor A cursor handle returned by #mdb_cursor_open()
1449 * @return A non-zero error value on failure and 0 on success. Some possible
1452 * <li>EINVAL - an invalid parameter was specified.
1455 int mdb_cursor_renew(MDB_txn *txn, MDB_cursor *cursor);
1457 /** @brief Return the cursor's transaction handle.
1459 * @param[in] cursor A cursor handle returned by #mdb_cursor_open()
1461 MDB_txn *mdb_cursor_txn(MDB_cursor *cursor);
1463 /** @brief Return the cursor's database handle.
1465 * @param[in] cursor A cursor handle returned by #mdb_cursor_open()
1467 MDB_dbi mdb_cursor_dbi(MDB_cursor *cursor);
1469 /** @brief Retrieve by cursor.
1471 * This function retrieves key/data pairs from the database. The address and length
1472 * of the key are returned in the object to which \b key refers (except for the
1473 * case of the #MDB_SET option, in which the \b key object is unchanged), and
1474 * the address and length of the data are returned in the object to which \b data
1476 * See #mdb_get() for restrictions on using the output values.
1477 * @param[in] cursor A cursor handle returned by #mdb_cursor_open()
1478 * @param[in,out] key The key for a retrieved item
1479 * @param[in,out] data The data of a retrieved item
1480 * @param[in] op A cursor operation #MDB_cursor_op
1481 * @return A non-zero error value on failure and 0 on success. Some possible
1484 * <li>#MDB_NOTFOUND - no matching key found.
1485 * <li>EINVAL - an invalid parameter was specified.
1488 int mdb_cursor_get(MDB_cursor *cursor, MDB_val *key, MDB_val *data,
1491 /** @brief Store by cursor.
1493 * This function stores key/data pairs into the database.
1494 * The cursor is positioned at the new item, or on failure usually near it.
1495 * @note Earlier documentation incorrectly said errors would leave the
1496 * state of the cursor unchanged.
1497 * @param[in] cursor A cursor handle returned by #mdb_cursor_open()
1498 * @param[in] key The key operated on.
1499 * @param[in] data The data operated on.
1500 * @param[in] flags Options for this operation. This parameter
1501 * must be set to 0 or one of the values described here.
1503 * <li>#MDB_CURRENT - replace the item at the current cursor position.
1504 * The \b key parameter must still be provided, and must match it.
1505 * If using sorted duplicates (#MDB_DUPSORT) the data item must still
1506 * sort into the same place. This is intended to be used when the
1507 * new data is the same size as the old. Otherwise it will simply
1508 * perform a delete of the old record followed by an insert.
1509 * <li>#MDB_NODUPDATA - enter the new key/data pair only if it does not
1510 * already appear in the database. This flag may only be specified
1511 * if the database was opened with #MDB_DUPSORT. The function will
1512 * return #MDB_KEYEXIST if the key/data pair already appears in the
1514 * <li>#MDB_NOOVERWRITE - enter the new key/data pair only if the key
1515 * does not already appear in the database. The function will return
1516 * #MDB_KEYEXIST if the key already appears in the database, even if
1517 * the database supports duplicates (#MDB_DUPSORT).
1518 * <li>#MDB_RESERVE - reserve space for data of the given size, but
1519 * don't copy the given data. Instead, return a pointer to the
1520 * reserved space, which the caller can fill in later - before
1521 * the next update operation or the transaction ends. This saves
1522 * an extra memcpy if the data is being generated later. This flag
1523 * must not be specified if the database was opened with #MDB_DUPSORT.
1524 * <li>#MDB_APPEND - append the given key/data pair to the end of the
1525 * database. No key comparisons are performed. This option allows
1526 * fast bulk loading when keys are already known to be in the
1527 * correct order. Loading unsorted keys with this flag will cause
1528 * a #MDB_KEYEXIST error.
1529 * <li>#MDB_APPENDDUP - as above, but for sorted dup data.
1530 * <li>#MDB_MULTIPLE - store multiple contiguous data elements in a
1531 * single request. This flag may only be specified if the database
1532 * was opened with #MDB_DUPFIXED. The \b data argument must be an
1533 * array of two MDB_vals. The mv_size of the first MDB_val must be
1534 * the size of a single data element. The mv_data of the first MDB_val
1535 * must point to the beginning of the array of contiguous data elements.
1536 * The mv_size of the second MDB_val must be the count of the number
1537 * of data elements to store. On return this field will be set to
1538 * the count of the number of elements actually written. The mv_data
1539 * of the second MDB_val is unused.
1541 * @return A non-zero error value on failure and 0 on success. Some possible
1544 * <li>#MDB_MAP_FULL - the database is full, see #mdb_env_set_mapsize().
1545 * <li>#MDB_TXN_FULL - the transaction has too many dirty pages.
1546 * <li>EACCES - an attempt was made to write in a read-only transaction.
1547 * <li>EINVAL - an invalid parameter was specified.
1550 int mdb_cursor_put(MDB_cursor *cursor, MDB_val *key, MDB_val *data,
1551 unsigned int flags);
1553 /** @brief Delete current key/data pair
1555 * This function deletes the key/data pair to which the cursor refers.
1556 * @param[in] cursor A cursor handle returned by #mdb_cursor_open()
1557 * @param[in] flags Options for this operation. This parameter
1558 * must be set to 0 or one of the values described here.
1560 * <li>#MDB_NODUPDATA - delete all of the data items for the current key.
1561 * This flag may only be specified if the database was opened with #MDB_DUPSORT.
1563 * @return A non-zero error value on failure and 0 on success. Some possible
1566 * <li>EACCES - an attempt was made to write in a read-only transaction.
1567 * <li>EINVAL - an invalid parameter was specified.
1570 int mdb_cursor_del(MDB_cursor *cursor, unsigned int flags);
1572 /** @brief Return count of duplicates for current key.
1574 * This call is only valid on databases that support sorted duplicate
1575 * data items #MDB_DUPSORT.
1576 * @param[in] cursor A cursor handle returned by #mdb_cursor_open()
1577 * @param[out] countp Address where the count will be stored
1578 * @return A non-zero error value on failure and 0 on success. Some possible
1581 * <li>EINVAL - cursor is not initialized, or an invalid parameter was specified.
1584 int mdb_cursor_count(MDB_cursor *cursor, mdb_size_t *countp);
1586 /** @brief Compare two data items according to a particular database.
1588 * This returns a comparison as if the two data items were keys in the
1589 * specified database.
1590 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
1591 * @param[in] dbi A database handle returned by #mdb_dbi_open()
1592 * @param[in] a The first item to compare
1593 * @param[in] b The second item to compare
1594 * @return < 0 if a < b, 0 if a == b, > 0 if a > b
1596 int mdb_cmp(MDB_txn *txn, MDB_dbi dbi, const MDB_val *a, const MDB_val *b);
1598 /** @brief Compare two data items according to a particular database.
1600 * This returns a comparison as if the two items were data items of
1601 * the specified database. The database must have the #MDB_DUPSORT flag.
1602 * @param[in] txn A transaction handle returned by #mdb_txn_begin()
1603 * @param[in] dbi A database handle returned by #mdb_dbi_open()
1604 * @param[in] a The first item to compare
1605 * @param[in] b The second item to compare
1606 * @return < 0 if a < b, 0 if a == b, > 0 if a > b
1608 int mdb_dcmp(MDB_txn *txn, MDB_dbi dbi, const MDB_val *a, const MDB_val *b);
1610 /** @brief A callback function used to print a message from the library.
1612 * @param[in] msg The string to be printed.
1613 * @param[in] ctx An arbitrary context pointer for the callback.
1614 * @return < 0 on failure, >= 0 on success.
1616 typedef int (MDB_msg_func)(const char *msg, void *ctx);
1618 /** @brief Dump the entries in the reader lock table.
1620 * @param[in] env An environment handle returned by #mdb_env_create()
1621 * @param[in] func A #MDB_msg_func function
1622 * @param[in] ctx Anything the message function needs
1623 * @return < 0 on failure, >= 0 on success.
1625 int mdb_reader_list(MDB_env *env, MDB_msg_func *func, void *ctx);
1627 /** @brief Check for stale entries in the reader lock table.
1629 * @param[in] env An environment handle returned by #mdb_env_create()
1630 * @param[out] dead Number of stale slots that were cleared
1631 * @return 0 on success, non-zero on failure.
1633 int mdb_reader_check(MDB_env *env, int *dead);
1639 /** @page tools LMDB Command Line Tools
1640 The following describes the command line tools that are available for LMDB.
1647 #endif /* _LMDB_H_ */