Lmdb internals -
Common header for all page types. Overflow records occupy a number of contiguous pages with no headers on any page after the first.
Data Fields
union {
pgno_t p_pgno
struct MDB_page * p_next
} mp_p
uint16_t mp_pad
uint16_t mp_flags
union {
struct {
indx_t pb_lower
indx_t pb_upper
} pb
uint32_t pb_pages
} mp_pb
indx_t mp_ptrs [1]
page number
for in-memory list of freed pages
Page Flags
lower bound of free space
upper bound of free space
number of overflow pages
dynamic size
Header for a single key/data pair within a page. Used in pages of type P_BRANCH and P_LEAF without P_LEAF2. We guarantee 2-byte alignment for 'MDB_node's.
Data Fields
unsigned short mn_lo
unsigned short mn_hi
unsigned short mn_flags
unsigned short mn_ksize
char mn_data [1]
lo and hi are used for data size on leaf nodes and for child pgno on branch nodes. On 64 bit platforms, flags is also used for pgno. (Branch nodes have no flags). They are in host byte order in case that lets some accesses be optimized into a 32-bit word access.
part of data size or pgno
Node Flags
key size
key and data are appended here
Information about a single database in the environment.
Data Fields
uint32_t md_pad
uint16_t md_flags
uint16_t md_depth
pgno_t md_branch_pages
pgno_t md_leaf_pages
pgno_t md_overflow_pages
size_t md_entries
pgno_t md_root
also ksize for LEAF2 pages
Database Flags
depth of this tree
number of internal pages
number of leaf pages
number of overflow pages
number of data items
the root page of this tree
Meta page content. A meta page is the start point for accessing a database snapshot. Pages 0-1 are meta pages. Transaction N writes meta page #(N % 2).
Data Fields
uint32_t mm_magic
uint32_t mm_version
void * mm_address
size_t mm_mapsize
MDB_db mm_dbs [2]
pgno_t mm_last_pg
txnid_t mm_txnid
Stamp identifying this as an LMDB file. It must be set to MDB_MAGIC.
Version number of this lock file. Must be set to MDB_DATA_VERSION.
address for fixed mapping
size of mmap region
first is free space, 2nd is main db
last used page in file
txnid that committed this page
Buffer for a stack-allocated meta page. The members define size and alignment, and silence type aliasing warnings. They are not used directly; that could mean incorrectly using several union members in parallel.
Data Fields
MDB_page mb_page
struct {
char mm_pad [PAGEHDRSZ]
MDB_meta mm_meta
} mb_metabuf
Auxiliary DB info. The information here is mostly static/read-only. There is only a single copy of this record in the environment.
Data Fields
MDB_val md_name
MDB_cmp_func * md_cmp
MDB_cmp_func * md_dcmp
MDB_rel_func * md_rel
void * md_relctx
name of the database
function for comparing keys
function for comparing data items
user relocate function
user-provided context for md_rel
Opaque structure for a transaction handle.
A database transaction. Every operation requires a transaction handle.
All database operations require a transaction handle. Transactions may be read-only or read-write.
Data Fields
MDB_txn * mt_parent
MDB_txn * mt_child
pgno_t mt_next_pgno
txnid_t mt_txnid
MDB_env * mt_env
MDB_IDL mt_free_pgs
MDB_page * mt_loose_pgs
int mt_loose_count
MDB_IDL mt_spill_pgs
union {
MDB_ID2L dirty_list
MDB_reader * reader
} mt_u
MDB_dbx * mt_dbxs
MDB_db * mt_dbs
unsigned int * mt_dbiseqs
MDB_cursor ** mt_cursors
unsigned char * mt_dbflags
MDB_dbi mt_numdbs
unsigned int mt_flags
unsigned int mt_dirty_room
parent of a nested txn
nested txn under this txn
next unallocated page
The ID of this transaction. IDs are integers incrementing from 1. Only committed write transactions increment the ID. If a transaction aborts, the ID may be re-used by the next writer.
the DB environment
The list of pages that became unused during this transaction.
The list of loose pages that became unused and may be reused in this transaction, linked through NEXT_LOOSE_PAGE(page).
The sorted list of dirty pages we temporarily wrote to disk because the dirty list was full. page numbers in here are shifted left by 1, deleted slots have the LSB set.
For write txns: Modified pages. Sorted when not MDB_WRITEMAP.
For read txns: This thread/txn's reader table slot, or NULL.
Array of records for each DB known in the environment.
Array of MDB_db records for each known DB
Array of sequence numbers for each DB handle
In write txns, array of cursors for each DB
Array of flags for each DB
Number of DB records in use. This number only ever increments; we don't decrement it when individual DB handles are closed.
Transaction Flags
dirty_list room: Array size - #dirty pages visible to this txn. Includes ancestor txns' dirty pages not hidden by other txns' dirty/spilled pages. Thus commit(nested txn) has room to merge dirty_list into mt_parent after freeing hidden mt_parent pages.
Opaque structure for navigating through a database.
Cursors are used for all DB operations. A cursor holds a path of (page pointer, key index) from the DB root to a position in the DB, plus other state. MDB_DUPSORT cursors include an xcursor to the current data item. Write txns track their cursors and keep them up to date when data moves. Exception: An xcursor's pointer to a P_SUBP page can be stale. (A node with F_DUPDATA but no F_SUBDATA contains a subpage).
Data Fields
MDB_cursor * mc_next
MDB_cursor * mc_backup
struct MDB_xcursor * mc_xcursor
MDB_txn * mc_txn
MDB_dbi mc_dbi
MDB_db * mc_db
MDB_dbx * mc_dbx
unsigned char * mc_dbflag
unsigned short mc_snum
unsigned short mc_top
unsigned int mc_flags
MDB_page * mc_pg [CURSOR_STACK]
indx_t mc_ki [CURSOR_STACK]
Next cursor on this DB in this txn
Backup of the original cursor if this cursor is a shadow
Context used for databases with MDB_DUPSORT, otherwise NULL
The transaction that owns this cursor
The database handle this cursor operates on
The database record for this cursor
The database auxiliary record for this cursor
The Transaction DB Flags for this database
number of pushed pages
index of top page, normally mc_snum-1
Cursor Flags
stack of pushed pages
stack of page indices
Context for sorted-dup records. We could have gone to a fully recursive design, with arbitrarily deep nesting of sub-databases. But for now we only handle these levels - main DB, optional sub-DB, sorted-duplicate DB.
Data Fields
MDB_cursor mx_cursor
MDB_db mx_db
MDB_dbx mx_dbx
unsigned char mx_dbflag
A sub-cursor for traversing the Dup DB
The database record for this Dup DB
The auxiliary DB record for this Dup DB
The Transaction DB Flags for this Dup DB
State of FreeDB old pages, stored in the MDB_env
Data Fields
pgno_t * mf_pghead
txnid_t mf_pglast
Reclaimed freeDB pages, or NULL before use
ID of last used record, or 0 if !mf_pghead
Opaque structure for a database environment.
The database environment.
A DB environment supports multiple databases, all residing in the same shared-memory map.
Data Fields
HANDLE me_fd
HANDLE me_lfd
HANDLE me_mfd
uint32_t me_flags
unsigned int me_psize
unsigned int me_os_psize
unsigned int me_maxreaders
unsigned int me_numreaders
MDB_dbi me_numdbs
MDB_dbi me_maxdbs
MDB_PID_T me_pid
char * me_path
char * me_map
MDB_txninfo * me_txns
MDB_meta * me_metas [2]
void * me_pbuf
MDB_txn * me_txn
MDB_txn * me_txn0
size_t me_mapsize
off_t me_size
pgno_t me_maxpg
MDB_dbx * me_dbxs
uint16_t * me_dbflags
unsigned int * me_dbiseqs
pthread_key_t me_txkey
txnid_t me_pgoldest
MDB_pgstate me_pgstate
MDB_page * me_dpages
MDB_IDL me_free_pgs
MDB_ID2L me_dirty_list
int me_maxfree_1pg
unsigned int me_nodemax
int me_live_reader
void * me_userctx
MDB_assert_func * me_assert_func
The main data file
The lock file
just for writing the meta pages
Environment Flags
DB page size, inited from me_os_psize
OS page size, from GET_PAGESIZE
size of the reader table
max numreaders set by this env
number of DBs opened
size of the DB table
process ID of this env
path to the DB files
the memory map of the data file
the memory map of the lock file or NULL
pointers to the two meta pages
scratch area for DUPSORT put()
current write transaction
prealloc'd write transaction
size of the data memory map
current file size
me_mapsize / me_psize
array of static DB info
array of flags from MDB_db.md_flags
array of dbi sequence numbers
thread-key for readers
ID of oldest reader last time we looked
state of old pages from freeDB
list of malloc'd blocks for re-use
IDL of pages that became unused in a write txn
ID2L of pages written during a write txn. Length MDB_IDL_UM_SIZE.
Max number of freelist items that can fit in a single overflow page
Max size of a node on a page
have liveness lock in reader table
User-settable context
Callback for assertion failures
Nested transaction
Data Fields
MDB_txn mnt_txn
MDB_pgstate mnt_pgstate
the transaction
parent transaction's saved freestate
State needed for a compacting copy.
Data Fields
pthread_mutex_t mc_mutex
pthread_cond_t mc_cond
char * mc_wbuf [2]
char * mc_over [2]
MDB_env * mc_env
MDB_txn * mc_txn
int mc_wlen [2]
int mc_olen [2]
pgno_t mc_next_pgno
HANDLE mc_fd
int mc_status
volatile int mc_new
int mc_toggle
A flag for opening a file and requesting synchronous data writes. This is only used when writing a meta page. It's not strictly needed; we could just do a normal write and then immediately perform a flush. But if this flag is available it saves us an extra system call.
Note:
If O_DSYNC is undefined but exists in /usr/include, preferably set some compiler flag to get the definition. Otherwise compile with the less efficient -DMDB_DSYNC=O_SYNC.
Function for flushing the data of a file. Define this to fsync if fdatasync() is not supported.
The maximum size of a database page. It is 32k or 64k, since value-PAGEBASE must fit in MDB_page.mp_upper.
LMDB will use database pages < OS pages if needed. That causes more I/O in write transactions: The OS must know (read) the whole page before writing a partial page.
Note that we don't currently support Huge pages. On Linux, regular data files cannot use Huge pages, and in general Huge pages aren't actually pageable. We rely on the OS demand-pager to read our data and page it out when memory pressure from other processes is high. So until OSs have actual paging support for Huge pages, they're not viable.
The minimum number of keys required in a database page. Setting this to a larger value will place a smaller bound on the maximum size of a data item. Data items larger than this size will be pushed into overflow pages instead of being stored directly in the B-tree node. This value used to default to 4. With a page size of 4096 bytes that meant that any item larger than 1024 bytes would go into an overflow page. That also meant that on average 2-3KB of each overflow page was wasted space. The value cannot be lower than 2 because then there would no longer be a tree structure. With this value, items larger than 2KB will go into overflow pages, and on average only 1KB will be wasted.
A stamp that identifies a file as an LMDB file. There's nothing special about this value other than that it is easily recognizable, and it will reflect any byte order mismatches.
The version number for a database's datafile format.
The version number for a database's lockfile format.
The max size of a key we can write, or 0 for dynamic max. Define this as 0 to compute the max from the page size. 511 is default for backwards compat: liblmdb <= 0.9.10 can break when modifying a DB with keys/dupsort data bigger than its max. MDB_DEVEL sets the default to 0.
Data items in an MDB_DUPSORT database are also limited to this size, since they're actually keys of a sub-DB. Keys and MDB_DUPSORT data items must fit on a node in a regular page.
The maximum size of a key we can write to the environment.
The maximum size of a data item. We only store a 32 bit value for node sizes.
An invalid page number. Mainly used to denote an empty tree.
Test if the flags f are set in a flag word w.
Round n up to an even number.
Default size of memory map. This is certainly too small for any actual applications. Apps should always set the size explicitly using mdb_env_set_mapsize().
Size of the page header, excluding dynamic data at the end
Address of first usable data byte in a page, after the header
ITS#7713, change PAGEBASE to handle 65536 byte pages
Number of nodes on a page
The amount of space remaining in the page
Value:
(1000L * ((env)->me_psize - PAGEHDRSZ - SIZELEFT(p)) / ((env)->me_psize - PAGEHDRSZ))
The percentage of space used in the page, in tenths of a percent.
The minimum page fill factor, in tenths of a percent. Pages emptier than this are candidates for merging.
Test if a page is a leaf page
Test if a page is a LEAF2 page
Test if a page is a branch page
Test if a page is an overflow page
Test if a page is a sub page
The number of overflow pages needed to store the given size.
Link in MDB_txn.mt_loose_pgs list
Size of the node header, excluding dynamic data at the end
Bit position of top word in page number, for shifting mn_flags
Size of a node in a branch page with a given key. This is just the node header plus the key, there is no data.
Size of a node in a leaf page with a given key and data. This is node header plus key plus data size.
Address of node i in page p
Address of the key for the node
Address of the data for a node
Value:
((node)->mn_lo | ((pgno_t) (node)->mn_hi << 16) | (PGNO_TOPWORD ? ((pgno_t) (node)->mn_flags << PGNO_TOPWORD) : 0))
Get the page number pointed to by a branch node
Value:
do { (node)->mn_lo = (pgno) & 0xffff; (node)->mn_hi = (pgno) >> 16; if (PGNO_TOPWORD) (node)->mn_flags = (pgno) >> PGNO_TOPWORD; } while(0)
Set the page number in a branch node
Get the size of the data in a leaf node
Value:
do { (node)->mn_lo = (size) & 0xffff; (node)->mn_hi = (size) >> 16;} while(0)
Set the size of the data for a leaf node
The size of a key in a node
Value:
do { unsigned short *s, *d; s = (unsigned short *)&(src); d = (unsigned short *)&(dst); *d++ = *s++; *d = *s; } while (0)
Copy a page number from src to dst
The address of a key in a LEAF2 page. LEAF2 pages are used for MDB_DUPFIXED sorted-duplicate sub-DBs. There are no node headers, keys are stored contiguously.
Value:
{ if ((keyptr) != NULL) { (keyptr)->mv_size = NODEKSZ(node); (keyptr)->mv_data = NODEKEY(node); } }
Set the node's key into keyptr, if requested.
Set the node's key into key.
mdb_dbi_open flags DB handle is valid, for me_dbflags
Value:
(MDB_REVERSEKEY|MDB_DUPSORT|MDB_INTEGERKEY|MDB_DUPFIXED|\ MDB_INTEGERDUP|MDB_REVERSEDUP|MDB_CREATE)
Handle for the DB used to track free pages.
Handle for the default DB.
Enough space for 2^32 nodes with minimum of 2 keys per node. I.e., plenty. At 4 keys per node, enough for 2^64 nodes, so there's probably no need to raise this on a 64 bit machine.
max number of pages to commit in one writev() call
max bytes to write in one call
Check txn and dbi arguments to a function
Check for misused dbi handles
newkey is not new
assert(3) variant in cursor context
assert(3) variant in transaction context
assert(3) variant in environment context
Value:
((expr) ? (void)0 : mdb_assert_fail(env, expr_txt, mdb_func_, __FILE__, __LINE__))
The name of the lock file in the DB environment
The name of the data file in the DB environment
The suffix of the lock file when no subdir is used
Only a subset of the Environment Flags flags can be changed at runtime. Changing other flags requires closing the environment and re-opening it with the new flags.
Value:
(MDB_FIXEDMAP|MDB_NOSUBDIR|MDB_RDONLY|MDB_WRITEMAP| \ MDB_NOTLS|MDB_NOLOCK|MDB_NORDAHEAD)
Compare two items pointing at size_t's of unknown alignment.
Do not spill pages to disk if txn is getting full, may fail instead
A page number in the database. Note that 64 bit page numbers are overkill, since pages themselves already represent 12-13 bits of addressable memory, and the OS will always limit applications to a maximum of 63 bits of address space.
Note:
In the MDB_node structure, we only store 48 bits of this value, which thus limits us to only 60 bits of addressable data.
A transaction ID. See struct MDB_txn.mt_txnid for details.
Used for offsets within a single page. Since memory pages are typically 4 or 8KB in size, 12-13 bits, this is plenty.
Allocate page numbers and memory for writing. Maintain me_pglast, me_pghead and mt_next_pgno.
If there are free pages available from older transactions, they are re-used first. Otherwise allocate a new page at mt_next_pgno. Do not modify the freedB, just merge freeDB records into me_pghead[] and move me_pglast to say which records were consumed. Only this function can create me_pghead and move me_pglast/mt_next_pgno.
Parameters:
mc cursor A cursor handle identifying the transaction and database for which we are allocating.
num the number of pages to allocate.
mp Address of the allocated page(s). Requests for multiple pages will always be satisfied by a single contiguous chunk of memory.
Returns:
0 on success, non-zero on failure.
Allocate and initialize new pages for a database.
Parameters:
mc a cursor on the database being added to.
flags flags defining what type of page is being allocated.
num the number of pages to allocate. This is usually 1, unless allocating overflow pages for a large record.
mp Address of a page, or NULL on failure.
Returns:
0 on success, non-zero on failure.
Touch a page: make it dirty and re-insert into tree with updated pgno.
Parameters:
mc cursor pointing to the page to be touched
Returns:
0 on success, non-zero on failure.
Find the address of the page corresponding to a given page number.
Parameters:
txn the transaction for this access.
pgno the page number for the page to retrieve.
ret address of a pointer where the page's address will be stored.
lvl dirty_list inheritance level of found page. 1=current txn, 0=mapped page.
Returns:
0 on success, non-zero on failure.
Finish mdb_page_search() / mdb_page_search_lowest(). The cursor is at the root page, set up the rest of it.
Search for the page a given key should be in. Push it and its parent pages on the cursor stack.
Parameters:
mc the cursor for this operation.
key the key to search for, or NULL for first/last page.
flags If MDB_PS_MODIFY is set, visited pages in the DB are touched (updated with new page numbers). If MDB_PS_FIRST or MDB_PS_LAST is set, find first or last leaf. This is used by mdb_cursor_first() and mdb_cursor_last(). If MDB_PS_ROOTONLY set, just fetch root node, no further lookups.
Returns:
0 on success, non-zero on failure.
Merge one page into another. The nodes from the page pointed to by csrc will be copied to the page pointed to by cdst and then the csrc page will be freed.
Parameters:
csrc Cursor pointing to the source page.
cdst Cursor pointing to the destination page.
Returns:
0 on success, non-zero on failure.
Split a page and insert a new node.
Parameters:
mc Cursor pointing to the page and desired insertion index. The cursor will be updated to point to the actual page and index where the node got inserted after the split.
newkey The key for the newly inserted node.
newdata The data for the newly inserted node.
newpgno The page number, if the new node is a branch node.
nflags The NODE_ADD_FLAGS for the new node.
Returns:
0 on success, non-zero on failure.
Read the environment parameters of a DB environment before mapping it into memory.
Parameters:
env the environment handle
meta address of where to store the meta information
Returns:
0 on success, non-zero on failure.
Check both meta pages to see which one is newer.
Parameters:
env the environment handle
Returns:
meta toggle (0 or 1).
Update the environment info to commit a transaction.
Parameters:
txn the transaction that's being committed
Returns:
0 on success, non-zero on failure.
Destroy resources from mdb_env_open(), clear our readers & DBIs
Search for key within a page, using binary search. Returns the smallest entry larger or equal to the key. If exactp is non-null, stores whether the found entry was an exact match in *exactp (1 or 0). Updates the cursor index with the index of the found entry. If no entry larger or equal to the key is found, returns NULL.
Add a node to the page pointed to by the cursor.
Parameters:
mc The cursor for this operation.
indx The index on the page where the new node should be added.
key The key for the new node.
data The data for the new node, if any.
pgno The page number, if adding a branch node.
flags Flags for the node.
Returns:
0 on success, non-zero on failure. Possible errors are:
ENOMEM - failed to allocate overflow pages for the node.
MDB_PAGE_FULL - there is insufficient room in the page. This error should never happen since all callers already calculate the page's free space before calling this function.
Delete the specified node from a page.
Parameters:
mc Cursor pointing to the node to delete.
ksize The size of a node. Only used if the page is part of a MDB_DUPFIXED database.
Compact the main page after deleting a node on a subpage.
Parameters:
mp The main page to operate on.
indx The index of the subpage on the main page.
Move a node from csrc to cdst.
Return the data associated with a given node.
Parameters:
txn The transaction for this operation.
leaf The node being read.
data Updated to point to the node's data.
Returns:
0 on success, non-zero on failure.
Calculate the size of a leaf node. The size depends on the environment's page size; if a data item is too large it will be put onto an overflow page and the node size will only include the key and not the data. Sizes are always rounded up to an even number of bytes, to guarantee 2-byte alignment of the MDB_node headers.
Parameters:
env The environment handle.
key The key for the node.
data The data for the node.
Returns:
The number of bytes needed to store the node.
Calculate the size of a branch node. The size should depend on the environment's page size but since we currently don't support spilling large keys onto overflow pages, it's simply the size of the MDB_node header plus the size of the key. Sizes are always rounded up to an even number of bytes, to guarantee 2-byte alignment of the MDB_node headers.
Parameters:
env The environment handle.
key The key for the node.
Returns:
The number of bytes needed to store the node.
Rebalance the tree after a delete operation.
Parameters:
mc Cursor pointing to the page where rebalancing should begin.
Returns:
0 on success, non-zero on failure.
Replace the key for a branch node with a new key.
Parameters:
mc Cursor pointing to the node to operate on.
key The new key to use.
Returns:
0 on success, non-zero on failure.
Pop a page off the top of the cursor's stack.
Push a page onto the top of the cursor's stack.
Complete a delete operation started by mdb_cursor_del().
Find a sibling for a page. Replaces the page at the top of the cursor's stack with the specified sibling, if one exists.
Parameters:
mc The cursor for this operation.
move_right Non-zero if the right sibling is requested, otherwise the left sibling.
Returns:
0 on success, non-zero on failure.
Move the cursor to the next data item.
Move the cursor to the previous data item.
Set the cursor on a specific data item.
Move the cursor to the first item in the database.
Move the cursor to the last item in the database.
Initialize a cursor for a given transaction and database.
Initial setup of a sorted-dups cursor. Sorted duplicates are implemented as a sub-database for the given key. The duplicate data items are actually keys of the sub-database. Operations on the duplicate data items are performed using a sub-cursor initialized when the sub-database is first accessed. This function does the preliminary setup of the sub-cursor, filling in the fields that depend only on the parent DB.
Parameters:
mc The main cursor whose sorted-dups cursor is to be initialized.
Final setup of a sorted-dups cursor. Sets up the fields that depend on the data from the main cursor.
Parameters:
mc The main cursor whose sorted-dups cursor is to be initialized.
node The data containing the MDB_db record for the sorted-dup database.
Add all the DB's pages to the free list.
Parameters:
mc Cursor on the DB to free.
subs non-Zero to check for sub-DBs in this DB.
Returns:
0 on success, non-zero on failure.
Set the default comparison functions for a database. Called immediately after a database is opened to set the defaults. The user can then override them with mdb_set_compare() or mdb_set_dupsort().
Parameters:
txn A transaction handle returned by mdb_txn_begin()
dbi A database handle returned by mdb_dbi_open()
Return the LMDB library version information. Return the library version info.
Return a string describing a given error code. This function is a superset of the ANSI C X3.159-1989 (ANSI C) strerror(3) function. If the error code is greater than or equal to 0, then the string returned by the system function strerror(3) is returned. If the error code is less than 0, an error string corresponding to the LMDB library error is returned. See Return Codes for a list of LMDB-specific error codes.
Parameters:
err The error code
Return values:
error message The description of the error
Compare two data items according to a particular database. This returns a comparison as if the two data items were keys in the specified database.
Parameters:
txn A transaction handle returned by mdb_txn_begin()
dbi A database handle returned by mdb_dbi_open()
a The first item to compare
b The second item to compare
Returns:
< 0 if a < b, 0 if a == b, > 0 if a > b
Compare two data items according to a particular database. This returns a comparison as if the two items were data items of the specified database. The database must have the MDB_DUPSORT flag.
Parameters:
txn A transaction handle returned by mdb_txn_begin()
dbi A database handle returned by mdb_dbi_open()
a The first item to compare
b The second item to compare
Returns:
< 0 if a < b, 0 if a == b, > 0 if a > b
Allocate memory for a page. Re-use old malloc'd pages first for singletons, otherwise just malloc.
Free a single page. Saves single pages to a list, for future reuse. (This is not used for multi-page overflow pages.)
Free a dirty page
Return all dirty pages to dpage list
Loosen or free a single page. Saves single pages to a list for future reuse in this same txn. It has been pulled from the freeDB and already resides on the dirty list, but has been deleted. Use these pages first before pulling again from the freeDB.
If the page wasn't dirtied in this txn, just add it to this txn's free list.
Set or clear P_KEEP in dirty, non-overflow, non-sub pages watched by txn.
Parameters:
mc A cursor handle for the current operation.
pflags Flags of the pages to update: P_DIRTY to set P_KEEP, P_DIRTY|P_KEEP to clear it.
all No shortcuts. Needed except after a full mdb_page_flush().
Returns:
0 on success, non-zero on failure.
Flush (some) dirty pages to the map, after clearing their dirty flag.
Parameters:
txn the transaction that's being committed
keep number of initial pages in dirty_list to keep dirty.
Returns:
0 on success, non-zero on failure.
Spill pages from the dirty list back to disk. This is intended to prevent running into MDB_TXN_FULL situations, but note that they may still occur in a few cases: 1) our estimate of the txn size could be too small. Currently this seems unlikely, except with a large number of MDB_MULTIPLE items. 2) child txns may run out of space if their parents dirtied a lot of pages and never spilled them. TODO: we probably should do a preemptive spill during mdb_txn_begin() of a child txn, if the parent's dirty_room is below a given threshold.
Otherwise, if not using nested txns, it is expected that apps will not run into MDB_TXN_FULL any more. The pages are flushed to disk the same way as for a txn commit, e.g. their P_DIRTY flag is cleared. If the txn never references them again, they can be left alone. If the txn only reads them, they can be used without any fuss. If the txn writes them again, they can be dirtied immediately without going thru all of the work of mdb_page_touch(). Such references are handled by mdb_page_unspill().
Also note, we never spill DB root pages, nor pages of active cursors, because we'll need these back again soon anyway. And in nested txns, we can't spill a page in a child txn if it was already spilled in a parent txn. That would alter the parent txns' data even though the child hasn't committed yet, and we'd have no way to undo it if the child aborted.
Parameters:
m0 cursor A cursor handle identifying the transaction and database for which we are checking space.
key For a put operation, the key being stored.
data For a put operation, the data being stored.
Returns:
0 on success, non-zero on failure.
Find oldest txnid still referenced. Expects txn->mt_txnid > 0.
Add a page to the txn's dirty list
Copy the used portions of a non-overflow page.
Parameters:
dst page to copy into
src page to copy from
psize size of a page
Pull a page off the txn's spill list, if present. If a page being referenced was spilled to disk in this txn, bring it back and make it dirty/writable again.
Parameters:
txn the transaction handle.
mp the page being referenced. It must not be dirty.
ret the writable page, if any. ret is unchanged if mp wasn't spilled.
Flush the data buffers to disk. Data is always written to disk when mdb_txn_commit() is called, but the operating system may keep it buffered. LMDB always flushes the OS buffers upon commit as well, unless the environment was opened with MDB_NOSYNC or in part MDB_NOMETASYNC.
Parameters:
env An environment handle returned by mdb_env_create()
force If non-zero, force a synchronous flush. Otherwise if the environment has the MDB_NOSYNC flag set the flushes will be omitted, and with MDB_MAPASYNC they will be asynchronous.
Returns:
A non-zero error value on failure and 0 on success. Some possible errors are:
EINVAL - an invalid parameter was specified.
EIO - an error occurred during synchronization.
Back up parent txn's cursors, then grab the originals for tracking
Close this write txn's cursors, give parent txn's cursors back to parent.
Parameters:
txn the transaction handle.
merge true to keep changes to parent cursors, false to revert.
Returns:
0 on success, non-zero on failure.
Common code for mdb_txn_reset() and mdb_txn_abort(). May be called twice for readonly txns: First reset it, then abort.
Parameters:
txn the transaction handle to reset
act why the transaction is being reset
Set or check a pid lock. Set returns 0 on success. Check returns 0 if the process is certainly dead, nonzero if it may be alive (the lock exists or an error happened so we do not know).
On Windows Pidset is a no-op, we merely check for the existence of the process with the given pid. On POSIX we use a single byte lock on the lockfile, set at an offset equal to the pid.
Common code for mdb_txn_begin() and mdb_txn_renew().
Parameters:
txn the transaction handle to initialize
Returns:
0 on success, non-zero on failure.
Renew a read-only transaction. This acquires a new reader lock for a transaction handle that had been released by mdb_txn_reset(). It must be called before a reset transaction may be used again.
Parameters:
txn A transaction handle returned by mdb_txn_begin()
Returns:
A non-zero error value on failure and 0 on success. Some possible errors are:
MDB_PANIC - a fatal error occurred earlier and the environment must be shut down.
EINVAL - an invalid parameter was specified.
Create a transaction for use with the environment. The transaction handle may be discarded using mdb_txn_abort() or mdb_txn_commit().
Note:
A transaction and its cursors must only be used by a single thread, and a thread may only have a single transaction at a time. If MDB_NOTLS is in use, this does not apply to read-only transactions.
Cursors may not span transactions.
Parameters:
env An environment handle returned by mdb_env_create()
parent If this parameter is non-NULL, the new transaction will be a nested transaction, with the transaction indicated by parent as its parent. Transactions may be nested to any level. A parent transaction and its cursors may not issue any other operations than mdb_txn_commit and mdb_txn_abort while it has active child transactions.
flags Special options for this transaction. This parameter must be set to 0 or by bitwise OR'ing together one or more of the values described here.
MDB_RDONLY This transaction will not perform any write operations.
txn Address where the new MDB_txn handle will be stored
Returns:
A non-zero error value on failure and 0 on success. Some possible errors are:
MDB_PANIC - a fatal error occurred earlier and the environment must be shut down.
MDB_MAP_RESIZED - another process wrote data beyond this MDB_env's mapsize and this environment's map must be resized as well. See mdb_env_set_mapsize().
MDB_READERS_FULL - a read-only transaction was requested and the reader lock table is full. See mdb_env_set_maxreaders().
ENOMEM - out of memory.
Returns the transaction's MDB_env.
Parameters:
txn A transaction handle returned by mdb_txn_begin()
Export or close DBI handles opened in this txn.
Reset a read-only transaction. Abort the transaction like mdb_txn_abort(), but keep the transaction handle. mdb_txn_renew() may reuse the handle. This saves allocation overhead if the process will start a new read-only transaction soon, and also locking overhead if MDB_NOTLS is in use. The reader table lock is released, but the table slot stays tied to its thread or MDB_txn. Use mdb_txn_abort() to discard a reset handle, and to free its lock table slot if MDB_NOTLS is in use. Cursors opened within the transaction must not be used again after this call, except with mdb_cursor_renew(). Reader locks generally don't interfere with writers, but they keep old versions of database pages allocated. Thus they prevent the old pages from being reused when writers commit new data, and so under heavy load the database size may grow much more rapidly than otherwise.
Parameters:
txn A transaction handle returned by mdb_txn_begin()
Abandon all the operations of the transaction instead of saving them. The transaction handle is freed. It and its cursors must not be used again after this call, except with mdb_cursor_renew().
Note:
Earlier documentation incorrectly said all cursors would be freed. Only write-transactions free cursors.
Parameters:
txn A transaction handle returned by mdb_txn_begin()
Save the freelist as of this transaction to the freeDB. This changes the freelist. Keep trying until it stabilizes.
Commit all the operations of a transaction into the database. The transaction handle is freed. It and its cursors must not be used again after this call, except with mdb_cursor_renew().
Note:
Earlier documentation incorrectly said all cursors would be freed. Only write-transactions free cursors.
Parameters:
txn A transaction handle returned by mdb_txn_begin()
Returns:
A non-zero error value on failure and 0 on success. Some possible errors are:
EINVAL - an invalid parameter was specified.
ENOSPC - no more disk space.
EIO - a low-level I/O error occurred while writing.
ENOMEM - out of memory.
Write the environment parameters of a freshly created DB environment.
Parameters:
env the environment handle
meta address of where to store the meta information
Returns:
0 on success, non-zero on failure.
Create an LMDB environment handle. This function allocates memory for a MDB_env structure. To release the allocated memory and discard the handle, call mdb_env_close(). Before the handle may be used, it must be opened using mdb_env_open(). Various other options may also need to be set before opening the handle, e.g. mdb_env_set_mapsize(), mdb_env_set_maxreaders(), mdb_env_set_maxdbs(), depending on usage requirements.
Parameters:
env The address where the new handle will be stored
Returns:
A non-zero error value on failure and 0 on success.
Set the size of the memory map to use for this environment. The size should be a multiple of the OS page size. The default is 10485760 bytes. The size of the memory map is also the maximum size of the database. The value should be chosen as large as possible, to accommodate future growth of the database. This function should be called after mdb_env_create() and before mdb_env_open(). It may be called at later times if no transactions are active in this process. Note that the library does not check for this condition, the caller must ensure it explicitly.
The new size takes effect immediately for the current process but will not be persisted to any others until a write transaction has been committed by the current process. Also, only mapsize increases are persisted into the environment.
If the mapsize is increased by another process, and data has grown beyond the range of the current mapsize, mdb_txn_begin() will return MDB_MAP_RESIZED. This function may be called with a size of zero to adopt the new size.
Any attempt to set a size smaller than the space already consumed by the environment will be silently changed to the current size of the used space.
Parameters:
env An environment handle returned by mdb_env_create()
size The size in bytes
Returns:
A non-zero error value on failure and 0 on success. Some possible errors are:
EINVAL - an invalid parameter was specified, or the environment has an active write transaction.
Set the maximum number of named databases for the environment. This function is only needed if multiple databases will be used in the environment. Simpler applications that use the environment as a single unnamed database can ignore this option. This function may only be called after mdb_env_create() and before mdb_env_open().
Currently a moderate number of slots are cheap but a huge number gets expensive: 7-120 words per transaction, and every mdb_dbi_open() does a linear search of the opened slots.
Parameters:
env An environment handle returned by mdb_env_create()
dbs The maximum number of databases
Returns:
A non-zero error value on failure and 0 on success. Some possible errors are:
EINVAL - an invalid parameter was specified, or the environment is already open.
Set the maximum number of threads/reader slots for the environment. This defines the number of slots in the lock table that is used to track readers in the the environment. The default is 126. Starting a read-only transaction normally ties a lock table slot to the current thread until the environment closes or the thread exits. If MDB_NOTLS is in use, mdb_txn_begin() instead ties the slot to the MDB_txn object until it or the MDB_env object is destroyed. This function may only be called after mdb_env_create() and before mdb_env_open().
Parameters:
env An environment handle returned by mdb_env_create()
readers The maximum number of reader lock table slots
Returns:
A non-zero error value on failure and 0 on success. Some possible errors are:
EINVAL - an invalid parameter was specified, or the environment is already open.
Get the maximum number of threads/reader slots for the environment.
Parameters:
env An environment handle returned by mdb_env_create()
readers Address of an integer to store the number of readers
Returns:
A non-zero error value on failure and 0 on success. Some possible errors are:
EINVAL - an invalid parameter was specified.
Further setup required for opening an LMDB environment
Release a reader thread's slot in the reader lock table. This function is called automatically when a thread exits.
Parameters:
ptr This points to the slot in the reader lock table.
Downgrade the exclusive lock on the region back to shared
Try to get exlusive lock, otherwise shared. Maintain *excl = -1: no/unknown lock, 0: shared, 1: exclusive.
Open and/or initialize the lock region for the environment.
Parameters:
env The LMDB environment.
lpath The pathname of the file used for the lock region.
mode The Unix permissions for the file, if we create it.
excl Resulting file lock type: -1 none, 0 shared, 1 exclusive
excl In -1, out lock type: -1 none, 0 shared, 1 exclusive
Returns:
0 on success, non-zero on failure.
Open an environment handle. If this function fails, mdb_env_close() must be called to discard the MDB_env handle.
Parameters:
env An environment handle returned by mdb_env_create()
path The directory in which the database files reside. This directory must already exist and be writable.
flags Special options for this environment. This parameter must be set to 0 or by bitwise OR'ing together one or more of the values described here. Flags set by mdb_env_set_flags() are also used.
MDB_FIXEDMAP use a fixed address for the mmap region. This flag must be specified when creating the environment, and is stored persistently in the environment. If successful, the memory map will always reside at the same virtual address and pointers used to reference data items in the database will be constant across multiple invocations. This option may not always work, depending on how the operating system has allocated memory to shared libraries and other uses. The feature is highly experimental.
MDB_NOSUBDIR By default, LMDB creates its environment in a directory whose pathname is given in path, and creates its data and lock files under that directory. With this option, path is used as-is for the database main data file. The database lock file is the path with '-lock' appended.
MDB_RDONLY Open the environment in read-only mode. No write operations will be allowed. LMDB will still modify the lock file - except on read-only filesystems, where LMDB does not use locks.
MDB_WRITEMAP Use a writeable memory map unless MDB_RDONLY is set. This is faster and uses fewer mallocs, but loses protection from application bugs like wild pointer writes and other bad updates into the database. Incompatible with nested transactions. Processes with and without MDB_WRITEMAP on the same environment do not cooperate well.
MDB_NOMETASYNC Flush system buffers to disk only once per transaction, omit the metadata flush. Defer that until the system flushes files to disk, or next non-MDB_RDONLY commit or mdb_env_sync(). This optimization maintains database integrity, but a system crash may undo the last committed transaction. I.e. it preserves the ACI (atomicity, consistency, isolation) but not D (durability) database property. This flag may be changed at any time using mdb_env_set_flags().
MDB_NOSYNC Don't flush system buffers to disk when committing a transaction. This optimization means a system crash can corrupt the database or lose the last transactions if buffers are not yet flushed to disk. The risk is governed by how often the system flushes dirty buffers to disk and how often mdb_env_sync() is called. However, if the filesystem preserves write order and the MDB_WRITEMAP flag is not used, transactions exhibit ACI (atomicity, consistency, isolation) properties and only lose D (durability). I.e. database integrity is maintained, but a system crash may undo the final transactions. Note that (MDB_NOSYNC | MDB_WRITEMAP) leaves the system with no hint for when to write transactions to disk, unless mdb_env_sync() is called. (MDB_MAPASYNC | MDB_WRITEMAP) may be preferable. This flag may be changed at any time using mdb_env_set_flags().
MDB_MAPASYNC When using MDB_WRITEMAP, use asynchronous flushes to disk. As with MDB_NOSYNC, a system crash can then corrupt the database or lose the last transactions. Calling mdb_env_sync() ensures on-disk database integrity until next commit. This flag may be changed at any time using mdb_env_set_flags().
MDB_NOTLS Don't use Thread-Local Storage. Tie reader locktable slots to MDB_txn objects instead of to threads. I.e. mdb_txn_reset() keeps the slot reseved for the MDB_txn object. A thread may use parallel read-only transactions. A read-only transaction may span threads if the user synchronizes its use. Applications that multiplex many user threads over individual OS threads need this option. Such an application must also serialize the write transactions in an OS thread, since LMDB's write locking is unaware of the user threads.
MDB_NOLOCK Don't do any locking. If concurrent access is anticipated, the caller must manage all concurrency itself. For proper operation the caller must enforce single-writer semantics, and must ensure that no readers are using old transactions while a writer is active. The simplest approach is to use an exclusive lock so that no readers may be active at all when a writer begins.
MDB_NORDAHEAD Turn off readahead. Most operating systems perform readahead on read requests by default. This option turns it off if the OS supports it. Turning it off may help random read performance when the DB is larger than RAM and system RAM is full. The option is not implemented on Windows.
MDB_NOMEMINIT Don't initialize malloc'd memory before writing to unused spaces in the data file. By default, memory for pages written to the data file is obtained using malloc. While these pages may be reused in subsequent transactions, freshly malloc'd pages will be initialized to zeroes before use. This avoids persisting leftover data from other code (that used the heap and subsequently freed the memory) into the data file. Note that many other system libraries may allocate and free memory from the heap for arbitrary uses. E.g., stdio may use the heap for file I/O buffers. This initialization step has a modest performance cost so some applications may want to disable it using this flag. This option can be a problem for applications which handle sensitive data like passwords, and it makes memory checkers like Valgrind noisy. This flag is not needed with MDB_WRITEMAP, which writes directly to the mmap instead of using malloc for pages. The initialization is also skipped if MDB_RESERVE is used; the caller is expected to overwrite all of the memory that was reserved in that case. This flag may be changed at any time using mdb_env_set_flags().
mode The UNIX permissions to set on created files. This parameter is ignored on Windows.
Returns:
A non-zero error value on failure and 0 on success. Some possible errors are:
MDB_VERSION_MISMATCH - the version of the LMDB library doesn't match the version that created the database environment.
MDB_INVALID - the environment file headers are corrupted.
ENOENT - the directory specified by the path parameter doesn't exist.
EACCES - the user didn't have permission to access the environment files.
EAGAIN - the environment was locked by another process.
Close the environment and release the memory map. Only a single thread may call this function. All transactions, databases, and cursors must already be closed before calling this function. Attempts to use any such handles after calling this function will cause a SIGSEGV. The environment handle will be freed and must not be used again after this call.
Parameters:
env An environment handle returned by mdb_env_create()
Compare two items pointing at aligned size_t's
Compare two items pointing at aligned unsigned int's
Compare two items pointing at unsigned ints of unknown alignment. Nodes and keys are guaranteed to be 2-byte aligned.
Compare two items lexically
Compare two items in reverse byte order
Search for the lowest key under the current branch page. This just bypasses a NUMKEYS check in the current page before calling mdb_page_search_root(), because the callers are all in situations where the current page is known to be underfilled.
Get items from a database. This function retrieves key/data pairs from the database. The address and length of the data associated with the specified key are returned in the structure to which data refers. If the database supports duplicate keys (MDB_DUPSORT) then the first data item for the key will be returned. Retrieval of other items requires the use of mdb_cursor_get().
Note:
The memory pointed to by the returned values is owned by the database. The caller need not dispose of the memory, and may not modify it in any way. For values returned in a read-only transaction any modification attempts will cause a SIGSEGV.
Values returned from the database are valid only until a subsequent update operation, or the end of the transaction.
Parameters:
txn A transaction handle returned by mdb_txn_begin()
dbi A database handle returned by mdb_dbi_open()
key The key to search for in the database
data The data corresponding to the key
Returns:
A non-zero error value on failure and 0 on success. Some possible errors are:
MDB_NOTFOUND - the key was not in the database.
EINVAL - an invalid parameter was specified.
Retrieve by cursor. This function retrieves key/data pairs from the database. The address and length of the key are returned in the object to which key refers (except for the case of the MDB_SET option, in which the key object is unchanged), and the address and length of the data are returned in the object to which data refers. See mdb_get() for restrictions on using the output values.
Parameters:
cursor A cursor handle returned by mdb_cursor_open()
key The key for a retrieved item
data The data of a retrieved item
op A cursor operation MDB_cursor_op
Returns:
A non-zero error value on failure and 0 on success. Some possible errors are:
MDB_NOTFOUND - no matching key found.
EINVAL - an invalid parameter was specified.
Touch all the pages in the cursor stack. Set mc_top. Makes sure all the pages are writable, before attempting a write operation.
Parameters:
mc The cursor to operate on.
Store by cursor. This function stores key/data pairs into the database. The cursor is positioned at the new item, or on failure usually near it.
Note:
Earlier documentation incorrectly said errors would leave the state of the cursor unchanged.
Parameters:
cursor A cursor handle returned by mdb_cursor_open()
key The key operated on.
data The data operated on.
flags Options for this operation. This parameter must be set to 0 or one of the values described here.
MDB_CURRENT - replace the item at the current cursor position. The key parameter must still be provided, and must match it. If using sorted duplicates (MDB_DUPSORT) the data item must still sort into the same place. This is intended to be used when the new data is the same size as the old. Otherwise it will simply perform a delete of the old record followed by an insert.
MDB_NODUPDATA - enter the new key/data pair only if it does not already appear in the database. This flag may only be specified if the database was opened with MDB_DUPSORT. The function will return MDB_KEYEXIST if the key/data pair already appears in the database.
MDB_NOOVERWRITE - enter the new key/data pair only if the key does not already appear in the database. The function will return MDB_KEYEXIST if the key already appears in the database, even if the database supports duplicates (MDB_DUPSORT).
MDB_RESERVE - reserve space for data of the given size, but don't copy the given data. Instead, return a pointer to the reserved space, which the caller can fill in later. This saves an extra memcpy if the data is being generated later.
MDB_APPEND - append the given key/data pair to the end of the database. No key comparisons are performed. This option allows fast bulk loading when keys are already known to be in the correct order. Loading unsorted keys with this flag will cause data corruption.
MDB_APPENDDUP - as above, but for sorted dup data.
MDB_MULTIPLE - store multiple contiguous data elements in a single request. This flag may only be specified if the database was opened with MDB_DUPFIXED. The data argument must be an array of two MDB_vals. The mv_size of the first MDB_val must be the size of a single data element. The mv_data of the first MDB_val must point to the beginning of the array of contiguous data elements. The mv_size of the second MDB_val must be the count of the number of data elements to store. On return this field will be set to the count of the number of elements actually written. The mv_data of the second MDB_val is unused.
Returns:
A non-zero error value on failure and 0 on success. Some possible errors are:
MDB_MAP_FULL - the database is full, see mdb_env_set_mapsize().
MDB_TXN_FULL - the transaction has too many dirty pages.
EACCES - an attempt was made to modify a read-only database.
EINVAL - an invalid parameter was specified.
Delete current key/data pair. This function deletes the key/data pair to which the cursor refers.
Parameters:
cursor A cursor handle returned by mdb_cursor_open()
flags Options for this operation. This parameter must be set to 0 or one of the values described here.
MDB_NODUPDATA - delete all of the data items for the current key. This flag may only be specified if the database was opened with MDB_DUPSORT.
Returns:
A non-zero error value on failure and 0 on success. Some possible errors are:
EACCES - an attempt was made to modify a read-only database.
EINVAL - an invalid parameter was specified.
Create a cursor handle. A cursor is associated with a specific transaction and database. A cursor cannot be used when its database handle is closed. Nor when its transaction has ended, except with mdb_cursor_renew(). It can be discarded with mdb_cursor_close(). A cursor in a write-transaction can be closed before its transaction ends, and will otherwise be closed when its transaction ends. A cursor in a read-only transaction must be closed explicitly, before or after its transaction ends. It can be reused with mdb_cursor_renew() before finally closing it.
Note:
Earlier documentation said that cursors in every transaction were closed when the transaction committed or aborted.
Parameters:
txn A transaction handle returned by mdb_txn_begin()
dbi A database handle returned by mdb_dbi_open()
cursor Address where the new MDB_cursor handle will be stored
Returns:
A non-zero error value on failure and 0 on success. Some possible errors are:
EINVAL - an invalid parameter was specified.
Renew a cursor handle. A cursor is associated with a specific transaction and database. Cursors that are only used in read-only transactions may be re-used, to avoid unnecessary malloc/free overhead. The cursor may be associated with a new read-only transaction, and referencing the same database handle as it was created with. This may be done whether the previous transaction is live or dead.
Parameters:
txn A transaction handle returned by mdb_txn_begin()
cursor A cursor handle returned by mdb_cursor_open()
Returns:
A non-zero error value on failure and 0 on success. Some possible errors are:
EINVAL - an invalid parameter was specified.
Return count of duplicates for current key. This call is only valid on databases that support sorted duplicate data items MDB_DUPSORT.
Parameters:
cursor A cursor handle returned by mdb_cursor_open()
countp Address where the count will be stored
Returns:
A non-zero error value on failure and 0 on success. Some possible errors are:
EINVAL - cursor is not initialized, or an invalid parameter was specified.
Close a cursor handle. The cursor handle will be freed and must not be used again after this call. Its transaction must still be live if it is a write-transaction.
Parameters:
cursor A cursor handle returned by mdb_cursor_open()
Return the cursor's transaction handle.
Parameters:
cursor A cursor handle returned by mdb_cursor_open()
Return the cursor's database handle.
Parameters:
cursor A cursor handle returned by mdb_cursor_open()
Copy the contents of a cursor.
Parameters:
csrc The cursor to copy from.
cdst The cursor to copy to.
Delete items from a database. This function removes key/data pairs from the database. If the database does not support sorted duplicate data items (MDB_DUPSORT) the data parameter is ignored. If the database supports sorted duplicates and the data parameter is NULL, all of the duplicate data items for the key will be deleted. Otherwise, if the data parameter is non-NULL only the matching data item will be deleted. This function will return MDB_NOTFOUND if the specified key/data pair is not in the database.
Parameters:
txn A transaction handle returned by mdb_txn_begin()
dbi A database handle returned by mdb_dbi_open()
key The key to delete from the database
data The data to delete
Returns:
A non-zero error value on failure and 0 on success. Some possible errors are:
EACCES - an attempt was made to write in a read-only transaction.
EINVAL - an invalid parameter was specified.
Store items into a database. This function stores key/data pairs in the database. The default behavior is to enter the new key/data pair, replacing any previously existing key if duplicates are disallowed, or adding a duplicate data item if duplicates are allowed (MDB_DUPSORT).
Parameters:
txn A transaction handle returned by mdb_txn_begin()
dbi A database handle returned by mdb_dbi_open()
key The key to store in the database
data The data to store
flags Special options for this operation. This parameter must be set to 0 or by bitwise OR'ing together one or more of the values described here.
MDB_NODUPDATA - enter the new key/data pair only if it does not already appear in the database. This flag may only be specified if the database was opened with MDB_DUPSORT. The function will return MDB_KEYEXIST if the key/data pair already appears in the database.
MDB_NOOVERWRITE - enter the new key/data pair only if the key does not already appear in the database. The function will return MDB_KEYEXIST if the key already appears in the database, even if the database supports duplicates (MDB_DUPSORT). The data parameter will be set to point to the existing item.
MDB_RESERVE - reserve space for data of the given size, but don't copy the given data. Instead, return a pointer to the reserved space, which the caller can fill in later - before the next update operation or the transaction ends. This saves an extra memcpy if the data is being generated later. LMDB does nothing else with this memory, the caller is expected to modify all of the space requested.
MDB_APPEND - append the given key/data pair to the end of the database. No key comparisons are performed. This option allows fast bulk loading when keys are already known to be in the correct order. Loading unsorted keys with this flag will cause data corruption.
MDB_APPENDDUP - as above, but for sorted dup data.
Returns:
A non-zero error value on failure and 0 on success. Some possible errors are:
MDB_MAP_FULL - the database is full, see mdb_env_set_mapsize().
MDB_TXN_FULL - the transaction has too many dirty pages.
EACCES - an attempt was made to write in a read-only transaction.
EINVAL - an invalid parameter was specified.
Dedicated writer thread for compacting copy.
Tell the writer thread there's a buffer ready to write
Depth-first tree traversal for compacting copy.
Copy environment with compaction.
Copy environment as-is.
Copy an LMDB environment to the specified path, with options. This function may be used to make a backup of an existing environment. No lockfile is created, since it gets recreated at need.
Note:
This call can trigger significant file size growth if run in parallel with write transactions, because it employs a read-only transaction. See long-lived transactions under Caveats.
Parameters:
env An environment handle returned by mdb_env_create(). It must have already been opened successfully.
path The directory in which the copy will reside. This directory must already exist and be writable but must otherwise be empty.
flags Special options for this operation. This parameter must be set to 0 or by bitwise OR'ing together one or more of the values described here.
MDB_CP_COMPACT - Perform compaction while copying: omit free pages and sequentially renumber all pages in output. This option consumes more CPU and runs more slowly than the default.
Returns:
A non-zero error value on failure and 0 on success.
Copy an LMDB environment to the specified path. This function may be used to make a backup of an existing environment. No lockfile is created, since it gets recreated at need.
Note:
This call can trigger significant file size growth if run in parallel with write transactions, because it employs a read-only transaction. See long-lived transactions under Caveats.
Parameters:
env An environment handle returned by mdb_env_create(). It must have already been opened successfully.
path The directory in which the copy will reside. This directory must already exist and be writable but must otherwise be empty.
Returns:
A non-zero error value on failure and 0 on success.
Set environment flags. This may be used to set some flags in addition to those from mdb_env_open(), or to unset these flags. If several threads change the flags at the same time, the result is undefined.
Parameters:
env An environment handle returned by mdb_env_create()
flags The flags to change, bitwise OR'ed together
onoff A non-zero value sets the flags, zero clears them.
Returns:
A non-zero error value on failure and 0 on success. Some possible errors are:
EINVAL - an invalid parameter was specified.
Get environment flags.
Parameters:
env An environment handle returned by mdb_env_create()
flags The address of an integer to store the flags
Returns:
A non-zero error value on failure and 0 on success. Some possible errors are:
EINVAL - an invalid parameter was specified.
Set application information associated with the MDB_env.
Parameters:
env An environment handle returned by mdb_env_create()
ctx An arbitrary pointer for whatever the application needs.
Returns:
A non-zero error value on failure and 0 on success.
Get the application information associated with the MDB_env.
Parameters:
env An environment handle returned by mdb_env_create()
Returns:
The pointer set by mdb_env_set_userctx().
Set or reset the assert() callback of the environment. Disabled if liblmdb is buillt with NDEBUG.
Note:
This hack should become obsolete as lmdb's error handling matures.
Parameters:
env An environment handle returned by mdb_env_create().
func An MDB_assert_func function, or 0.
Returns:
A non-zero error value on failure and 0 on success.
Return the path that was used in mdb_env_open().
Parameters:
env An environment handle returned by mdb_env_create()
path Address of a string pointer to contain the path. This is the actual string in the environment, not a copy. It should not be altered in any way.
Returns:
A non-zero error value on failure and 0 on success. Some possible errors are:
EINVAL - an invalid parameter was specified.
Return the filedescriptor for the given environment.
Parameters:
env An environment handle returned by mdb_env_create()
fd Address of a mdb_filehandle_t to contain the descriptor.
Returns:
A non-zero error value on failure and 0 on success. Some possible errors are:
EINVAL - an invalid parameter was specified.
Common code for mdb_stat() and mdb_env_stat().
Parameters:
env the environment to operate in.
db the MDB_db record containing the stats to return.
arg the address of an MDB_stat structure to receive the stats.
Returns:
0, this function always succeeds.
Return statistics about the LMDB environment.
Parameters:
env An environment handle returned by mdb_env_create()
stat The address of an MDB_stat structure where the statistics will be copied
Return information about the LMDB environment.
Parameters:
env An environment handle returned by mdb_env_create()
stat The address of an MDB_envinfo structure where the information will be copied
Open a database in the environment. A database handle denotes the name and parameters of a database, independently of whether such a database exists. The database handle may be discarded by calling mdb_dbi_close(). The old database handle is returned if the database was already open. The handle may only be closed once. The database handle will be private to the current transaction until the transaction is successfully committed. If the transaction is aborted the handle will be closed automatically. After a successful commit the handle will reside in the shared environment, and may be used by other transactions. This function must not be called from multiple concurrent transactions. A transaction that uses this function must finish (either commit or abort) before any other transaction may use this function.
To use named databases (with name != NULL), mdb_env_set_maxdbs() must be called before opening the environment. Database names are kept as keys in the unnamed database.
Parameters:
txn A transaction handle returned by mdb_txn_begin()
name The name of the database to open. If only a single database is needed in the environment, this value may be NULL.
flags Special options for this database. This parameter must be set to 0 or by bitwise OR'ing together one or more of the values described here.
MDB_REVERSEKEY Keys are strings to be compared in reverse order, from the end of the strings to the beginning. By default, Keys are treated as strings and compared from beginning to end.
MDB_DUPSORT Duplicate keys may be used in the database. (Or, from another perspective, keys may have multiple data items, stored in sorted order.) By default keys must be unique and may have only a single data item.
MDB_INTEGERKEY Keys are binary integers in native byte order. Setting this option requires all keys to be the same size, typically sizeof(int) or sizeof(size_t).
MDB_DUPFIXED This flag may only be used in combination with MDB_DUPSORT. This option tells the library that the data items for this database are all the same size, which allows further optimizations in storage and retrieval. When all data items are the same size, the MDB_GET_MULTIPLE and MDB_NEXT_MULTIPLE cursor operations may be used to retrieve multiple items at once.
MDB_INTEGERDUP This option specifies that duplicate data items are also integers, and should be sorted as such.
MDB_REVERSEDUP This option specifies that duplicate data items should be compared as strings in reverse order.
MDB_CREATE Create the named database if it doesn't exist. This option is not allowed in a read-only transaction or a read-only environment.
dbi Address where the new MDB_dbi handle will be stored
Returns:
A non-zero error value on failure and 0 on success. Some possible errors are:
MDB_NOTFOUND - the specified database doesn't exist in the environment and MDB_CREATE was not specified.
MDB_DBS_FULL - too many databases have been opened. See mdb_env_set_maxdbs().
Retrieve statistics for a database.
Parameters:
txn A transaction handle returned by mdb_txn_begin()
dbi A database handle returned by mdb_dbi_open()
stat The address of an MDB_stat structure where the statistics will be copied
Returns:
A non-zero error value on failure and 0 on success. Some possible errors are:
EINVAL - an invalid parameter was specified.
Close a database handle. Normally unnecessary. Use with care: This call is not mutex protected. Handles should only be closed by a single thread, and only if no other threads are going to reference the database handle or one of its cursors any further. Do not close a handle if an existing transaction has modified its database. Doing so can cause misbehavior from database corruption to errors like MDB_BAD_VALSIZE (since the DB name is gone).
Closing a database handle is not necessary, but lets mdb_dbi_open() reuse the handle value. Usually it's better to set a bigger mdb_env_set_maxdbs(), unless that value would be large.
Parameters:
env An environment handle returned by mdb_env_create()
dbi A database handle returned by mdb_dbi_open()
Retrieve the DB flags for a database handle.
Parameters:
txn A transaction handle returned by mdb_txn_begin()
dbi A database handle returned by mdb_dbi_open()
flags Address where the flags will be returned.
Returns:
A non-zero error value on failure and 0 on success.
Empty or delete+close a database. See mdb_dbi_close() for restrictions about closing the DB handle.
Parameters:
txn A transaction handle returned by mdb_txn_begin()
dbi A database handle returned by mdb_dbi_open()
del 0 to empty the DB, 1 to delete it from the environment and close the DB handle.
Returns:
A non-zero error value on failure and 0 on success.
Set a custom key comparison function for a database. The comparison function is called whenever it is necessary to compare a key specified by the application with a key currently stored in the database. If no comparison function is specified, and no special key flags were specified with mdb_dbi_open(), the keys are compared lexically, with shorter keys collating before longer keys.
Warning:
This function must be called before any data access functions are used, otherwise data corruption may occur. The same comparison function must be used by every program accessing the database, every time the database is used.
Parameters:
txn A transaction handle returned by mdb_txn_begin()
dbi A database handle returned by mdb_dbi_open()
cmp A MDB_cmp_func function
Returns:
A non-zero error value on failure and 0 on success. Some possible errors are:
EINVAL - an invalid parameter was specified.
Set a custom data comparison function for a MDB_DUPSORT database. This comparison function is called whenever it is necessary to compare a data item specified by the application with a data item currently stored in the database. This function only takes effect if the database was opened with the MDB_DUPSORT flag. If no comparison function is specified, and no special key flags were specified with mdb_dbi_open(), the data items are compared lexically, with shorter items collating before longer items.
Warning:
This function must be called before any data access functions are used, otherwise data corruption may occur. The same comparison function must be used by every program accessing the database, every time the database is used.
Parameters:
txn A transaction handle returned by mdb_txn_begin()
dbi A database handle returned by mdb_dbi_open()
cmp A MDB_cmp_func function
Returns:
A non-zero error value on failure and 0 on success. Some possible errors are:
EINVAL - an invalid parameter was specified.
Set a relocation function for a MDB_FIXEDMAP database.
Todo
The relocation function is called whenever it is necessary to move the data of an item to a different position in the database (e.g. through tree balancing operations, shifts as a result of adds or deletes, etc.). It is intended to allow address/position-dependent data items to be stored in a database in an environment opened with the MDB_FIXEDMAP option. Currently the relocation feature is unimplemented and setting this function has no effect.
Parameters:
txn A transaction handle returned by mdb_txn_begin()
dbi A database handle returned by mdb_dbi_open()
rel A MDB_rel_func function
Returns:
A non-zero error value on failure and 0 on success. Some possible errors are:
EINVAL - an invalid parameter was specified.
Set a context pointer for a MDB_FIXEDMAP database's relocation function. See mdb_set_relfunc and MDB_rel_func for more details.
Parameters:
txn A transaction handle returned by mdb_txn_begin()
dbi A database handle returned by mdb_dbi_open()
ctx An arbitrary pointer for whatever the application needs. It will be passed to the callback function set by mdb_set_relfunc as its relctx parameter whenever the callback is invoked.
Returns:
A non-zero error value on failure and 0 on success. Some possible errors are:
EINVAL - an invalid parameter was specified.
Get the maximum size of keys and MDB_DUPSORT data we can write. Depends on the compile-time constant MDB_MAXKEYSIZE. Default 511. See MDB_val.
Parameters:
env An environment handle returned by mdb_env_create()
Returns:
The maximum size of a key we can write
Dump the entries in the reader lock table.
Parameters:
env An environment handle returned by mdb_env_create()
func A MDB_msg_func function
ctx Anything the message function needs
Returns:
< 0 on failure, >= 0 on success.
Insert pid into list if not already present. return -1 if already present.
Check for stale entries in the reader lock table.
Parameters:
env An environment handle returned by mdb_env_create()
dead Number of stale slots that were cleared
Returns:
0 on success, non-zero on failure.
Initial value:
= {
"MDB_KEYEXIST: Key/data pair already exists",
"MDB_NOTFOUND: No matching key/data pair found",
"MDB_PAGE_NOTFOUND: Requested page not found",
"MDB_CORRUPTED: Located page was wrong type",
"MDB_PANIC: Update of meta page failed",
"MDB_VERSION_MISMATCH: Database environment version mismatch",
"MDB_INVALID: File is not an LMDB file",
"MDB_MAP_FULL: Environment mapsize limit reached",
"MDB_DBS_FULL: Environment maxdbs limit reached",
"MDB_READERS_FULL: Environment maxreaders limit reached",
"MDB_TLS_FULL: Thread-local storage keys full - too many environments open",
"MDB_TXN_FULL: Transaction has too many dirty pages - transaction too big",
"MDB_CURSOR_FULL: Internal error - cursor stack limit reached",
"MDB_PAGE_FULL: Internal error - page has no more space",
"MDB_MAP_RESIZED: Database contents grew beyond environment mapsize",
"MDB_INCOMPATIBLE: Operation and DB incompatible, or DB flags changed",
"MDB_BAD_RSLOT: Invalid reuse of reader locktable slot",
"MDB_BAD_TXN: Transaction cannot recover - it must be aborted",
"MDB_BAD_VALSIZE: Unsupported size of key/DB name/data, or wrong DUPFIXED size",
"MDB_BAD_DBI: The specified DBI handle was closed/changed unexpectedly",
}
Table of descriptions for LMDB Return Codes
Generated automatically by Doxygen for LMDB from the source code.