#include <assert.h>
#include <errno.h>
+#include <limits.h>
#include <stddef.h>
#include <stdint.h>
#include <stdio.h>
#ifndef _WIN32
#include <pthread.h>
+#ifdef __APPLE__
+#include <semaphore.h>
+#endif
+#endif
+
+#ifndef BYTE_ORDER
+#define BYTE_ORDER __BYTE_ORDER
+#endif
+#ifndef LITTLE_ENDIAN
+#define LITTLE_ENDIAN __LITTLE_ENDIAN
+#endif
+#ifndef BIG_ENDIAN
+#define BIG_ENDIAN __BIG_ENDIAN
#endif
#include "mdb.h"
#include "midl.h"
+#if (BYTE_ORDER == LITTLE_ENDIAN) == (BYTE_ORDER == BIG_ENDIAN)
+# error "Unknown or unsupported endianness (BYTE_ORDER)"
+#elif (-6 & 5) || CHAR_BIT != 8 || UINT_MAX < 0xffffffff || ULONG_MAX % 0xFFFF
+# error "Two's complement, reasonably sized integer types, please"
+#endif
+
/** @defgroup internal MDB Internals
* @{
*/
#define pthread_mutex_t HANDLE
#define pthread_key_t DWORD
#define pthread_self() GetCurrentThreadId()
-#define pthread_key_create(x,y) *(x) = TlsAlloc()
+#define pthread_key_create(x,y) (*(x) = TlsAlloc())
#define pthread_key_delete(x) TlsFree(x)
#define pthread_getspecific(x) TlsGetValue(x)
#define pthread_setspecific(x,y) TlsSetValue(x,y)
#define pthread_mutex_unlock(x) ReleaseMutex(x)
#define pthread_mutex_lock(x) WaitForSingleObject(x, INFINITE)
-#define LOCK_MUTEX_R(env) pthread_mutex_lock(env->me_rmutex)
-#define UNLOCK_MUTEX_R(env) pthread_mutex_unlock(env->me_rmutex)
-#define LOCK_MUTEX_W(env) pthread_mutex_lock(env->me_wmutex)
-#define UNLOCK_MUTEX_W(env) pthread_mutex_unlock(env->me_wmutex)
+#define LOCK_MUTEX_R(env) pthread_mutex_lock((env)->me_rmutex)
+#define UNLOCK_MUTEX_R(env) pthread_mutex_unlock((env)->me_rmutex)
+#define LOCK_MUTEX_W(env) pthread_mutex_lock((env)->me_wmutex)
+#define UNLOCK_MUTEX_W(env) pthread_mutex_unlock((env)->me_wmutex)
#define getpid() GetCurrentProcessId()
-#define fdatasync(fd) !FlushFileBuffers(fd)
+#define fdatasync(fd) (!FlushFileBuffers(fd))
#define ErrCode() GetLastError()
-#define GetPageSize(x) {SYSTEM_INFO si; GetSystemInfo(&si); (x) = si.dwPageSize;}
+#define GET_PAGESIZE(x) {SYSTEM_INFO si; GetSystemInfo(&si); (x) = si.dwPageSize;}
#define close(fd) CloseHandle(fd)
#define munmap(ptr,len) UnmapViewOfFile(ptr)
+#else
+#ifdef __APPLE__
+#define LOCK_MUTEX_R(env) sem_wait((env)->me_rmutex)
+#define UNLOCK_MUTEX_R(env) sem_post((env)->me_rmutex)
+#define LOCK_MUTEX_W(env) sem_wait((env)->me_wmutex)
+#define UNLOCK_MUTEX_W(env) sem_post((env)->me_wmutex)
+#define fdatasync(fd) fsync(fd)
#else
/** Lock the reader mutex.
*/
-#define LOCK_MUTEX_R(env) pthread_mutex_lock(&env->me_txns->mti_mutex)
+#define LOCK_MUTEX_R(env) pthread_mutex_lock(&(env)->me_txns->mti_mutex)
/** Unlock the reader mutex.
*/
-#define UNLOCK_MUTEX_R(env) pthread_mutex_unlock(&env->me_txns->mti_mutex)
+#define UNLOCK_MUTEX_R(env) pthread_mutex_unlock(&(env)->me_txns->mti_mutex)
/** Lock the writer mutex.
* Only a single write transaction is allowed at a time. Other writers
* will block waiting for this mutex.
*/
-#define LOCK_MUTEX_W(env) pthread_mutex_lock(&env->me_txns->mti_wmutex)
+#define LOCK_MUTEX_W(env) pthread_mutex_lock(&(env)->me_txns->mti_wmutex)
/** Unlock the writer mutex.
*/
-#define UNLOCK_MUTEX_W(env) pthread_mutex_unlock(&env->me_txns->mti_wmutex)
+#define UNLOCK_MUTEX_W(env) pthread_mutex_unlock(&(env)->me_txns->mti_wmutex)
+#endif /* __APPLE__ */
/** Get the error code for the last failed system function.
*/
* Mainly used to initialize file variables and signify that they are
* unused.
*/
-#define INVALID_HANDLE_VALUE -1
+#define INVALID_HANDLE_VALUE (-1)
/** Get the size of a memory page for the system.
* This is the basic size that the platform's memory manager uses, and is
* fundamental to the use of memory-mapped files.
*/
-#define GetPageSize(x) (x) = sysconf(_SC_PAGE_SIZE)
+#define GET_PAGESIZE(x) ((x) = sysconf(_SC_PAGE_SIZE))
+#endif
+
+#if defined(_WIN32) || defined(__APPLE__)
+#define MNAME_LEN 32
#endif
/** @} */
* @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.
*/
-typedef ULONG pgno_t;
+typedef ID pgno_t;
+
+ /** A transaction ID.
+ * See struct MDB_txn.mt_txnid for details.
+ */
+typedef ID txnid_t;
/** @defgroup debug Debug Macros
* @{
/** A default memory page size.
* The actual size is platform-dependent, but we use this for
* boot-strapping. We probably should not be using this any more.
- * The #GetPageSize() macro is used to get the actual size.
+ * The #GET_PAGESIZE() macro is used to get the actual size.
*
* Note that we don't currently support Huge pages. On Linux,
* regular data files cannot use Huge pages, and in general
*/
#define DKEY(x) mdb_dkey(x, kbuf)
#else
-#define DKBUF
+#define DKBUF typedef int dummy_kbuf /* so we can put ';' after */
#define DKEY(x)
#endif
#define LAZY_RWLOCK_WRLOCK(x)
/** Grab the DB table read lock */
#define LAZY_RWLOCK_RDLOCK(x)
- /** Declare the DB table rwlock */
+ /** Declare the DB table rwlock. Should not be followed by ';'. */
#define LAZY_RWLOCK_DEF(x)
/** Initialize the DB table rwlock */
#define LAZY_RWLOCK_INIT(x,y)
#define LAZY_RWLOCK_UNLOCK(x) pthread_rwlock_unlock(x)
#define LAZY_RWLOCK_WRLOCK(x) pthread_rwlock_wrlock(x)
#define LAZY_RWLOCK_RDLOCK(x) pthread_rwlock_rdlock(x)
-#define LAZY_RWLOCK_DEF(x) pthread_rwlock_t x
+#define LAZY_RWLOCK_DEF(x) pthread_rwlock_t x;
#define LAZY_RWLOCK_INIT(x,y) pthread_rwlock_init(x,y)
#define LAZY_RWLOCK_DESTROY(x) pthread_rwlock_destroy(x)
#endif
* started from so we can avoid overwriting any data used in that
* particular version.
*/
- ULONG mrb_txnid;
+ txnid_t mrb_txnid;
/** The process ID of the process owning this reader txn. */
pid_t mrb_pid;
/** The thread ID of the thread owning this txn. */
uint32_t mtb_magic;
/** Version number of this lock file. Must be set to #MDB_VERSION. */
uint32_t mtb_version;
-#ifdef _WIN32
- char mtb_rmname[32];
+#if defined(_WIN32) || defined(__APPLE__)
+ char mtb_rmname[MNAME_LEN];
#else
/** Mutex protecting access to this table.
* This is the reader lock that #LOCK_MUTEX_R acquires.
* This is recorded here only for convenience; the value can always
* be determined by reading the main database meta pages.
*/
- ULONG mtb_txnid;
+ txnid_t mtb_txnid;
/** The number of slots that have been used in the reader table.
* This always records the maximum count, it is not decremented
* when readers release their slots.
*/
- uint32_t mtb_numreaders;
+ unsigned mtb_numreaders;
/** The ID of the most recent meta page in the database.
* This is recorded here only for convenience; the value can always
* be determined by reading the main database meta pages.
char pad[(sizeof(MDB_txbody)+CACHELINE-1) & ~(CACHELINE-1)];
} mt1;
union {
-#ifdef _WIN32
- char mt2_wmname[32];
+#if defined(_WIN32) || defined(__APPLE__)
+ char mt2_wmname[MNAME_LEN];
#define mti_wmname mt2.mt2_wmname
#else
pthread_mutex_t mt2_wmutex;
/** @} */
/** Common header for all page types.
- * Overflow pages occupy a number of contiguous pages with no
+ * Overflow records occupy a number of contiguous pages with no
* headers on any page after the first.
*/
typedef struct MDB_page {
#define mp_pgno mp_p.p_pgno
#define mp_next mp_p.p_next
- union padded {
+ union {
pgno_t p_pgno; /**< page number */
void * p_next; /**< for in-memory list of freed structs */
} mp_p;
+/** @defgroup mdb_page Page Flags
+ * @ingroup internal
+ * Flags for the page headers.
+ * @{
+ */
#define P_BRANCH 0x01 /**< branch page */
#define P_LEAF 0x02 /**< leaf page */
#define P_OVERFLOW 0x04 /**< overflow page */
#define P_META 0x08 /**< meta page */
#define P_DIRTY 0x10 /**< dirty page */
#define P_LEAF2 0x20 /**< for #MDB_DUPFIXED records */
- uint32_t mp_flags;
+/** @} */
+ uint32_t mp_flags; /**< @ref mdb_page */
#define mp_lower mp_pb.pb.pb_lower
#define mp_upper mp_pb.pb.pb_upper
#define mp_pages mp_pb.pb_pages
- union page_bounds {
+ union {
struct {
indx_t pb_lower; /**< lower bound of free space */
indx_t pb_upper; /**< upper bound of free space */
typedef struct MDB_node {
/** 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 ignore 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.
*/
- unsigned short mn_lo;
- unsigned short mn_hi; /**< part of dsize or pgno */
- unsigned short mn_flags; /**< flags for special node types */
+#define mn_lo mn_offset[BYTE_ORDER!=LITTLE_ENDIAN]
+#define mn_hi mn_offset[BYTE_ORDER==LITTLE_ENDIAN] /**< part of dsize or pgno */
+ unsigned short mn_offset[2]; /**< storage for #mn_lo and #mn_hi */
+/** @defgroup mdb_node Node Flags
+ * @ingroup internal
+ * Flags for node headers.
+ * @{
+ */
#define F_BIGDATA 0x01 /**< data put on overflow page */
#define F_SUBDATA 0x02 /**< data is a sub-database */
#define F_DUPDATA 0x04 /**< data has duplicates */
+/** @} */
+ unsigned short mn_flags; /**< @ref mdb_node */
unsigned short mn_ksize; /**< key size */
char mn_data[1]; /**< key and data are appended here */
} MDB_node;
/** Size of the node header, excluding dynamic data at the end */
#define NODESIZE offsetof(MDB_node, mn_data)
+ /** Bit position of top word in page number, for shifting mn_flags */
+#define PGNO_TOPWORD ((pgno_t)-1 > 0xffffffffu ? 32 : 0)
+
/** 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.
*/
#define NODEDATA(node) (void *)((char *)(node)->mn_data + (node)->mn_ksize)
/** Get the page number pointed to by a branch node */
-#if LONG_MAX == 0x7fffffff
-#define NODEPGNO(node) ((node)->mn_lo | ((node)->mn_hi << 16))
- /** Set the page number in a branch node */
-#define SETPGNO(node,pgno) do { \
- (node)->mn_lo = (pgno) & 0xffff; (node)->mn_hi = (pgno) >> 16;} while(0)
-#else
-#define NODEPGNO(node) ((node)->mn_lo | ((node)->mn_hi << 16) | ((unsigned long)(node)->mn_flags << 32))
+#define NODEPGNO(node) \
+ ((node)->mn_lo | ((pgno_t) (node)->mn_hi << 16) | \
+ (PGNO_TOPWORD ? ((pgno_t) (node)->mn_flags << PGNO_TOPWORD) : 0))
/** Set the page number in a branch node */
#define SETPGNO(node,pgno) do { \
(node)->mn_lo = (pgno) & 0xffff; (node)->mn_hi = (pgno) >> 16; \
- (node)->mn_flags = (pgno) >> 32; } while(0)
-#endif
+ if (PGNO_TOPWORD) (node)->mn_flags = (pgno) >> PGNO_TOPWORD; } while(0)
/** Get the size of the data in a leaf node */
#define NODEDSZ(node) ((node)->mn_lo | ((unsigned)(node)->mn_hi << 16))
#define LEAF2KEY(p, i, ks) ((char *)(p) + PAGEHDRSZ + ((i)*(ks)))
/** Set the \b node's key into \b key, if requested. */
-#define MDB_SET_KEY(node, key) if (key!=NULL) {(key)->mv_size = NODEKSZ(node); (key)->mv_data = NODEKEY(node);}
+#define MDB_SET_KEY(node, key) { if ((key) != NULL) { \
+ (key)->mv_size = NODEKSZ(node); (key)->mv_data = NODEKEY(node); } }
/** Information about a single database in the environment. */
typedef struct MDB_db {
uint32_t md_pad; /**< also ksize for LEAF2 pages */
uint16_t md_flags; /**< @ref mdb_open */
uint16_t md_depth; /**< depth of this tree */
- ULONG md_branch_pages; /**< number of internal pages */
- ULONG md_leaf_pages; /**< number of leaf pages */
- ULONG md_overflow_pages; /**< number of overflow pages */
- ULONG md_entries; /**< number of data items */
+ pgno_t md_branch_pages; /**< number of internal pages */
+ pgno_t md_leaf_pages; /**< number of leaf pages */
+ pgno_t md_overflow_pages; /**< number of overflow pages */
+ size_t md_entries; /**< number of data items */
pgno_t md_root; /**< the root page of this tree */
} MDB_db;
/** Any persistent environment flags. @ref mdb_env */
#define mm_flags mm_dbs[0].md_flags
pgno_t mm_last_pg; /**< last used page in file */
- ULONG mm_txnid; /**< txnid that committed this page */
+ txnid_t mm_txnid; /**< txnid that committed this page */
} MDB_meta;
/** Auxiliary DB info.
MDB_cmp_func *md_cmp; /**< function for comparing keys */
MDB_cmp_func *md_dcmp; /**< function for comparing data items */
MDB_rel_func *md_rel; /**< user relocate function */
+ void *md_relctx; /**< user-provided context for md_rel */
MDB_dbi md_parent; /**< parent DB of a sub-DB */
unsigned int md_dirty; /**< TRUE if DB was written in this txn */
} MDB_dbx;
* Only committed write transactions increment the ID. If a transaction
* aborts, the ID may be re-used by the next writer.
*/
- ULONG mt_txnid;
+ txnid_t mt_txnid;
MDB_env *mt_env; /**< the DB environment */
/** The list of pages that became unused during this transaction.
* This is an #IDL.
/** Number of DB records in use. This number only ever increments;
* we don't decrement it when individual DB handles are closed.
*/
- unsigned int mt_numdbs;
+ MDB_dbi mt_numdbs;
+/** @defgroup mdb_txn Transaction Flags
+ * @ingroup internal
+ * @{
+ */
#define MDB_TXN_RDONLY 0x01 /**< read-only transaction */
#define MDB_TXN_ERROR 0x02 /**< an error has occurred */
- unsigned int mt_flags;
+/** @} */
+ unsigned int mt_flags; /**< @ref mdb_txn */
/** Tracks which of the two meta pages was used at the start
* of this transaction.
*/
MDB_dbx *mc_dbx;
unsigned short mc_snum; /**< number of pushed pages */
unsigned short mc_top; /**< index of top page, mc_snum-1 */
- unsigned int mc_flags;
+/** @defgroup mdb_cursor Cursor Flags
+ * @ingroup internal
+ * Cursor state flags.
+ * @{
+ */
#define C_INITIALIZED 0x01 /**< cursor has been initialized and is valid */
#define C_EOF 0x02 /**< No more data */
-#define C_XDIRTY 0x04 /**< @deprecated mc_xcursor needs to be flushed */
+/** @} */
+ unsigned int mc_flags; /**< @ref mdb_cursor */
MDB_page *mc_pg[CURSOR_STACK]; /**< stack of pushed pages */
indx_t mc_ki[CURSOR_STACK]; /**< stack of page indices */
};
*/
struct MDB_oldpages *mo_next;
/** The ID of the transaction in which these pages were freed. */
- ULONG mo_txnid;
+ txnid_t mo_txnid;
/** An #IDL of the pages */
pgno_t mo_pages[1]; /* dynamic */
} MDB_oldpages;
HANDLE me_mfd; /**< just for writing the meta pages */
/** Failed to update the meta page. Probably an I/O error. */
#define MDB_FATAL_ERROR 0x80000000U
- uint32_t me_flags;
+ uint32_t me_flags; /**< @ref mdb_env */
uint32_t me_extrapad; /**< unused for now */
unsigned int me_maxreaders; /**< size of the reader table */
- unsigned int me_numdbs; /**< number of DBs opened */
- unsigned int me_maxdbs; /**< size of the DB table */
+ MDB_dbi me_numdbs; /**< number of DBs opened */
+ MDB_dbi me_maxdbs; /**< size of the DB table */
char *me_path; /**< path to the DB files */
char *me_map; /**< the memory map of the data file */
MDB_txninfo *me_txns; /**< the memory map of the lock file */
size_t me_mapsize; /**< size of the data memory map */
off_t me_size; /**< current file size */
pgno_t me_maxpg; /**< me_mapsize / me_psize */
- unsigned int me_psize; /**< size of a page, from #GetPageSize */
+ unsigned int me_psize; /**< size of a page, from #GET_PAGESIZE */
unsigned int me_db_toggle; /**< which DB table is current */
MDB_dbx *me_dbxs; /**< array of static DB info */
MDB_db *me_dbs[2]; /**< two arrays of MDB_db info */
/** ID2L of pages that were written during a write txn */
ID2 me_dirty_list[MDB_IDL_UM_SIZE];
/** rwlock for the DB tables, if #LAZY_LOCKS is false */
- LAZY_RWLOCK_DEF(me_dblock);
+ LAZY_RWLOCK_DEF(me_dblock)
#ifdef _WIN32
HANDLE me_rmutex; /* Windows mutexes don't reside in shared mem */
HANDLE me_wmutex;
#endif
+#ifdef __APPLE__
+ sem_t *me_rmutex; /* Apple doesn't support shared mutexes */
+ sem_t *me_wmutex;
+#endif
};
/** max number of pages to commit in one writev() call */
#define MDB_COMMIT_PAGES 64
-static MDB_page *mdb_alloc_page(MDB_cursor *mc, int num);
-static int mdb_touch(MDB_cursor *mc);
+static MDB_page *mdb_page_alloc(MDB_cursor *mc, int num);
+static MDB_page *mdb_page_new(MDB_cursor *mc, uint32_t flags, int num);
+static int mdb_page_touch(MDB_cursor *mc);
-static int mdb_search_page_root(MDB_cursor *mc,
+static int mdb_page_get(MDB_txn *txn, pgno_t pgno, MDB_page **mp);
+static int mdb_page_search_root(MDB_cursor *mc,
MDB_val *key, int modify);
-static int mdb_search_page(MDB_cursor *mc,
+static int mdb_page_search(MDB_cursor *mc,
MDB_val *key, int modify);
+static int mdb_page_merge(MDB_cursor *csrc, MDB_cursor *cdst);
+static int mdb_page_split(MDB_cursor *mc, MDB_val *newkey, MDB_val *newdata,
+ pgno_t newpgno);
static int mdb_env_read_header(MDB_env *env, MDB_meta *meta);
static int mdb_env_read_meta(MDB_env *env, int *which);
static int mdb_env_write_meta(MDB_txn *txn);
-static int mdb_get_page(MDB_txn *txn, pgno_t pgno, MDB_page **mp);
-static MDB_node *mdb_search_node(MDB_cursor *mc, MDB_val *key, int *exactp);
-static int mdb_add_node(MDB_cursor *mc, indx_t indx,
+static MDB_node *mdb_node_search(MDB_cursor *mc, MDB_val *key, int *exactp);
+static int mdb_node_add(MDB_cursor *mc, indx_t indx,
MDB_val *key, MDB_val *data, pgno_t pgno, uint8_t flags);
-static void mdb_del_node(MDB_page *mp, indx_t indx, int ksize);
-static int mdb_del0(MDB_cursor *mc, MDB_node *leaf);
-static int mdb_read_data(MDB_txn *txn, MDB_node *leaf, MDB_val *data);
+static void mdb_node_del(MDB_page *mp, indx_t indx, int ksize);
+static int mdb_node_move(MDB_cursor *csrc, MDB_cursor *cdst);
+static int mdb_node_read(MDB_txn *txn, MDB_node *leaf, MDB_val *data);
+static size_t mdb_leaf_size(MDB_env *env, MDB_val *key, MDB_val *data);
+static size_t mdb_branch_size(MDB_env *env, MDB_val *key);
static int mdb_rebalance(MDB_cursor *mc);
static int mdb_update_key(MDB_page *mp, indx_t indx, MDB_val *key);
-static int mdb_move_node(MDB_cursor *csrc, MDB_cursor *cdst);
-static int mdb_merge(MDB_cursor *csrc, MDB_cursor *cdst);
-static int mdb_split(MDB_cursor *mc, MDB_val *newkey, MDB_val *newdata,
- pgno_t newpgno);
-static MDB_page *mdb_new_page(MDB_cursor *mc, uint32_t flags, int num);
-static void cursor_pop_page(MDB_cursor *mc);
-static int cursor_push_page(MDB_cursor *mc, MDB_page *mp);
+static void mdb_cursor_pop(MDB_cursor *mc);
+static int mdb_cursor_push(MDB_cursor *mc, MDB_page *mp);
-static int mdb_sibling(MDB_cursor *mc, int move_right);
+static int mdb_cursor_del0(MDB_cursor *mc, MDB_node *leaf);
+static int mdb_cursor_sibling(MDB_cursor *mc, int move_right);
static int mdb_cursor_next(MDB_cursor *mc, MDB_val *key, MDB_val *data, MDB_cursor_op op);
static int mdb_cursor_prev(MDB_cursor *mc, MDB_val *key, MDB_val *data, MDB_cursor_op op);
static int mdb_cursor_set(MDB_cursor *mc, MDB_val *key, MDB_val *data, MDB_cursor_op op,
static int mdb_cursor_first(MDB_cursor *mc, MDB_val *key, MDB_val *data);
static int mdb_cursor_last(MDB_cursor *mc, MDB_val *key, MDB_val *data);
-static void mdb_cursor_init(MDB_cursor *mc, MDB_txn *txn, MDB_dbi dbi);
+static void mdb_cursor_init(MDB_cursor *mc, MDB_txn *txn, MDB_dbi dbi, MDB_xcursor *mx);
static void mdb_xcursor_init0(MDB_cursor *mc);
static void mdb_xcursor_init1(MDB_cursor *mc, MDB_node *node);
-static size_t mdb_leaf_size(MDB_env *env, MDB_val *key, MDB_val *data);
-static size_t mdb_branch_size(MDB_env *env, MDB_val *key);
-
static void mdb_default_cmp(MDB_txn *txn, MDB_dbi dbi);
/** @cond */
-static MDB_cmp_func memncmp, memnrcmp, intcmp, cintcmp;
+static MDB_cmp_func mdb_cmp_memn, mdb_cmp_memnr, mdb_cmp_int, mdb_cmp_cint, mdb_cmp_long;
/** @endcond */
#ifdef _WIN32
return txn->mt_dbxs[dbi].md_cmp(a, b);
}
-/** Compare two data items according to a particular database.
- * This returns a comparison as if the two items were data items of
- * a sorted duplicates #MDB_DUPSORT database.
- * @param[in] txn A transaction handle returned by #mdb_txn_begin()
- * @param[in] dbi A database handle returned by #mdb_open()
- * @param[in] a The first item to compare
- * @param[in] b The second item to compare
- * @return < 0 if a < b, 0 if a == b, > 0 if a > b
- */
int
mdb_dcmp(MDB_txn *txn, MDB_dbi dbi, const MDB_val *a, const MDB_val *b)
{
* will always be satisfied by a single contiguous chunk of memory.
*/
static MDB_page *
-mdb_alloc_page(MDB_cursor *mc, int num)
+mdb_page_alloc(MDB_cursor *mc, int num)
{
MDB_txn *txn = mc->mc_txn;
MDB_page *np;
/* See if there's anything in the free DB */
MDB_cursor m2;
MDB_node *leaf;
- ULONG *kptr, oldest;
+ txnid_t *kptr, oldest;
- mdb_cursor_init(&m2, txn, FREE_DBI);
- mdb_search_page(&m2, NULL, 0);
+ mdb_cursor_init(&m2, txn, FREE_DBI, NULL);
+ mdb_page_search(&m2, NULL, 0);
leaf = NODEPTR(m2.mc_pg[m2.mc_top], 0);
- kptr = (ULONG *)NODEKEY(leaf);
+ kptr = (txnid_t *)NODEKEY(leaf);
{
unsigned int i;
oldest = txn->mt_txnid - 1;
for (i=0; i<txn->mt_env->me_txns->mti_numreaders; i++) {
- ULONG mr = txn->mt_env->me_txns->mti_readers[i].mr_txnid;
+ txnid_t mr = txn->mt_env->me_txns->mti_readers[i].mr_txnid;
if (mr && mr < oldest)
oldest = mr;
}
MDB_val data;
pgno_t *idl;
- mdb_read_data(txn, leaf, &data);
- idl = (ULONG *)data.mv_data;
+ mdb_node_read(txn, leaf, &data);
+ idl = (ID *) data.mv_data;
mop = malloc(sizeof(MDB_oldpages) + MDB_IDL_SIZEOF(idl) - sizeof(pgno_t));
mop->mo_next = txn->mt_env->me_pghead;
mop->mo_txnid = *kptr;
#if DEBUG > 1
{
unsigned int i;
- DPRINTF("IDL read txn %lu root %lu num %lu",
+ DPRINTF("IDL read txn %zu root %zu num %zu",
mop->mo_txnid, txn->mt_dbs[FREE_DBI].md_root, idl[0]);
for (i=0; i<idl[0]; i++) {
- DPRINTF("IDL %lu", idl[i+1]);
+ DPRINTF("IDL %zu", idl[i+1]);
}
}
#endif
* @return 0 on success, non-zero on failure.
*/
static int
-mdb_touch(MDB_cursor *mc)
+mdb_page_touch(MDB_cursor *mc)
{
MDB_page *mp = mc->mc_pg[mc->mc_top];
pgno_t pgno;
if (!F_ISSET(mp->mp_flags, P_DIRTY)) {
MDB_page *np;
- if ((np = mdb_alloc_page(mc, 1)) == NULL)
+ if ((np = mdb_page_alloc(mc, 1)) == NULL)
return ENOMEM;
- DPRINTF("touched db %u page %lu -> %lu", mc->mc_dbi, mp->mp_pgno, np->mp_pgno);
+ DPRINTF("touched db %u page %zu -> %zu", mc->mc_dbi, mp->mp_pgno, np->mp_pgno);
assert(mp->mp_pgno != np->mp_pgno);
mdb_midl_append(mc->mc_txn->mt_free_pgs, mp->mp_pgno);
pgno = np->mp_pgno;
rc = mdb_txn_renew0(txn);
if (rc == MDB_SUCCESS) {
- DPRINTF("renew txn %lu%c %p on mdbenv %p, root page %lu",
- txn->mt_txnid, (txn->mt_flags & MDB_TXN_RDONLY) ? 'r' : 'w', txn,
- (void *)txn->mt_env, txn->mt_dbs[MAIN_DBI].md_root);
+ DPRINTF("renew txn %zu%c %p on mdbenv %p, root page %zu",
+ txn->mt_txnid, (txn->mt_flags & MDB_TXN_RDONLY) ? 'r' : 'w',
+ (void *)txn, (void *)txn->mt_env, txn->mt_dbs[MAIN_DBI].md_root);
}
return rc;
}
free(txn);
else {
*ret = txn;
- DPRINTF("begin txn %lu%c %p on mdbenv %p, root page %lu",
- txn->mt_txnid, (txn->mt_flags & MDB_TXN_RDONLY) ? 'r' : 'w', txn,
- (void *) env, txn->mt_dbs[MAIN_DBI].md_root);
+ DPRINTF("begin txn %zu%c %p on mdbenv %p, root page %zu",
+ txn->mt_txnid, (txn->mt_flags & MDB_TXN_RDONLY) ? 'r' : 'w',
+ (void *) txn, (void *) env, txn->mt_dbs[MAIN_DBI].md_root);
}
return rc;
} else {
MDB_oldpages *mop;
MDB_page *dp;
+ MDB_dbi dbi;
unsigned int i;
/* return all dirty pages to dpage list */
}
env->me_txn = NULL;
- for (i=2; i<env->me_numdbs; i++)
- env->me_dbxs[i].md_dirty = 0;
+ for (dbi=2; dbi<env->me_numdbs; dbi++)
+ env->me_dbxs[dbi].md_dirty = 0;
/* The writer mutex was locked in mdb_txn_begin. */
UNLOCK_MUTEX_W(env);
}
if (txn == NULL)
return;
- DPRINTF("reset txn %lu%c %p on mdbenv %p, root page %lu",
- txn->mt_txnid, (txn->mt_flags & MDB_TXN_RDONLY) ? 'r' : 'w', txn,
- (void *)txn->mt_env, txn->mt_dbs[MAIN_DBI].md_root);
+ DPRINTF("reset txn %zu%c %p on mdbenv %p, root page %zu",
+ txn->mt_txnid, (txn->mt_flags & MDB_TXN_RDONLY) ? 'r' : 'w',
+ (void *) txn, (void *)txn->mt_env, txn->mt_dbs[MAIN_DBI].md_root);
mdb_txn_reset0(txn);
}
if (txn == NULL)
return;
- DPRINTF("abort txn %lu%c %p on mdbenv %p, root page %lu",
- txn->mt_txnid, (txn->mt_flags & MDB_TXN_RDONLY) ? 'r' : 'w', txn,
- (void *)txn->mt_env, txn->mt_dbs[MAIN_DBI].md_root);
+ DPRINTF("abort txn %zu%c %p on mdbenv %p, root page %zu",
+ txn->mt_txnid, (txn->mt_flags & MDB_TXN_RDONLY) ? 'r' : 'w',
+ (void *)txn, (void *)txn->mt_env, txn->mt_dbs[MAIN_DBI].md_root);
mdb_txn_reset0(txn);
free(txn);
if (!txn->mt_u.dirty_list[0].mid)
goto done;
- DPRINTF("committing txn %lu %p on mdbenv %p, root page %lu",
- txn->mt_txnid, txn, (void *)env, txn->mt_dbs[MAIN_DBI].md_root);
+ DPRINTF("committing txn %zu %p on mdbenv %p, root page %zu",
+ txn->mt_txnid, (void *)txn, (void *)env, txn->mt_dbs[MAIN_DBI].md_root);
- mdb_cursor_init(&mc, txn, FREE_DBI);
+ mdb_cursor_init(&mc, txn, FREE_DBI, NULL);
/* should only be one record now */
if (env->me_pghead) {
/* make sure first page of freeDB is touched and on freelist */
- mdb_search_page(&mc, NULL, 1);
+ mdb_page_search(&mc, NULL, 1);
}
/* save to free list */
if (!MDB_IDL_IS_ZERO(txn->mt_free_pgs)) {
MDB_val key, data;
- ULONG i;
+ pgno_t i;
/* make sure last page of freeDB is touched and on freelist */
key.mv_size = MAXKEYSIZE+1;
key.mv_data = NULL;
- mdb_search_page(&mc, &key, 1);
+ mdb_page_search(&mc, &key, 1);
mdb_midl_sort(txn->mt_free_pgs);
#if DEBUG > 1
{
unsigned int i;
- ULONG *idl = txn->mt_free_pgs;
- DPRINTF("IDL write txn %lu root %lu num %lu",
+ ID *idl = txn->mt_free_pgs;
+ DPRINTF("IDL write txn %zu root %zu num %zu",
txn->mt_txnid, txn->mt_dbs[FREE_DBI].md_root, idl[0]);
for (i=0; i<idl[0]; i++) {
- DPRINTF("IDL %lu", idl[i+1]);
+ DPRINTF("IDL %zu", idl[i+1]);
}
}
#endif
/* write to last page of freeDB */
key.mv_size = sizeof(pgno_t);
- key.mv_data = (char *)&txn->mt_txnid;
+ key.mv_data = &txn->mt_txnid;
data.mv_data = txn->mt_free_pgs;
/* The free list can still grow during this call,
* despite the pre-emptive touches above. So check
mop = env->me_pghead;
key.mv_size = sizeof(pgno_t);
- key.mv_data = (char *)&mop->mo_txnid;
+ key.mv_data = &mop->mo_txnid;
data.mv_size = MDB_IDL_SIZEOF(mop->mo_pages);
data.mv_data = mop->mo_pages;
mdb_cursor_put(&mc, &key, &data, 0);
* touched so this is all in-place and cannot fail.
*/
{
+ MDB_dbi i;
MDB_val data;
data.mv_size = sizeof(MDB_db);
- mdb_cursor_init(&mc, txn, MAIN_DBI);
+ mdb_cursor_init(&mc, txn, MAIN_DBI, NULL);
for (i = 2; i < txn->mt_numdbs; i++) {
if (txn->mt_dbxs[i].md_dirty) {
data.mv_data = &txn->mt_dbs[i];
for (; i<=txn->mt_u.dirty_list[0].mid; i++) {
size_t wsize;
dp = txn->mt_u.dirty_list[i].mptr;
- DPRINTF("committing page %lu", dp->mp_pgno);
+ DPRINTF("committing page %zu", dp->mp_pgno);
size = dp->mp_pgno * env->me_psize;
ov.Offset = size & 0xffffffff;
ov.OffsetHigh = size >> 16;
lseek(env->me_fd, dp->mp_pgno * env->me_psize, SEEK_SET);
next = dp->mp_pgno;
}
- DPRINTF("committing page %lu", dp->mp_pgno);
+ DPRINTF("committing page %zu", dp->mp_pgno);
iov[n].iov_len = env->me_psize;
if (IS_OVERFLOW(dp)) iov[n].iov_len *= dp->mp_pages;
iov[n].iov_base = dp;
{
int toggle = !env->me_db_toggle;
MDB_db *ip, *jp;
+ MDB_dbi i;
ip = &env->me_dbs[toggle][2];
jp = &txn->mt_dbs[2];
p = (MDB_page *)page;
if (!F_ISSET(p->mp_flags, P_META)) {
- DPRINTF("page %lu not a meta page", p->mp_pgno);
+ DPRINTF("page %zu not a meta page", p->mp_pgno);
return EINVAL;
}
DPUTS("writing new meta page");
- GetPageSize(psize);
+ GET_PAGESIZE(psize);
meta->mm_magic = MDB_MAGIC;
meta->mm_version = MDB_VERSION;
assert(txn->mt_env != NULL);
toggle = !txn->mt_toggle;
- DPRINTF("writing meta page %d for root page %lu",
+ DPRINTF("writing meta page %d for root page %zu",
toggle, txn->mt_dbs[MAIN_DBI].md_root);
env = txn->mt_env;
}
int
-mdb_env_set_maxdbs(MDB_env *env, int dbs)
+mdb_env_set_maxdbs(MDB_env *env, MDB_dbi dbs)
{
if (env->me_map)
return EINVAL;
}
int
-mdb_env_set_maxreaders(MDB_env *env, int readers)
+mdb_env_set_maxreaders(MDB_env *env, unsigned int readers)
{
- if (env->me_map)
+ if (env->me_map || readers < 1)
return EINVAL;
env->me_maxreaders = readers;
return MDB_SUCCESS;
}
int
-mdb_env_get_maxreaders(MDB_env *env, int *readers)
+mdb_env_get_maxreaders(MDB_env *env, unsigned int *readers)
{
if (!env || !readers)
return EINVAL;
DPRINTF("opened database version %u, pagesize %u",
env->me_metas[toggle]->mm_version, env->me_psize);
DPRINTF("depth: %u", env->me_metas[toggle]->mm_dbs[MAIN_DBI].md_depth);
- DPRINTF("entries: %lu", env->me_metas[toggle]->mm_dbs[MAIN_DBI].md_entries);
- DPRINTF("branch pages: %lu", env->me_metas[toggle]->mm_dbs[MAIN_DBI].md_branch_pages);
- DPRINTF("leaf pages: %lu", env->me_metas[toggle]->mm_dbs[MAIN_DBI].md_leaf_pages);
- DPRINTF("overflow pages: %lu", env->me_metas[toggle]->mm_dbs[MAIN_DBI].md_overflow_pages);
- DPRINTF("root: %lu", env->me_metas[toggle]->mm_dbs[MAIN_DBI].md_root);
+ DPRINTF("entries: %zu", env->me_metas[toggle]->mm_dbs[MAIN_DBI].md_entries);
+ DPRINTF("branch pages: %zu", env->me_metas[toggle]->mm_dbs[MAIN_DBI].md_branch_pages);
+ DPRINTF("leaf pages: %zu", env->me_metas[toggle]->mm_dbs[MAIN_DBI].md_leaf_pages);
+ DPRINTF("overflow pages: %zu", env->me_metas[toggle]->mm_dbs[MAIN_DBI].md_overflow_pages);
+ DPRINTF("root: %zu", env->me_metas[toggle]->mm_dbs[MAIN_DBI].md_root);
return MDB_SUCCESS;
}
#ifndef _WIN32
-/* Windows doesn't support destructor callbacks for thread-specific storage */
+/** Release a reader thread's slot in the reader lock table.
+ * This function is called automatically when a thread exits.
+ * Windows doesn't support destructor callbacks for thread-specific storage,
+ * so this function is not compiled there.
+ * @param[in] ptr This points to the slot in the reader lock table.
+ */
static void
mdb_env_reader_dest(void *ptr)
{
}
#endif
-/* downgrade the exclusive lock on the region back to shared */
+/** Downgrade the exclusive lock on the region back to shared */
static void
mdb_env_share_locks(MDB_env *env)
{
}
#endif
}
+#if defined(_WIN32) || defined(__APPLE__)
+/*
+ * hash_64 - 64 bit Fowler/Noll/Vo-0 FNV-1a hash code
+ *
+ * @(#) $Revision: 5.1 $
+ * @(#) $Id: hash_64a.c,v 5.1 2009/06/30 09:01:38 chongo Exp $
+ * @(#) $Source: /usr/local/src/cmd/fnv/RCS/hash_64a.c,v $
+ *
+ * http://www.isthe.com/chongo/tech/comp/fnv/index.html
+ *
+ ***
+ *
+ * Please do not copyright this code. This code is in the public domain.
+ *
+ * LANDON CURT NOLL DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE,
+ * INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO
+ * EVENT SHALL LANDON CURT NOLL BE LIABLE FOR ANY SPECIAL, INDIRECT OR
+ * CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF
+ * USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR
+ * OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR
+ * PERFORMANCE OF THIS SOFTWARE.
+ *
+ * By:
+ * chongo <Landon Curt Noll> /\oo/\
+ * http://www.isthe.com/chongo/
+ *
+ * Share and Enjoy! :-)
+ */
+
+typedef unsigned long long mdb_hash_t;
+#define MDB_HASH_INIT ((mdb_hash_t)0xcbf29ce484222325ULL)
+
+/** perform a 64 bit Fowler/Noll/Vo FNV-1a hash on a buffer
+ * @param[in] str string to hash
+ * @param[in] hval initial value for hash
+ * @return 64 bit hash
+ *
+ * NOTE: To use the recommended 64 bit FNV-1a hash, use MDB_HASH_INIT as the
+ * hval arg on the first call.
+ */
+static inline mdb_hash_t
+mdb_hash_str(char *str, mdb_hash_t hval)
+{
+ unsigned char *s = (unsigned char *)str; /* unsigned string */
+ /*
+ * FNV-1a hash each octet of the string
+ */
+ while (*s) {
+ /* xor the bottom with the current octet */
+ hval ^= (mdb_hash_t)*s++;
+
+ /* multiply by the 64 bit FNV magic prime mod 2^64 */
+ hval += (hval << 1) + (hval << 4) + (hval << 5) +
+ (hval << 7) + (hval << 8) + (hval << 40);
+ }
+ /* return our new hash value */
+ return hval;
+}
+
+/** Hash the string and output the hash in hex.
+ * @param[in] str string to hash
+ * @param[out] hexbuf an array of 17 chars to hold the hash
+ */
+static void
+mdb_hash_hex(char *str, char *hexbuf)
+{
+ int i;
+ mdb_hash_t h = mdb_hash_str(str, MDB_HASH_INIT);
+ for (i=0; i<8; i++) {
+ hexbuf += sprintf(hexbuf, "%02x", (unsigned int)h & 0xff);
+ h >>= 8;
+ }
+}
+#endif
+/** Open and/or initialize the lock region for the environment.
+ * @param[in] env The MDB environment.
+ * @param[in] lpath The pathname of the file used for the lock region.
+ * @param[in] mode The Unix permissions for the file, if we create it.
+ * @param[out] excl Set to true if we got an exclusive lock on the region.
+ * @return 0 on success, non-zero on failure.
+ */
static int
mdb_env_setup_locks(MDB_env *env, char *lpath, int mode, int *excl)
{
#endif
if (*excl) {
#ifdef _WIN32
- char *ptr;
+ char hexbuf[17];
if (!mdb_sec_inited) {
InitializeSecurityDescriptor(&mdb_null_sd,
SECURITY_DESCRIPTOR_REVISION);
mdb_all_sa.lpSecurityDescriptor = &mdb_null_sd;
mdb_sec_inited = 1;
}
- /* FIXME: only using up to 20 characters of the env path here,
- * probably not enough to assure uniqueness...
- */
- sprintf(env->me_txns->mti_rmname, "Global\\MDBr%.20s", lpath);
- ptr = env->me_txns->mti_rmname + sizeof("Global\\MDBr");
- while ((ptr = strchr(ptr, '\\')))
- *ptr++ = '/';
+ mdb_hash_hex(lpath, hexbuf);
+ sprintf(env->me_txns->mti_rmname, "Global\\MDBr%s", hexbuf);
env->me_rmutex = CreateMutex(&mdb_all_sa, FALSE, env->me_txns->mti_rmname);
if (!env->me_rmutex) {
rc = ErrCode();
goto fail;
}
- sprintf(env->me_txns->mti_rmname, "Global\\MDBw%.20s", lpath);
- ptr = env->me_txns->mti_rmname + sizeof("Global\\MDBw");
- while ((ptr = strchr(ptr, '\\')))
- *ptr++ = '/';
- env->me_wmutex = CreateMutex(&mdb_all_sa, FALSE, env->me_txns->mti_rmname);
+ sprintf(env->me_txns->mti_wmname, "Global\\MDBw%s", hexbuf);
+ env->me_wmutex = CreateMutex(&mdb_all_sa, FALSE, env->me_txns->mti_wmname);
if (!env->me_wmutex) {
rc = ErrCode();
goto fail;
}
-#else
+#else /* _WIN32 */
+#ifdef __APPLE__
+ char hexbuf[17];
+ mdb_hash_hex(lpath, hexbuf);
+ sprintf(env->me_txns->mti_rmname, "MDBr%s", hexbuf);
+ if (sem_unlink(env->me_txns->mti_rmname)) {
+ rc = ErrCode();
+ if (rc != ENOENT && rc != EINVAL)
+ goto fail;
+ }
+ env->me_rmutex = sem_open(env->me_txns->mti_rmname, O_CREAT, mode, 1);
+ if (!env->me_rmutex) {
+ rc = ErrCode();
+ goto fail;
+ }
+ sprintf(env->me_txns->mti_wmname, "MDBw%s", hexbuf);
+ if (sem_unlink(env->me_txns->mti_wmname)) {
+ rc = ErrCode();
+ if (rc != ENOENT && rc != EINVAL)
+ goto fail;
+ }
+ env->me_wmutex = sem_open(env->me_txns->mti_wmname, O_CREAT, mode, 1);
+ if (!env->me_wmutex) {
+ rc = ErrCode();
+ goto fail;
+ }
+#else /* __APPLE__ */
pthread_mutexattr_t mattr;
pthread_mutexattr_init(&mattr);
}
pthread_mutex_init(&env->me_txns->mti_mutex, &mattr);
pthread_mutex_init(&env->me_txns->mti_wmutex, &mattr);
-#endif
+#endif /* __APPLE__ */
+#endif /* _WIN32 */
env->me_txns->mti_version = MDB_VERSION;
env->me_txns->mti_magic = MDB_MAGIC;
env->me_txns->mti_txnid = 0;
rc = ErrCode();
goto fail;
}
+#endif
+#ifdef __APPLE__
+ env->me_rmutex = sem_open(env->me_txns->mti_rmname, 0);
+ if (!env->me_rmutex) {
+ rc = ErrCode();
+ goto fail;
+ }
+ env->me_wmutex = sem_open(env->me_txns->mti_wmname, 0);
+ if (!env->me_wmutex) {
+ rc = ErrCode();
+ goto fail;
+ }
#endif
}
return MDB_SUCCESS;
free(env);
}
-/* only for aligned ints */
+/** Compare two items pointing at aligned size_t's */
static int
-intcmp(const MDB_val *a, const MDB_val *b)
+mdb_cmp_long(const MDB_val *a, const MDB_val *b)
{
- if (a->mv_size == sizeof(long))
- {
- unsigned long *la, *lb;
- la = a->mv_data;
- lb = b->mv_data;
- return *la - *lb;
- } else {
- unsigned int *ia, *ib;
- ia = a->mv_data;
- ib = b->mv_data;
- return *ia - *ib;
- }
+ return (*(size_t *)a->mv_data < *(size_t *)b->mv_data) ? -1 :
+ *(size_t *)a->mv_data > *(size_t *)b->mv_data;
}
-/* ints must always be the same size */
+/** Compare two items pointing at aligned int's */
static int
-cintcmp(const MDB_val *a, const MDB_val *b)
+mdb_cmp_int(const MDB_val *a, const MDB_val *b)
{
-#if __BYTE_ORDER == __LITTLE_ENDIAN
+ return (*(unsigned int *)a->mv_data < *(unsigned int *)b->mv_data) ? -1 :
+ *(unsigned int *)a->mv_data > *(unsigned int *)b->mv_data;
+}
+
+/** Compare two items pointing at ints of unknown alignment.
+ * Nodes and keys are guaranteed to be 2-byte aligned.
+ */
+static int
+mdb_cmp_cint(const MDB_val *a, const MDB_val *b)
+{
+#if BYTE_ORDER == LITTLE_ENDIAN
unsigned short *u, *c;
int x;
- u = a->mv_data + a->mv_size;
- c = b->mv_data + a->mv_size;
+ u = (unsigned short *) ((char *) a->mv_data + a->mv_size);
+ c = (unsigned short *) ((char *) b->mv_data + a->mv_size);
do {
x = *--u - *--c;
} while(!x && u > (unsigned short *)a->mv_data);
#endif
}
+/** Compare two items lexically */
static int
-memncmp(const MDB_val *a, const MDB_val *b)
+mdb_cmp_memn(const MDB_val *a, const MDB_val *b)
{
- int diff, len_diff;
+ int diff;
+ ssize_t len_diff;
unsigned int len;
len = a->mv_size;
- len_diff = a->mv_size - b->mv_size;
- if (len_diff > 0)
+ len_diff = (ssize_t) a->mv_size - (ssize_t) b->mv_size;
+ if (len_diff > 0) {
len = b->mv_size;
+ len_diff = 1;
+ }
+
diff = memcmp(a->mv_data, b->mv_data, len);
- return diff ? diff : len_diff;
+ return diff ? diff : len_diff<0 ? -1 : len_diff;
}
+/** Compare two items in reverse byte order */
static int
-memnrcmp(const MDB_val *a, const MDB_val *b)
+mdb_cmp_memnr(const MDB_val *a, const MDB_val *b)
{
const unsigned char *p1, *p2, *p1_lim;
- int diff, len_diff;
+ ssize_t len_diff;
+ int diff;
- if (b->mv_size == 0)
- return a->mv_size != 0;
- if (a->mv_size == 0)
- return -1;
+ p1_lim = (const unsigned char *)a->mv_data;
+ p1 = (const unsigned char *)a->mv_data + a->mv_size;
+ p2 = (const unsigned char *)b->mv_data + b->mv_size;
- p1 = (const unsigned char *)a->mv_data + a->mv_size - 1;
- p2 = (const unsigned char *)b->mv_data + b->mv_size - 1;
-
- len_diff = a->mv_size - b->mv_size;
- if (len_diff < 0)
- p1_lim = p1 - a->mv_size;
- else
- p1_lim = p1 - b->mv_size;
+ len_diff = (ssize_t) a->mv_size - (ssize_t) b->mv_size;
+ if (len_diff > 0) {
+ p1_lim += len_diff;
+ len_diff = 1;
+ }
while (p1 > p1_lim) {
- diff = *p1 - *p2;
+ diff = *--p1 - *--p2;
if (diff)
return diff;
- p1--;
- p2--;
}
- return len_diff;
+ return len_diff<0 ? -1 : len_diff;
}
-/* Search for key within a leaf page, using binary search.
+/** 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).
- * If kip is non-null, stores the index of the found entry in *kip.
+ * Updates the cursor index with the index of the found entry.
* If no entry larger or equal to the key is found, returns NULL.
*/
static MDB_node *
-mdb_search_node(MDB_cursor *mc, MDB_val *key, int *exactp)
+mdb_node_search(MDB_cursor *mc, MDB_val *key, int *exactp)
{
unsigned int i = 0, nkeys;
int low, high;
nkeys = NUMKEYS(mp);
- DPRINTF("searching %u keys in %s page %lu",
+ DPRINTF("searching %u keys in %s page %zu",
nkeys, IS_LEAF(mp) ? "leaf" : "branch",
mp->mp_pgno);
low = IS_LEAF(mp) ? 0 : 1;
high = nkeys - 1;
cmp = mc->mc_dbx->md_cmp;
+
+ /* Branch pages have no data, so if using integer keys,
+ * alignment is guaranteed. Use faster mdb_cmp_int.
+ */
+ if (cmp == mdb_cmp_cint && IS_BRANCH(mp)) {
+ if (NODEPTR(mp, 1)->mn_ksize == sizeof(size_t))
+ cmp = mdb_cmp_long;
+ else
+ cmp = mdb_cmp_int;
+ }
+
if (IS_LEAF2(mp)) {
nodekey.mv_size = mc->mc_db->md_pad;
node = NODEPTR(mp, 0); /* fake */
- }
- while (low <= high) {
- i = (low + high) >> 1;
-
- if (IS_LEAF2(mp)) {
+ while (low <= high) {
+ i = (low + high) >> 1;
nodekey.mv_data = LEAF2KEY(mp, i, nodekey.mv_size);
- } else {
- node = NODEPTR(mp, i);
-
- nodekey.mv_size = node->mn_ksize;
- nodekey.mv_data = NODEKEY(node);
+ rc = cmp(key, &nodekey);
+ DPRINTF("found leaf index %u [%s], rc = %i",
+ i, DKEY(&nodekey), rc);
+ if (rc == 0)
+ break;
+ if (rc > 0)
+ low = i + 1;
+ else
+ high = i - 1;
}
+ } else {
+ while (low <= high) {
+ i = (low + high) >> 1;
- rc = cmp(key, &nodekey);
+ node = NODEPTR(mp, i);
+ nodekey.mv_size = NODEKSZ(node);
+ nodekey.mv_data = NODEKEY(node);
+ rc = cmp(key, &nodekey);
#if DEBUG
- if (IS_LEAF(mp))
- DPRINTF("found leaf index %u [%s], rc = %i",
- i, DKEY(&nodekey), rc);
- else
- DPRINTF("found branch index %u [%s -> %lu], rc = %i",
- i, DKEY(&nodekey), NODEPGNO(node), rc);
+ if (IS_LEAF(mp))
+ DPRINTF("found leaf index %u [%s], rc = %i",
+ i, DKEY(&nodekey), rc);
+ else
+ DPRINTF("found branch index %u [%s -> %zu], rc = %i",
+ i, DKEY(&nodekey), NODEPGNO(node), rc);
#endif
-
- if (rc == 0)
- break;
- if (rc > 0)
- low = i + 1;
- else
- high = i - 1;
+ if (rc == 0)
+ break;
+ if (rc > 0)
+ low = i + 1;
+ else
+ high = i - 1;
+ }
}
if (rc > 0) { /* Found entry is less than the key. */
return node;
}
+/** Pop a page off the top of the cursor's stack. */
static void
-cursor_pop_page(MDB_cursor *mc)
+mdb_cursor_pop(MDB_cursor *mc)
{
MDB_page *top;
if (mc->mc_snum)
mc->mc_top--;
- DPRINTF("popped page %lu off db %u cursor %p", top->mp_pgno,
+ DPRINTF("popped page %zu off db %u cursor %p", top->mp_pgno,
mc->mc_dbi, (void *) mc);
}
}
+/** Push a page onto the top of the cursor's stack. */
static int
-cursor_push_page(MDB_cursor *mc, MDB_page *mp)
+mdb_cursor_push(MDB_cursor *mc, MDB_page *mp)
{
- DPRINTF("pushing page %lu on db %u cursor %p", mp->mp_pgno,
+ DPRINTF("pushing page %zu on db %u cursor %p", mp->mp_pgno,
mc->mc_dbi, (void *) mc);
if (mc->mc_snum >= CURSOR_STACK) {
return MDB_SUCCESS;
}
+/** Find the address of the page corresponding to a given page number.
+ * @param[in] txn the transaction for this access.
+ * @param[in] pgno the page number for the page to retrieve.
+ * @param[out] ret address of a pointer where the page's address will be stored.
+ * @return 0 on success, non-zero on failure.
+ */
static int
-mdb_get_page(MDB_txn *txn, pgno_t pgno, MDB_page **ret)
+mdb_page_get(MDB_txn *txn, pgno_t pgno, MDB_page **ret)
{
MDB_page *p = NULL;
}
*ret = p;
if (!p) {
- DPRINTF("page %lu not found", pgno);
+ DPRINTF("page %zu not found", pgno);
assert(p != NULL);
}
return (p != NULL) ? MDB_SUCCESS : MDB_PAGE_NOTFOUND;
}
+/** Search for the page a given key should be in.
+ * Pushes parent pages on the cursor stack. This function continues a
+ * search on a cursor that has already been initialized. (Usually by
+ * #mdb_page_search() but also by #mdb_node_move().)
+ * @param[in,out] mc the cursor for this operation.
+ * @param[in] key the key to search for. If NULL, search for the lowest
+ * page. (This is used by #mdb_cursor_first().)
+ * @param[in] modify If true, visited pages are updated with new page numbers.
+ * @return 0 on success, non-zero on failure.
+ */
static int
-mdb_search_page_root(MDB_cursor *mc, MDB_val *key, int modify)
+mdb_page_search_root(MDB_cursor *mc, MDB_val *key, int modify)
{
MDB_page *mp = mc->mc_pg[mc->mc_top];
DKBUF;
while (IS_BRANCH(mp)) {
MDB_node *node;
+ indx_t i;
- DPRINTF("branch page %lu has %u keys", mp->mp_pgno, NUMKEYS(mp));
+ DPRINTF("branch page %zu has %u keys", mp->mp_pgno, NUMKEYS(mp));
assert(NUMKEYS(mp) > 1);
- DPRINTF("found index 0 to page %lu", NODEPGNO(NODEPTR(mp, 0)));
+ DPRINTF("found index 0 to page %zu", NODEPGNO(NODEPTR(mp, 0)));
if (key == NULL) /* Initialize cursor to first page. */
- mc->mc_ki[mc->mc_top] = 0;
+ i = 0;
else if (key->mv_size > MAXKEYSIZE && key->mv_data == NULL) {
/* cursor to last page */
- mc->mc_ki[mc->mc_top] = NUMKEYS(mp)-1;
+ i = NUMKEYS(mp)-1;
} else {
int exact;
- node = mdb_search_node(mc, key, &exact);
+ node = mdb_node_search(mc, key, &exact);
if (node == NULL)
- mc->mc_ki[mc->mc_top] = NUMKEYS(mp) - 1;
- else if (!exact) {
- assert(mc->mc_ki[mc->mc_top] > 0);
- mc->mc_ki[mc->mc_top]--;
+ i = NUMKEYS(mp) - 1;
+ else {
+ i = mc->mc_ki[mc->mc_top];
+ if (!exact) {
+ assert(i > 0);
+ i--;
+ }
}
}
if (key)
DPRINTF("following index %u for key [%s]",
- mc->mc_ki[mc->mc_top], DKEY(key));
- assert(mc->mc_ki[mc->mc_top] < NUMKEYS(mp));
- node = NODEPTR(mp, mc->mc_ki[mc->mc_top]);
+ i, DKEY(key));
+ assert(i < NUMKEYS(mp));
+ node = NODEPTR(mp, i);
- if ((rc = mdb_get_page(mc->mc_txn, NODEPGNO(node), &mp)))
+ if ((rc = mdb_page_get(mc->mc_txn, NODEPGNO(node), &mp)))
return rc;
- if ((rc = cursor_push_page(mc, mp)))
+ mc->mc_ki[mc->mc_top] = i;
+ if ((rc = mdb_cursor_push(mc, mp)))
return rc;
if (modify) {
- if ((rc = mdb_touch(mc)) != 0)
+ if ((rc = mdb_page_touch(mc)) != 0)
return rc;
mp = mc->mc_pg[mc->mc_top];
}
return MDB_CORRUPTED;
}
- DPRINTF("found leaf page %lu for key [%s]", mp->mp_pgno,
+ DPRINTF("found leaf page %zu for key [%s]", mp->mp_pgno,
key ? DKEY(key) : NULL);
return MDB_SUCCESS;
}
-/* Search for the page a given key should be in.
- * Pushes parent pages on the cursor stack.
- * If key is NULL, search for the lowest page (used by mdb_cursor_first).
- * If modify is true, visited pages are updated with new page numbers.
+/** Search for the page a given key should be in.
+ * Pushes parent pages on the cursor stack. This function just sets up
+ * the search; it finds the root page for \b mc's database and sets this
+ * as the root of the cursor's stack. Then #mdb_page_search_root() is
+ * called to complete the search.
+ * @param[in,out] mc the cursor for this operation.
+ * @param[in] key the key to search for. If NULL, search for the lowest
+ * page. (This is used by #mdb_cursor_first().)
+ * @param[in] modify If true, visited pages are updated with new page numbers.
+ * @return 0 on success, non-zero on failure.
*/
static int
-mdb_search_page(MDB_cursor *mc, MDB_val *key, int modify)
+mdb_page_search(MDB_cursor *mc, MDB_val *key, int modify)
{
int rc;
pgno_t root;
return MDB_NOTFOUND;
}
- if ((rc = mdb_get_page(mc->mc_txn, root, &mc->mc_pg[0])))
+ if ((rc = mdb_page_get(mc->mc_txn, root, &mc->mc_pg[0])))
return rc;
mc->mc_snum = 1;
mc->mc_top = 0;
- DPRINTF("db %u root page %lu has flags 0x%X",
+ DPRINTF("db %u root page %zu has flags 0x%X",
mc->mc_dbi, root, mc->mc_pg[0]->mp_flags);
if (modify) {
/* For sub-databases, update main root first */
if (mc->mc_dbi > MAIN_DBI && !mc->mc_dbx->md_dirty) {
MDB_cursor mc2;
- mdb_cursor_init(&mc2, mc->mc_txn, MAIN_DBI);
- rc = mdb_search_page(&mc2, &mc->mc_dbx->md_name, 1);
+ mdb_cursor_init(&mc2, mc->mc_txn, MAIN_DBI, NULL);
+ rc = mdb_page_search(&mc2, &mc->mc_dbx->md_name, 1);
if (rc)
return rc;
mc->mc_dbx->md_dirty = 1;
}
if (!F_ISSET(mc->mc_pg[0]->mp_flags, P_DIRTY)) {
- if ((rc = mdb_touch(mc)))
+ if ((rc = mdb_page_touch(mc)))
return rc;
mc->mc_db->md_root = mc->mc_pg[0]->mp_pgno;
}
}
- return mdb_search_page_root(mc, key, modify);
+ return mdb_page_search_root(mc, key, modify);
}
+/** Return the data associated with a given node.
+ * @param[in] txn The transaction for this operation.
+ * @param[in] leaf The node being read.
+ * @param[out] data Updated to point to the node's data.
+ * @return 0 on success, non-zero on failure.
+ */
static int
-mdb_read_data(MDB_txn *txn, MDB_node *leaf, MDB_val *data)
+mdb_node_read(MDB_txn *txn, MDB_node *leaf, MDB_val *data)
{
- MDB_page *omp; /* overflow mpage */
+ MDB_page *omp; /* overflow page */
pgno_t pgno;
int rc;
*/
data->mv_size = NODEDSZ(leaf);
memcpy(&pgno, NODEDATA(leaf), sizeof(pgno));
- if ((rc = mdb_get_page(txn, pgno, &omp))) {
- DPRINTF("read overflow page %lu failed", pgno);
+ if ((rc = mdb_page_get(txn, pgno, &omp))) {
+ DPRINTF("read overflow page %zu failed", pgno);
return rc;
}
data->mv_data = METADATA(omp);
return EINVAL;
}
- mdb_cursor_init(&mc, txn, dbi);
- if (txn->mt_dbs[dbi].md_flags & MDB_DUPSORT) {
- mc.mc_xcursor = &mx;
- mdb_xcursor_init0(&mc);
- } else {
- mc.mc_xcursor = NULL;
- }
+ mdb_cursor_init(&mc, txn, dbi, &mx);
return mdb_cursor_set(&mc, key, data, MDB_SET, &exact);
}
+/** Find a sibling for a page.
+ * Replaces the page at the top of the cursor's stack with the
+ * specified sibling, if one exists.
+ * @param[in] mc The cursor for this operation.
+ * @param[in] move_right Non-zero if the right sibling is requested,
+ * otherwise the left sibling.
+ * @return 0 on success, non-zero on failure.
+ */
static int
-mdb_sibling(MDB_cursor *mc, int move_right)
+mdb_cursor_sibling(MDB_cursor *mc, int move_right)
{
int rc;
MDB_node *indx;
return MDB_NOTFOUND; /* root has no siblings */
}
- cursor_pop_page(mc);
- DPRINTF("parent page is page %lu, index %u",
+ mdb_cursor_pop(mc);
+ DPRINTF("parent page is page %zu, index %u",
mc->mc_pg[mc->mc_top]->mp_pgno, mc->mc_ki[mc->mc_top]);
if (move_right ? (mc->mc_ki[mc->mc_top] + 1u >= NUMKEYS(mc->mc_pg[mc->mc_top]))
: (mc->mc_ki[mc->mc_top] == 0)) {
DPRINTF("no more keys left, moving to %s sibling",
move_right ? "right" : "left");
- if ((rc = mdb_sibling(mc, move_right)) != MDB_SUCCESS)
+ if ((rc = mdb_cursor_sibling(mc, move_right)) != MDB_SUCCESS)
return rc;
} else {
if (move_right)
assert(IS_BRANCH(mc->mc_pg[mc->mc_top]));
indx = NODEPTR(mc->mc_pg[mc->mc_top], mc->mc_ki[mc->mc_top]);
- if ((rc = mdb_get_page(mc->mc_txn, NODEPGNO(indx), &mp)))
+ if ((rc = mdb_page_get(mc->mc_txn, NODEPGNO(indx), &mp)))
return rc;;
- cursor_push_page(mc, mp);
+ mdb_cursor_push(mc, mp);
return MDB_SUCCESS;
}
+/** Move the cursor to the next data item. */
static int
mdb_cursor_next(MDB_cursor *mc, MDB_val *key, MDB_val *data, MDB_cursor_op op)
{
}
}
- DPRINTF("cursor_next: top page is %lu in cursor %p", mp->mp_pgno, (void *) mc);
+ DPRINTF("cursor_next: top page is %zu in cursor %p", mp->mp_pgno, (void *) mc);
if (mc->mc_ki[mc->mc_top] + 1u >= NUMKEYS(mp)) {
DPUTS("=====> move to next sibling page");
- if (mdb_sibling(mc, 1) != MDB_SUCCESS) {
+ if (mdb_cursor_sibling(mc, 1) != MDB_SUCCESS) {
mc->mc_flags |= C_EOF;
return MDB_NOTFOUND;
}
mp = mc->mc_pg[mc->mc_top];
- DPRINTF("next page is %lu, key index %u", mp->mp_pgno, mc->mc_ki[mc->mc_top]);
+ DPRINTF("next page is %zu, key index %u", mp->mp_pgno, mc->mc_ki[mc->mc_top]);
} else
mc->mc_ki[mc->mc_top]++;
- DPRINTF("==> cursor points to page %lu with %u keys, key index %u",
+ DPRINTF("==> cursor points to page %zu with %u keys, key index %u",
mp->mp_pgno, NUMKEYS(mp), mc->mc_ki[mc->mc_top]);
if (IS_LEAF2(mp)) {
mdb_xcursor_init1(mc, leaf);
}
if (data) {
- if ((rc = mdb_read_data(mc->mc_txn, leaf, data) != MDB_SUCCESS))
+ if ((rc = mdb_node_read(mc->mc_txn, leaf, data) != MDB_SUCCESS))
return rc;
if (F_ISSET(leaf->mn_flags, F_DUPDATA)) {
return MDB_SUCCESS;
}
+/** Move the cursor to the previous data item. */
static int
mdb_cursor_prev(MDB_cursor *mc, MDB_val *key, MDB_val *data, MDB_cursor_op op)
{
}
}
- DPRINTF("cursor_prev: top page is %lu in cursor %p", mp->mp_pgno, (void *) mc);
+ DPRINTF("cursor_prev: top page is %zu in cursor %p", mp->mp_pgno, (void *) mc);
if (mc->mc_ki[mc->mc_top] == 0) {
DPUTS("=====> move to prev sibling page");
- if (mdb_sibling(mc, 0) != MDB_SUCCESS) {
+ if (mdb_cursor_sibling(mc, 0) != MDB_SUCCESS) {
mc->mc_flags &= ~C_INITIALIZED;
return MDB_NOTFOUND;
}
mp = mc->mc_pg[mc->mc_top];
mc->mc_ki[mc->mc_top] = NUMKEYS(mp) - 1;
- DPRINTF("prev page is %lu, key index %u", mp->mp_pgno, mc->mc_ki[mc->mc_top]);
+ DPRINTF("prev page is %zu, key index %u", mp->mp_pgno, mc->mc_ki[mc->mc_top]);
} else
mc->mc_ki[mc->mc_top]--;
mc->mc_flags &= ~C_EOF;
- DPRINTF("==> cursor points to page %lu with %u keys, key index %u",
+ DPRINTF("==> cursor points to page %zu with %u keys, key index %u",
mp->mp_pgno, NUMKEYS(mp), mc->mc_ki[mc->mc_top]);
if (IS_LEAF2(mp)) {
mdb_xcursor_init1(mc, leaf);
}
if (data) {
- if ((rc = mdb_read_data(mc->mc_txn, leaf, data) != MDB_SUCCESS))
+ if ((rc = mdb_node_read(mc->mc_txn, leaf, data) != MDB_SUCCESS))
return rc;
if (F_ISSET(leaf->mn_flags, F_DUPDATA)) {
return MDB_SUCCESS;
}
+/** Set the cursor on a specific data item. */
static int
mdb_cursor_set(MDB_cursor *mc, MDB_val *key, MDB_val *data,
MDB_cursor_op op, int *exactp)
{
int rc;
+ MDB_page *mp;
MDB_node *leaf;
DKBUF;
if (mc->mc_flags & C_INITIALIZED) {
MDB_val nodekey;
- if (mc->mc_pg[mc->mc_top]->mp_flags & P_LEAF2) {
+ mp = mc->mc_pg[mc->mc_top];
+ if (mp->mp_flags & P_LEAF2) {
nodekey.mv_size = mc->mc_db->md_pad;
- nodekey.mv_data = LEAF2KEY(mc->mc_pg[mc->mc_top], 0, nodekey.mv_size);
+ nodekey.mv_data = LEAF2KEY(mp, 0, nodekey.mv_size);
} else {
- leaf = NODEPTR(mc->mc_pg[mc->mc_top], 0);
+ leaf = NODEPTR(mp, 0);
MDB_SET_KEY(leaf, &nodekey);
}
rc = mc->mc_dbx->md_cmp(key, &nodekey);
* was the one we wanted.
*/
mc->mc_ki[mc->mc_top] = 0;
-set1:
+ leaf = NODEPTR(mp, 0);
if (exactp)
*exactp = 1;
- leaf = NODEPTR(mc->mc_pg[mc->mc_top], mc->mc_ki[mc->mc_top]);
- goto set3;
+ goto set1;
}
if (rc > 0) {
unsigned int i;
- if (NUMKEYS(mc->mc_pg[mc->mc_top]) > 1) {
- if (mc->mc_pg[mc->mc_top]->mp_flags & P_LEAF2) {
- nodekey.mv_data = LEAF2KEY(mc->mc_pg[mc->mc_top],
- NUMKEYS(mc->mc_pg[mc->mc_top])-1, nodekey.mv_size);
+ unsigned int nkeys = NUMKEYS(mp);
+ if (nkeys > 1) {
+ if (mp->mp_flags & P_LEAF2) {
+ nodekey.mv_data = LEAF2KEY(mp,
+ nkeys-1, nodekey.mv_size);
} else {
- leaf = NODEPTR(mc->mc_pg[mc->mc_top], NUMKEYS(mc->mc_pg[mc->mc_top])-1);
+ leaf = NODEPTR(mp, nkeys-1);
MDB_SET_KEY(leaf, &nodekey);
}
rc = mc->mc_dbx->md_cmp(key, &nodekey);
if (rc == 0) {
/* last node was the one we wanted */
- mc->mc_ki[mc->mc_top] = NUMKEYS(mc->mc_pg[mc->mc_top])-1;
+ mc->mc_ki[mc->mc_top] = nkeys-1;
+ leaf = NODEPTR(mp, nkeys-1);
+ if (exactp)
+ *exactp = 1;
goto set1;
}
if (rc < 0) {
break;
if (i == mc->mc_top) {
/* There are no other pages */
- mc->mc_ki[mc->mc_top] = NUMKEYS(mc->mc_pg[mc->mc_top]);
+ mc->mc_ki[mc->mc_top] = nkeys;
return MDB_NOTFOUND;
}
}
}
- rc = mdb_search_page(mc, key, 0);
+ rc = mdb_page_search(mc, key, 0);
if (rc != MDB_SUCCESS)
return rc;
- assert(IS_LEAF(mc->mc_pg[mc->mc_top]));
+ mp = mc->mc_pg[mc->mc_top];
+ assert(IS_LEAF(mp));
set2:
- leaf = mdb_search_node(mc, key, exactp);
+ leaf = mdb_node_search(mc, key, exactp);
if (exactp != NULL && !*exactp) {
/* MDB_SET specified and not an exact match. */
return MDB_NOTFOUND;
if (leaf == NULL) {
DPUTS("===> inexact leaf not found, goto sibling");
- if ((rc = mdb_sibling(mc, 1)) != MDB_SUCCESS)
+ if ((rc = mdb_cursor_sibling(mc, 1)) != MDB_SUCCESS)
return rc; /* no entries matched */
- mc->mc_ki[mc->mc_top] = 0;
- assert(IS_LEAF(mc->mc_pg[mc->mc_top]));
- leaf = NODEPTR(mc->mc_pg[mc->mc_top], 0);
+ mp = mc->mc_pg[mc->mc_top];
+ assert(IS_LEAF(mp));
+ leaf = NODEPTR(mp, 0);
}
-set3:
+set1:
mc->mc_flags |= C_INITIALIZED;
mc->mc_flags &= ~C_EOF;
- if (IS_LEAF2(mc->mc_pg[mc->mc_top])) {
+ if (IS_LEAF2(mp)) {
key->mv_size = mc->mc_db->md_pad;
- key->mv_data = LEAF2KEY(mc->mc_pg[mc->mc_top], mc->mc_ki[mc->mc_top], key->mv_size);
+ key->mv_data = LEAF2KEY(mp, mc->mc_ki[mc->mc_top], key->mv_size);
return MDB_SUCCESS;
}
}
} else if (op == MDB_GET_BOTH || op == MDB_GET_BOTH_RANGE) {
MDB_val d2;
- if ((rc = mdb_read_data(mc->mc_txn, leaf, &d2)) != MDB_SUCCESS)
+ if ((rc = mdb_node_read(mc->mc_txn, leaf, &d2)) != MDB_SUCCESS)
return rc;
rc = mc->mc_dbx->md_dcmp(data, &d2);
if (rc) {
}
} else {
- if ((rc = mdb_read_data(mc->mc_txn, leaf, data)) != MDB_SUCCESS)
+ if (mc->mc_xcursor)
+ mc->mc_xcursor->mx_cursor.mc_flags = 0;
+ if ((rc = mdb_node_read(mc->mc_txn, leaf, data)) != MDB_SUCCESS)
return rc;
}
}
return rc;
}
+/** Move the cursor to the first item in the database. */
static int
mdb_cursor_first(MDB_cursor *mc, MDB_val *key, MDB_val *data)
{
int rc;
MDB_node *leaf;
- rc = mdb_search_page(mc, NULL, 0);
+ rc = mdb_page_search(mc, NULL, 0);
if (rc != MDB_SUCCESS)
return rc;
assert(IS_LEAF(mc->mc_pg[mc->mc_top]));
} else {
if (mc->mc_xcursor)
mc->mc_xcursor->mx_cursor.mc_flags = 0;
- if ((rc = mdb_read_data(mc->mc_txn, leaf, data)) != MDB_SUCCESS)
+ if ((rc = mdb_node_read(mc->mc_txn, leaf, data)) != MDB_SUCCESS)
return rc;
}
}
return MDB_SUCCESS;
}
+/** Move the cursor to the last item in the database. */
static int
mdb_cursor_last(MDB_cursor *mc, MDB_val *key, MDB_val *data)
{
lkey.mv_size = MAXKEYSIZE+1;
lkey.mv_data = NULL;
- rc = mdb_search_page(mc, &lkey, 0);
+ rc = mdb_page_search(mc, &lkey, 0);
if (rc != MDB_SUCCESS)
return rc;
assert(IS_LEAF(mc->mc_pg[mc->mc_top]));
if (rc)
return rc;
} else {
- if ((rc = mdb_read_data(mc->mc_txn, leaf, data)) != MDB_SUCCESS)
+ if (mc->mc_xcursor)
+ mc->mc_xcursor->mx_cursor.mc_flags = 0;
+ if ((rc = mdb_node_read(mc->mc_txn, leaf, data)) != MDB_SUCCESS)
return rc;
}
}
return rc;
}
+/** Touch all the pages in the cursor stack.
+ * Makes sure all the pages are writable, before attempting a write operation.
+ * @param[in] mc The cursor to operate on.
+ */
static int
mdb_cursor_touch(MDB_cursor *mc)
{
if (mc->mc_dbi > MAIN_DBI && !mc->mc_dbx->md_dirty) {
MDB_cursor mc2;
- mdb_cursor_init(&mc2, mc->mc_txn, MAIN_DBI);
- rc = mdb_search_page(&mc2, &mc->mc_dbx->md_name, 1);
+ mdb_cursor_init(&mc2, mc->mc_txn, MAIN_DBI, NULL);
+ rc = mdb_page_search(&mc2, &mc->mc_dbx->md_name, 1);
if (rc)
return rc;
mc->mc_dbx->md_dirty = 1;
}
for (mc->mc_top = 0; mc->mc_top < mc->mc_snum; mc->mc_top++) {
if (!F_ISSET(mc->mc_pg[mc->mc_top]->mp_flags, P_DIRTY)) {
- rc = mdb_touch(mc);
+ rc = mdb_page_touch(mc);
if (rc)
return rc;
if (!mc->mc_top) {
MDB_page *np;
/* new database, write a root leaf page */
DPUTS("allocating new root leaf page");
- if ((np = mdb_new_page(mc, P_LEAF, 1)) == NULL) {
+ if ((np = mdb_page_new(mc, P_LEAF, 1)) == NULL) {
return ENOMEM;
}
mc->mc_snum = 0;
- cursor_push_page(mc, np);
+ mdb_cursor_push(mc, np);
mc->mc_db->md_root = np->mp_pgno;
mc->mc_db->md_depth++;
mc->mc_dbx->md_dirty = 1;
memcpy(NODEDATA(leaf), &dummy, sizeof(dummy));
goto put_sub;
}
- mdb_del_node(mc->mc_pg[mc->mc_top], mc->mc_ki[mc->mc_top], 0);
+ mdb_node_del(mc->mc_pg[mc->mc_top], mc->mc_ki[mc->mc_top], 0);
do_sub = 1;
rdata = &xdata;
xdata.mv_size = sizeof(MDB_db);
memcpy(NODEDATA(leaf), data->mv_data, data->mv_size);
goto done;
}
- mdb_del_node(mc->mc_pg[mc->mc_top], mc->mc_ki[mc->mc_top], 0);
+ mdb_node_del(mc->mc_pg[mc->mc_top], mc->mc_ki[mc->mc_top], 0);
} else {
DPRINTF("inserting key at index %i", mc->mc_ki[mc->mc_top]);
}
new_sub:
nsize = IS_LEAF2(mc->mc_pg[mc->mc_top]) ? key->mv_size : mdb_leaf_size(mc->mc_txn->mt_env, key, rdata);
if (SIZELEFT(mc->mc_pg[mc->mc_top]) < nsize) {
- rc = mdb_split(mc, key, rdata, P_INVALID);
+ rc = mdb_page_split(mc, key, rdata, P_INVALID);
} else {
/* There is room already in this leaf page. */
- rc = mdb_add_node(mc, mc->mc_ki[mc->mc_top], key, rdata, 0, 0);
+ rc = mdb_node_add(mc, mc->mc_ki[mc->mc_top], key, rdata, 0, 0);
}
if (rc != MDB_SUCCESS)
}
/* add all the child DB's pages to the free list */
- rc = mdb_search_page(&mc->mc_xcursor->mx_cursor, NULL, 0);
+ rc = mdb_page_search(&mc->mc_xcursor->mx_cursor, NULL, 0);
if (rc == MDB_SUCCESS) {
MDB_node *ni;
MDB_cursor *mx;
mc->mc_db->md_entries -=
mx->mc_db->md_entries;
- cursor_pop_page(mx);
+ mdb_cursor_pop(mx);
while (mx->mc_snum > 1) {
for (i=0; i<NUMKEYS(mx->mc_pg[mx->mc_top]); i++) {
MDB_page *mp;
pgno_t pg;
ni = NODEPTR(mx->mc_pg[mx->mc_top], i);
pg = NODEPGNO(ni);
- if ((rc = mdb_get_page(mc->mc_txn, pg, &mp)))
+ if ((rc = mdb_page_get(mc->mc_txn, pg, &mp)))
return rc;
/* free it */
mdb_midl_append(mc->mc_txn->mt_free_pgs, pg);
}
- rc = mdb_sibling(mx, 1);
+ rc = mdb_cursor_sibling(mx, 1);
if (rc)
break;
}
}
}
- return mdb_del0(mc, leaf);
+ return mdb_cursor_del0(mc, leaf);
}
-/* Allocate a page and initialize it
+/** Allocate and initialize new pages for a database.
+ * @param[in] mc a cursor on the database being added to.
+ * @param[in] flags flags defining what type of page is being allocated.
+ * @param[in] num the number of pages to allocate. This is usually 1,
+ * unless allocating overflow pages for a large record.
+ * @return Address of a page, or NULL on failure.
*/
static MDB_page *
-mdb_new_page(MDB_cursor *mc, uint32_t flags, int num)
+mdb_page_new(MDB_cursor *mc, uint32_t flags, int num)
{
MDB_page *np;
- if ((np = mdb_alloc_page(mc, num)) == NULL)
+ if ((np = mdb_page_alloc(mc, num)) == NULL)
return NULL;
- DPRINTF("allocated new mpage %lu, page size %u",
+ DPRINTF("allocated new mpage %zu, page size %u",
np->mp_pgno, mc->mc_txn->mt_env->me_psize);
np->mp_flags = flags | P_DIRTY;
np->mp_lower = PAGEHDRSZ;
return np;
}
+/** 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.
+ * @param[in] env The environment handle.
+ * @param[in] key The key for the node.
+ * @param[in] data The data for the node.
+ * @return The number of bytes needed to store the node.
+ */
static size_t
mdb_leaf_size(MDB_env *env, MDB_val *key, MDB_val *data)
{
return sz + sizeof(indx_t);
}
+/** 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.
+ * @param[in] env The environment handle.
+ * @param[in] key The key for the node.
+ * @return The number of bytes needed to store the node.
+ */
static size_t
mdb_branch_size(MDB_env *env, MDB_val *key)
{
return sz + sizeof(indx_t);
}
+/** Add a node to the page pointed to by the cursor.
+ * @param[in] mc The cursor for this operation.
+ * @param[in] indx The index on the page where the new node should be added.
+ * @param[in] key The key for the new node.
+ * @param[in] data The data for the new node, if any.
+ * @param[in] pgno The page number, if adding a branch node.
+ * @param[in] flags Flags for the node.
+ * @return 0 on success, non-zero on failure. Possible errors are:
+ * <ul>
+ * <li>ENOMEM - failed to allocate overflow pages for the node.
+ * <li>ENOSPC - 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.
+ * </ul>
+ */
static int
-mdb_add_node(MDB_cursor *mc, indx_t indx,
+mdb_node_add(MDB_cursor *mc, indx_t indx,
MDB_val *key, MDB_val *data, pgno_t pgno, uint8_t flags)
{
unsigned int i;
assert(mp->mp_upper >= mp->mp_lower);
- DPRINTF("add to %s page %lu index %i, data size %zu key size %zu [%s]",
+ DPRINTF("add to %s page %zu index %i, data size %zu key size %zu [%s]",
IS_LEAF(mp) ? "leaf" : "branch",
mp->mp_pgno, indx, data ? data->mv_size : 0,
key ? key->mv_size : 0, key ? DKEY(key) : NULL);
DPRINTF("data size is %zu, put on overflow page",
data->mv_size);
node_size += sizeof(pgno_t);
- if ((ofp = mdb_new_page(mc, P_OVERFLOW, ovpages)) == NULL)
+ if ((ofp = mdb_page_new(mc, P_OVERFLOW, ovpages)) == NULL)
return ENOMEM;
- DPRINTF("allocated overflow page %lu", ofp->mp_pgno);
+ DPRINTF("allocated overflow page %zu", ofp->mp_pgno);
flags |= F_BIGDATA;
} else {
node_size += data->mv_size;
node_size += node_size & 1;
if (node_size + sizeof(indx_t) > SIZELEFT(mp)) {
- DPRINTF("not enough room in page %lu, got %u ptrs",
+ DPRINTF("not enough room in page %zu, got %u ptrs",
mp->mp_pgno, NUMKEYS(mp));
DPRINTF("upper - lower = %u - %u = %u", mp->mp_upper, mp->mp_lower,
mp->mp_upper - mp->mp_lower);
return MDB_SUCCESS;
}
+/** Delete the specified node from a page.
+ * @param[in] mp The page to operate on.
+ * @param[in] indx The index of the node to delete.
+ * @param[in] ksize The size of a node. Only used if the page is
+ * part of a #MDB_DUPFIXED database.
+ */
static void
-mdb_del_node(MDB_page *mp, indx_t indx, int ksize)
+mdb_node_del(MDB_page *mp, indx_t indx, int ksize)
{
unsigned int sz;
indx_t i, j, numkeys, ptr;
MDB_node *node;
char *base;
- DPRINTF("delete node %u on %s page %lu", indx,
+ DPRINTF("delete node %u on %s page %zu", indx,
IS_LEAF(mp) ? "leaf" : "branch", mp->mp_pgno);
assert(indx < NUMKEYS(mp));
mp->mp_upper += sz;
}
+/** 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.
+ * @param[in] mc The main cursor whose sorted-dups cursor is to be initialized.
+ */
static void
mdb_xcursor_init0(MDB_cursor *mc)
{
mx->mx_dbx.md_dirty = 0;
}
+/** Final setup of a sorted-dups cursor.
+ * Sets up the fields that depend on the data from the main cursor.
+ * @param[in] mc The main cursor whose sorted-dups cursor is to be initialized.
+ * @param[in] node The data containing the #MDB_db record for the
+ * sorted-dup database.
+ */
static void
mdb_xcursor_init1(MDB_cursor *mc, MDB_node *node)
{
mx->mx_dbx.md_name.mv_size = node->mn_ksize;
mx->mx_cursor.mc_snum = 0;
mx->mx_cursor.mc_flags = 0;
+ if (mx->mx_dbx.md_cmp == mdb_cmp_int && mx->mx_db.md_pad == sizeof(size_t))
+ mx->mx_dbx.md_cmp = mdb_cmp_long;
}
+/** Initialize a cursor for a given transaction and database. */
static void
-mdb_cursor_init(MDB_cursor *mc, MDB_txn *txn, MDB_dbi dbi)
+mdb_cursor_init(MDB_cursor *mc, MDB_txn *txn, MDB_dbi dbi, MDB_xcursor *mx)
{
mc->mc_dbi = dbi;
mc->mc_txn = txn;
mc->mc_dbx = &txn->mt_dbxs[dbi];
mc->mc_snum = 0;
mc->mc_flags = 0;
+ if (txn->mt_dbs[dbi].md_flags & MDB_DUPSORT) {
+ assert(mx != NULL);
+ mc->mc_xcursor = mx;
+ mdb_xcursor_init0(mc);
+ } else {
+ mc->mc_xcursor = NULL;
+ }
}
int
mdb_cursor_open(MDB_txn *txn, MDB_dbi dbi, MDB_cursor **ret)
{
MDB_cursor *mc;
+ MDB_xcursor *mx = NULL;
size_t size = sizeof(MDB_cursor);
if (txn == NULL || ret == NULL || !dbi || dbi >= txn->mt_numdbs)
size += sizeof(MDB_xcursor);
if ((mc = malloc(size)) != NULL) {
- mdb_cursor_init(mc, txn, dbi);
if (txn->mt_dbs[dbi].md_flags & MDB_DUPSORT) {
- MDB_xcursor *mx = (MDB_xcursor *)(mc + 1);
- mc->mc_xcursor = mx;
- mdb_xcursor_init0(mc);
+ mx = (MDB_xcursor *)(mc + 1);
}
+ mdb_cursor_init(mc, txn, dbi, mx);
} else {
return ENOMEM;
}
/* Return the count of duplicate data items for the current key */
int
-mdb_cursor_count(MDB_cursor *mc, unsigned long *countp)
+mdb_cursor_count(MDB_cursor *mc, size_t *countp)
{
MDB_node *leaf;
}
}
+/** Replace the key for a node with a new key.
+ * @param[in] mp The page containing the node to operate on.
+ * @param[in] indx The index of the node to operate on.
+ * @param[in] key The new key to use.
+ * @return 0 on success, non-zero on failure.
+ */
static int
mdb_update_key(MDB_page *mp, indx_t indx, MDB_val *key)
{
node = NODEPTR(mp, indx);
ptr = mp->mp_ptrs[indx];
- DPRINTF("update key %u (ofs %u) [%.*s] to [%s] on page %lu",
+ DPRINTF("update key %u (ofs %u) [%.*s] to [%s] on page %zu",
indx, ptr,
(int)node->mn_ksize, (char *)NODEKEY(node),
DKEY(key),
return MDB_SUCCESS;
}
-/* Move a node from csrc to cdst.
+/** Move a node from csrc to cdst.
*/
static int
-mdb_move_node(MDB_cursor *csrc, MDB_cursor *cdst)
+mdb_node_move(MDB_cursor *csrc, MDB_cursor *cdst)
{
int rc;
MDB_node *srcnode;
DKBUF;
/* Mark src and dst as dirty. */
- if ((rc = mdb_touch(csrc)) ||
- (rc = mdb_touch(cdst)))
+ if ((rc = mdb_page_touch(csrc)) ||
+ (rc = mdb_page_touch(cdst)))
return rc;
if (IS_LEAF2(csrc->mc_pg[csrc->mc_top])) {
unsigned int snum = csrc->mc_snum;
MDB_node *s2;
/* must find the lowest key below src */
- mdb_search_page_root(csrc, NULL, 0);
+ mdb_page_search_root(csrc, NULL, 0);
s2 = NODEPTR(csrc->mc_pg[csrc->mc_top], 0);
key.mv_size = NODEKSZ(s2);
key.mv_data = NODEKEY(s2);
data.mv_size = NODEDSZ(srcnode);
data.mv_data = NODEDATA(srcnode);
}
- DPRINTF("moving %s node %u [%s] on page %lu to node %u on page %lu",
+ DPRINTF("moving %s node %u [%s] on page %zu to node %u on page %zu",
IS_LEAF(csrc->mc_pg[csrc->mc_top]) ? "leaf" : "branch",
csrc->mc_ki[csrc->mc_top],
DKEY(&key),
/* Add the node to the destination page.
*/
- rc = mdb_add_node(cdst, cdst->mc_ki[cdst->mc_top], &key, &data, NODEPGNO(srcnode),
+ rc = mdb_node_add(cdst, cdst->mc_ki[cdst->mc_top], &key, &data, NODEPGNO(srcnode),
srcnode->mn_flags);
if (rc != MDB_SUCCESS)
return rc;
/* Delete the node from the source page.
*/
- mdb_del_node(csrc->mc_pg[csrc->mc_top], csrc->mc_ki[csrc->mc_top], key.mv_size);
+ mdb_node_del(csrc->mc_pg[csrc->mc_top], csrc->mc_ki[csrc->mc_top], key.mv_size);
/* Update the parent separators.
*/
key.mv_size = NODEKSZ(srcnode);
key.mv_data = NODEKEY(srcnode);
}
- DPRINTF("update separator for source page %lu to [%s]",
+ DPRINTF("update separator for source page %zu to [%s]",
csrc->mc_pg[csrc->mc_top]->mp_pgno, DKEY(&key));
if ((rc = mdb_update_key(csrc->mc_pg[csrc->mc_top-1], csrc->mc_ki[csrc->mc_top-1],
&key)) != MDB_SUCCESS)
key.mv_size = NODEKSZ(srcnode);
key.mv_data = NODEKEY(srcnode);
}
- DPRINTF("update separator for destination page %lu to [%s]",
+ DPRINTF("update separator for destination page %zu to [%s]",
cdst->mc_pg[cdst->mc_top]->mp_pgno, DKEY(&key));
if ((rc = mdb_update_key(cdst->mc_pg[cdst->mc_top-1], cdst->mc_ki[cdst->mc_top-1],
&key)) != MDB_SUCCESS)
return MDB_SUCCESS;
}
+/** Merge one page into another.
+ * The nodes from the page pointed to by \b csrc will
+ * be copied to the page pointed to by \b cdst and then
+ * the \b csrc page will be freed.
+ * @param[in] csrc Cursor pointing to the source page.
+ * @param[in] cdst Cursor pointing to the destination page.
+ */
static int
-mdb_merge(MDB_cursor *csrc, MDB_cursor *cdst)
+mdb_page_merge(MDB_cursor *csrc, MDB_cursor *cdst)
{
int rc;
indx_t i, j;
MDB_node *srcnode;
MDB_val key, data;
- DPRINTF("merging page %lu into %lu", csrc->mc_pg[csrc->mc_top]->mp_pgno, cdst->mc_pg[cdst->mc_top]->mp_pgno);
+ DPRINTF("merging page %zu into %zu", csrc->mc_pg[csrc->mc_top]->mp_pgno,
+ cdst->mc_pg[cdst->mc_top]->mp_pgno);
assert(csrc->mc_snum > 1); /* can't merge root page */
assert(cdst->mc_snum > 1);
/* Mark dst as dirty. */
- if ((rc = mdb_touch(cdst)))
+ if ((rc = mdb_page_touch(cdst)))
return rc;
/* Move all nodes from src to dst.
key.mv_size = csrc->mc_db->md_pad;
key.mv_data = METADATA(csrc->mc_pg[csrc->mc_top]);
for (i = 0; i < NUMKEYS(csrc->mc_pg[csrc->mc_top]); i++, j++) {
- rc = mdb_add_node(cdst, j, &key, NULL, 0, 0);
+ rc = mdb_node_add(cdst, j, &key, NULL, 0, 0);
if (rc != MDB_SUCCESS)
return rc;
key.mv_data = (char *)key.mv_data + key.mv_size;
key.mv_data = NODEKEY(srcnode);
data.mv_size = NODEDSZ(srcnode);
data.mv_data = NODEDATA(srcnode);
- rc = mdb_add_node(cdst, j, &key, &data, NODEPGNO(srcnode), srcnode->mn_flags);
+ rc = mdb_node_add(cdst, j, &key, &data, NODEPGNO(srcnode), srcnode->mn_flags);
if (rc != MDB_SUCCESS)
return rc;
}
}
- DPRINTF("dst page %lu now has %u keys (%.1f%% filled)",
+ DPRINTF("dst page %zu now has %u keys (%.1f%% filled)",
cdst->mc_pg[cdst->mc_top]->mp_pgno, NUMKEYS(cdst->mc_pg[cdst->mc_top]), (float)PAGEFILL(cdst->mc_txn->mt_env, cdst->mc_pg[cdst->mc_top]) / 10);
/* Unlink the src page from parent and add to free list.
*/
- mdb_del_node(csrc->mc_pg[csrc->mc_top-1], csrc->mc_ki[csrc->mc_top-1], 0);
+ mdb_node_del(csrc->mc_pg[csrc->mc_top-1], csrc->mc_ki[csrc->mc_top-1], 0);
if (csrc->mc_ki[csrc->mc_top-1] == 0) {
key.mv_size = 0;
if ((rc = mdb_update_key(csrc->mc_pg[csrc->mc_top-1], 0, &key)) != MDB_SUCCESS)
csrc->mc_db->md_leaf_pages--;
else
csrc->mc_db->md_branch_pages--;
- cursor_pop_page(csrc);
+ mdb_cursor_pop(csrc);
return mdb_rebalance(csrc);
}
+/** Copy the contents of a cursor.
+ * @param[in] csrc The cursor to copy from.
+ * @param[out] cdst The cursor to copy to.
+ */
static void
mdb_cursor_copy(const MDB_cursor *csrc, MDB_cursor *cdst)
{
}
}
+/** Rebalance the tree after a delete operation.
+ * @param[in] mc Cursor pointing to the page where rebalancing
+ * should begin.
+ * @return 0 on success, non-zero on failure.
+ */
static int
mdb_rebalance(MDB_cursor *mc)
{
unsigned int ptop;
MDB_cursor mn;
- DPRINTF("rebalancing %s page %lu (has %u keys, %.1f%% full)",
+ DPRINTF("rebalancing %s page %zu (has %u keys, %.1f%% full)",
IS_LEAF(mc->mc_pg[mc->mc_top]) ? "leaf" : "branch",
mc->mc_pg[mc->mc_top]->mp_pgno, NUMKEYS(mc->mc_pg[mc->mc_top]), (float)PAGEFILL(mc->mc_txn->mt_env, mc->mc_pg[mc->mc_top]) / 10);
if (PAGEFILL(mc->mc_txn->mt_env, mc->mc_pg[mc->mc_top]) >= FILL_THRESHOLD) {
- DPRINTF("no need to rebalance page %lu, above fill threshold",
+ DPRINTF("no need to rebalance page %zu, above fill threshold",
mc->mc_pg[mc->mc_top]->mp_pgno);
return MDB_SUCCESS;
}
DPUTS("collapsing root page!");
mdb_midl_append(mc->mc_txn->mt_free_pgs, mc->mc_pg[mc->mc_top]->mp_pgno);
mc->mc_db->md_root = NODEPGNO(NODEPTR(mc->mc_pg[mc->mc_top], 0));
- if ((rc = mdb_get_page(mc->mc_txn, mc->mc_db->md_root,
+ if ((rc = mdb_page_get(mc->mc_txn, mc->mc_db->md_root,
&mc->mc_pg[mc->mc_top])))
return rc;
mc->mc_db->md_depth--;
DPUTS("reading right neighbor");
mn.mc_ki[ptop]++;
node = NODEPTR(mc->mc_pg[ptop], mn.mc_ki[ptop]);
- if ((rc = mdb_get_page(mc->mc_txn, NODEPGNO(node), &mn.mc_pg[mn.mc_top])))
+ if ((rc = mdb_page_get(mc->mc_txn, NODEPGNO(node), &mn.mc_pg[mn.mc_top])))
return rc;
mn.mc_ki[mn.mc_top] = 0;
mc->mc_ki[mc->mc_top] = NUMKEYS(mc->mc_pg[mc->mc_top]);
DPUTS("reading left neighbor");
mn.mc_ki[ptop]--;
node = NODEPTR(mc->mc_pg[ptop], mn.mc_ki[ptop]);
- if ((rc = mdb_get_page(mc->mc_txn, NODEPGNO(node), &mn.mc_pg[mn.mc_top])))
+ if ((rc = mdb_page_get(mc->mc_txn, NODEPGNO(node), &mn.mc_pg[mn.mc_top])))
return rc;
mn.mc_ki[mn.mc_top] = NUMKEYS(mn.mc_pg[mn.mc_top]) - 1;
mc->mc_ki[mc->mc_top] = 0;
}
- DPRINTF("found neighbor page %lu (%u keys, %.1f%% full)",
+ DPRINTF("found neighbor page %zu (%u keys, %.1f%% full)",
mn.mc_pg[mn.mc_top]->mp_pgno, NUMKEYS(mn.mc_pg[mn.mc_top]), (float)PAGEFILL(mc->mc_txn->mt_env, mn.mc_pg[mn.mc_top]) / 10);
/* If the neighbor page is above threshold and has at least two
* Otherwise we should try to merge them.
*/
if (PAGEFILL(mc->mc_txn->mt_env, mn.mc_pg[mn.mc_top]) >= FILL_THRESHOLD && NUMKEYS(mn.mc_pg[mn.mc_top]) >= 2)
- return mdb_move_node(&mn, mc);
+ return mdb_node_move(&mn, mc);
else { /* FIXME: if (has_enough_room()) */
if (mc->mc_ki[ptop] == 0)
- return mdb_merge(&mn, mc);
+ return mdb_page_merge(&mn, mc);
else
- return mdb_merge(mc, &mn);
+ return mdb_page_merge(mc, &mn);
}
}
+/** Complete a delete operation started by #mdb_cursor_del(). */
static int
-mdb_del0(MDB_cursor *mc, MDB_node *leaf)
+mdb_cursor_del0(MDB_cursor *mc, MDB_node *leaf)
{
int rc;
memcpy(&pg, NODEDATA(leaf), sizeof(pg));
ovpages = OVPAGES(NODEDSZ(leaf), mc->mc_txn->mt_env->me_psize);
for (i=0; i<ovpages; i++) {
- DPRINTF("freed ov page %lu", pg);
+ DPRINTF("freed ov page %zu", pg);
mdb_midl_append(mc->mc_txn->mt_free_pgs, pg);
pg++;
}
}
- mdb_del_node(mc->mc_pg[mc->mc_top], mc->mc_ki[mc->mc_top], mc->mc_db->md_pad);
+ mdb_node_del(mc->mc_pg[mc->mc_top], mc->mc_ki[mc->mc_top], mc->mc_db->md_pad);
mc->mc_db->md_entries--;
rc = mdb_rebalance(mc);
if (rc != MDB_SUCCESS)
return EINVAL;
}
- mdb_cursor_init(&mc, txn, dbi);
- if (txn->mt_dbs[dbi].md_flags & MDB_DUPSORT) {
- mc.mc_xcursor = &mx;
- mdb_xcursor_init0(&mc);
- } else {
- mc.mc_xcursor = NULL;
- }
+ mdb_cursor_init(&mc, txn, dbi, &mx);
exact = 0;
if (data) {
return rc;
}
-/* Split page <mc->top>, and insert <key,(data|newpgno)> in either left or
- * right sibling, at index <mc->ki> (as if unsplit). Updates mc->top and
- * mc->ki with the actual values after split, ie if mc->top and mc->ki
- * refer to a node in the new right sibling page.
+/** Split a page and insert a new node.
+ * @param[in,out] 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.
+ * @param[in] newkey The key for the newly inserted node.
+ * @param[in] newdata The data for the newly inserted node.
+ * @param[in] newpgno The page number, if the new node is a branch node.
+ * @return 0 on success, non-zero on failure.
*/
static int
-mdb_split(MDB_cursor *mc, MDB_val *newkey, MDB_val *newdata, pgno_t newpgno)
+mdb_page_split(MDB_cursor *mc, MDB_val *newkey, MDB_val *newdata, pgno_t newpgno)
{
uint8_t flags;
int rc = MDB_SUCCESS, ins_new = 0;
mp = mc->mc_pg[mc->mc_top];
newindx = mc->mc_ki[mc->mc_top];
- DPRINTF("-----> splitting %s page %lu and adding [%s] at index %i",
+ DPRINTF("-----> splitting %s page %zu and adding [%s] at index %i",
IS_LEAF(mp) ? "leaf" : "branch", mp->mp_pgno,
DKEY(newkey), mc->mc_ki[mc->mc_top]);
if (mc->mc_snum < 2) {
- if ((pp = mdb_new_page(mc, P_BRANCH, 1)) == NULL)
+ if ((pp = mdb_page_new(mc, P_BRANCH, 1)) == NULL)
return ENOMEM;
/* shift current top to make room for new parent */
mc->mc_pg[1] = mc->mc_pg[0];
mc->mc_db->md_depth++;
/* Add left (implicit) pointer. */
- if ((rc = mdb_add_node(mc, 0, NULL, NULL, mp->mp_pgno, 0)) != MDB_SUCCESS) {
+ if ((rc = mdb_node_add(mc, 0, NULL, NULL, mp->mp_pgno, 0)) != MDB_SUCCESS) {
/* undo the pre-push */
mc->mc_pg[0] = mc->mc_pg[1];
mc->mc_ki[0] = mc->mc_ki[1];
ptop = 0;
} else {
ptop = mc->mc_top-1;
- DPRINTF("parent branch page is %lu", mc->mc_pg[ptop]->mp_pgno);
+ DPRINTF("parent branch page is %zu", mc->mc_pg[ptop]->mp_pgno);
}
/* Create a right sibling. */
- if ((rp = mdb_new_page(mc, mp->mp_flags, 1)) == NULL)
+ if ((rp = mdb_page_new(mc, mp->mp_flags, 1)) == NULL)
return ENOMEM;
mdb_cursor_copy(mc, &mn);
mn.mc_pg[mn.mc_top] = rp;
mn.mc_ki[ptop] = mc->mc_ki[ptop]+1;
- DPRINTF("new right sibling: page %lu", rp->mp_pgno);
+ DPRINTF("new right sibling: page %zu", rp->mp_pgno);
nkeys = NUMKEYS(mp);
split_indx = nkeys / 2 + 1;
if (SIZELEFT(mn.mc_pg[ptop]) < mdb_branch_size(mc->mc_txn->mt_env, &sepkey)) {
mn.mc_snum--;
mn.mc_top--;
- rc = mdb_split(&mn, &sepkey, NULL, rp->mp_pgno);
+ rc = mdb_page_split(&mn, &sepkey, NULL, rp->mp_pgno);
/* Right page might now have changed parent.
* Check if left page also changed parent.
}
} else {
mn.mc_top--;
- rc = mdb_add_node(&mn, mn.mc_ki[ptop], &sepkey, NULL, rp->mp_pgno, 0);
+ rc = mdb_node_add(&mn, mn.mc_ki[ptop], &sepkey, NULL, rp->mp_pgno, 0);
mn.mc_top++;
}
if (IS_LEAF2(rp)) {
rkey.mv_size = 0;
}
- rc = mdb_add_node(mc, j, &rkey, &rdata, pgno, flags);
+ rc = mdb_node_add(mc, j, &rkey, &rdata, pgno, flags);
}
/* reset back to original page */
if ((flags & (MDB_NOOVERWRITE|MDB_NODUPDATA)) != flags)
return EINVAL;
- mdb_cursor_init(&mc, txn, dbi);
- if (txn->mt_dbs[dbi].md_flags & MDB_DUPSORT) {
- mc.mc_xcursor = &mx;
- mdb_xcursor_init0(&mc);
- } else {
- mc.mc_xcursor = NULL;
- }
+ mdb_cursor_init(&mc, txn, dbi, &mx);
return mdb_cursor_put(&mc, key, data, flags);
}
return MDB_SUCCESS;
}
+/** Common code for #mdb_stat() and #mdb_env_stat().
+ * @param[in] env the environment to operate in.
+ * @param[in] db the #MDB_db record containing the stats to return.
+ * @param[out] arg the address of an #MDB_stat structure to receive the stats.
+ * @return 0, this function always succeeds.
+ */
static int
mdb_stat0(MDB_env *env, MDB_db *db, MDB_stat *arg)
{
return mdb_stat0(env, &env->me_metas[toggle]->mm_dbs[MAIN_DBI], arg);
}
+/** 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().
+ * @param[in] txn A transaction handle returned by #mdb_txn_begin()
+ * @param[in] dbi A database handle returned by #mdb_open()
+ */
static void
mdb_default_cmp(MDB_txn *txn, MDB_dbi dbi)
{
if (txn->mt_dbs[dbi].md_flags & MDB_REVERSEKEY)
- txn->mt_dbxs[dbi].md_cmp = memnrcmp;
+ txn->mt_dbxs[dbi].md_cmp = mdb_cmp_memnr;
else if (txn->mt_dbs[dbi].md_flags & MDB_INTEGERKEY)
- txn->mt_dbxs[dbi].md_cmp = cintcmp;
+ txn->mt_dbxs[dbi].md_cmp = mdb_cmp_cint;
else
- txn->mt_dbxs[dbi].md_cmp = memncmp;
+ txn->mt_dbxs[dbi].md_cmp = mdb_cmp_memn;
if (txn->mt_dbs[dbi].md_flags & MDB_DUPSORT) {
if (txn->mt_dbs[dbi].md_flags & MDB_INTEGERDUP) {
if (txn->mt_dbs[dbi].md_flags & MDB_DUPFIXED)
- txn->mt_dbxs[dbi].md_dcmp = intcmp;
+ txn->mt_dbxs[dbi].md_dcmp = mdb_cmp_int;
else
- txn->mt_dbxs[dbi].md_dcmp = cintcmp;
+ txn->mt_dbxs[dbi].md_dcmp = mdb_cmp_cint;
} else if (txn->mt_dbs[dbi].md_flags & MDB_REVERSEDUP) {
- txn->mt_dbxs[dbi].md_dcmp = memnrcmp;
+ txn->mt_dbxs[dbi].md_dcmp = mdb_cmp_memnr;
} else {
- txn->mt_dbxs[dbi].md_dcmp = memncmp;
+ txn->mt_dbxs[dbi].md_dcmp = mdb_cmp_memn;
}
} else {
txn->mt_dbxs[dbi].md_dcmp = NULL;
memset(&dummy, 0, sizeof(dummy));
dummy.md_root = P_INVALID;
dummy.md_flags = flags & 0xffff;
- mdb_cursor_init(&mc, txn, MAIN_DBI);
+ mdb_cursor_init(&mc, txn, MAIN_DBI, NULL);
rc = mdb_cursor_put(&mc, &key, &data, F_SUBDATA);
dirty = 1;
}
return MDB_SUCCESS;
}
+int mdb_set_relctx(MDB_txn *txn, MDB_dbi dbi, void *ctx)
+{
+ if (txn == NULL || !dbi || dbi >= txn->mt_numdbs)
+ return EINVAL;
+
+ txn->mt_dbxs[dbi].md_relctx = ctx;
+ return MDB_SUCCESS;
+}
+
/** @} */