Disk Based Hashtables (DBH) 64 bit ********************************** Name **** Disk Based Hashtables (DBH) 64 bit -- Library to create and manage hash tables residing on disk. Associations are made between keys and values so that for a given a key the value can be found and loaded into memory quickly. Being disk based allows for large and persistent hashes. 64 bit support allows for hashtables with sizes over 4 Gigabytes on 32 bit systems. Quantified key generation allows for minimum access time on balanced multidimensional trees. Stability Level *************** Stable, unless otherwise indicated Synopsis ******** #include #define DBH_CREATE #define DBH_DATA (dbh) #define DBH_DATA_SPACE (dbh) #define DBH_ERASED_SPACE (dbh) #define DBH_FILE_VERSION #define DBH_FORMAT_SPACE (dbh) #define DBH_KEY (dbh) #define DBH_KEYLENGTH (dbh) #define DBH_MAXIMUM_RECORD_SIZE (dbh) #define DBH_PARALLEL_SAFE #define DBH_PATH (dbh) #define DBH_READ_ONLY #define DBH_RECORDS (dbh) #define DBH_RECORD_SIZE (dbh) #define DBH_THREAD_SAFE #define DBH_TOTAL_SPACE (dbh) #define DBH_VERSION void (*DBHashFunc) (DBHashTable *dbh); void (*DBHashFunc2) (DBHashTable *dbh, void *data); struct DBHashTable; #define FILE_POINTER DBHashTable * dbh_new (const char *path, unsigned char *key_length, int flags); DBHashTable * dbh_open (const char *path); DBHashTable * dbh_open_ro (const char *path); DBHashTable * dbh_create (const char *path, unsigned char key_length); int dbh_close (DBHashTable *dbh); int dbh_destroy (DBHashTable *dbh); int dbh_erase (DBHashTable *dbh); int dbh_unerase (DBHashTable *dbh); FILE_POINTER dbh_update (DBHashTable *dbh); FILE_POINTER dbh_load (DBHashTable *dbh); unsigned char dbh_load_address (DBHashTable *dbh, FILE_POINTER currentseek); FILE_POINTER dbh_load_child (DBHashTable *dbh, unsigned char key_index); FILE_POINTER dbh_load_parent (DBHashTable *dbh); void dbh_set_data (DBHashTable *dbh, void *data, FILE_POINTER size); void dbh_set_key (DBHashTable *dbh, unsigned char *key); void dbh_set_recordsize (DBHashTable *dbh, int record_size); int dbh_set_size (DBHashTable *dbh, FILE_POINTER size); int dbh_settempdir (DBHashTable *dbh, char *temp_dir); dbh_lock_t; int dbh_clear_locks (DBHashTable *dbh); int dbh_set_lock_timeout (int seconds); int dbh_get_lock_timeout (void); int dbh_set_parallel_lock_attempt_limit (DBHashTable *dbh, int limit); int dbh_set_parallel_lock_timeout (DBHashTable *dbh, int seconds); int dbh_lock_read (DBHashTable *dbh); int dbh_unlock_read (DBHashTable *dbh); int dbh_lock_write (DBHashTable *dbh); int dbh_unlock_write (DBHashTable *dbh); int dbh_mutex_lock (DBHashTable *dbh); int dbh_mutex_unlock (DBHashTable *dbh); FILE_POINTER dbh_find (DBHashTable *dbh, int n); int dbh_fanout (DBHashTable *dbh, DBHashFunc operate, unsigned char *key1, unsigned char *key2, unsigned char ignore_portion); int dbh_sweep (DBHashTable *dbh, DBHashFunc operate, unsigned char *key1, unsigned char *key2, unsigned char ignore_portion); int dbh_foreach (DBHashTable *dbh, DBHashFunc2 operate, void *data); int dbh_foreach_fanout (DBHashTable *dbh, DBHashFunc operate); int dbh_foreach_sweep (DBHashTable *dbh, DBHashFunc operate); void dbh_exit_fanout (DBHashTable *dbh); void dbh_exit_sweep (DBHashTable *dbh); int dbh_prune (DBHashTable *dbh, unsigned char *key, unsigned char subtree_length); int dbh_unprune (DBHashTable *dbh, unsigned char *key, unsigned char subtree_length); void dbh_regen_fanout (DBHashTable **dbh); void dbh_regen_sweep (DBHashTable **dbh); void dbh_genkey (unsigned char *key, unsigned char length, unsigned int n); void dbh_genkey0 (unsigned char *key, unsigned char length, unsigned int n); void dbh_genkey2 (unsigned char *key, unsigned char length, unsigned int n); void dbh_orderkey (unsigned char *key, unsigned char length, unsigned int n, unsigned char base); struct dbh_header_t; int dbh_info (DBHashTable *dbh); int dbh_writeheader (DBHashTable *dbh); Description *********** A DBHashTable provides associations between keys and values which is optimized so that given a key, the associated value can be found very quickly. Note that only one hash record is loaded from disk to memory at any given moment for a DBHashTable. Both keys and values should be copied into the DBHashTable record, so they need not exist for the lifetime of the DBHashTable. This means that the use of static strings and temporary strings (i.e. those created in buffers and those returned by GTK+ widgets) should be copied with `dbh_set_key()' (*note dbh_set_key []::) and `dbh_set_data()' (*note dbh_set_data []::) into the DBHashTable record before being inserted. You must be careful to ensure that copied key length matches the defined key length of the DBHashTable, and also that the copied data does not exceed the maximum length of the DBHashTable record (1024 bytes by default, and expandable by `dbh_set_size()' (*note dbh_set_size []::) ). If the DBHashTable record length is to be variable, be sure to set the appropriate length before each `dbh_update()' (*note dbh_update []::), with `dbh_set_recordsize()' (*note dbh_set_recordsize []::), otherwise the record length need only be set before the first `dbh_update()' (*note dbh_update []::). To create a DBHashTable, use `dbh_new()' (*note dbh_new []::). A DBHashTable may be opened (either new or existing) in read-only mode, parallel-safe mode or thread-safe mode. To insert a key and value into a DBHashTable, use `dbh_update()' (*note dbh_update []::). The DBHashTable will not be modified until this command is given. All changes to the current DBHashTable record only reside in memory. `dbh_update()' (*note dbh_update []::) is necessary to commit the changes to the DBHashTable. To lookup a value corresponding to a given key, use `dbh_load()' (*note dbh_load []::). To erase and unerase a key and value, use `dbh_erase()' (*note dbh_erase []::) and `dbh_unerase()' (*note dbh_unerase []::). To call a function for each key and value pair (using a sweep route) use `dbh_foreach_sweep()' (*note dbh_foreach_sweep []::) and `dbh_sweep()' (*note dbh_sweep []::). To call a function for each key and value pair (using a fanout route) use `dbh_foreach_fanout()' (*note dbh_foreach_fanout []::) and `dbh_foreach_fanout()' (*note dbh_foreach_fanout []::). To destroy a DBHashTable use `dbh_destroy()' (*note dbh_destroy []::). This is dbh version 2, incompatible with dbh version 1 files. The main difference between the two version is the handling of file pointers. In version 1, file pointers were 32 bits in length, while in version 2, file pointers are 64 bits in length. This allows for DBHashTables with sizes greater than 2 GBytes. `Quantified numbers' are an alternate way to view the set of `natural numbers' {1, 2, 3, ...} where order is defined in two levels. In `natural numbers' there is only one level of order (defined by the > boolean operator). In `quantified numbers' the first level of order is defined by the `cuanta' or quantity. The `cuanta' is obtained by adding all the digits of the `quantified number'. Thus, for example, 10022, 5, 32, and 11111 are all equal at the first level of order since they all add up to 5. The second level or order may be obtained in different manners. In functions `dbh_genkey()' (*note dbh_genkey []::) and `dbh_genkey2()' (*note dbh_genkey2 []::) the corresponding order of the `natural numbers' from which they are associated is _not conserved_. In `dbh_orderkey()' (*note dbh_orderkey []::) the corresponding order of the `natural numbers' from which they are associated _is conserved_, but at a price. The base, or maximum value each digit may reach, must be defined. This effectively puts a limit on the number of keys which may be generated for a given number of digits. When a `DBHashTable' (*note struct DBHashTable::) is constructed with `quantified' keys, the maximum amount of disk access instructions generated to access any given record is equal to the `cuanta' of the quantified number represented by the key. This allows a `DBHashTable' (*note struct DBHashTable::) to be constructed with minimum access time across all records. Details ******* DBH_CREATE ========== #define DBH_CREATE Bit flag for `dbh_new()' (*note dbh_new []::) to create a new dbh file on disk, overwriting any file with the same name and cleansing all locks. DBH_DATA() ========== #define DBH_DATA(dbh) This macro returns a pointer to the current `DBHashTable' (*note struct DBHashTable::) data area. `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *) DBH_DATA_SPACE() ================ #define DBH_DATA_SPACE(dbh) This macro returns the amount of bytes taken up by valid data in the `DBHashTable' (*note struct DBHashTable::). `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *) DBH_ERASED_SPACE() ================== #define DBH_ERASED_SPACE(dbh) This macro returns the amount of bytes taken up by erased data in the `DBHashTable' (*note struct DBHashTable::). `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *) DBH_FILE_VERSION ================ #define DBH_FILE_VERSION "DBH_2.0/64bit" Disk Based Hashtables library file version compatibility DBH_FORMAT_SPACE() ================== #define DBH_FORMAT_SPACE(dbh) This macro returns the total amount of bytes taken up by the format of the `DBHashTable' (*note struct DBHashTable::). `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *) DBH_KEY() ========= #define DBH_KEY(dbh) This macro returns a pointer to the current `DBHashTable' (*note struct DBHashTable::) key area. `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *) DBH_KEYLENGTH() =============== #define DBH_KEYLENGTH(dbh) This macro returns the keylenth in bytes associated to the `DBHashTable' (*note struct DBHashTable::). The value is fixed when the `DBHashTable' (*note struct DBHashTable::) is created with `dbh_new' (*note dbh_new []::). `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *) DBH_MAXIMUM_RECORD_SIZE() ========================= #define DBH_MAXIMUM_RECORD_SIZE(dbh) This macro returns the maximum allocated space for data in the current `DBHashTable' (*note struct DBHashTable::) record. `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *) DBH_PARALLEL_SAFE ================= #define DBH_PARALLEL_SAFE Bit flag for `dbh_new()' (*note dbh_new []::) to use if more than one heavy weight process will be accessing the same `DBHashTable' (*note struct DBHashTable::) in write mode. If no process will be writing to the `DBHashTable' (*note struct DBHashTable::), then `DBH_READ_ONLY' (*note DBH_READ_ONLY::) is enough and faster since each process will hold a separate memory allocation for the `DBHashTable' (*note struct DBHashTable::) pointer. DBH_PATH() ========== #define DBH_PATH(dbh) This macro returns a pointer to a string containing the path to the current `DBHashTable' (*note struct DBHashTable::). `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *) DBH_READ_ONLY ============= #define DBH_READ_ONLY Bit flag for `dbh_new()' (*note dbh_new []::) to open an existing dbh file on disk in read only mode. DBH_RECORDS() ============= #define DBH_RECORDS(dbh) This macro returns the number of records in the `DBHashTable' (*note struct DBHashTable::). `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *) DBH_RECORD_SIZE() ================= #define DBH_RECORD_SIZE(dbh) This macro returns the size of the current record loaded in memory. If no record has been loaded, then the return value is not defined. `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *) DBH_THREAD_SAFE =============== #define DBH_THREAD_SAFE Bit flag for `dbh_new()' (*note dbh_new []::) to use if more than one thread will be accessing the same `DBHashTable' (*note struct DBHashTable::) in write mode in parallel. DBH function calls which may be racing each other in different threads should be enclosed within a `dbh_mutex_lock()' (*note dbh_mutex_lock []::) and `dbh_mutex_unlock()' (*note dbh_mutex_unlock []::). Each DBH table opened with the `DBH_THREAD_SAFE' (*note DBH_THREAD_SAFE::) attribute will have a specific mutex for this function. If threads are to access the same `DBHashTable' (*note struct DBHashTable::) in read mode only, then `DBH_READ_ONLY' (*note DBH_READ_ONLY::) and separate memory allocation for each thread's `DBHashTable' (*note struct DBHashTable::) pointer is more than enough and faster. When `DBH_THREAD_SAFE' (*note DBH_THREAD_SAFE::) is specified, `dbh_new()' (*note dbh_new []::) is automatically mutex locked until function completes. The function `dbh_close()' (*note dbh_close []::) is also automatically locked until completion on tables opened with the `DBH_THREAD_SAFE' (*note DBH_THREAD_SAFE::) attribute. DBH_TOTAL_SPACE() ================= #define DBH_TOTAL_SPACE(dbh) This macro returns the total amount of bytes taken up by the `DBHashTable' (*note struct DBHashTable::). `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *) DBH_VERSION =========== #define DBH_VERSION "5.0.15" Disk Based Hashtables library version DBHashFunc () ============= void (*DBHashFunc) (DBHashTable *dbh); Pointer to function to apply during `dbh_sweep()' (*note dbh_sweep []::), `dbh_fanout()' (*note dbh_fanout []::), `dbh_foreach_sweep()' (*note dbh_foreach_sweep []::) and `dbh_foreach_fanout()' (*note dbh_foreach_fanout []::). This function will be applied to all data records involved in the sweep or fanout process `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *) DBHashFunc2 () ============== void (*DBHashFunc2) (DBHashTable *dbh, void *data); `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *) Pointer to function to apply during `dbh_sweep()' (*note dbh_sweep []::), `dbh_fanout()' (*note dbh_fanout []::), `dbh_foreach_sweep()' (*note dbh_foreach_sweep []::) and `dbh_foreach_fanout()' (*note dbh_foreach_fanout []::). `data' : pointer to other data to be passed to function This function will be applied to all data records involved in the sweep or fanout process struct DBHashTable ================== struct DBHashTable { unsigned char branches; FILE_POINTER bytes_userdata; unsigned char *key; void *data; int fd; dbh_header_t *head_info; char *path; }; `DBHashTable' (*note struct DBHashTable::) is a data structure containing the record information for an open `DBHashTable' (*note struct DBHashTable::) file. unsigned [Cross reference to non-existant ID "char"] `branches'; Maximum toplevel branches `FILE_POINTER' (*note FILE_POINTER::) `bytes_userdata'; size of data record unsigned [Cross reference to non-existant ID "char"] *`key'; access key [Cross reference to non-existant ID "void"] *`data'; record data pointer [Cross reference to non-existant ID "int"] `fd'; file descriptor `dbh_header_t' (*note struct dbh_header_t::) *`head_info'; nonvolatile header information [Cross reference to non-existant ID "char"] *`path'; file path FILE_POINTER ============ #define FILE_POINTER Architecture independent 64 bit integer type dbh_new () ========== DBHashTable * dbh_new (const char *path, unsigned char *key_length, int flags); Open or create an existing DBH table. Flag is bitwise or of the following: `DBH_CREATE' (*note DBH_CREATE::), `DBH_READ_ONLY' (*note DBH_READ_ONLY::), `DBH_THREAD_SAFE' (*note DBH_THREAD_SAFE::), `DBH_PARALLEL_SAFE' (*note DBH_PARALLEL_SAFE::). (since 4.7.6) `path' : Path on disk where DBHashTable resides. `key_length' : A pointer to store the length of the key to access the DBHashTable. `flags' : Bitwise or of DBH_CREATE, DBH_READ_ONLY, DBH_THREAD_SAFE, DBH_PARALLEL_SAFE _Returns_ : A pointer to the newly opened DBHashTable, or NULL if it fails. dbh_open () =========== DBHashTable * dbh_open (const char *path); *Warning* `dbh_open' is deprecated and should not be used in newly-written code. Use `dbh_new()' (*note dbh_new []::) instead Open an existing hash in read-write mode. `path' : Path on disk where DBHashTable resides. _Returns_ : A pointer to the newly opened `DBHashTable' (*note struct DBHashTable::), or NULL if it fails. dbh_open_ro () ============== DBHashTable * dbh_open_ro (const char *path); *Warning* `dbh_open_ro' is deprecated and should not be used in newly-written code. Use `dbh_new()' (*note dbh_new []::) instead Open an existing hash in read-only mode. `path' : Path on disk where DBHashTable resides. _Returns_ : A pointer to the newly opened read-only DBHashTable, or NULL if it fails. dbh_create () ============= DBHashTable * dbh_create (const char *path, unsigned char key_length); *Warning* `dbh_create' is deprecated and should not be used in newly-written code. Use `dbh_new()' (*note dbh_new []::) instead Create a new hash file (overwriting old version). Creates and opens for writing a new `DBHashTable' (*note struct DBHashTable::). This function will overwrite any file with the specified path, including any previous DBH file. The `key_length' is fixed. If you want variable length, use a g_hash_table to associate quantified keys generated by [Cross reference to non-existant ID "genkey"], and create an extra DBHashTable to save the g_hash. Quantified keys assure that large DBHashes are spread out optimally. `path' : Path on disk where DBHashTable will reside. `key_length' : The length of the key to access the DBHashTable. _Returns_ : A pointer to the newly created and opened `DBHashTable' (*note struct DBHashTable::), or NULL if it fails. dbh_close () ============ int dbh_close (DBHashTable *dbh); Close hash file (thus flushing io buffer). `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). _Returns_ : 0 on error, 1 otherwise. dbh_destroy () ============== int dbh_destroy (DBHashTable *dbh); Close an open DBHashTable and erase file from disk. Convenience function that does a close and rm. `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). _Returns_ : 0 if error, 1 otherwise dbh_erase () ============ int dbh_erase (DBHashTable *dbh); Mark the record currently loaded into memory as erased. If no record is currently loaded, behaviour is undefined. `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). _Returns_ : 0 on error, 1 otherwise. dbh_unerase () ============== int dbh_unerase (DBHashTable *dbh); This is the opposite of `dbh_erase()' (*note dbh_erase []::). Mark the record currently loaded into memory as unerased. If no record is currently loaded, behaviour is undefined. `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). _Returns_ : 0 on error, 1 otherwise. dbh_update () ============= FILE_POINTER dbh_update (DBHashTable *dbh); Update the current record in memory to the disk based hash. Update function will update erased records as well as unerased records, but if an erased record is updated, it is automatically unerased. `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). _Returns_ : 0 on error, byte offset of loaded record otherwise. dbh_load () =========== FILE_POINTER dbh_load (DBHashTable *dbh); Load a record using the currently set key. This function will also load erased values, except that it will return 0. `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). _Returns_ : 0 on error, byte offset of loaded record otherwise. dbh_load_address () =================== unsigned char dbh_load_address (DBHashTable *dbh, FILE_POINTER currentseek); Load a record from hash table directly from byte offset `currentseek' `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). `currentseek' : A byte offset. _Returns_ : 0 on error, number of branches otherwise. dbh_load_child () ================= FILE_POINTER dbh_load_child (DBHashTable *dbh, unsigned char key_index); Load the first child of the currently loaded record, on branch identified by `key_index'. Since the number of childs (or branches) of each record is variable, this may be tricky. Top level records have `DBH_KEYLENGTH' (*note DBH_KEYLENGTH[]::) branches. Lower level records have less. Each byte of a key represents a branch on top level records. `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). `key_index' : branch number on which to return the child. _Returns_ : 0 on error, byte offset of loaded record otherwise. dbh_load_parent () ================== FILE_POINTER dbh_load_parent (DBHashTable *dbh); Load the parent of the currently loaded record. `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). _Returns_ : 0 on error, byte offset of loaded record otherwise. dbh_set_data () =============== void dbh_set_data (DBHashTable *dbh, void *data, FILE_POINTER size); This function copies the user data into the current `DBHashTable' (*note struct DBHashTable::) record and along with function `dbh_set_key()' (*note dbh_set_key []::), makes the current `DBHashTable' (*note struct DBHashTable::) record ready for the `dbh_update()' (*note dbh_update []::) function to commit to the actual `DBHashTable' (*note struct DBHashTable::) on disk. `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). `data' : Pointer to the data to copy to the current `DBHashTable' (*note struct DBHashTable::) record `size' : The amount of bytes to copy to the current `DBHashTable' (*note struct DBHashTable::) record dbh_set_key () ============== void dbh_set_key (DBHashTable *dbh, unsigned char *key); This function sets the key of the current DBHashTable record. `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). `key' : The key to set as the current DBHashTable record key. dbh_set_recordsize () ===================== void dbh_set_recordsize (DBHashTable *dbh, int record_size); This sets the recordsize of the the data in the current `DBHashTable' (*note struct DBHashTable::) record. It is called implicitly by calling `dbh_set_data()' (*note dbh_set_data []::). It is very important to call this function. Unpredictable results will follow if record_size is not set. `DBHashTable' (*note struct DBHashTable::) records are variable in length, so use this function at least once if you are planning to use fixed length records. This function is not needed if `dbh_set_data()' (*note dbh_set_data []::) is used to set the record data. `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). `record_size' : The amount of bytes in the current `DBHashTable' (*note struct DBHashTable::) record. dbh_set_size () =============== int dbh_set_size (DBHashTable *dbh, FILE_POINTER size); Defines the maximum amount of memory to be allocated to the `DBHashTable' (*note struct DBHashTable::) records. This is nonvolatile information which need to be set only once. The default is 1Kbyte. `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). `size' : size in bytes. _Returns_ : 0 on error, 1 otherwise. dbh_settempdir () ================= int dbh_settempdir (DBHashTable *dbh, char *temp_dir); Sets the temporary directory to be used by `dbh_regen_sweep()' (*note dbh_regen_sweep []::) or `dbh_regen_fanout()' (*note dbh_regen_fanout []::). It is usually best to set temporary directory on the same filesystem device. The default value for the temporary directory is the directory where `dbh' is located. To reset to default value, send NULL as the `temp_dir' `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). `temp_dir' : path to temporary directory to use. _Returns_ : 0 if error, 1 otherwise dbh_lock_t ========== typedef struct { pid_t write_lock; int write_lock_count; int read_lock_count; } dbh_lock_t; [Cross reference to non-existant ID "pid-t"] `write_lock'; PID of process holding the write lock or zero. [Cross reference to non-existant ID "int"] `write_lock_count'; Number of write locks the PID hold (write locks are recursive). [Cross reference to non-existant ID "int"] `read_lock_count'; Number of read locks on DBH table. dbh_clear_locks () ================== int dbh_clear_locks (DBHashTable *dbh); `Returns': 0 if error, 1 otherwise Clear dbh file locks associated to `DBHashTable' (*note struct DBHashTable::) Use this function to clean up persistent file locks (since 4.7.6) `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). dbh_set_lock_timeout () ======================= int dbh_set_lock_timeout (int seconds); Sets the default time for obtaining a read/write lock in parallel safe mode. The default value is zero, which means there is no timeout. If there is no timeout, file locking will block until lock is secured. Locks may persist beyond program life and may be stale if program crashed before unlocking was performed. Does not affect currently open dbh files. If the value for a currently open dbh file is to be modified, use `dbh_set_parallel_lock_timeout()' (*note dbh_set_parallel_lock_timeout []::) as well. `seconds' : Timeout default for obtaining a read/write lock in parallel safe mode. _Returns_ : 0 if error, 1 otherwise dbh_get_lock_timeout () ======================= int dbh_get_lock_timeout (void); Gets the default time for obtaining a read/write lock in parallel safe mode. The default value is zero, which means there is no timeout. If there is no timeout, file locking will block until lock is secured. Locks may persist beyond program life and may be stale if program crashed before unlocking was performed. _Returns_ : the default timeout in seconds to secure a read/write lock in parallel safe mode. dbh_set_parallel_lock_attempt_limit () ====================================== int dbh_set_parallel_lock_attempt_limit (DBHashTable *dbh, int limit); *Warning* `dbh_set_parallel_lock_attempt_limit' is deprecated and should not be used in newly-written code. Use `dbh_set_parallel_lock_timeout()' (*note dbh_set_parallel_lock_timeout []::) instead. As of 5.0.10, this function is inoperative. Sets the limit on the attempts to lock a parallel protected dbh file lock before considering the lock to be stale. Stale locks may occur when the calling program crashes while the lock is set in either read or write mode. Lock will persist in shared memory beyond program crash. Lock may be removed manually, or a lock attempt limit on the number of tries specified to remove the lock automatically. Each lock attempt limit is equal to 1/10th of a second (1E+08 nanoseconds). If limit is set to zero, then lock attempts will continue indefinitely. `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). `limit' : Number of attempts to lock a parallel protected file lock before removing lock. _Returns_ : 0 if error, 1 otherwise dbh_set_parallel_lock_timeout () ================================ int dbh_set_parallel_lock_timeout (DBHashTable *dbh, int seconds); `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). `seconds' : Number of second to try to lock a parallel protected file before failing. A value of zero means function will block until lock is obtained. _Returns_ : 0 if error, 1 otherwise dbh_lock_read () ================ int dbh_lock_read (DBHashTable *dbh); Attempts to get a read lock on the dbh file. A file may have any number of readlocks as long as no write lock is set. If `dbh_set_parallel_lock_timeout()' (*note dbh_set_parallel_lock_timeout []::) is set to zero (that's the default) this function will block until lock is secured. `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). _Returns_ : 0 if error, 1 otherwise dbh_unlock_read () ================== int dbh_unlock_read (DBHashTable *dbh); Releases a read lock on the dbh file. `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). _Returns_ : 0 if error, 1 otherwise dbh_lock_write () ================= int dbh_lock_write (DBHashTable *dbh); Attempts to get a write lock on the dbh file. A file can only have one write lock, and when write lock is set, no read locks may be secured. If `dbh_set_parallel_lock_timeout()' (*note dbh_set_parallel_lock_timeout []::) is set to zero (that's the default) this function will block until lock is secured. `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). _Returns_ : 0 if error, 1 otherwise dbh_unlock_write () =================== int dbh_unlock_write (DBHashTable *dbh); Releases a write lock on the dbh file. `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). _Returns_ : 0 if error, 1 otherwise dbh_mutex_lock () ================= int dbh_mutex_lock (DBHashTable *dbh); Lock the DBHashTable mutex. This is only valid if table was opened with the DBH_THREAD_SAFE flag, Otherwise the function does nothing. `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). _Returns_ : 0 if error, 1 otherwise dbh_mutex_unlock () =================== int dbh_mutex_unlock (DBHashTable *dbh); Unlock the DBHashTable mutex. This is only valid if table was opened with the DBH_THREAD_SAFE flag, Otherwise the function does nothing. `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). _Returns_ : 0 if error, 1 otherwise dbh_find () =========== FILE_POINTER dbh_find (DBHashTable *dbh, int n); Find the top level subtree FILE_POINTER for the currently loaded record, but ignoring the last `n' branches. `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). `n' : Number of branches to ignore on top record. _Returns_ : 0 on error, byte offset of loaded record otherwise. dbh_fanout () ============= int dbh_fanout (DBHashTable *dbh, DBHashFunc operate, unsigned char *key1, unsigned char *key2, unsigned char ignore_portion); Apply a function to subtree members of the hash, following a fanout trajectory (horizontally through records). In order for `dbh_fanout()' (*note dbh_fanout []::) to be extremely fast, you should prepare the `DBHashTable' (*note struct DBHashTable::) for the trajectory with `dbh_regen_fanout()' (*note dbh_regen_fanout []::) first. This allows for extremely efficient use of hardware and operating system caches. `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). `operate' : The function to apply to each selected member of the `DBHashTable' (*note struct DBHashTable::) `key1' : The key from which to start the fanout or NULL if you don't care. Make sure it is a top level node of a subtree with `dbh_find()' (*note dbh_find []::) first. `key2' : The key which will trigger an exit condition from the sweep, or NULL if don't care. `ignore_portion' : The ignored trailing bytes of key1 which will define the magnitud of the subtree to be sweeped, or zero if don't care. _Returns_ : 0 on error, 1 otherwise. dbh_sweep () ============ int dbh_sweep (DBHashTable *dbh, DBHashFunc operate, unsigned char *key1, unsigned char *key2, unsigned char ignore_portion); Apply a function to subtree members of the hash, following a sweep trajectory (vertically through branches). In order for `dbh_sweep()' (*note dbh_sweep []::) to be extremely fast, you should prepare the `DBHashTable' (*note struct DBHashTable::) for the trajectory with `dbh_regen_sweep()' (*note dbh_regen_sweep []::) first. This allows for extremely efficient use of hardware and operating system caches. `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). `operate' : The function to apply to each selected member of the `DBHashTable' (*note struct DBHashTable::) `key1' : The key from which to start the sweep or NULL if you don't care. Make sure it is a top level node of a subtree with `dbh_find()' (*note dbh_find []::) first. `key2' : The key which will trigger an exit condition from the sweep, or NULL if don't care. `ignore_portion' : The ignored trailing bytes of key1 which will define the magnitud of the subtree to be sweeped, or zero if don't care. _Returns_ : 0 on error, 1 otherwise. dbh_foreach () ============== int dbh_foreach (DBHashTable *dbh, DBHashFunc2 operate, void *data); Apply a function to each member of the hash, following a sweep trajectory. Sweep is done by traversing the `DBHashTable' (*note struct DBHashTable::) in a vertical direction through all branches. In order for `dbh_foreach_sweep()' (*note dbh_foreach_sweep []::) to be extremely fast, you should prepare the `DBHashTable' (*note struct DBHashTable::) for the trajectory with `dbh_regen_sweep()' (*note dbh_regen_sweep []::) first. This allows for extremely efficient use of hardware and operating system caches. `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). `operate' : A `DBHashFunc2()' (*note DBHashFunc2 []::) to execute on all records `data' : pointer to data passed to `DBHashFunc2()' (*note DBHashFunc2 []::) _Returns_ : 0 on error, 1 otherwise. dbh_foreach_fanout () ===================== int dbh_foreach_fanout (DBHashTable *dbh, DBHashFunc operate); Apply a function to each member of the hash, following a fanout trajectory (horizontally through records). `dbh_foreach_fanout()' (*note dbh_foreach_fanout []::) is done by traversing the `DBHashTable' (*note struct DBHashTable::) in a horizontal direction through all records. In order for `dbh_foreach_fanout()' (*note dbh_foreach_fanout []::) to be extremely fast, you should prepare the `DBHashTable' (*note struct DBHashTable::) for the trajectory with `dbh_regen_fanout()' (*note dbh_regen_fanout []::) first. This allows for extremely efficient use of hardware and operating system caches. `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). `operate' : A `DBHashFunc()' (*note DBHashFunc []::) to execute on all records _Returns_ : 0 on error, 1 otherwise. dbh_foreach_sweep () ==================== int dbh_foreach_sweep (DBHashTable *dbh, DBHashFunc operate); Apply a function to each member of the hash, following a sweep trajectory. Sweep is done by traversing the `DBHashTable' (*note struct DBHashTable::) in a vertical direction through all branches. In order for `dbh_foreach_sweep()' (*note dbh_foreach_sweep []::) to be extremely fast, you should prepare the `DBHashTable' (*note struct DBHashTable::) for the trajectory with `dbh_regen_sweep()' (*note dbh_regen_sweep []::) first. This allows for extremely efficient use of hardware and operating system caches. `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). `operate' : A `DBHashFunc()' (*note DBHashFunc []::) to execute on all records _Returns_ : 0 on error, 1 otherwise. dbh_exit_fanout () ================== void dbh_exit_fanout (DBHashTable *dbh); Calling this function from within a `DBHashFunc' (*note DBHashFunc []::) will cause an exit of a currently running fanout. `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). dbh_exit_sweep () ================= void dbh_exit_sweep (DBHashTable *dbh); Calling this function from within a `DBHashFunc' (*note DBHashFunc []::) will cause an exit of a currently running sweep. `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). dbh_prune () ============ int dbh_prune (DBHashTable *dbh, unsigned char *key, unsigned char subtree_length); Erases a whole subtree from the record currently loaded into memory. Records are not really removed fisically, but rather marked erased so they may be recovered (if not overwritten later on). Records are permanently removed after `DBHashTable' (*note struct DBHashTable::) is reconstructed with `dbh_regen_sweep()' (*note dbh_regen_sweep []::) or `dbh_regen_fanout()' (*note dbh_regen_fanout []::). `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). `key' : key of top level record of subtree to erase. `subtree_length' : number of branches to erase. _Returns_ : 0 on error, 1 otherwise. dbh_unprune () ============== int dbh_unprune (DBHashTable *dbh, unsigned char *key, unsigned char subtree_length); Does the opposite of `dbh_prune()' (*note dbh_prune []::), marking entire subtree as unerased. May fail to work if records have been overwritten since the `dbh_prune()' (*note dbh_prune []::) instruction was issued. `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). `key' : key of top level record of subtree to erase. `subtree_length' : number of branches to erase. _Returns_ : 0 on error, 1 otherwise. dbh_regen_fanout () =================== void dbh_regen_fanout (DBHashTable **dbh); Regenerate the `DBHashTable' (*note struct DBHashTable::), eliminating erased records and optimizing disk access and speed for fanout access. This is done by creating a new `DBHashTable' (*note struct DBHashTable::) where the physical structure matches the logical fanout structure. The temporary directory where the new `DBHashTable' (*note struct DBHashTable::) is created may be set with `dbh_settempdir()' (*note dbh_settempdir []::). Current `DBHashTable' (*note struct DBHashTable::) is closed before removed. New `DBHashTable' (*note struct DBHashTable::) is opened after renamed. `dbh' : A pointer to a `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). _Returns_ : void. dbh_regen_sweep () ================== void dbh_regen_sweep (DBHashTable **dbh); Regenerate the `DBHashTable' (*note struct DBHashTable::), eliminating erased records and optimizing disk access and speed for sweep access. This is done by creating a new `DBHashTable' (*note struct DBHashTable::) where the physical structure matches the logical sweep structure. The temporary directory where the new `DBHashTable' (*note struct DBHashTable::) is created may be set with `dbh_settempdir()' (*note dbh_settempdir []::). Current `DBHashTable' (*note struct DBHashTable::) is closed before removed. New `DBHashTable' (*note struct DBHashTable::) is opened after renamed. `dbh' : A pointer to a `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). _Returns_ : void. dbh_genkey () ============= void dbh_genkey (unsigned char *key, unsigned char length, unsigned int n); Obtain a key from a secuential series of natural numbers (positive integers without zero) which does not conserve the order of the natural numbers, but which are optimized for construction of a balanced hash tree. These keys are expressed in quantified numbers. Digits are offset to the `0' symbol (+48). `key' : The address where to put the generated key `length' : The key length `n' : The natural number from which to generate the key dbh_genkey0 () ============== void dbh_genkey0 (unsigned char *key, unsigned char length, unsigned int n); Obtain a key from a secuential series of natural numbers (positive integers without zero) which does not conserve the order of the natural numbers, but which are optimized for construction of a balanced hash tree. These keys are expressed in quantified numbers. Digits are not offset. `key' : The address where to put the generated key `length' : The key length `n' : The natural number from which to generate the key dbh_genkey2 () ============== void dbh_genkey2 (unsigned char *key, unsigned char length, unsigned int n); Obtain a key from a secuential series of natural numbers (positive integers without zero) which does not conserve the order of the natural numbers, but which are optimized for construction of a balanced hash tree. These keys are expressed in quantified numbers. Digits are offset to the `A' symbol (+65). `key' : The address where to put the generated key `length' : The key length `n' : The natural number from which to generate the key dbh_orderkey () =============== void dbh_orderkey (unsigned char *key, unsigned char length, unsigned int n, unsigned char base); Obtain a key from a secuential series of natural numbers (positive integers without zero) which conserves the order of the natural numbers. This function generates a key that belongs to a finite subset of the quantified numbers, but which preserves the order of the natural numbers (up to the supreme, of course) `key' : The address where to put the generated key `length' : The key length `n' : The natural number for which to generate the key `base' : The number system base to use. This will equal the maximum number of nodes per branch. This --along with the keylength-- will also define a maximum number of records for the DBHashTable struct dbh_header_t =================== struct dbh_header_t { unsigned char n_limit; unsigned char user_chars[5]; FILE_POINTER bof; FILE_POINTER erased_space; FILE_POINTER data_space; FILE_POINTER total_space; FILE_POINTER records; FILE_POINTER record_length; FILE_POINTER user_filepointer[6]; char version[16]; char copyright[128]; }; `dbh_header_t' (*note struct dbh_header_t::) is the structural information written at the first 256 bytes of a `DBHashTable' (*note struct DBHashTable::) file. unsigned [Cross reference to non-existant ID "char"] `n_limit'; Maximum toplevel branches unsigned [Cross reference to non-existant ID "char"] `user_chars'[5]; Five unsigned chars available to user `FILE_POINTER' (*note FILE_POINTER::) `bof'; File pointer to root of tree `FILE_POINTER' (*note FILE_POINTER::) `erased_space'; Amount of bytes marked as erased `FILE_POINTER' (*note FILE_POINTER::) `data_space'; Amount of bytes ocuppied by data `FILE_POINTER' (*note FILE_POINTER::) `total_space'; Amount of bytes ocuppied by data and format `FILE_POINTER' (*note FILE_POINTER::) `records'; Number of records `FILE_POINTER' (*note FILE_POINTER::) `record_length'; Maximum record length `FILE_POINTER' (*note FILE_POINTER::) `user_filepointer'[6]; Six 64-bit filepointers available to user [Cross reference to non-existant ID "char"] `version'[16]; DBHashTable version compatibility information [Cross reference to non-existant ID "char"] `copyright'[128]; DBH sourcecode distribution copyright and download information dbh_info () =========== int dbh_info (DBHashTable *dbh); Prints header information to stdout. `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). _Returns_ : 0 if error, 1 otherwise dbh_writeheader () ================== int dbh_writeheader (DBHashTable *dbh); Write out the DBHashTable header information. It is advisable to call this function inmediately after creation of a new DBHashTable to force a buffer flush. `dbh' : A `DBHashTable' (*note struct DBHashTable::) pointer (`DBHashTable' (*note struct DBHashTable::) *). _Returns_ : 0 if error, 1 otherwise See Also ******** [Cross reference to non-existant ID "GHashTables"]  Local Variables: coding: US-ASCII End: