DBH_CREATE

#define             DBH_CREATE

Bit flag for 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 data area.

DBH_DATA_SPACE()

#define             DBH_DATA_SPACE(dbh)

This macro returns the amount of bytes taken up by valid data in the DBHashTable.

DBH_ERASED_SPACE()

#define             DBH_ERASED_SPACE(dbh)

This macro returns the amount of bytes taken up by erased data in the 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.

DBH_KEY()

#define             DBH_KEY(dbh)

This macro returns a pointer to the current DBHashTable key area.

DBH_KEYLENGTH()

#define             DBH_KEYLENGTH(dbh)

This macro returns the keylenth in bytes associated to the DBHashTable. The value is fixed when the DBHashTable is created with dbh_new.

DBH_MAXIMUM_RECORD_SIZE()

#define             DBH_MAXIMUM_RECORD_SIZE(dbh)

This macro returns the maximum allocated space for data in the current DBHashTable record.

DBH_PARALLEL_SAFE

#define             DBH_PARALLEL_SAFE

Bit flag for dbh_new() to use if more than one heavy weight process will be accessing the same DBHashTable in write mode. If no process will be writing to the DBHashTable, then DBH_READ_ONLY is enough and faster since each process will hold a separate memory allocation for the DBHashTable pointer.

DBH_PATH()

#define             DBH_PATH(dbh)

This macro returns a pointer to a string containing the path to the current DBHashTable.

DBH_READ_ONLY

#define             DBH_READ_ONLY

Bit flag for 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.

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_THREAD_SAFE

#define             DBH_THREAD_SAFE

Bit flag for dbh_new() to use if more than one thread will be accessing the same 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() and dbh_mutex_unlock(). Each DBH table opened with the DBH_THREAD_SAFE attribute will have a specific mutex for this function. If threads are to access the same DBHashTable in read mode only, then DBH_READ_ONLY and separate memory allocation for each thread's DBHashTable pointer is more than enough and faster.

When DBH_THREAD_SAFE is specified, dbh_new() is automatically mutex locked until function completes. The function dbh_close() is also automatically locked until completion on tables opened with the 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.

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(), dbh_fanout(), dbh_foreach_sweep() and dbh_foreach_fanout().

This function will be applied to all data records involved in the sweep or fanout process

DBHashFunc2 ()

void                (*DBHashFunc2)                      (DBHashTable *dbh,
                                                         void *data);

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 is a data structure containing the record information for an open DBHashTable file.

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, DBH_READ_ONLY, DBH_THREAD_SAFE, DBH_PARALLEL_SAFE. (since 4.7.6)

dbh_open ()

DBHashTable *       dbh_open                            (const char *path);

Open an existing hash in read-write mode.

dbh_open_ro ()

DBHashTable *       dbh_open_ro                         (const char *path);

Open an existing hash in read-only mode.

dbh_create ()

DBHashTable *       dbh_create                          (const char *path,
                                                         unsigned char key_length);

Create a new hash file (overwriting old version). Creates and opens for writing a new 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 genkey(), and create an extra DBHashTable to save the g_hash. Quantified keys assure that large DBHashes are spread out optimally.

dbh_close ()

int                 dbh_close                           (DBHashTable *dbh);

Close hash file (thus flushing io buffer).

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_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_unerase ()

int                 dbh_unerase                         (DBHashTable *dbh);

This is the opposite of dbh_erase(). Mark the record currently loaded into memory as unerased. If no record is currently loaded, behaviour is undefined.

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_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_load_address ()

unsigned char       dbh_load_address                    (DBHashTable *dbh,
                                                         FILE_POINTER currentseek);

Load a record from hash table directly from byte offset currentseek

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 branches. Lower level records have less. Each byte of a key represents a branch on top level records.

dbh_load_parent ()

FILE_POINTER        dbh_load_parent                     (DBHashTable *dbh);

Load the parent of the currently loaded record.

dbh_set_data ()

void                dbh_set_data                        (DBHashTable *dbh,
                                                         void *data,
                                                         FILE_POINTER size);

This function copies the user data into the current DBHashTable record and along with function dbh_set_key(), makes the current DBHashTable record ready for the dbh_update() function to commit to the actual DBHashTable on disk.

dbh_set_key ()

void                dbh_set_key                         (DBHashTable *dbh,
                                                         unsigned char *key);

This function sets the key of the current DBHashTable record.

dbh_set_recordsize ()

void                dbh_set_recordsize                  (DBHashTable *dbh,
                                                         int record_size);

This sets the recordsize of the the data in the current DBHashTable record. It is called implicitly by calling dbh_set_data(). It is very important to call this function. Unpredictable results will follow if record_size is not set. 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() is used to set the record data.

dbh_set_size ()

int                 dbh_set_size                        (DBHashTable *dbh,
                                                         FILE_POINTER size);

Defines the maximum amount of memory to be allocated to the DBHashTable records. This is nonvolatile information which need to be set only once. The default is 1Kbyte.

dbh_settempdir ()

int                 dbh_settempdir                      (DBHashTable *dbh,
                                                         char *temp_dir);

Sets the temporary directory to be used by dbh_regen_sweep() or 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_lock_t

typedef struct {
    pid_t write_lock;
    int write_lock_count;
    int read_lock_count;
} dbh_lock_t;

dbh_clear_locks ()

int                 dbh_clear_locks                     (DBHashTable *dbh);

Returns: 0 if error, 1 otherwise

Clear dbh file locks associated to DBHashTable Use this function to clean up persistent file locks

(since 4.7.6)

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() as well.

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.

dbh_set_parallel_lock_attempt_limit ()

int                 dbh_set_parallel_lock_attempt_limit (DBHashTable *dbh,
                                                         int limit);

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_set_parallel_lock_timeout ()

int                 dbh_set_parallel_lock_timeout       (DBHashTable *dbh,
                                                         int seconds);

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() is set to zero (that's the default) this function will block until lock is secured.

dbh_unlock_read ()

int                 dbh_unlock_read                     (DBHashTable *dbh);

Releases a read lock on the dbh file.

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() is set to zero (that's the default) this function will block until lock is secured.

dbh_unlock_write ()

int                 dbh_unlock_write                    (DBHashTable *dbh);

Releases a write lock on the dbh file.

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_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_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_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() to be extremely fast, you should prepare the DBHashTable for the trajectory with dbh_regen_fanout() first. This allows for extremely efficient use of hardware and operating system caches.

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() to be extremely fast, you should prepare the DBHashTable for the trajectory with dbh_regen_sweep() first. This allows for extremely efficient use of hardware and operating system caches.

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 in a vertical direction through all branches.

In order for dbh_foreach_sweep() to be extremely fast, you should prepare the DBHashTable for the trajectory with dbh_regen_sweep() first. This allows for extremely efficient use of hardware and operating system caches.

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() is done by traversing the DBHashTable in a horizontal direction through all records.

In order for dbh_foreach_fanout() to be extremely fast, you should prepare the DBHashTable for the trajectory with dbh_regen_fanout() first. This allows for extremely efficient use of hardware and operating system caches.

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 in a vertical direction through all branches.

In order for dbh_foreach_sweep() to be extremely fast, you should prepare the DBHashTable for the trajectory with dbh_regen_sweep() first. This allows for extremely efficient use of hardware and operating system caches.

dbh_exit_fanout ()

void                dbh_exit_fanout                     (DBHashTable *dbh);

Calling this function from within a DBHashFunc will cause an exit of a currently running fanout.

dbh_exit_sweep ()

void                dbh_exit_sweep                      (DBHashTable *dbh);

Calling this function from within a DBHashFunc will cause an exit of a currently running sweep.

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 is reconstructed with dbh_regen_sweep() or dbh_regen_fanout().

dbh_unprune ()

int                 dbh_unprune                         (DBHashTable *dbh,
                                                         unsigned char *key,
                                                         unsigned char subtree_length);

Does the opposite of dbh_prune(), marking entire subtree as unerased. May fail to work if records have been overwritten since the dbh_prune() instruction was issued.

dbh_regen_fanout ()

void                dbh_regen_fanout                    (DBHashTable **dbh);

Regenerate the DBHashTable, eliminating erased records and optimizing disk access and speed for fanout access. This is done by creating a new DBHashTable where the physical structure matches the logical fanout structure. The temporary directory where the new DBHashTable is created may be set with dbh_settempdir(). Current DBHashTable is closed before removed. New DBHashTable is opened after renamed.

dbh_regen_sweep ()

void                dbh_regen_sweep                     (DBHashTable **dbh);

Regenerate the DBHashTable, eliminating erased records and optimizing disk access and speed for sweep access. This is done by creating a new DBHashTable where the physical structure matches the logical sweep structure. The temporary directory where the new DBHashTable is created may be set with dbh_settempdir(). Current DBHashTable is closed before removed. New DBHashTable is opened after renamed.

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).

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.

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).

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)

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 is the structural information written at the first 256 bytes of a DBHashTable file.

dbh_info ()

int                 dbh_info                            (DBHashTable *dbh);

Prints header information to stdout.

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.