PureDB is a portable and tiny set of libraries for creating and reading constant databases.
It manages data files that contains text or binary key/data pairs of arbitrary sizes. Lookups are very fast (normally only one disk access to match a hash value), overhead is low (a database is 1028 bytes plus only 16 extra bytes per record), multiple concurrent read access are supported, and databases can be up to 4 Gb long, and they are portable accross architectures.
This is a standalone distribution of Pure-FTPd's built-in database.
Compiling PureDB is a simple matter of:
./configure
make installStatic libraries, shared libraries and links are installed in
/usr/local/lib/libpuredb_read* and /usr/local/lib/libpuredb_write*.
Header files are installed as /usr/local/include/puredb_read.h and
/usr/local/include/puredb_write.h.
The library that creates databases is different from the library that reads them, because these different tasks are usually handled by separate applications. However, you can safely link both libraries together.
To compile a program with puredb, you have to include the following headers:
#include <puredb_read.h>and/or
#include <puredb_write.h>If your application only reads PureDB databases, just include the first header. If it only writes databases, just include the second one. It it does both, include both.
The same thing goes for linker options: you have to link against
libpuredb_read and/or libpuredb_write:
cc -o myapp1 myapp1.c -lpuredb_read
cc -o myapp2 myapp2.c -lpuredb_write
cc -o myapp3 myapp3.c -lpuredb_read -lpuredb_writeCreating a new database is usually a 4-step operation:
-
Create the database files and initialize the internal structures with
puredbw_open(). -
Feed key/data pairs with
puredbw_add()orpuredbw_add_s(). -
Complete and close the database files with
puredbw_close(). -
Free the internal structures with
puredbw_free().
Here are the functions:
int puredbw_open(PureDBW * const dbw,
const char * const file_index,
const char * const file_data,
const char * const file_final);This function takes a point to an already allocated PureDBW structure, and
three file names. file_index and file_data are temporary files, needed to
create the database. They will be automatically deleted, and the final
database will atomically be stored in file_final.
Return value: 0 if everything is ok, a negative value if something went wrong.
int puredbw_add(PureDBW * const dbw,
const char * const key, const size_t key_len,
const char * const content, const size_t content_len);This function stores a new key/data pair in the database. key is a pointer
to the key, that is key_len long. Same thing for content and content_len.
These buffers can handle binary data, and can have any size up to 4 Gb.
Return value: 0 if everything is ok, a negative value if something went wrong.
int puredbw_add_s(PureDBW * const dbw,
const char * const key, const char * const content);This function is a shortcut to puredbw_add(), designed to store \0-terminated
strings. It's equivalent to call puredbw_add() with strlen(key) and
strlen(content) as parameters 3 and 5.
Return value: 0 if everything is ok, a negative value if something went wrong.
int puredbw_close(PureDBW * const dbw);This function performs a quick sort of the hashed values, writes them to the disk, merges index and data files, rename the result to the final file name and delete the old files. You must call this after having inserted all values in the database.
Return value: 0 if everything is ok, a negative value if something went wrong.
void puredbw_free(PureDBW * const dbw);This function frees all memory chunks allocated by puredbw_open(),
puredbw_add() and puredbw_add_s(). You must call this either after a
puredbw_close(), or after something went wrong if you decide to abort.
Here's an example, that creates a new database, and inserts three key/data pairs into it.
#include <stdio.h>
#include <stdlib.h>
#include <puredb_write.h>
int main(void)
{
PureDBW dbw;
if (puredbw_open(&dbw, "puredb.index", "puredb.data", "puredb.pdb") != 0) {
perror("Can't create the database");
} else if (puredbw_add_s(&dbw, "key", "content") != 0 ||
puredbw_add_s(&dbw, "key2", "content2") != 0 ||
puredbw_add_s(&dbw, "key42", "content42") != 0) {
perror("Error while inserting key/data pairs");
}
if (puredbw_close(&dbw) != 0) {
perror("Error while closing the database");
}
puredbw_free(&dbw);
return 0;
}Reading the content of a database is usually a 5-step operation:
-
Open the database files and initialize the internal structures with
puredb_open(). -
Perform a lookup for a key with
puredb_find()orpuredb_find_s(). -
If the key is found, read the associated data with
puredb_read().
[process the data]
- Free the data with
puredb_read_free().
[repeat steps 2, 3 and 4 to read more key/data pairs]
- Close the database and free the internal structures with
puredb_close().
Here are the functions:
int puredb_open(PureDB * const db, const char *dbfile);This function opens an existing database, stored in a file named dbfile, and
initializes a preallocated PureDB structure.
Return value: 0 if everything is ok, a negative value if something went wrong.
int puredb_find(PureDB * const db, const char * const tofind,
const size_t tofind_len, off_t * const retpos,
size_t * const retlen);This function looks the database for a key matching tofind, whoose length is
tofind_len. After a successful match, retpos contains the offset to the
first byte of the matching data, and retlen is the length of the data.
Return value: 0 if the key was found.
-1 if the key was not found.
-2 if the database is corrupted.
-3 if a system error occured.
int puredb_find_s(PureDB * const db, const char * const tofind,
off_t * const retpos, size_t * const retlen);This function is a shortcut to puredb_find() for text keys, which computes
strlen(tofind) as a key length.
Return value: 0 if the key was found.
-1 if the key was not found.
-2 if the database is corrupted.
-3 if a system error occured.
void *puredb_read(PureDB * const db, const off_t offset, const size_t len);This function reads len bytes in the database file, starting at offset. A
large enough buffer is allocated, filled and returned. It is guaranteed to
be terminated by an extra \0, so it's safe to process C-strings returned by
that function.
Return value: the address of a buffer with the data, or NULL if something
went wrong (no memory, corrupted file, no permission, etc) .
void puredb_read_free(void *data);Frees a buffer allocated by puredb_read().
int puredb_getfd(PureDB * const db);Returns the file descriptor opened for the database, just in case you want to read the data by yourself. It can be interesting if the data is too large to be stored in memory.
Return value: a file descriptor. or -1 if none was allocated (error).
off_t puredb_getsize(PureDB * const db);This function returns the size of the file handling the database, in bytes.
Return value: the size of the file.
int puredb_close(PureDB * const db);This function closes the database and frees all related internal structures. Don't forget to call this even if something went wrong, and you decide to abort.
Here's an example, that reads a previously created database.
#include <stdio.h>
#include <stdlib.h>
#include <puredb_read.h>
int main(void)
{
PureDB db;
off_t retpos;
size_t retlen;
char *data;
if (puredb_open(&db, "puredb.pdb") != 0) {
perror("Can't open the database");
} else if (puredb_find_s(&db, "key42", &retpos, &retlen) != 0) {
fprintf(stderr, "The key wasn't found\n");
} else if ((data = puredb_read(&db, retpos, retlen)) != NULL) {
printf("The maching data are: [%s]\n", data);
puredb_read_free(data);
}
if (puredb_close(&db) != 0) {
perror("The database couldn't be properly closed");
}
return 0;
}PureDB home page is: https://github.com/jedisct1/PureDB
Thank you,
-Frank DENIS <j at pureftpd dot org>.