Advanced Programming in the UNIX® Environment

ptg10805159

Section 20.8 Source Code 767

348 /*

349 * Now read the actual index record. We read it into the key

350 * buffer that we malloced when we opened the database.

351 */

352 if ((i = read(db->idxfd, db->idxbuf, db->idxlen)) != db->idxlen)

353 err_dump("_db_readidx: read error of index record");

354 if (db->idxbuf[db->idxlen-1] != NEWLINE) /* sanity check */

355 err_dump("_db_readidx: missing newline");

356 db->idxbuf[db->idxlen-1]=0;/*replace newline with null */

357 /*

358 * Find the separators in the index record.

359 */

360 if ((ptr1 = strchr(db->idxbuf, SEP)) == NULL)

361 err_dump("_db_readidx: missing first separator");

362 *ptr1++=0;/*replace SEP with null */

363 if ((ptr2 = strchr(ptr1, SEP)) == NULL)

364 err_dump("_db_readidx: missing second separator");

365 *ptr2++=0;/*replace SEP with null */

366 if (strchr(ptr2, SEP) != NULL)

367 err_dump("_db_readidx: too many separators");

368 /*

369 * Get the starting offset and length of the data record.

370 */

371 if ((db->datoff = atol(ptr1)) < 0)

372 err_dump("_db_readidx: starting offset < 0");

373 if ((db->datlen = atol(ptr2)) <= 0 || db->datlen > DATLEN_MAX)

374 err_dump("_db_readidx: invalid length");

375 return(db->ptrval); /* return offset of next key in chain */

376 }

[348 – 356] We read the variable-length index recordinto theidxbuffield in theDB

structure. Therecordshould be terminated with a newline, which we

replace with a null byte. If the index file is corrupt, we terminate and drop

core by callingerr_dump.

[357 – 367] We separate the index recordinto three fields: the key,the offset of the

corresponding data record, and the length of the data record. Thestrchr

function finds the first occurrence of the specified character in the given

string. Here we look for the character that separates fields in the record

(SEP,which we define to be a colon).

[368 – 376] We convert the data recordoffset and length into integers and storethem in

theDBstructure. Then we return the offset of the next record in the hash

chain. Note that we do not read the data record; that task is left to the caller.

In thedb_fetchfunction, for example, we don’t read the data recorduntil

_db_find_and_lockhas read the index recordthat matches the key that

we’relooking for.