Advanced Programming in the UNIX® Environment

ptg10805159

Section 20.8 Source Code 775

597 /*

598 * _db_find_and_lock locked the hash chain for us; read

599 * the chain ptr to the first index record on hash chain.

600 */

601 ptrval=_db_readptr(db, db->chainoff);

602 if (_db_findfree(db, keylen, datlen) < 0) {

603 /*

604 * Can’t find an empty record big enough. Append the

605 * new record to the ends of the index and data files.

606 */

607 _db_writedat(db, data, 0, SEEK_END);

608 _db_writeidx(db, key, 0, SEEK_END, ptrval);

609 /*

610 * db->idxoff was set by _db_writeidx. The new

611 * record goes to the front of the hash chain.

612 */

613 _db_writeptr(db, db->chainoff, db->idxoff);

614 db->cnt_stor1++;

615 } else {

616 /*

617 * Reuse an empty record. _db_findfree removed it from

618 * the free list and set both db->datoff and db->idxoff.

619 * Reused record goes to the front of the hash chain.

620 */

621 _db_writedat(db, data, db->datoff, SEEK_SET);

622 _db_writeidx(db, key, db->idxoff, SEEK_SET, ptrval);

623 _db_writeptr(db, db->chainoff, db->idxoff);

624 db->cnt_stor2++;

625 }

[597 – 601] After we call_db_find_and_lock,the code divides into four cases. In the

first two, no recordwas found, so we areadding a new record. Weread the

offset of the first entry on the hash list.

[602 – 614] Case 1: we call_db_findfreeto search the free list for a deleted record

with the same size key and same size data. If no such record is found, we

have to append the new record to the ends of the index and data files. We

call_db_writedatto write the data part,_db_writeidxto write the

index part, and_db_writeptrto place the new record on the front of the

hash chain. We increment a count (cnt_stor1) of the number of times we

executed this case to allow us to characterize the behavior of the database.

[615 – 625] Case 2:_db_findfreefound an empty recordwith the correct sizes and

removed it from the free list (we’ll see the implementation of

_db_findfreeshortly). Wewrite the data and index portions of the new

recordand add the record to the front of the hash chain as we did in case 1.

Thecnt_stor2field counts how many times we’ve executed this case.