📄 hash_dup.c
字号:
/*- * See the file LICENSE for redistribution information. * * Copyright (c) 1996-2002 * Sleepycat Software. All rights reserved. *//* * Copyright (c) 1990, 1993, 1994 * The Regents of the University of California. All rights reserved. * * This code is derived from software contributed to Berkeley by * Margo Seltzer. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */#include "db_config.h"#ifndef lintstatic const char revid[] = "$Id: hash_dup.c,v 11.76 2002/08/06 05:34:40 bostic Exp $";#endif /* not lint *//* * PACKAGE: hashing * * DESCRIPTION: * Manipulation of duplicates for the hash package. */#ifndef NO_SYSTEM_INCLUDES#include <sys/types.h>#include <string.h>#endif#include "db_int.h"#include "dbinc/db_page.h"#include "dbinc/hash.h"#include "dbinc/btree.h"static int __ham_c_chgpg __P((DBC *, db_pgno_t, u_int32_t, db_pgno_t, u_int32_t));static int __ham_check_move __P((DBC *, u_int32_t));static int __ham_dcursor __P((DBC *, db_pgno_t, u_int32_t));static int __ham_move_offpage __P((DBC *, PAGE *, u_int32_t, db_pgno_t));/* * Called from hash_access to add a duplicate key. nval is the new * value that we want to add. The flags correspond to the flag values * to cursor_put indicating where to add the new element. * There are 4 cases. * Case 1: The existing duplicate set already resides on a separate page. * We return and let the common code handle this. * Case 2: The element is small enough to just be added to the existing set. * Case 3: The element is large enough to be a big item, so we're going to * have to push the set onto a new page. * Case 4: The element is large enough to push the duplicate set onto a * separate page. * * PUBLIC: int __ham_add_dup __P((DBC *, DBT *, u_int32_t, db_pgno_t *)); */int__ham_add_dup(dbc, nval, flags, pgnop) DBC *dbc; DBT *nval; u_int32_t flags; db_pgno_t *pgnop;{ DB *dbp; DBT pval, tmp_val; DB_MPOOLFILE *mpf; HASH_CURSOR *hcp; u_int32_t add_bytes, new_size; int cmp, ret; u_int8_t *hk; dbp = dbc->dbp; mpf = dbp->mpf; hcp = (HASH_CURSOR *)dbc->internal; DB_ASSERT(flags != DB_CURRENT); add_bytes = nval->size + (F_ISSET(nval, DB_DBT_PARTIAL) ? nval->doff : 0); add_bytes = DUP_SIZE(add_bytes); if ((ret = __ham_check_move(dbc, add_bytes)) != 0) return (ret); /* * Check if resulting duplicate set is going to need to go * onto a separate duplicate page. If so, convert the * duplicate set and add the new one. After conversion, * hcp->dndx is the first free ndx or the index of the * current pointer into the duplicate set. */ hk = H_PAIRDATA(dbp, hcp->page, hcp->indx); /* Add the len bytes to the current singleton. */ if (HPAGE_PTYPE(hk) != H_DUPLICATE) add_bytes += DUP_SIZE(0); new_size = LEN_HKEYDATA(dbp, hcp->page, dbp->pgsize, H_DATAINDEX(hcp->indx)) + add_bytes; /* * We convert to off-page duplicates if the item is a big item, * the addition of the new item will make the set large, or * if there isn't enough room on this page to add the next item. */ if (HPAGE_PTYPE(hk) != H_OFFDUP && (HPAGE_PTYPE(hk) == H_OFFPAGE || ISBIG(hcp, new_size) || add_bytes > P_FREESPACE(dbp, hcp->page))) { if ((ret = __ham_dup_convert(dbc)) != 0) return (ret); return (hcp->opd->c_am_put(hcp->opd, NULL, nval, flags, NULL)); } /* There are two separate cases here: on page and off page. */ if (HPAGE_PTYPE(hk) != H_OFFDUP) { if (HPAGE_PTYPE(hk) != H_DUPLICATE) { pval.flags = 0; pval.data = HKEYDATA_DATA(hk); pval.size = LEN_HDATA(dbp, hcp->page, dbp->pgsize, hcp->indx); if ((ret = __ham_make_dup(dbp->dbenv, &pval, &tmp_val, &dbc->my_rdata.data, &dbc->my_rdata.ulen)) != 0 || (ret = __ham_replpair(dbc, &tmp_val, 1)) != 0) return (ret); hk = H_PAIRDATA(dbp, hcp->page, hcp->indx); HPAGE_PTYPE(hk) = H_DUPLICATE; /* * Update the cursor position since we now are in * duplicates. */ F_SET(hcp, H_ISDUP); hcp->dup_off = 0; hcp->dup_len = pval.size; hcp->dup_tlen = DUP_SIZE(hcp->dup_len); } /* Now make the new entry a duplicate. */ if ((ret = __ham_make_dup(dbp->dbenv, nval, &tmp_val, &dbc->my_rdata.data, &dbc->my_rdata.ulen)) != 0) return (ret); tmp_val.dlen = 0; switch (flags) { /* On page. */ case DB_KEYFIRST: case DB_KEYLAST: case DB_NODUPDATA: if (dbp->dup_compare != NULL) { __ham_dsearch(dbc, nval, &tmp_val.doff, &cmp, flags); /* dup dups are not supported w/ sorted dups */ if (cmp == 0) return (__db_duperr(dbp, flags)); } else { hcp->dup_tlen = LEN_HDATA(dbp, hcp->page, dbp->pgsize, hcp->indx); hcp->dup_len = nval->size; F_SET(hcp, H_ISDUP); if (flags == DB_KEYFIRST) hcp->dup_off = tmp_val.doff = 0; else hcp->dup_off = tmp_val.doff = hcp->dup_tlen; } break; case DB_BEFORE: tmp_val.doff = hcp->dup_off; break; case DB_AFTER: tmp_val.doff = hcp->dup_off + DUP_SIZE(hcp->dup_len); break; } /* Add the duplicate. */ ret = __ham_replpair(dbc, &tmp_val, 0); if (ret == 0) ret = mpf->set(mpf, hcp->page, DB_MPOOL_DIRTY); if (ret != 0) return (ret); /* Now, update the cursor if necessary. */ switch (flags) { case DB_AFTER: hcp->dup_off += DUP_SIZE(hcp->dup_len); hcp->dup_len = nval->size; hcp->dup_tlen += (db_indx_t)DUP_SIZE(nval->size); break; case DB_KEYFIRST: case DB_KEYLAST: case DB_BEFORE: hcp->dup_tlen += (db_indx_t)DUP_SIZE(nval->size); hcp->dup_len = nval->size; break; } ret = __ham_c_update(dbc, tmp_val.size, 1, 1); return (ret); } /* * If we get here, then we're on duplicate pages; set pgnop and * return so the common code can handle it. */ memcpy(pgnop, HOFFDUP_PGNO(H_PAIRDATA(dbp, hcp->page, hcp->indx)), sizeof(db_pgno_t)); return (ret);}/* * Convert an on-page set of duplicates to an offpage set of duplicates. * * PUBLIC: int __ham_dup_convert __P((DBC *)); */int__ham_dup_convert(dbc) DBC *dbc;{ BOVERFLOW bo; DB *dbp; DBC **hcs; DBT dbt; DB_LSN lsn; DB_MPOOLFILE *mpf; HASH_CURSOR *hcp; HOFFPAGE ho; PAGE *dp; db_indx_t i, len, off; int c, ret, t_ret; u_int8_t *p, *pend; dbp = dbc->dbp; mpf = dbp->mpf; hcp = (HASH_CURSOR *)dbc->internal; /* * Create a new page for the duplicates. */ if ((ret = __db_new(dbc, dbp->dup_compare == NULL ? P_LRECNO : P_LDUP, &dp)) != 0) return (ret); P_INIT(dp, dbp->pgsize, dp->pgno, PGNO_INVALID, PGNO_INVALID, LEAFLEVEL, TYPE(dp)); /* * Get the list of cursors that may need to be updated. */ if ((ret = __ham_get_clist(dbp, PGNO(hcp->page), (u_int32_t)hcp->indx, &hcs)) != 0) goto err; /* * Now put the duplicates onto the new page. */ dbt.flags = 0; switch (HPAGE_PTYPE(H_PAIRDATA(dbp, hcp->page, hcp->indx))) { case H_KEYDATA: /* Simple case, one key on page; move it to dup page. */ dbt.size = LEN_HDATA(dbp, hcp->page, dbp->pgsize, hcp->indx); dbt.data = HKEYDATA_DATA(H_PAIRDATA(dbp, hcp->page, hcp->indx)); ret = __db_pitem(dbc, dp, 0, BKEYDATA_SIZE(dbt.size), NULL, &dbt); goto finish; case H_OFFPAGE: /* Simple case, one key on page; move it to dup page. */ memcpy(&ho, P_ENTRY(dbp, hcp->page, H_DATAINDEX(hcp->indx)), HOFFPAGE_SIZE); UMRW_SET(bo.unused1); B_TSET(bo.type, ho.type, 0); UMRW_SET(bo.unused2); bo.pgno = ho.pgno; bo.tlen = ho.tlen; dbt.size = BOVERFLOW_SIZE; dbt.data = &bo; ret = __db_pitem(dbc, dp, 0, dbt.size, &dbt, NULL);finish: if (ret == 0) { if ((ret = mpf->set(mpf, dp, DB_MPOOL_DIRTY)) != 0) break; /* Update any other cursors. */ if (hcs != NULL && DBC_LOGGING(dbc) && IS_SUBTRANSACTION(dbc->txn)) { if ((ret = __ham_chgpg_log(dbp, dbc->txn, &lsn, 0, DB_HAM_DUP, PGNO(hcp->page), PGNO(dp), hcp->indx, 0)) != 0) break; } for (c = 0; hcs != NULL && hcs[c] != NULL; c++) if ((ret = __ham_dcursor(hcs[c], PGNO(dp), 0)) != 0) break; } break; case H_DUPLICATE: p = HKEYDATA_DATA(H_PAIRDATA(dbp, hcp->page, hcp->indx)); pend = p + LEN_HDATA(dbp, hcp->page, dbp->pgsize, hcp->indx); /* * We need to maintain the duplicate cursor position. * Keep track of where we are in the duplicate set via * the offset, and when it matches the one in the cursor, * set the off-page duplicate cursor index to the current * index. */ for (off = 0, i = 0; p < pend; i++) { memcpy(&len, p, sizeof(db_indx_t)); dbt.size = len; p += sizeof(db_indx_t); dbt.data = p; p += len + sizeof(db_indx_t); if ((ret = __db_pitem(dbc, dp, i, BKEYDATA_SIZE(dbt.size), NULL, &dbt)) != 0) break; /* Update any other cursors */ if (hcs != NULL && DBC_LOGGING(dbc) && IS_SUBTRANSACTION(dbc->txn)) { if ((ret = __ham_chgpg_log(dbp, dbc->txn, &lsn, 0, DB_HAM_DUP, PGNO(hcp->page), PGNO(dp), hcp->indx, i)) != 0) break; } for (c = 0; hcs != NULL && hcs[c] != NULL; c++) if (((HASH_CURSOR *)(hcs[c]->internal))->dup_off == off && (ret = __ham_dcursor(hcs[c], PGNO(dp), i)) != 0) goto err; off += len + 2 * sizeof(db_indx_t); } break; default: ret = __db_pgfmt(dbp->dbenv, (u_long)hcp->pgno); break; } /* * Now attach this to the source page in place of the old duplicate * item. */ if (ret == 0) ret = __ham_move_offpage(dbc, hcp->page, (u_int32_t)H_DATAINDEX(hcp->indx), PGNO(dp));err: if (ret == 0) ret = mpf->set(mpf, hcp->page, DB_MPOOL_DIRTY); if ((t_ret = mpf->put(mpf, dp, ret == 0 ? DB_MPOOL_DIRTY : 0)) != 0 && ret == 0) ret = t_ret; if (ret == 0) hcp->dup_tlen = hcp->dup_off = hcp->dup_len = 0; if (hcs != NULL) __os_free(dbp->dbenv, hcs); return (ret);}/* * __ham_make_dup * * Take a regular dbt and make it into a duplicate item with all the partial * information set appropriately. If the incoming dbt is a partial, assume * we are creating a new entry and make sure that we do any initial padding. * * PUBLIC: int __ham_make_dup __P((DB_ENV *, * PUBLIC: const DBT *, DBT *d, void **, u_int32_t *)); */int__ham_make_dup(dbenv, notdup, duplicate, bufp, sizep) DB_ENV *dbenv; const DBT *notdup; DBT *duplicate; void **bufp; u_int32_t *sizep;{ db_indx_t tsize, item_size; int ret; u_int8_t *p; item_size = (db_indx_t)notdup->size; if (F_ISSET(notdup, DB_DBT_PARTIAL)) item_size += notdup->doff; tsize = DUP_SIZE(item_size); if ((ret = __ham_init_dbt(dbenv, duplicate, tsize, bufp, sizep)) != 0) return (ret); duplicate->dlen = 0; duplicate->flags = notdup->flags; F_SET(duplicate, DB_DBT_PARTIAL); p = duplicate->data; memcpy(p, &item_size, sizeof(db_indx_t)); p += sizeof(db_indx_t); if (F_ISSET(notdup, DB_DBT_PARTIAL)) { memset(p, 0, notdup->doff); p += notdup->doff; } memcpy(p, notdup->data, notdup->size); p += notdup->size; memcpy(p, &item_size, sizeof(db_indx_t)); duplicate->doff = 0; duplicate->dlen = notdup->size; return (0);}/* * __ham_check_move -- * * Check if we can do whatever we need to on this page. If not, * then we'll have to move the current element to a new page. */static int
⌨️ 快捷键说明
复制代码
Ctrl + C
搜索代码
Ctrl + F
全屏模式
F11
切换主题
Ctrl + Shift + D
显示快捷键
?
增大字号
Ctrl + =
减小字号
Ctrl + -