/*
* The GiST PickSplit method
*
* New linear algorithm, see 'New Linear Node Splitting Algorithm for R-tree',
* C.H.Ang and T.C.Tan
*/
Datum
gist_box_picksplit(PG_FUNCTION_ARGS)
{
GistEntryVector *entryvec = (GistEntryVector *) PG_GETARG_POINTER(0);
GIST_SPLITVEC *v = (GIST_SPLITVEC *) PG_GETARG_POINTER(1);
OffsetNumber i;
OffsetNumber *listL,
*listR,
*listB,
*listT;
BOX *unionL,
*unionR,
*unionB,
*unionT;
int posL,
posR,
posB,
posT;
BOX pageunion;
BOX *cur;
char direction = ' ';
bool allisequal = true;
OffsetNumber maxoff;
int nbytes;
posL = posR = posB = posT = 0;
maxoff = entryvec->n - 1;
cur = DatumGetBoxP(entryvec->vector[FirstOffsetNumber].key);
memcpy((void *) &pageunion, (void *) cur, sizeof(BOX));
/* find MBR */
for (i = OffsetNumberNext(FirstOffsetNumber); i <= maxoff; i = OffsetNumberNext(i))
{
cur = DatumGetBoxP(entryvec->vector[i].key);
if (allisequal == true && (
pageunion.high.x != cur->high.x ||
pageunion.high.y != cur->high.y ||
pageunion.low.x != cur->low.x ||
pageunion.low.y != cur->low.y
))
allisequal = false;
adjustBox(&pageunion, cur);
}
if (allisequal)
{
/*
* All entries are the same
*/
fallbackSplit(entryvec, v);
PG_RETURN_POINTER(v);
}
nbytes = (maxoff + 2) * sizeof(OffsetNumber);
listL = (OffsetNumber *) palloc(nbytes);
listR = (OffsetNumber *) palloc(nbytes);
listB = (OffsetNumber *) palloc(nbytes);
listT = (OffsetNumber *) palloc(nbytes);
unionL = (BOX *) palloc(sizeof(BOX));
unionR = (BOX *) palloc(sizeof(BOX));
unionB = (BOX *) palloc(sizeof(BOX));
unionT = (BOX *) palloc(sizeof(BOX));
#define ADDLIST( list, unionD, pos, num ) do { \
if ( pos ) { \
if ( (unionD)->high.x < cur->high.x ) (unionD)->high.x = cur->high.x; \
if ( (unionD)->low.x > cur->low.x ) (unionD)->low.x = cur->low.x; \
if ( (unionD)->high.y < cur->high.y ) (unionD)->high.y = cur->high.y; \
if ( (unionD)->low.y > cur->low.y ) (unionD)->low.y = cur->low.y; \
} else { \
memcpy( (void*)(unionD), (void*) cur, sizeof( BOX ) ); \
} \
(list)[pos] = num; \
(pos)++; \
} while(0)
for (i = FirstOffsetNumber; i <= maxoff; i = OffsetNumberNext(i))
{
cur = DatumGetBoxP(entryvec->vector[i].key);
if (cur->low.x - pageunion.low.x < pageunion.high.x - cur->high.x)
ADDLIST(listL, unionL, posL, i);
else
ADDLIST(listR, unionR, posR, i);
if (cur->low.y - pageunion.low.y < pageunion.high.y - cur->high.y)
ADDLIST(listB, unionB, posB, i);
else
ADDLIST(listT, unionT, posT, i);
}
#define LIMIT_RATIO 0.1
#define _IS_BADRATIO(x,y) ( (y) == 0 || (float)(x)/(float)(y) < LIMIT_RATIO )
#define IS_BADRATIO(x,y) ( _IS_BADRATIO((x),(y)) || _IS_BADRATIO((y),(x)) )
/* bad disposition, try to split by centers of boxes */
if (IS_BADRATIO(posR, posL) && IS_BADRATIO(posT, posB))
{
double avgCenterX = 0.0,
avgCenterY = 0.0;
double CenterX,
CenterY;
for (i = FirstOffsetNumber; i <= maxoff; i = OffsetNumberNext(i))
{
cur = DatumGetBoxP(entryvec->vector[i].key);
avgCenterX += ((double) cur->high.x + (double) cur->low.x) / 2.0;
avgCenterY += ((double) cur->high.y + (double) cur->low.y) / 2.0;
}
avgCenterX /= maxoff;
avgCenterY /= maxoff;
posL = posR = posB = posT = 0;
for (i = FirstOffsetNumber; i <= maxoff; i = OffsetNumberNext(i))
{
cur = DatumGetBoxP(entryvec->vector[i].key);
CenterX = ((double) cur->high.x + (double) cur->low.x) / 2.0;
CenterY = ((double) cur->high.y + (double) cur->low.y) / 2.0;
if (CenterX < avgCenterX)
ADDLIST(listL, unionL, posL, i);
else if (CenterX == avgCenterX)
{
if (posL > posR)
ADDLIST(listR, unionR, posR, i);
else
ADDLIST(listL, unionL, posL, i);
}
else
ADDLIST(listR, unionR, posR, i);
if (CenterY < avgCenterY)
ADDLIST(listB, unionB, posB, i);
else if (CenterY == avgCenterY)
{
if (posB > posT)
ADDLIST(listT, unionT, posT, i);
else
ADDLIST(listB, unionB, posB, i);
}
else
ADDLIST(listT, unionT, posT, i);
}
if (IS_BADRATIO(posR, posL) && IS_BADRATIO(posT, posB))
{
fallbackSplit(entryvec, v);
PG_RETURN_POINTER(v);
}
}
/* which split more optimal? */
if (Max(posL, posR) < Max(posB, posT))
direction = 'x';
else if (Max(posL, posR) > Max(posB, posT))
direction = 'y';
else
{
Datum interLR = DirectFunctionCall2(rt_box_inter,
BoxPGetDatum(unionL),
BoxPGetDatum(unionR));
Datum interBT = DirectFunctionCall2(rt_box_inter,
BoxPGetDatum(unionB),
BoxPGetDatum(unionT));
double sizeLR,
sizeBT;
sizeLR = size_box(interLR);
sizeBT = size_box(interBT);
if (sizeLR < sizeBT)
direction = 'x';
else
direction = 'y';
}
if (direction == 'x')
chooseLR(v,
listL, posL, unionL,
listR, posR, unionR);
else
chooseLR(v,
listB, posB, unionB,
listT, posT, unionT);
PG_RETURN_POINTER(v);
}
/*
** The GiST PickSplit method for segments
** We use Guttman's poly time split algorithm
*/
GIST_SPLITVEC *
gseg_picksplit(GistEntryVector *entryvec,
GIST_SPLITVEC *v)
{
OffsetNumber i,
j;
SEG *datum_alpha,
*datum_beta;
SEG *datum_l,
*datum_r;
SEG *union_d,
*union_dl,
*union_dr;
SEG *inter_d;
bool firsttime;
float size_alpha,
size_beta,
size_union,
size_inter;
float size_waste,
waste;
float size_l,
size_r;
int nbytes;
OffsetNumber seed_1 = 1,
seed_2 = 2;
OffsetNumber *left,
*right;
OffsetNumber maxoff;
#ifdef GIST_DEBUG
fprintf(stderr, "picksplit\n");
#endif
maxoff = entryvec->n - 2;
nbytes = (maxoff + 2) * sizeof(OffsetNumber);
v->spl_left = (OffsetNumber *) palloc(nbytes);
v->spl_right = (OffsetNumber *) palloc(nbytes);
firsttime = true;
waste = 0.0;
for (i = FirstOffsetNumber; i < maxoff; i = OffsetNumberNext(i))
{
datum_alpha = (SEG *) DatumGetPointer(entryvec->vector[i].key);
for (j = OffsetNumberNext(i); j <= maxoff; j = OffsetNumberNext(j))
{
datum_beta = (SEG *) DatumGetPointer(entryvec->vector[j].key);
/* compute the wasted space by unioning these guys */
/* size_waste = size_union - size_inter; */
union_d = seg_union(datum_alpha, datum_beta);
rt_seg_size(union_d, &size_union);
inter_d = seg_inter(datum_alpha, datum_beta);
rt_seg_size(inter_d, &size_inter);
size_waste = size_union - size_inter;
/*
* are these a more promising split that what we've already seen?
*/
if (size_waste > waste || firsttime)
{
waste = size_waste;
seed_1 = i;
seed_2 = j;
firsttime = false;
}
}
}
left = v->spl_left;
v->spl_nleft = 0;
right = v->spl_right;
v->spl_nright = 0;
datum_alpha = (SEG *) DatumGetPointer(entryvec->vector[seed_1].key);
datum_l = seg_union(datum_alpha, datum_alpha);
rt_seg_size(datum_l, &size_l);
datum_beta = (SEG *) DatumGetPointer(entryvec->vector[seed_2].key);
datum_r = seg_union(datum_beta, datum_beta);
rt_seg_size(datum_r, &size_r);
/*
* Now split up the regions between the two seeds. An important property
* of this split algorithm is that the split vector v has the indices of
* items to be split in order in its left and right vectors. We exploit
* this property by doing a merge in the code that actually splits the
* page.
*
* For efficiency, we also place the new index tuple in this loop. This is
* handled at the very end, when we have placed all the existing tuples
* and i == maxoff + 1.
*/
maxoff = OffsetNumberNext(maxoff);
for (i = FirstOffsetNumber; i <= maxoff; i = OffsetNumberNext(i))
{
/*
* If we've already decided where to place this item, just put it on
* the right list. Otherwise, we need to figure out which page needs
* the least enlargement in order to store the item.
*/
if (i == seed_1)
{
*left++ = i;
v->spl_nleft++;
continue;
}
else if (i == seed_2)
{
*right++ = i;
v->spl_nright++;
continue;
}
/* okay, which page needs least enlargement? */
datum_alpha = (SEG *) DatumGetPointer(entryvec->vector[i].key);
union_dl = seg_union(datum_l, datum_alpha);
union_dr = seg_union(datum_r, datum_alpha);
rt_seg_size(union_dl, &size_alpha);
rt_seg_size(union_dr, &size_beta);
/* pick which page to add it to */
if (size_alpha - size_l < size_beta - size_r)
{
datum_l = union_dl;
size_l = size_alpha;
*left++ = i;
v->spl_nleft++;
}
else
{
datum_r = union_dr;
size_r = size_alpha;
*right++ = i;
v->spl_nright++;
}
}
*left = *right = FirstOffsetNumber; /* sentinel value, see dosplit() */
v->spl_ldatum = PointerGetDatum(datum_l);
v->spl_rdatum = PointerGetDatum(datum_r);
return v;
}
/*
* Helper function to perform deletion of index entries from a bucket.
*
* This function expects that the caller has acquired a cleanup lock on the
* primary bucket page, and will return with a write lock again held on the
* primary bucket page. The lock won't necessarily be held continuously,
* though, because we'll release it when visiting overflow pages.
*
* It would be very bad if this function cleaned a page while some other
* backend was in the midst of scanning it, because hashgettuple assumes
* that the next valid TID will be greater than or equal to the current
* valid TID. There can't be any concurrent scans in progress when we first
* enter this function because of the cleanup lock we hold on the primary
* bucket page, but as soon as we release that lock, there might be. We
* handle that by conspiring to prevent those scans from passing our cleanup
* scan. To do that, we lock the next page in the bucket chain before
* releasing the lock on the previous page. (This type of lock chaining is
* not ideal, so we might want to look for a better solution at some point.)
*
* We need to retain a pin on the primary bucket to ensure that no concurrent
* split can start.
*/
void
hashbucketcleanup(Relation rel, Bucket cur_bucket, Buffer bucket_buf,
BlockNumber bucket_blkno, BufferAccessStrategy bstrategy,
uint32 maxbucket, uint32 highmask, uint32 lowmask,
double *tuples_removed, double *num_index_tuples,
bool split_cleanup,
IndexBulkDeleteCallback callback, void *callback_state)
{
BlockNumber blkno;
Buffer buf;
Bucket new_bucket PG_USED_FOR_ASSERTS_ONLY = InvalidBucket;
bool bucket_dirty = false;
blkno = bucket_blkno;
buf = bucket_buf;
if (split_cleanup)
new_bucket = _hash_get_newbucket_from_oldbucket(rel, cur_bucket,
lowmask, maxbucket);
/* Scan each page in bucket */
for (;;)
{
HashPageOpaque opaque;
OffsetNumber offno;
OffsetNumber maxoffno;
Buffer next_buf;
Page page;
OffsetNumber deletable[MaxOffsetNumber];
int ndeletable = 0;
bool retain_pin = false;
bool clear_dead_marking = false;
vacuum_delay_point();
page = BufferGetPage(buf);
opaque = (HashPageOpaque) PageGetSpecialPointer(page);
/* Scan each tuple in page */
maxoffno = PageGetMaxOffsetNumber(page);
for (offno = FirstOffsetNumber;
offno <= maxoffno;
offno = OffsetNumberNext(offno))
{
ItemPointer htup;
IndexTuple itup;
Bucket bucket;
bool kill_tuple = false;
itup = (IndexTuple) PageGetItem(page,
PageGetItemId(page, offno));
htup = &(itup->t_tid);
/*
* To remove the dead tuples, we strictly want to rely on results
* of callback function. refer btvacuumpage for detailed reason.
*/
if (callback && callback(htup, callback_state))
{
kill_tuple = true;
if (tuples_removed)
*tuples_removed += 1;
}
else if (split_cleanup)
{
/* delete the tuples that are moved by split. */
bucket = _hash_hashkey2bucket(_hash_get_indextuple_hashkey(itup),
maxbucket,
highmask,
lowmask);
/* mark the item for deletion */
if (bucket != cur_bucket)
{
/*
* We expect tuples to either belong to curent bucket or
* new_bucket. This is ensured because we don't allow
* further splits from bucket that contains garbage. See
* comments in _hash_expandtable.
*/
Assert(bucket == new_bucket);
kill_tuple = true;
}
}
if (kill_tuple)
{
/* mark the item for deletion */
deletable[ndeletable++] = offno;
}
else
{
/* we're keeping it, so count it */
if (num_index_tuples)
*num_index_tuples += 1;
}
}
/* retain the pin on primary bucket page till end of bucket scan */
if (blkno == bucket_blkno)
retain_pin = true;
else
retain_pin = false;
blkno = opaque->hasho_nextblkno;
/*
* Apply deletions, advance to next page and write page if needed.
*/
if (ndeletable > 0)
{
/* No ereport(ERROR) until changes are logged */
START_CRIT_SECTION();
PageIndexMultiDelete(page, deletable, ndeletable);
bucket_dirty = true;
/*
* Let us mark the page as clean if vacuum removes the DEAD tuples
* from an index page. We do this by clearing LH_PAGE_HAS_DEAD_TUPLES
* flag.
*/
if (tuples_removed && *tuples_removed > 0 &&
opaque->hasho_flag & LH_PAGE_HAS_DEAD_TUPLES)
{
opaque->hasho_flag &= ~LH_PAGE_HAS_DEAD_TUPLES;
clear_dead_marking = true;
}
MarkBufferDirty(buf);
/* XLOG stuff */
if (RelationNeedsWAL(rel))
{
xl_hash_delete xlrec;
XLogRecPtr recptr;
xlrec.clear_dead_marking = clear_dead_marking;
xlrec.is_primary_bucket_page = (buf == bucket_buf) ? true : false;
XLogBeginInsert();
XLogRegisterData((char *) &xlrec, SizeOfHashDelete);
/*
* bucket buffer needs to be registered to ensure that we can
* acquire a cleanup lock on it during replay.
*/
if (!xlrec.is_primary_bucket_page)
XLogRegisterBuffer(0, bucket_buf, REGBUF_STANDARD | REGBUF_NO_IMAGE);
XLogRegisterBuffer(1, buf, REGBUF_STANDARD);
XLogRegisterBufData(1, (char *) deletable,
ndeletable * sizeof(OffsetNumber));
recptr = XLogInsert(RM_HASH_ID, XLOG_HASH_DELETE);
PageSetLSN(BufferGetPage(buf), recptr);
}
END_CRIT_SECTION();
}
/* bail out if there are no more pages to scan. */
if (!BlockNumberIsValid(blkno))
break;
next_buf = _hash_getbuf_with_strategy(rel, blkno, HASH_WRITE,
LH_OVERFLOW_PAGE,
bstrategy);
/*
* release the lock on previous page after acquiring the lock on next
* page
*/
if (retain_pin)
LockBuffer(buf, BUFFER_LOCK_UNLOCK);
else
_hash_relbuf(rel, buf);
buf = next_buf;
}
/*
* lock the bucket page to clear the garbage flag and squeeze the bucket.
* if the current buffer is same as bucket buffer, then we already have
* lock on bucket page.
*/
if (buf != bucket_buf)
{
_hash_relbuf(rel, buf);
LockBuffer(bucket_buf, BUFFER_LOCK_EXCLUSIVE);
}
/*
* Clear the garbage flag from bucket after deleting the tuples that are
* moved by split. We purposefully clear the flag before squeeze bucket,
* so that after restart, vacuum shouldn't again try to delete the moved
* by split tuples.
*/
if (split_cleanup)
{
HashPageOpaque bucket_opaque;
Page page;
page = BufferGetPage(bucket_buf);
bucket_opaque = (HashPageOpaque) PageGetSpecialPointer(page);
/* No ereport(ERROR) until changes are logged */
START_CRIT_SECTION();
bucket_opaque->hasho_flag &= ~LH_BUCKET_NEEDS_SPLIT_CLEANUP;
MarkBufferDirty(bucket_buf);
/* XLOG stuff */
if (RelationNeedsWAL(rel))
{
XLogRecPtr recptr;
XLogBeginInsert();
XLogRegisterBuffer(0, bucket_buf, REGBUF_STANDARD);
recptr = XLogInsert(RM_HASH_ID, XLOG_HASH_SPLIT_CLEANUP);
PageSetLSN(page, recptr);
}
END_CRIT_SECTION();
}
/*
* If we have deleted anything, try to compact free space. For squeezing
* the bucket, we must have a cleanup lock, else it can impact the
* ordering of tuples for a scan that has started before it.
*/
if (bucket_dirty && IsBufferCleanupOK(bucket_buf))
_hash_squeezebucket(rel, cur_bucket, bucket_blkno, bucket_buf,
bstrategy);
else
LockBuffer(bucket_buf, BUFFER_LOCK_UNLOCK);
}
/*
* find entry with lowest penalty
*/
OffsetNumber
gistchoose(Relation r, Page p, IndexTuple it, /* it has compressed entry */
GISTSTATE *giststate)
{
OffsetNumber maxoff;
OffsetNumber i;
OffsetNumber which;
float sum_grow,
which_grow[INDEX_MAX_KEYS];
GISTENTRY entry,
identry[INDEX_MAX_KEYS];
bool isnull[INDEX_MAX_KEYS];
maxoff = PageGetMaxOffsetNumber(p);
*which_grow = -1.0;
which = InvalidOffsetNumber;
sum_grow = 1;
gistDeCompressAtt(giststate, r,
it, NULL, (OffsetNumber) 0,
identry, isnull);
Assert(maxoff >= FirstOffsetNumber);
Assert(!GistPageIsLeaf(p));
for (i = FirstOffsetNumber; i <= maxoff && sum_grow; i = OffsetNumberNext(i))
{
int j;
IndexTuple itup = (IndexTuple) PageGetItem(p, PageGetItemId(p, i));
if (!GistPageIsLeaf(p) && GistTupleIsInvalid(itup))
{
ereport(LOG,
(errmsg("index \"%s\" needs VACUUM or REINDEX to finish crash recovery",
RelationGetRelationName(r))));
continue;
}
sum_grow = 0;
for (j = 0; j < r->rd_att->natts; j++)
{
Datum datum;
float usize;
bool IsNull;
datum = index_getattr(itup, j + 1, giststate->tupdesc, &IsNull);
gistdentryinit(giststate, j, &entry, datum, r, p, i,
FALSE, IsNull);
usize = gistpenalty(giststate, j, &entry, IsNull,
&identry[j], isnull[j]);
if (which_grow[j] < 0 || usize < which_grow[j])
{
which = i;
which_grow[j] = usize;
if (j < r->rd_att->natts - 1 && i == FirstOffsetNumber)
which_grow[j + 1] = -1;
sum_grow += which_grow[j];
}
else if (which_grow[j] == usize)
sum_grow += usize;
else
{
sum_grow = 1;
break;
}
}
}
if (which == InvalidOffsetNumber)
which = FirstOffsetNumber;
return which;
}
/*
* Updates the stack so that child->parent is the correct parent of the
* child. child->parent must be exclusively locked on entry, and will
* remain so at exit, but it might not be the same page anymore.
*/
static void
gistFindCorrectParent(Relation r, GISTInsertStack *child)
{
GISTInsertStack *parent = child->parent;
gistcheckpage(r, parent->buffer);
parent->page = (Page) BufferGetPage(parent->buffer);
/* here we don't need to distinguish between split and page update */
if (child->downlinkoffnum == InvalidOffsetNumber ||
parent->lsn != PageGetLSN(parent->page))
{
/* parent is changed, look child in right links until found */
OffsetNumber i,
maxoff;
ItemId iid;
IndexTuple idxtuple;
GISTInsertStack *ptr;
while (true)
{
maxoff = PageGetMaxOffsetNumber(parent->page);
for (i = FirstOffsetNumber; i <= maxoff; i = OffsetNumberNext(i))
{
iid = PageGetItemId(parent->page, i);
idxtuple = (IndexTuple) PageGetItem(parent->page, iid);
if (ItemPointerGetBlockNumber(&(idxtuple->t_tid)) == child->blkno)
{
/* yes!!, found */
child->downlinkoffnum = i;
return;
}
}
parent->blkno = GistPageGetOpaque(parent->page)->rightlink;
UnlockReleaseBuffer(parent->buffer);
if (parent->blkno == InvalidBlockNumber)
{
/*
* End of chain and still didn't find parent. It's a very-very
* rare situation when root splited.
*/
break;
}
parent->buffer = ReadBuffer(r, parent->blkno);
LockBuffer(parent->buffer, GIST_EXCLUSIVE);
gistcheckpage(r, parent->buffer);
parent->page = (Page) BufferGetPage(parent->buffer);
}
/*
* awful!!, we need search tree to find parent ... , but before we
* should release all old parent
*/
ptr = child->parent->parent; /* child->parent already released
* above */
while (ptr)
{
ReleaseBuffer(ptr->buffer);
ptr = ptr->parent;
}
/* ok, find new path */
ptr = parent = gistFindPath(r, child->blkno, &child->downlinkoffnum);
/* read all buffers as expected by caller */
/* note we don't lock them or gistcheckpage them here! */
while (ptr)
{
ptr->buffer = ReadBuffer(r, ptr->blkno);
ptr->page = (Page) BufferGetPage(ptr->buffer);
ptr = ptr->parent;
}
/* install new chain of parents to stack */
child->parent = parent;
/* make recursive call to normal processing */
LockBuffer(child->parent->buffer, GIST_EXCLUSIVE);
gistFindCorrectParent(r, child);
}
return;
}
/*
** The GiST PickSplit method for _intments
** We use Guttman's poly time split algorithm
*/
Datum
g_int_picksplit(PG_FUNCTION_ARGS)
{
GistEntryVector *entryvec = (GistEntryVector *) PG_GETARG_POINTER(0);
GIST_SPLITVEC *v = (GIST_SPLITVEC *) PG_GETARG_POINTER(1);
OffsetNumber i,
j;
ArrayType *datum_alpha,
*datum_beta;
ArrayType *datum_l,
*datum_r;
ArrayType *union_d,
*union_dl,
*union_dr;
ArrayType *inter_d;
bool firsttime;
float size_alpha,
size_beta,
size_union,
size_inter;
float size_waste,
waste;
float size_l,
size_r;
int nbytes;
OffsetNumber seed_1 = 0,
seed_2 = 0;
OffsetNumber *left,
*right;
OffsetNumber maxoff;
SPLITCOST *costvector;
#ifdef GIST_DEBUG
elog(DEBUG3, "--------picksplit %d", entryvec->n);
#endif
maxoff = entryvec->n - 2;
nbytes = (maxoff + 2) * sizeof(OffsetNumber);
v->spl_left = (OffsetNumber *) palloc(nbytes);
v->spl_right = (OffsetNumber *) palloc(nbytes);
firsttime = true;
waste = 0.0;
for (i = FirstOffsetNumber; i < maxoff; i = OffsetNumberNext(i))
{
datum_alpha = GETENTRY(entryvec, i);
for (j = OffsetNumberNext(i); j <= maxoff; j = OffsetNumberNext(j))
{
datum_beta = GETENTRY(entryvec, j);
/* compute the wasted space by unioning these guys */
/* size_waste = size_union - size_inter; */
union_d = inner_int_union(datum_alpha, datum_beta);
rt__int_size(union_d, &size_union);
inter_d = inner_int_inter(datum_alpha, datum_beta);
rt__int_size(inter_d, &size_inter);
size_waste = size_union - size_inter;
pfree(union_d);
if (inter_d != (ArrayType *) NULL)
pfree(inter_d);
/*
* are these a more promising split that what we've already seen?
*/
if (size_waste > waste || firsttime)
{
waste = size_waste;
seed_1 = i;
seed_2 = j;
firsttime = false;
}
}
}
left = v->spl_left;
v->spl_nleft = 0;
right = v->spl_right;
v->spl_nright = 0;
if (seed_1 == 0 || seed_2 == 0)
{
seed_1 = 1;
seed_2 = 2;
}
datum_alpha = GETENTRY(entryvec, seed_1);
datum_l = copy_intArrayType(datum_alpha);
rt__int_size(datum_l, &size_l);
datum_beta = GETENTRY(entryvec, seed_2);
datum_r = copy_intArrayType(datum_beta);
rt__int_size(datum_r, &size_r);
maxoff = OffsetNumberNext(maxoff);
/*
* sort entries
*/
costvector = (SPLITCOST *) palloc(sizeof(SPLITCOST) * maxoff);
for (i = FirstOffsetNumber; i <= maxoff; i = OffsetNumberNext(i))
{
costvector[i - 1].pos = i;
datum_alpha = GETENTRY(entryvec, i);
union_d = inner_int_union(datum_l, datum_alpha);
rt__int_size(union_d, &size_alpha);
pfree(union_d);
union_d = inner_int_union(datum_r, datum_alpha);
rt__int_size(union_d, &size_beta);
pfree(union_d);
costvector[i - 1].cost = Abs((size_alpha - size_l) - (size_beta - size_r));
}
qsort((void *) costvector, maxoff, sizeof(SPLITCOST), comparecost);
/*
* Now split up the regions between the two seeds. An important property
* of this split algorithm is that the split vector v has the indices of
* items to be split in order in its left and right vectors. We exploit
* this property by doing a merge in the code that actually splits the
* page.
*
* For efficiency, we also place the new index tuple in this loop. This is
* handled at the very end, when we have placed all the existing tuples
* and i == maxoff + 1.
*/
for (j = 0; j < maxoff; j++)
{
i = costvector[j].pos;
/*
* If we've already decided where to place this item, just put it on
* the right list. Otherwise, we need to figure out which page needs
* the least enlargement in order to store the item.
*/
if (i == seed_1)
{
*left++ = i;
v->spl_nleft++;
continue;
}
else if (i == seed_2)
{
*right++ = i;
v->spl_nright++;
continue;
}
/* okay, which page needs least enlargement? */
datum_alpha = GETENTRY(entryvec, i);
union_dl = inner_int_union(datum_l, datum_alpha);
union_dr = inner_int_union(datum_r, datum_alpha);
rt__int_size(union_dl, &size_alpha);
rt__int_size(union_dr, &size_beta);
/* pick which page to add it to */
if (size_alpha - size_l < size_beta - size_r + WISH_F(v->spl_nleft, v->spl_nright, 0.01))
{
if (datum_l)
pfree(datum_l);
if (union_dr)
pfree(union_dr);
datum_l = union_dl;
size_l = size_alpha;
*left++ = i;
v->spl_nleft++;
}
else
{
if (datum_r)
pfree(datum_r);
if (union_dl)
pfree(union_dl);
datum_r = union_dr;
size_r = size_beta;
*right++ = i;
v->spl_nright++;
}
}
pfree(costvector);
*right = *left = FirstOffsetNumber;
datum_l->flags &= ~LEAFKEY;
datum_r->flags &= ~LEAFKEY;
v->spl_ldatum = PointerGetDatum(datum_l);
v->spl_rdatum = PointerGetDatum(datum_r);
PG_RETURN_POINTER(v);
}
/*
* _hash_splitbucket -- split 'obucket' into 'obucket' and 'nbucket'
*
* This routine is used to partition the tuples between old and new bucket and
* is used to finish the incomplete split operations. To finish the previously
* interrupted split operation, the caller needs to fill htab. If htab is set,
* then we skip the movement of tuples that exists in htab, otherwise NULL
* value of htab indicates movement of all the tuples that belong to the new
* bucket.
*
* We are splitting a bucket that consists of a base bucket page and zero
* or more overflow (bucket chain) pages. We must relocate tuples that
* belong in the new bucket.
*
* The caller must hold cleanup locks on both buckets to ensure that
* no one else is trying to access them (see README).
*
* The caller must hold a pin, but no lock, on the metapage buffer.
* The buffer is returned in the same state. (The metapage is only
* touched if it becomes necessary to add or remove overflow pages.)
*
* Split needs to retain pin on primary bucket pages of both old and new
* buckets till end of operation. This is to prevent vacuum from starting
* while a split is in progress.
*
* In addition, the caller must have created the new bucket's base page,
* which is passed in buffer nbuf, pinned and write-locked. The lock will be
* released here and pin must be released by the caller. (The API is set up
* this way because we must do _hash_getnewbuf() before releasing the metapage
* write lock. So instead of passing the new bucket's start block number, we
* pass an actual buffer.)
*/
static void
_hash_splitbucket(Relation rel,
Buffer metabuf,
Bucket obucket,
Bucket nbucket,
Buffer obuf,
Buffer nbuf,
HTAB *htab,
uint32 maxbucket,
uint32 highmask,
uint32 lowmask)
{
Buffer bucket_obuf;
Buffer bucket_nbuf;
Page opage;
Page npage;
HashPageOpaque oopaque;
HashPageOpaque nopaque;
OffsetNumber itup_offsets[MaxIndexTuplesPerPage];
IndexTuple itups[MaxIndexTuplesPerPage];
Size all_tups_size = 0;
int i;
uint16 nitups = 0;
bucket_obuf = obuf;
opage = BufferGetPage(obuf);
oopaque = (HashPageOpaque) PageGetSpecialPointer(opage);
bucket_nbuf = nbuf;
npage = BufferGetPage(nbuf);
nopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
/*
* Partition the tuples in the old bucket between the old bucket and the
* new bucket, advancing along the old bucket's overflow bucket chain and
* adding overflow pages to the new bucket as needed. Outer loop iterates
* once per page in old bucket.
*/
for (;;)
{
BlockNumber oblkno;
OffsetNumber ooffnum;
OffsetNumber omaxoffnum;
/* Scan each tuple in old page */
omaxoffnum = PageGetMaxOffsetNumber(opage);
for (ooffnum = FirstOffsetNumber;
ooffnum <= omaxoffnum;
ooffnum = OffsetNumberNext(ooffnum))
{
IndexTuple itup;
Size itemsz;
Bucket bucket;
bool found = false;
/* skip dead tuples */
if (ItemIdIsDead(PageGetItemId(opage, ooffnum)))
continue;
/*
* Before inserting a tuple, probe the hash table containing TIDs
* of tuples belonging to new bucket, if we find a match, then
* skip that tuple, else fetch the item's hash key (conveniently
* stored in the item) and determine which bucket it now belongs
* in.
*/
itup = (IndexTuple) PageGetItem(opage,
PageGetItemId(opage, ooffnum));
if (htab)
(void) hash_search(htab, &itup->t_tid, HASH_FIND, &found);
if (found)
continue;
bucket = _hash_hashkey2bucket(_hash_get_indextuple_hashkey(itup),
maxbucket, highmask, lowmask);
if (bucket == nbucket)
{
IndexTuple new_itup;
/*
* make a copy of index tuple as we have to scribble on it.
*/
new_itup = CopyIndexTuple(itup);
/*
* mark the index tuple as moved by split, such tuples are
* skipped by scan if there is split in progress for a bucket.
*/
new_itup->t_info |= INDEX_MOVED_BY_SPLIT_MASK;
/*
* insert the tuple into the new bucket. if it doesn't fit on
* the current page in the new bucket, we must allocate a new
* overflow page and place the tuple on that page instead.
*/
itemsz = IndexTupleDSize(*new_itup);
itemsz = MAXALIGN(itemsz);
if (PageGetFreeSpaceForMultipleTuples(npage, nitups + 1) < (all_tups_size + itemsz))
{
/*
* Change the shared buffer state in critical section,
* otherwise any error could make it unrecoverable.
*/
START_CRIT_SECTION();
_hash_pgaddmultitup(rel, nbuf, itups, itup_offsets, nitups);
MarkBufferDirty(nbuf);
/* log the split operation before releasing the lock */
log_split_page(rel, nbuf);
END_CRIT_SECTION();
/* drop lock, but keep pin */
LockBuffer(nbuf, BUFFER_LOCK_UNLOCK);
/* be tidy */
for (i = 0; i < nitups; i++)
pfree(itups[i]);
nitups = 0;
all_tups_size = 0;
/* chain to a new overflow page */
nbuf = _hash_addovflpage(rel, metabuf, nbuf, (nbuf == bucket_nbuf) ? true : false);
npage = BufferGetPage(nbuf);
nopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
}
itups[nitups++] = new_itup;
all_tups_size += itemsz;
}
else
{
/*
* the tuple stays on this page, so nothing to do.
*/
Assert(bucket == obucket);
}
}
oblkno = oopaque->hasho_nextblkno;
/* retain the pin on the old primary bucket */
if (obuf == bucket_obuf)
LockBuffer(obuf, BUFFER_LOCK_UNLOCK);
else
_hash_relbuf(rel, obuf);
/* Exit loop if no more overflow pages in old bucket */
if (!BlockNumberIsValid(oblkno))
{
/*
* Change the shared buffer state in critical section, otherwise
* any error could make it unrecoverable.
*/
START_CRIT_SECTION();
_hash_pgaddmultitup(rel, nbuf, itups, itup_offsets, nitups);
MarkBufferDirty(nbuf);
/* log the split operation before releasing the lock */
log_split_page(rel, nbuf);
END_CRIT_SECTION();
if (nbuf == bucket_nbuf)
LockBuffer(nbuf, BUFFER_LOCK_UNLOCK);
else
_hash_relbuf(rel, nbuf);
/* be tidy */
for (i = 0; i < nitups; i++)
pfree(itups[i]);
break;
}
/* Else, advance to next old page */
obuf = _hash_getbuf(rel, oblkno, HASH_READ, LH_OVERFLOW_PAGE);
opage = BufferGetPage(obuf);
oopaque = (HashPageOpaque) PageGetSpecialPointer(opage);
}
/*
* We're at the end of the old bucket chain, so we're done partitioning
* the tuples. Mark the old and new buckets to indicate split is
* finished.
*
* To avoid deadlocks due to locking order of buckets, first lock the old
* bucket and then the new bucket.
*/
LockBuffer(bucket_obuf, BUFFER_LOCK_EXCLUSIVE);
opage = BufferGetPage(bucket_obuf);
oopaque = (HashPageOpaque) PageGetSpecialPointer(opage);
LockBuffer(bucket_nbuf, BUFFER_LOCK_EXCLUSIVE);
npage = BufferGetPage(bucket_nbuf);
nopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
START_CRIT_SECTION();
oopaque->hasho_flag &= ~LH_BUCKET_BEING_SPLIT;
nopaque->hasho_flag &= ~LH_BUCKET_BEING_POPULATED;
/*
* After the split is finished, mark the old bucket to indicate that it
* contains deletable tuples. We will clear split-cleanup flag after
* deleting such tuples either at the end of split or at the next split
* from old bucket or at the time of vacuum.
*/
oopaque->hasho_flag |= LH_BUCKET_NEEDS_SPLIT_CLEANUP;
/*
* now write the buffers, here we don't release the locks as caller is
* responsible to release locks.
*/
MarkBufferDirty(bucket_obuf);
MarkBufferDirty(bucket_nbuf);
if (RelationNeedsWAL(rel))
{
XLogRecPtr recptr;
xl_hash_split_complete xlrec;
xlrec.old_bucket_flag = oopaque->hasho_flag;
xlrec.new_bucket_flag = nopaque->hasho_flag;
XLogBeginInsert();
XLogRegisterData((char *) &xlrec, SizeOfHashSplitComplete);
XLogRegisterBuffer(0, bucket_obuf, REGBUF_STANDARD);
XLogRegisterBuffer(1, bucket_nbuf, REGBUF_STANDARD);
recptr = XLogInsert(RM_HASH_ID, XLOG_HASH_SPLIT_COMPLETE);
PageSetLSN(BufferGetPage(bucket_obuf), recptr);
PageSetLSN(BufferGetPage(bucket_nbuf), recptr);
}
END_CRIT_SECTION();
/*
* If possible, clean up the old bucket. We might not be able to do this
* if someone else has a pin on it, but if not then we can go ahead. This
* isn't absolutely necessary, but it reduces bloat; if we don't do it
* now, VACUUM will do it eventually, but maybe not until new overflow
* pages have been allocated. Note that there's no need to clean up the
* new bucket.
*/
if (IsBufferCleanupOK(bucket_obuf))
{
LockBuffer(bucket_nbuf, BUFFER_LOCK_UNLOCK);
hashbucketcleanup(rel, obucket, bucket_obuf,
BufferGetBlockNumber(bucket_obuf), NULL,
maxbucket, highmask, lowmask, NULL, NULL, true,
NULL, NULL);
}
else
{
LockBuffer(bucket_nbuf, BUFFER_LOCK_UNLOCK);
LockBuffer(bucket_obuf, BUFFER_LOCK_UNLOCK);
}
}
Datum
readindex(PG_FUNCTION_ARGS)
{
FuncCallContext *funcctx;
readindexinfo *info;
MIRROREDLOCK_BUFMGR_DECLARE;
if (SRF_IS_FIRSTCALL())
{
Oid irelid = PG_GETARG_OID(0);
TupleDesc tupdesc;
MemoryContext oldcontext;
AttrNumber outattnum;
Relation irel;
TupleDesc itupdesc;
int i;
AttrNumber attno;
irel = index_open(irelid, AccessShareLock);
itupdesc = RelationGetDescr(irel);
outattnum = FIXED_COLUMN + itupdesc->natts;
funcctx = SRF_FIRSTCALL_INIT();
oldcontext = MemoryContextSwitchTo(funcctx->multi_call_memory_ctx);
tupdesc = CreateTemplateTupleDesc(outattnum, false);
attno = 1;
TupleDescInitEntry(tupdesc, attno++, "ictid", TIDOID, -1, 0);
TupleDescInitEntry(tupdesc, attno++, "hctid", TIDOID, -1, 0);
TupleDescInitEntry(tupdesc, attno++, "aotid", TEXTOID, -1, 0);
TupleDescInitEntry(tupdesc, attno++, "istatus", TEXTOID, -1, 0);
TupleDescInitEntry(tupdesc, attno++, "hstatus", TEXTOID, -1, 0);
for (i = 0; i < itupdesc->natts; i++)
{
Form_pg_attribute attr = itupdesc->attrs[i];
TupleDescInitEntry(tupdesc, attno++, NameStr(attr->attname), attr->atttypid, attr->atttypmod, 0);
}
funcctx->tuple_desc = BlessTupleDesc(tupdesc);
info = (readindexinfo *) palloc(sizeof(readindexinfo));
funcctx->user_fctx = (void *) info;
info->outattnum = outattnum;
info->irel = irel;
info->hrel = relation_open(irel->rd_index->indrelid, AccessShareLock);
if (info->hrel->rd_rel != NULL &&
(info->hrel->rd_rel->relstorage == 'a' ||
info->hrel->rd_rel->relstorage == 'c'))
{
relation_close(info->hrel, AccessShareLock);
info->hrel = NULL;
}
info->num_pages = RelationGetNumberOfBlocks(irel);
info->blkno = BTREE_METAPAGE + 1;
info->page = NULL;
MemoryContextSwitchTo(oldcontext);
}
funcctx = SRF_PERCALL_SETUP();
info = (readindexinfo *) funcctx->user_fctx;
while (info->blkno < info->num_pages)
{
Datum values[255];
bool nulls[255];
ItemPointerData itid;
HeapTuple tuple;
Datum result;
if (info->page == NULL)
{
MIRROREDLOCK_BUFMGR_LOCK;
info->buf = ReadBuffer(info->irel, info->blkno);
info->page = BufferGetPage(info->buf);
info->opaque = (BTPageOpaque) PageGetSpecialPointer(info->page);
info->minoff = P_FIRSTDATAKEY(info->opaque);
info->maxoff = PageGetMaxOffsetNumber(info->page);
info->offnum = info->minoff;
MIRROREDLOCK_BUFMGR_UNLOCK;
}
if (!P_ISLEAF(info->opaque) || info->offnum > info->maxoff)
{
ReleaseBuffer(info->buf);
info->page = NULL;
info->blkno++;
continue;
}
MemSet(nulls, false, info->outattnum * sizeof(bool));
ItemPointerSet(&itid, info->blkno, info->offnum);
values[0] = ItemPointerGetDatum(&itid);
readindextuple(info, values, nulls);
info->offnum = OffsetNumberNext(info->offnum);
tuple = heap_form_tuple(funcctx->tuple_desc, values, nulls);
result = HeapTupleGetDatum(tuple);
SRF_RETURN_NEXT(funcctx, result);
}
if (info->hrel != NULL)
relation_close(info->hrel, AccessShareLock);
index_close(info->irel, AccessShareLock);
SRF_RETURN_DONE(funcctx);
}
/*
* Write the index tuples contained in *collector into the index's
* pending list.
*
* Function guarantees that all these tuples will be inserted consecutively,
* preserving order
*/
void
ginHeapTupleFastInsert(GinState *ginstate, GinTupleCollector *collector)
{
Relation index = ginstate->index;
Buffer metabuffer;
Page metapage;
GinMetaPageData *metadata = NULL;
Buffer buffer = InvalidBuffer;
Page page = NULL;
ginxlogUpdateMeta data;
bool separateList = false;
bool needCleanup = false;
int cleanupSize;
bool needWal;
if (collector->ntuples == 0)
return;
needWal = RelationNeedsWAL(index);
data.node = index->rd_node;
data.ntuples = 0;
data.newRightlink = data.prevTail = InvalidBlockNumber;
metabuffer = ReadBuffer(index, GIN_METAPAGE_BLKNO);
metapage = BufferGetPage(metabuffer);
if (collector->sumsize + collector->ntuples * sizeof(ItemIdData) > GinListPageSize)
{
/*
* Total size is greater than one page => make sublist
*/
separateList = true;
}
else
{
LockBuffer(metabuffer, GIN_EXCLUSIVE);
metadata = GinPageGetMeta(metapage);
if (metadata->head == InvalidBlockNumber ||
collector->sumsize + collector->ntuples * sizeof(ItemIdData) > metadata->tailFreeSize)
{
/*
* Pending list is empty or total size is greater than freespace
* on tail page => make sublist
*
* We unlock metabuffer to keep high concurrency
*/
separateList = true;
LockBuffer(metabuffer, GIN_UNLOCK);
}
}
if (separateList)
{
/*
* We should make sublist separately and append it to the tail
*/
GinMetaPageData sublist;
memset(&sublist, 0, sizeof(GinMetaPageData));
makeSublist(index, collector->tuples, collector->ntuples, &sublist);
if (needWal)
XLogBeginInsert();
/*
* metapage was unlocked, see above
*/
LockBuffer(metabuffer, GIN_EXCLUSIVE);
metadata = GinPageGetMeta(metapage);
if (metadata->head == InvalidBlockNumber)
{
/*
* Main list is empty, so just insert sublist as main list
*/
START_CRIT_SECTION();
metadata->head = sublist.head;
metadata->tail = sublist.tail;
metadata->tailFreeSize = sublist.tailFreeSize;
metadata->nPendingPages = sublist.nPendingPages;
metadata->nPendingHeapTuples = sublist.nPendingHeapTuples;
}
else
{
/*
* Merge lists
*/
data.prevTail = metadata->tail;
data.newRightlink = sublist.head;
buffer = ReadBuffer(index, metadata->tail);
LockBuffer(buffer, GIN_EXCLUSIVE);
page = BufferGetPage(buffer);
Assert(GinPageGetOpaque(page)->rightlink == InvalidBlockNumber);
START_CRIT_SECTION();
GinPageGetOpaque(page)->rightlink = sublist.head;
MarkBufferDirty(buffer);
metadata->tail = sublist.tail;
metadata->tailFreeSize = sublist.tailFreeSize;
metadata->nPendingPages += sublist.nPendingPages;
metadata->nPendingHeapTuples += sublist.nPendingHeapTuples;
if (needWal)
XLogRegisterBuffer(1, buffer, REGBUF_STANDARD);
}
}
else
{
/*
* Insert into tail page. Metapage is already locked
*/
OffsetNumber l,
off;
int i,
tupsize;
char *ptr;
char *collectordata;
buffer = ReadBuffer(index, metadata->tail);
LockBuffer(buffer, GIN_EXCLUSIVE);
page = BufferGetPage(buffer);
off = (PageIsEmpty(page)) ? FirstOffsetNumber :
OffsetNumberNext(PageGetMaxOffsetNumber(page));
collectordata = ptr = (char *) palloc(collector->sumsize);
data.ntuples = collector->ntuples;
if (needWal)
XLogBeginInsert();
START_CRIT_SECTION();
/*
* Increase counter of heap tuples
*/
Assert(GinPageGetOpaque(page)->maxoff <= metadata->nPendingHeapTuples);
GinPageGetOpaque(page)->maxoff++;
metadata->nPendingHeapTuples++;
for (i = 0; i < collector->ntuples; i++)
{
tupsize = IndexTupleSize(collector->tuples[i]);
l = PageAddItem(page, (Item) collector->tuples[i], tupsize, off, false, false);
if (l == InvalidOffsetNumber)
elog(ERROR, "failed to add item to index page in \"%s\"",
RelationGetRelationName(index));
memcpy(ptr, collector->tuples[i], tupsize);
ptr += tupsize;
off++;
}
Assert((ptr - collectordata) <= collector->sumsize);
if (needWal)
{
XLogRegisterBuffer(1, buffer, REGBUF_STANDARD);
XLogRegisterBufData(1, collectordata, collector->sumsize);
}
metadata->tailFreeSize = PageGetExactFreeSpace(page);
MarkBufferDirty(buffer);
}
/*
* Write metabuffer, make xlog entry
*/
MarkBufferDirty(metabuffer);
if (needWal)
{
XLogRecPtr recptr;
memcpy(&data.metadata, metadata, sizeof(GinMetaPageData));
XLogRegisterBuffer(0, metabuffer, REGBUF_WILL_INIT);
XLogRegisterData((char *) &data, sizeof(ginxlogUpdateMeta));
recptr = XLogInsert(RM_GIN_ID, XLOG_GIN_UPDATE_META_PAGE);
PageSetLSN(metapage, recptr);
if (buffer != InvalidBuffer)
{
PageSetLSN(page, recptr);
}
}
if (buffer != InvalidBuffer)
UnlockReleaseBuffer(buffer);
/*
* Force pending list cleanup when it becomes too long. And,
* ginInsertCleanup could take significant amount of time, so we prefer to
* call it when it can do all the work in a single collection cycle. In
* non-vacuum mode, it shouldn't require maintenance_work_mem, so fire it
* while pending list is still small enough to fit into
* gin_pending_list_limit.
*
* ginInsertCleanup() should not be called inside our CRIT_SECTION.
*/
cleanupSize = GinGetPendingListCleanupSize(index);
if (metadata->nPendingPages * GIN_PAGE_FREESIZE > cleanupSize * 1024L)
needCleanup = true;
UnlockReleaseBuffer(metabuffer);
END_CRIT_SECTION();
if (needCleanup)
ginInsertCleanup(ginstate, true, NULL);
}
/*
* hashgettuple() -- Get the next tuple in the scan.
*/
Datum
hashgettuple(PG_FUNCTION_ARGS)
{
IndexScanDesc scan = (IndexScanDesc) PG_GETARG_POINTER(0);
ScanDirection dir = (ScanDirection) PG_GETARG_INT32(1);
HashScanOpaque so = (HashScanOpaque) scan->opaque;
Relation rel = scan->indexRelation;
Buffer buf;
Page page;
OffsetNumber offnum;
ItemPointer current;
bool res;
/* Hash indexes are always lossy since we store only the hash code */
scan->xs_recheck = true;
/*
* We hold pin but not lock on current buffer while outside the hash AM.
* Reacquire the read lock here.
*/
if (BufferIsValid(so->hashso_curbuf))
_hash_chgbufaccess(rel, so->hashso_curbuf, HASH_NOLOCK, HASH_READ);
/*
* If we've already initialized this scan, we can just advance it in the
* appropriate direction. If we haven't done so yet, we call a routine to
* get the first item in the scan.
*/
current = &(so->hashso_curpos);
if (ItemPointerIsValid(current))
{
/*
* An insertion into the current index page could have happened while
* we didn't have read lock on it. Re-find our position by looking
* for the TID we previously returned. (Because we hold share lock on
* the bucket, no deletions or splits could have occurred; therefore
* we can expect that the TID still exists in the current index page,
* at an offset >= where we were.)
*/
OffsetNumber maxoffnum;
buf = so->hashso_curbuf;
Assert(BufferIsValid(buf));
page = BufferGetPage(buf);
maxoffnum = PageGetMaxOffsetNumber(page);
for (offnum = ItemPointerGetOffsetNumber(current);
offnum <= maxoffnum;
offnum = OffsetNumberNext(offnum))
{
IndexTuple itup;
itup = (IndexTuple) PageGetItem(page, PageGetItemId(page, offnum));
if (ItemPointerEquals(&(so->hashso_heappos), &(itup->t_tid)))
break;
}
if (offnum > maxoffnum)
elog(ERROR, "failed to re-find scan position within index \"%s\"",
RelationGetRelationName(rel));
ItemPointerSetOffsetNumber(current, offnum);
/*
* Check to see if we should kill the previously-fetched tuple.
*/
if (scan->kill_prior_tuple)
{
/*
* Yes, so mark it by setting the LP_DEAD state in the item flags.
*/
ItemIdMarkDead(PageGetItemId(page, offnum));
/*
* Since this can be redone later if needed, it's treated the same
* as a commit-hint-bit status update for heap tuples: we mark the
* buffer dirty but don't make a WAL log entry.
*/
SetBufferCommitInfoNeedsSave(buf);
}
/*
* Now continue the scan.
*/
res = _hash_next(scan, dir);
}
else
res = _hash_first(scan, dir);
/*
* Skip killed tuples if asked to.
*/
if (scan->ignore_killed_tuples)
{
while (res)
{
offnum = ItemPointerGetOffsetNumber(current);
page = BufferGetPage(so->hashso_curbuf);
if (!ItemIdIsDead(PageGetItemId(page, offnum)))
break;
res = _hash_next(scan, dir);
}
}
/* Release read lock on current buffer, but keep it pinned */
if (BufferIsValid(so->hashso_curbuf))
_hash_chgbufaccess(rel, so->hashso_curbuf, HASH_READ, HASH_NOLOCK);
/* Return current heap TID on success */
scan->xs_ctup.t_self = so->hashso_heappos;
PG_RETURN_BOOL(res);
}
/*
* Bulk deletion of all index entries pointing to a set of heap tuples.
* The set of target tuples is specified via a callback routine that tells
* whether any given heap tuple (identified by ItemPointer) is being deleted.
*
* Result: a palloc'd struct containing statistical info for VACUUM displays.
*/
Datum
hashbulkdelete(PG_FUNCTION_ARGS)
{
IndexVacuumInfo *info = (IndexVacuumInfo *) PG_GETARG_POINTER(0);
IndexBulkDeleteResult *stats = (IndexBulkDeleteResult *) PG_GETARG_POINTER(1);
IndexBulkDeleteCallback callback = (IndexBulkDeleteCallback) PG_GETARG_POINTER(2);
void *callback_state = (void *) PG_GETARG_POINTER(3);
Relation rel = info->index;
double tuples_removed;
double num_index_tuples;
double orig_ntuples;
Bucket orig_maxbucket;
Bucket cur_maxbucket;
Bucket cur_bucket;
Buffer metabuf;
HashMetaPage metap;
HashMetaPageData local_metapage;
tuples_removed = 0;
num_index_tuples = 0;
/*
* Read the metapage to fetch original bucket and tuple counts. Also, we
* keep a copy of the last-seen metapage so that we can use its
* hashm_spares[] values to compute bucket page addresses. This is a bit
* hokey but perfectly safe, since the interesting entries in the spares
* array cannot change under us; and it beats rereading the metapage for
* each bucket.
*/
metabuf = _hash_getbuf(rel, HASH_METAPAGE, HASH_READ, LH_META_PAGE);
metap = HashPageGetMeta(BufferGetPage(metabuf));
orig_maxbucket = metap->hashm_maxbucket;
orig_ntuples = metap->hashm_ntuples;
memcpy(&local_metapage, metap, sizeof(local_metapage));
_hash_relbuf(rel, metabuf);
/* Scan the buckets that we know exist */
cur_bucket = 0;
cur_maxbucket = orig_maxbucket;
loop_top:
while (cur_bucket <= cur_maxbucket)
{
BlockNumber bucket_blkno;
BlockNumber blkno;
bool bucket_dirty = false;
/* Get address of bucket's start page */
bucket_blkno = BUCKET_TO_BLKNO(&local_metapage, cur_bucket);
/* Exclusive-lock the bucket so we can shrink it */
_hash_getlock(rel, bucket_blkno, HASH_EXCLUSIVE);
/* Shouldn't have any active scans locally, either */
if (_hash_has_active_scan(rel, cur_bucket))
elog(ERROR, "hash index has active scan during VACUUM");
/* Scan each page in bucket */
blkno = bucket_blkno;
while (BlockNumberIsValid(blkno))
{
Buffer buf;
Page page;
HashPageOpaque opaque;
OffsetNumber offno;
OffsetNumber maxoffno;
OffsetNumber deletable[MaxOffsetNumber];
int ndeletable = 0;
vacuum_delay_point();
buf = _hash_getbuf_with_strategy(rel, blkno, HASH_WRITE,
LH_BUCKET_PAGE | LH_OVERFLOW_PAGE,
info->strategy);
page = BufferGetPage(buf);
opaque = (HashPageOpaque) PageGetSpecialPointer(page);
Assert(opaque->hasho_bucket == cur_bucket);
/* Scan each tuple in page */
maxoffno = PageGetMaxOffsetNumber(page);
for (offno = FirstOffsetNumber;
offno <= maxoffno;
offno = OffsetNumberNext(offno))
{
IndexTuple itup;
ItemPointer htup;
itup = (IndexTuple) PageGetItem(page,
PageGetItemId(page, offno));
htup = &(itup->t_tid);
if (callback(htup, callback_state))
{
/* mark the item for deletion */
deletable[ndeletable++] = offno;
tuples_removed += 1;
}
else
num_index_tuples += 1;
}
/*
* Apply deletions and write page if needed, advance to next page.
*/
blkno = opaque->hasho_nextblkno;
if (ndeletable > 0)
{
PageIndexMultiDelete(page, deletable, ndeletable);
_hash_wrtbuf(rel, buf);
bucket_dirty = true;
}
else
_hash_relbuf(rel, buf);
}
/* If we deleted anything, try to compact free space */
if (bucket_dirty)
_hash_squeezebucket(rel, cur_bucket, bucket_blkno,
info->strategy);
/* Release bucket lock */
_hash_droplock(rel, bucket_blkno, HASH_EXCLUSIVE);
/* Advance to next bucket */
cur_bucket++;
}
/* Write-lock metapage and check for split since we started */
metabuf = _hash_getbuf(rel, HASH_METAPAGE, HASH_WRITE, LH_META_PAGE);
metap = HashPageGetMeta(BufferGetPage(metabuf));
if (cur_maxbucket != metap->hashm_maxbucket)
{
/* There's been a split, so process the additional bucket(s) */
cur_maxbucket = metap->hashm_maxbucket;
memcpy(&local_metapage, metap, sizeof(local_metapage));
_hash_relbuf(rel, metabuf);
goto loop_top;
}
/* Okay, we're really done. Update tuple count in metapage. */
if (orig_maxbucket == metap->hashm_maxbucket &&
orig_ntuples == metap->hashm_ntuples)
{
/*
* No one has split or inserted anything since start of scan, so
* believe our count as gospel.
*/
metap->hashm_ntuples = num_index_tuples;
}
else
{
/*
* Otherwise, our count is untrustworthy since we may have
* double-scanned tuples in split buckets. Proceed by dead-reckoning.
* (Note: we still return estimated_count = false, because using this
* count is better than not updating reltuples at all.)
*/
if (metap->hashm_ntuples > tuples_removed)
metap->hashm_ntuples -= tuples_removed;
else
metap->hashm_ntuples = 0;
num_index_tuples = metap->hashm_ntuples;
}
_hash_wrtbuf(rel, metabuf);
/* return statistics */
if (stats == NULL)
stats = (IndexBulkDeleteResult *) palloc0(sizeof(IndexBulkDeleteResult));
stats->estimated_count = false;
stats->num_index_tuples = num_index_tuples;
stats->tuples_removed += tuples_removed;
/* hashvacuumcleanup will fill in num_pages */
PG_RETURN_POINTER(stats);
}
Datum geography_gist_picksplit(PG_FUNCTION_ARGS)
{
GistEntryVector *entryvec = (GistEntryVector*) PG_GETARG_POINTER(0);
GIST_SPLITVEC *v = (GIST_SPLITVEC*) PG_GETARG_POINTER(1);
OffsetNumber i;
/* One union box for each half of the space. */
GIDX **box_union;
/* One offset number list for each half of the space. */
OffsetNumber **list;
/* One position index for each half of the space. */
int *pos;
GIDX *box_pageunion;
GIDX *box_current;
int direction = -1;
bool all_entries_equal = true;
OffsetNumber max_offset;
int nbytes, ndims_pageunion, d;
int posmax = -1;
POSTGIS_DEBUG(4, "[GIST] 'picksplit' function called");
/*
** First calculate the bounding box and maximum number of dimensions in this page.
*/
max_offset = entryvec->n - 1;
box_current = (GIDX*) DatumGetPointer(entryvec->vector[FirstOffsetNumber].key);
box_pageunion = gidx_copy(box_current);
/* Calculate the containing box (box_pageunion) for the whole page we are going to split. */
for ( i = OffsetNumberNext(FirstOffsetNumber); i <= max_offset; i = OffsetNumberNext(i) )
{
box_current = (GIDX*) DatumGetPointer(entryvec->vector[i].key);
if ( all_entries_equal == true && ! gidx_equals (box_pageunion, box_current) )
all_entries_equal = false;
gidx_merge( &box_pageunion, box_current );
}
POSTGIS_DEBUGF(3, "[GIST] box_pageunion: %s", gidx_to_string(box_pageunion));
/* Every box in the page is the same! So, we split and just put half the boxes in each child. */
if ( all_entries_equal )
{
POSTGIS_DEBUG(4, "[GIST] picksplit finds all entries equal!");
geography_gist_picksplit_fallback(entryvec, v);
PG_RETURN_POINTER(v);
}
/* Initialize memory structures. */
nbytes = (max_offset + 2) * sizeof(OffsetNumber);
ndims_pageunion = GIDX_NDIMS(box_pageunion);
POSTGIS_DEBUGF(4, "[GIST] ndims_pageunion == %d", ndims_pageunion);
pos = palloc(2*ndims_pageunion * sizeof(int));
list = palloc(2*ndims_pageunion * sizeof(OffsetNumber*));
box_union = palloc(2*ndims_pageunion * sizeof(GIDX*));
for ( d = 0; d < ndims_pageunion; d++ )
{
list[BELOW(d)] = (OffsetNumber*) palloc(nbytes);
list[ABOVE(d)] = (OffsetNumber*) palloc(nbytes);
box_union[BELOW(d)] = gidx_new(ndims_pageunion);
box_union[ABOVE(d)] = gidx_new(ndims_pageunion);
pos[BELOW(d)] = 0;
pos[ABOVE(d)] = 0;
}
/*
** Assign each entry in the node to the volume partitions it belongs to,
** such as "above the x/y plane, left of the y/z plane, below the x/z plane".
** Each entry thereby ends up in three of the six partitions.
*/
POSTGIS_DEBUG(4, "[GIST] 'picksplit' calculating best split axis");
for ( i = FirstOffsetNumber; i <= max_offset; i = OffsetNumberNext(i) )
{
box_current = (GIDX*) DatumGetPointer(entryvec->vector[i].key);
for ( d = 0; d < ndims_pageunion; d++ )
{
if ( GIDX_GET_MIN(box_current,d)-GIDX_GET_MIN(box_pageunion,d) < GIDX_GET_MAX(box_pageunion,d)-GIDX_GET_MAX(box_current,d) )
{
geography_gist_picksplit_addlist(list[BELOW(d)], &(box_union[BELOW(d)]), box_current, &(pos[BELOW(d)]), i);
}
else
{
geography_gist_picksplit_addlist(list[ABOVE(d)], &(box_union[ABOVE(d)]), box_current, &(pos[ABOVE(d)]), i);
}
}
}
/*
** "Bad disposition", too many entries fell into one octant of the space, so no matter which
** plane we choose to split on, we're going to end up with a mostly full node. Where the
** data is pretty homogeneous (lots of duplicates) entries that are equidistant from the
** sides of the page union box can occasionally all end up in one place, leading
** to this condition.
*/
if ( geography_gist_picksplit_badratios(pos,ndims_pageunion) == TRUE )
{
/*
** Instead we split on center points and see if we do better.
** First calculate the average center point for each axis.
*/
double *avgCenter = palloc(ndims_pageunion * sizeof(double));
for ( d = 0; d < ndims_pageunion; d++ )
{
avgCenter[d] = 0.0;
}
POSTGIS_DEBUG(4, "[GIST] picksplit can't find good split axis, trying center point method");
for ( i = FirstOffsetNumber; i <= max_offset; i = OffsetNumberNext(i) )
{
box_current = (GIDX*) DatumGetPointer(entryvec->vector[i].key);
for ( d = 0; d < ndims_pageunion; d++ )
{
avgCenter[d] += (GIDX_GET_MAX(box_current,d) + GIDX_GET_MIN(box_current,d)) / 2.0;
}
}
for ( d = 0; d < ndims_pageunion; d++ )
{
avgCenter[d] /= max_offset;
pos[BELOW(d)] = pos[ABOVE(d)] = 0; /* Re-initialize our counters. */
POSTGIS_DEBUGF(4, "[GIST] picksplit average center point[%d] = %.12g", d, avgCenter[d]);
}
/* For each of our entries... */
for ( i = FirstOffsetNumber; i <= max_offset; i = OffsetNumberNext(i) )
{
double center;
box_current = (GIDX*) DatumGetPointer(entryvec->vector[i].key);
for ( d = 0; d < ndims_pageunion; d++ )
{
center = (GIDX_GET_MIN(box_current,d)+GIDX_GET_MAX(box_current,d))/2.0;
if ( center < avgCenter[d] )
geography_gist_picksplit_addlist(list[BELOW(d)], &(box_union[BELOW(d)]), box_current, &(pos[BELOW(d)]), i);
else if ( FPeq(center, avgCenter[d]) )
if ( pos[BELOW(d)] > pos[ABOVE(d)] )
geography_gist_picksplit_addlist(list[ABOVE(d)], &(box_union[ABOVE(d)]), box_current, &(pos[ABOVE(d)]), i);
else
geography_gist_picksplit_addlist(list[BELOW(d)], &(box_union[BELOW(d)]), box_current, &(pos[BELOW(d)]), i);
else
geography_gist_picksplit_addlist(list[ABOVE(d)], &(box_union[ABOVE(d)]), box_current, &(pos[ABOVE(d)]), i);
}
}
/* Do we have a good disposition now? If not, screw it, just cut the node in half. */
if ( geography_gist_picksplit_badratios(pos,ndims_pageunion) == TRUE )
{
POSTGIS_DEBUG(4, "[GIST] picksplit still cannot find a good split! just cutting the node in half");
geography_gist_picksplit_fallback(entryvec, v);
PG_RETURN_POINTER(v);
}
}
/*
** Now, what splitting plane gives us the most even ratio of
** entries in our child pages? Since each split region has been apportioned entries
** against the same number of total entries, the axis that has the smallest maximum
** number of entries in its regions is the most evenly distributed.
** TODO: what if the distributions are equal in two or more axes?
*/
for ( d = 0; d < ndims_pageunion; d++ )
{
int posd = Max(pos[ABOVE(d)],pos[BELOW(d)]);
if ( posd > posmax )
{
direction = d;
posmax = posd;
}
}
if ( direction == -1 || posmax == -1 )
{
/* ERROR OUT HERE */
elog(ERROR, "Error in building split, unable to determine split direction.");
}
POSTGIS_DEBUGF(3, "[GIST] 'picksplit' splitting on axis %d", direction);
geography_gist_picksplit_constructsplit(v, list[BELOW(direction)],
pos[BELOW(direction)],
&(box_union[BELOW(direction)]),
list[ABOVE(direction)],
pos[ABOVE(direction)],
&(box_union[ABOVE(direction)]) );
POSTGIS_DEBUGF(4, "[GIST] spl_ldatum: %s", gidx_to_string((GIDX*)v->spl_ldatum));
POSTGIS_DEBUGF(4, "[GIST] spl_rdatum: %s", gidx_to_string((GIDX*)v->spl_rdatum));
POSTGIS_DEBUGF(4, "[GIST] axis %d: parent range (%.12g, %.12g) left range (%.12g, %.12g), right range (%.12g, %.12g)",
direction,
GIDX_GET_MIN(box_pageunion, direction), GIDX_GET_MAX(box_pageunion, direction),
GIDX_GET_MIN((GIDX*)v->spl_ldatum, direction), GIDX_GET_MAX((GIDX*)v->spl_ldatum, direction),
GIDX_GET_MIN((GIDX*)v->spl_rdatum, direction), GIDX_GET_MAX((GIDX*)v->spl_rdatum, direction) );
PG_RETURN_POINTER(v);
}
/*
* _bt_readpage() -- Load data from current index page into so->currPos
*
* Caller must have pinned and read-locked so->currPos.buf; the buffer's state
* is not changed here. Also, currPos.moreLeft and moreRight must be valid;
* they are updated as appropriate. All other fields of so->currPos are
* initialized from scratch here.
*
* We scan the current page starting at offnum and moving in the indicated
* direction. All items matching the scan keys are loaded into currPos.items.
* moreLeft or moreRight (as appropriate) is cleared if _bt_checkkeys reports
* that there can be no more matching tuples in the current scan direction.
*
* Returns true if any matching items found on the page, false if none.
*/
static bool
_bt_readpage(IndexScanDesc scan, ScanDirection dir, OffsetNumber offnum)
{
BTScanOpaque so = (BTScanOpaque) scan->opaque;
Page page;
BTPageOpaque opaque;
OffsetNumber minoff;
OffsetNumber maxoff;
int itemIndex;
bool continuescan;
/* we must have the buffer pinned and locked */
Assert(BufferIsValid(so->currPos.buf));
page = BufferGetPage(so->currPos.buf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
minoff = P_FIRSTDATAKEY(opaque);
maxoff = PageGetMaxOffsetNumber(page);
/*
* we must save the page's right-link while scanning it; this tells us
* where to step right to after we're done with these items. There is no
* corresponding need for the left-link, since splits always go right.
*/
so->currPos.nextPage = opaque->btpo_next;
if (ScanDirectionIsForward(dir))
{
/* load items[] in ascending order */
itemIndex = 0;
offnum = Max(offnum, minoff);
while (offnum <= maxoff)
{
if (_bt_checkkeys(scan, page, offnum, dir, &continuescan))
{
/* tuple passes all scan key conditions, so remember it */
/* _bt_checkkeys put the heap ptr into scan->xs_ctup.t_self */
so->currPos.items[itemIndex].heapTid = scan->xs_ctup.t_self;
so->currPos.items[itemIndex].indexOffset = offnum;
itemIndex++;
}
if (!continuescan)
{
/* there can't be any more matches, so stop */
so->currPos.moreRight = false;
break;
}
offnum = OffsetNumberNext(offnum);
}
Assert(itemIndex <= MaxIndexTuplesPerPage);
so->currPos.firstItem = 0;
so->currPos.lastItem = itemIndex - 1;
so->currPos.itemIndex = 0;
}
else
{
/* load items[] in descending order */
itemIndex = MaxIndexTuplesPerPage;
offnum = Min(offnum, maxoff);
while (offnum >= minoff)
{
if (_bt_checkkeys(scan, page, offnum, dir, &continuescan))
{
/* tuple passes all scan key conditions, so remember it */
/* _bt_checkkeys put the heap ptr into scan->xs_ctup.t_self */
itemIndex--;
so->currPos.items[itemIndex].heapTid = scan->xs_ctup.t_self;
so->currPos.items[itemIndex].indexOffset = offnum;
}
if (!continuescan)
{
/* there can't be any more matches, so stop */
so->currPos.moreLeft = false;
break;
}
offnum = OffsetNumberPrev(offnum);
}
Assert(itemIndex >= 0);
so->currPos.firstItem = itemIndex;
so->currPos.lastItem = MaxIndexTuplesPerPage - 1;
so->currPos.itemIndex = MaxIndexTuplesPerPage - 1;
}
return (so->currPos.firstItem <= so->currPos.lastItem);
}
/*
* _hash_squeezebucket(rel, bucket)
*
* Try to squeeze the tuples onto pages occurring earlier in the
* bucket chain in an attempt to free overflow pages. When we start
* the "squeezing", the page from which we start taking tuples (the
* "read" page) is the last bucket in the bucket chain and the page
* onto which we start squeezing tuples (the "write" page) is the
* first page in the bucket chain. The read page works backward and
* the write page works forward; the procedure terminates when the
* read page and write page are the same page.
*
* At completion of this procedure, it is guaranteed that all pages in
* the bucket are nonempty, unless the bucket is totally empty (in
* which case all overflow pages will be freed). The original implementation
* required that to be true on entry as well, but it's a lot easier for
* callers to leave empty overflow pages and let this guy clean it up.
*
* Caller must hold exclusive lock on the target bucket. This allows
* us to safely lock multiple pages in the bucket.
*
* Since this function is invoked in VACUUM, we provide an access strategy
* parameter that controls fetches of the bucket pages.
*/
void
_hash_squeezebucket(Relation rel,
Bucket bucket,
BlockNumber bucket_blkno,
BufferAccessStrategy bstrategy)
{
BlockNumber wblkno;
BlockNumber rblkno;
Buffer wbuf;
Buffer rbuf;
Page wpage;
Page rpage;
HashPageOpaque wopaque;
HashPageOpaque ropaque;
bool wbuf_dirty;
/*
* start squeezing into the base bucket page.
*/
wblkno = bucket_blkno;
wbuf = _hash_getbuf_with_strategy(rel,
wblkno,
HASH_WRITE,
LH_BUCKET_PAGE,
bstrategy);
wpage = BufferGetPage(wbuf);
wopaque = (HashPageOpaque) PageGetSpecialPointer(wpage);
/*
* if there aren't any overflow pages, there's nothing to squeeze.
*/
if (!BlockNumberIsValid(wopaque->hasho_nextblkno))
{
_hash_relbuf(rel, wbuf);
return;
}
/*
* Find the last page in the bucket chain by starting at the base bucket
* page and working forward. Note: we assume that a hash bucket chain is
* usually smaller than the buffer ring being used by VACUUM, else using
* the access strategy here would be counterproductive.
*/
rbuf = InvalidBuffer;
ropaque = wopaque;
do
{
rblkno = ropaque->hasho_nextblkno;
if (rbuf != InvalidBuffer)
_hash_relbuf(rel, rbuf);
rbuf = _hash_getbuf_with_strategy(rel,
rblkno,
HASH_WRITE,
LH_OVERFLOW_PAGE,
bstrategy);
rpage = BufferGetPage(rbuf);
ropaque = (HashPageOpaque) PageGetSpecialPointer(rpage);
Assert(ropaque->hasho_bucket == bucket);
} while (BlockNumberIsValid(ropaque->hasho_nextblkno));
/*
* squeeze the tuples.
*/
wbuf_dirty = false;
for (;;)
{
OffsetNumber roffnum;
OffsetNumber maxroffnum;
OffsetNumber deletable[MaxOffsetNumber];
int ndeletable = 0;
/* Scan each tuple in "read" page */
maxroffnum = PageGetMaxOffsetNumber(rpage);
for (roffnum = FirstOffsetNumber;
roffnum <= maxroffnum;
roffnum = OffsetNumberNext(roffnum))
{
IndexTuple itup;
Size itemsz;
itup = (IndexTuple) PageGetItem(rpage,
PageGetItemId(rpage, roffnum));
itemsz = IndexTupleDSize(*itup);
itemsz = MAXALIGN(itemsz);
/*
* Walk up the bucket chain, looking for a page big enough for
* this item. Exit if we reach the read page.
*/
while (PageGetFreeSpace(wpage) < itemsz)
{
Assert(!PageIsEmpty(wpage));
wblkno = wopaque->hasho_nextblkno;
Assert(BlockNumberIsValid(wblkno));
if (wbuf_dirty)
_hash_wrtbuf(rel, wbuf);
else
_hash_relbuf(rel, wbuf);
/* nothing more to do if we reached the read page */
if (rblkno == wblkno)
{
if (ndeletable > 0)
{
/* Delete tuples we already moved off read page */
PageIndexMultiDelete(rpage, deletable, ndeletable);
_hash_wrtbuf(rel, rbuf);
}
else
_hash_relbuf(rel, rbuf);
return;
}
wbuf = _hash_getbuf_with_strategy(rel,
wblkno,
HASH_WRITE,
LH_OVERFLOW_PAGE,
bstrategy);
wpage = BufferGetPage(wbuf);
wopaque = (HashPageOpaque) PageGetSpecialPointer(wpage);
Assert(wopaque->hasho_bucket == bucket);
wbuf_dirty = false;
}
/*
* we have found room so insert on the "write" page, being careful
* to preserve hashkey ordering. (If we insert many tuples into
* the same "write" page it would be worth qsort'ing instead of
* doing repeated _hash_pgaddtup.)
*/
(void) _hash_pgaddtup(rel, wbuf, itemsz, itup);
wbuf_dirty = true;
/* remember tuple for deletion from "read" page */
deletable[ndeletable++] = roffnum;
}
/*
* If we reach here, there are no live tuples on the "read" page ---
* it was empty when we got to it, or we moved them all. So we can
* just free the page without bothering with deleting tuples
* individually. Then advance to the previous "read" page.
*
* Tricky point here: if our read and write pages are adjacent in the
* bucket chain, our write lock on wbuf will conflict with
* _hash_freeovflpage's attempt to update the sibling links of the
* removed page. However, in that case we are done anyway, so we can
* simply drop the write lock before calling _hash_freeovflpage.
*/
rblkno = ropaque->hasho_prevblkno;
Assert(BlockNumberIsValid(rblkno));
/* are we freeing the page adjacent to wbuf? */
if (rblkno == wblkno)
{
/* yes, so release wbuf lock first */
if (wbuf_dirty)
_hash_wrtbuf(rel, wbuf);
else
_hash_relbuf(rel, wbuf);
/* free this overflow page (releases rbuf) */
_hash_freeovflpage(rel, rbuf, bstrategy);
/* done */
return;
}
/* free this overflow page, then get the previous one */
_hash_freeovflpage(rel, rbuf, bstrategy);
rbuf = _hash_getbuf_with_strategy(rel,
rblkno,
HASH_WRITE,
LH_OVERFLOW_PAGE,
bstrategy);
rpage = BufferGetPage(rbuf);
ropaque = (HashPageOpaque) PageGetSpecialPointer(rpage);
Assert(ropaque->hasho_bucket == bucket);
}
/* NOTREACHED */
}
/*
* Extract all item values from a BRIN index page
*
* Usage: SELECT * FROM brin_page_items(get_raw_page('idx', 1), 'idx'::regclass);
*/
Datum
brin_page_items(PG_FUNCTION_ARGS)
{
brin_page_state *state;
FuncCallContext *fctx;
if (!superuser())
ereport(ERROR,
(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
(errmsg("must be superuser to use raw page functions"))));
if (SRF_IS_FIRSTCALL())
{
bytea *raw_page = PG_GETARG_BYTEA_P(0);
Oid indexRelid = PG_GETARG_OID(1);
Page page;
TupleDesc tupdesc;
MemoryContext mctx;
Relation indexRel;
AttrNumber attno;
/* minimally verify the page we got */
page = verify_brin_page(raw_page, BRIN_PAGETYPE_REGULAR, "regular");
/* create a function context for cross-call persistence */
fctx = SRF_FIRSTCALL_INIT();
/* switch to memory context appropriate for multiple function calls */
mctx = MemoryContextSwitchTo(fctx->multi_call_memory_ctx);
/* Build a tuple descriptor for our result type */
if (get_call_result_type(fcinfo, NULL, &tupdesc) != TYPEFUNC_COMPOSITE)
elog(ERROR, "return type must be a row type");
indexRel = index_open(indexRelid, AccessShareLock);
state = palloc(offsetof(brin_page_state, columns) +
sizeof(brin_column_state) * RelationGetDescr(indexRel)->natts);
state->bdesc = brin_build_desc(indexRel);
state->page = page;
state->offset = FirstOffsetNumber;
state->unusedItem = false;
state->done = false;
state->dtup = NULL;
/*
* Initialize output functions for all indexed datatypes; simplifies
* calling them later.
*/
for (attno = 1; attno <= state->bdesc->bd_tupdesc->natts; attno++)
{
Oid output;
bool isVarlena;
BrinOpcInfo *opcinfo;
int i;
brin_column_state *column;
opcinfo = state->bdesc->bd_info[attno - 1];
column = palloc(offsetof(brin_column_state, outputFn) +
sizeof(FmgrInfo) * opcinfo->oi_nstored);
column->nstored = opcinfo->oi_nstored;
for (i = 0; i < opcinfo->oi_nstored; i++)
{
getTypeOutputInfo(opcinfo->oi_typids[i], &output, &isVarlena);
fmgr_info(output, &column->outputFn[i]);
}
state->columns[attno - 1] = column;
}
index_close(indexRel, AccessShareLock);
fctx->user_fctx = state;
fctx->tuple_desc = BlessTupleDesc(tupdesc);
MemoryContextSwitchTo(mctx);
}
fctx = SRF_PERCALL_SETUP();
state = fctx->user_fctx;
if (!state->done)
{
HeapTuple result;
Datum values[7];
bool nulls[7];
/*
* This loop is called once for every attribute of every tuple in the
* page. At the start of a tuple, we get a NULL dtup; that's our
* signal for obtaining and decoding the next one. If that's not the
* case, we output the next attribute.
*/
if (state->dtup == NULL)
{
BrinTuple *tup;
MemoryContext mctx;
ItemId itemId;
/* deformed tuple must live across calls */
mctx = MemoryContextSwitchTo(fctx->multi_call_memory_ctx);
/* verify item status: if there's no data, we can't decode */
itemId = PageGetItemId(state->page, state->offset);
if (ItemIdIsUsed(itemId))
{
tup = (BrinTuple *) PageGetItem(state->page,
PageGetItemId(state->page,
state->offset));
state->dtup = brin_deform_tuple(state->bdesc, tup);
state->attno = 1;
state->unusedItem = false;
}
else
state->unusedItem = true;
MemoryContextSwitchTo(mctx);
}
else
state->attno++;
MemSet(nulls, 0, sizeof(nulls));
if (state->unusedItem)
{
values[0] = UInt16GetDatum(state->offset);
nulls[1] = true;
nulls[2] = true;
nulls[3] = true;
nulls[4] = true;
nulls[5] = true;
nulls[6] = true;
}
else
{
int att = state->attno - 1;
values[0] = UInt16GetDatum(state->offset);
values[1] = UInt32GetDatum(state->dtup->bt_blkno);
values[2] = UInt16GetDatum(state->attno);
values[3] = BoolGetDatum(state->dtup->bt_columns[att].bv_allnulls);
values[4] = BoolGetDatum(state->dtup->bt_columns[att].bv_hasnulls);
values[5] = BoolGetDatum(state->dtup->bt_placeholder);
if (!state->dtup->bt_columns[att].bv_allnulls)
{
BrinValues *bvalues = &state->dtup->bt_columns[att];
StringInfoData s;
bool first;
int i;
initStringInfo(&s);
appendStringInfoChar(&s, '{');
first = true;
for (i = 0; i < state->columns[att]->nstored; i++)
{
char *val;
if (!first)
appendStringInfoString(&s, " .. ");
first = false;
val = OutputFunctionCall(&state->columns[att]->outputFn[i],
bvalues->bv_values[i]);
appendStringInfoString(&s, val);
pfree(val);
}
appendStringInfoChar(&s, '}');
values[6] = CStringGetTextDatum(s.data);
pfree(s.data);
}
else
{
nulls[6] = true;
}
}
result = heap_form_tuple(fctx->tuple_desc, values, nulls);
/*
* If the item was unused, jump straight to the next one; otherwise,
* the only cleanup needed here is to set our signal to go to the next
* tuple in the following iteration, by freeing the current one.
*/
if (state->unusedItem)
state->offset = OffsetNumberNext(state->offset);
else if (state->attno >= state->bdesc->bd_tupdesc->natts)
{
pfree(state->dtup);
state->dtup = NULL;
state->offset = OffsetNumberNext(state->offset);
}
/*
* If we're beyond the end of the page, set flag to end the function in
* the following iteration.
*/
if (state->offset > PageGetMaxOffsetNumber(state->page))
state->done = true;
SRF_RETURN_NEXT(fctx, HeapTupleGetDatum(result));
}
brin_free_desc(state->bdesc);
SRF_RETURN_DONE(fctx);
}
/*----------
* Add an item to a disk page from the sort output.
*
* We must be careful to observe the page layout conventions of nbtsearch.c:
* - rightmost pages start data items at P_HIKEY instead of at P_FIRSTKEY.
* - on non-leaf pages, the key portion of the first item need not be
* stored, we should store only the link.
*
* A leaf page being built looks like:
*
* +----------------+---------------------------------+
* | PageHeaderData | linp0 linp1 linp2 ... |
* +-----------+----+---------------------------------+
* | ... linpN | |
* +-----------+--------------------------------------+
* | ^ last |
* | |
* +-------------+------------------------------------+
* | | itemN ... |
* +-------------+------------------+-----------------+
* | ... item3 item2 item1 | "special space" |
* +--------------------------------+-----------------+
*
* Contrast this with the diagram in bufpage.h; note the mismatch
* between linps and items. This is because we reserve linp0 as a
* placeholder for the pointer to the "high key" item; when we have
* filled up the page, we will set linp0 to point to itemN and clear
* linpN. On the other hand, if we find this is the last (rightmost)
* page, we leave the items alone and slide the linp array over.
*
* 'last' pointer indicates the last offset added to the page.
*----------
*/
static void
_bt_buildadd(BTWriteState *wstate, BTPageState *state, IndexTuple itup)
{
Page npage;
BlockNumber nblkno;
OffsetNumber last_off;
Size pgspc;
Size itupsz;
/*
* This is a handy place to check for cancel interrupts during the btree
* load phase of index creation.
*/
CHECK_FOR_INTERRUPTS();
npage = state->btps_page;
nblkno = state->btps_blkno;
last_off = state->btps_lastoff;
pgspc = PageGetFreeSpace(npage);
itupsz = IndexTupleDSize(*itup);
itupsz = MAXALIGN(itupsz);
/*
* Check whether the item can fit on a btree page at all. (Eventually, we
* ought to try to apply TOAST methods if not.) We actually need to be
* able to fit three items on every page, so restrict any one item to 1/3
* the per-page available space. Note that at this point, itupsz doesn't
* include the ItemId.
*
* NOTE: similar code appears in _bt_insertonpg() to defend against
* oversize items being inserted into an already-existing index. But
* during creation of an index, we don't go through there.
*/
if (itupsz > BTMaxItemSize(npage))
ereport(ERROR,
(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
errmsg("index row size %zu exceeds maximum %zu for index \"%s\"",
itupsz, BTMaxItemSize(npage),
RelationGetRelationName(wstate->index)),
errhint("Values larger than 1/3 of a buffer page cannot be indexed.\n"
"Consider a function index of an MD5 hash of the value, "
"or use full text indexing."),
errtableconstraint(wstate->heap,
RelationGetRelationName(wstate->index))));
/*
* Check to see if page is "full". It's definitely full if the item won't
* fit. Otherwise, compare to the target freespace derived from the
* fillfactor. However, we must put at least two items on each page, so
* disregard fillfactor if we don't have that many.
*/
if (pgspc < itupsz || (pgspc < state->btps_full && last_off > P_FIRSTKEY))
{
/*
* Finish off the page and write it out.
*/
Page opage = npage;
BlockNumber oblkno = nblkno;
ItemId ii;
ItemId hii;
IndexTuple oitup;
/* Create new page of same level */
npage = _bt_blnewpage(state->btps_level);
/* and assign it a page position */
nblkno = wstate->btws_pages_alloced++;
/*
* We copy the last item on the page into the new page, and then
* rearrange the old page so that the 'last item' becomes its high key
* rather than a true data item. There had better be at least two
* items on the page already, else the page would be empty of useful
* data.
*/
Assert(last_off > P_FIRSTKEY);
ii = PageGetItemId(opage, last_off);
oitup = (IndexTuple) PageGetItem(opage, ii);
_bt_sortaddtup(npage, ItemIdGetLength(ii), oitup, P_FIRSTKEY);
/*
* Move 'last' into the high key position on opage
*/
hii = PageGetItemId(opage, P_HIKEY);
*hii = *ii;
ItemIdSetUnused(ii); /* redundant */
((PageHeader) opage)->pd_lower -= sizeof(ItemIdData);
/*
* Link the old page into its parent, using its minimum key. If we
* don't have a parent, we have to create one; this adds a new btree
* level.
*/
if (state->btps_next == NULL)
state->btps_next = _bt_pagestate(wstate, state->btps_level + 1);
Assert(state->btps_minkey != NULL);
ItemPointerSet(&(state->btps_minkey->t_tid), oblkno, P_HIKEY);
_bt_buildadd(wstate, state->btps_next, state->btps_minkey);
pfree(state->btps_minkey);
/*
* Save a copy of the minimum key for the new page. We have to copy
* it off the old page, not the new one, in case we are not at leaf
* level.
*/
state->btps_minkey = CopyIndexTuple(oitup);
/*
* Set the sibling links for both pages.
*/
{
BTPageOpaque oopaque = (BTPageOpaque) PageGetSpecialPointer(opage);
BTPageOpaque nopaque = (BTPageOpaque) PageGetSpecialPointer(npage);
oopaque->btpo_next = nblkno;
nopaque->btpo_prev = oblkno;
nopaque->btpo_next = P_NONE; /* redundant */
}
/*
* Write out the old page. We never need to touch it again, so we can
* free the opage workspace too.
*/
_bt_blwritepage(wstate, opage, oblkno);
/*
* Reset last_off to point to new page
*/
last_off = P_FIRSTKEY;
}
/*
* If the new item is the first for its page, stash a copy for later. Note
* this will only happen for the first item on a level; on later pages,
* the first item for a page is copied from the prior page in the code
* above.
*/
if (last_off == P_HIKEY)
{
Assert(state->btps_minkey == NULL);
state->btps_minkey = CopyIndexTuple(itup);
}
/*
* Add the new item into the current page.
*/
last_off = OffsetNumberNext(last_off);
_bt_sortaddtup(npage, itupsz, itup, last_off);
state->btps_page = npage;
state->btps_blkno = nblkno;
state->btps_lastoff = last_off;
}
/*
* _hash_splitbucket -- split 'obucket' into 'obucket' and 'nbucket'
*
* We are splitting a bucket that consists of a base bucket page and zero
* or more overflow (bucket chain) pages. We must relocate tuples that
* belong in the new bucket, and compress out any free space in the old
* bucket.
*
* The caller must hold exclusive locks on both buckets to ensure that
* no one else is trying to access them (see README).
*
* The caller must hold a pin, but no lock, on the metapage buffer.
* The buffer is returned in the same state. (The metapage is only
* touched if it becomes necessary to add or remove overflow pages.)
*/
static void
_hash_splitbucket(Relation rel,
Buffer metabuf,
Bucket obucket,
Bucket nbucket,
BlockNumber start_oblkno,
BlockNumber start_nblkno,
uint32 maxbucket,
uint32 highmask,
uint32 lowmask)
{
Bucket bucket;
Buffer obuf;
Buffer nbuf;
BlockNumber oblkno;
BlockNumber nblkno;
bool null;
Datum datum;
HashPageOpaque oopaque;
HashPageOpaque nopaque;
IndexTuple itup;
Size itemsz;
OffsetNumber ooffnum;
OffsetNumber noffnum;
OffsetNumber omaxoffnum;
Page opage;
Page npage;
TupleDesc itupdesc = RelationGetDescr(rel);
MIRROREDLOCK_BUFMGR_MUST_ALREADY_BE_HELD;
/*
* It should be okay to simultaneously write-lock pages from each bucket,
* since no one else can be trying to acquire buffer lock on pages of
* either bucket.
*/
oblkno = start_oblkno;
obuf = _hash_getbuf(rel, oblkno, HASH_WRITE);
_hash_checkpage(rel, obuf, LH_BUCKET_PAGE);
opage = BufferGetPage(obuf);
oopaque = (HashPageOpaque) PageGetSpecialPointer(opage);
nblkno = start_nblkno;
nbuf = _hash_getbuf(rel, nblkno, HASH_WRITE);
npage = BufferGetPage(nbuf);
/* initialize the new bucket's primary page */
_hash_pageinit(npage, BufferGetPageSize(nbuf));
nopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
nopaque->hasho_prevblkno = InvalidBlockNumber;
nopaque->hasho_nextblkno = InvalidBlockNumber;
nopaque->hasho_bucket = nbucket;
nopaque->hasho_flag = LH_BUCKET_PAGE;
nopaque->hasho_filler = HASHO_FILL;
/*
* Partition the tuples in the old bucket between the old bucket and the
* new bucket, advancing along the old bucket's overflow bucket chain and
* adding overflow pages to the new bucket as needed.
*/
ooffnum = FirstOffsetNumber;
omaxoffnum = PageGetMaxOffsetNumber(opage);
for (;;)
{
/*
* at each iteration through this loop, each of these variables should
* be up-to-date: obuf opage oopaque ooffnum omaxoffnum
*/
/* check if we're at the end of the page */
if (ooffnum > omaxoffnum)
{
/* at end of page, but check for an(other) overflow page */
oblkno = oopaque->hasho_nextblkno;
if (!BlockNumberIsValid(oblkno))
break;
/*
* we ran out of tuples on this particular page, but we have more
* overflow pages; advance to next page.
*/
_hash_wrtbuf(rel, obuf);
obuf = _hash_getbuf(rel, oblkno, HASH_WRITE);
_hash_checkpage(rel, obuf, LH_OVERFLOW_PAGE);
opage = BufferGetPage(obuf);
oopaque = (HashPageOpaque) PageGetSpecialPointer(opage);
ooffnum = FirstOffsetNumber;
omaxoffnum = PageGetMaxOffsetNumber(opage);
continue;
}
/*
* Re-hash the tuple to determine which bucket it now belongs in.
*
* It is annoying to call the hash function while holding locks, but
* releasing and relocking the page for each tuple is unappealing too.
*/
itup = (IndexTuple) PageGetItem(opage, PageGetItemId(opage, ooffnum));
datum = index_getattr(itup, 1, itupdesc, &null);
Assert(!null);
bucket = _hash_hashkey2bucket(_hash_datum2hashkey(rel, datum),
maxbucket, highmask, lowmask);
if (bucket == nbucket)
{
/*
* insert the tuple into the new bucket. if it doesn't fit on the
* current page in the new bucket, we must allocate a new overflow
* page and place the tuple on that page instead.
*/
itemsz = IndexTupleDSize(*itup);
itemsz = MAXALIGN(itemsz);
if (PageGetFreeSpace(npage) < itemsz)
{
/* write out nbuf and drop lock, but keep pin */
_hash_chgbufaccess(rel, nbuf, HASH_WRITE, HASH_NOLOCK);
/* chain to a new overflow page */
nbuf = _hash_addovflpage(rel, metabuf, nbuf);
_hash_checkpage(rel, nbuf, LH_OVERFLOW_PAGE);
npage = BufferGetPage(nbuf);
/* we don't need nopaque within the loop */
}
noffnum = OffsetNumberNext(PageGetMaxOffsetNumber(npage));
if (PageAddItem(npage, (Item) itup, itemsz, noffnum, LP_USED)
== InvalidOffsetNumber)
elog(ERROR, "failed to add index item to \"%s\"",
RelationGetRelationName(rel));
/*
* now delete the tuple from the old bucket. after this section
* of code, 'ooffnum' will actually point to the ItemId to which
* we would point if we had advanced it before the deletion
* (PageIndexTupleDelete repacks the ItemId array). this also
* means that 'omaxoffnum' is exactly one less than it used to be,
* so we really can just decrement it instead of calling
* PageGetMaxOffsetNumber.
*/
PageIndexTupleDelete(opage, ooffnum);
omaxoffnum = OffsetNumberPrev(omaxoffnum);
}
else
{
/*
* the tuple stays on this page. we didn't move anything, so we
* didn't delete anything and therefore we don't have to change
* 'omaxoffnum'.
*/
Assert(bucket == obucket);
ooffnum = OffsetNumberNext(ooffnum);
}
}
/*
* We're at the end of the old bucket chain, so we're done partitioning
* the tuples. Before quitting, call _hash_squeezebucket to ensure the
* tuples remaining in the old bucket (including the overflow pages) are
* packed as tightly as possible. The new bucket is already tight.
*/
_hash_wrtbuf(rel, obuf);
_hash_wrtbuf(rel, nbuf);
_hash_squeezebucket(rel, obucket, start_oblkno);
}
/*
* _hash_step() -- step to the next valid item in a scan in the bucket.
*
* If no valid record exists in the requested direction, return
* false. Else, return true and set the hashso_curpos for the
* scan to the right thing.
*
* 'bufP' points to the current buffer, which is pinned and read-locked.
* On success exit, we have pin and read-lock on whichever page
* contains the right item; on failure, we have released all buffers.
*/
bool
_hash_step(IndexScanDesc scan, Buffer *bufP, ScanDirection dir)
{
Relation rel = scan->indexRelation;
HashScanOpaque so = (HashScanOpaque) scan->opaque;
ItemPointer current;
Buffer buf;
Page page;
HashPageOpaque opaque;
OffsetNumber maxoff;
OffsetNumber offnum;
BlockNumber blkno;
IndexTuple itup;
current = &(so->hashso_curpos);
buf = *bufP;
_hash_checkpage(rel, buf, LH_BUCKET_PAGE | LH_OVERFLOW_PAGE);
page = BufferGetPage(buf);
opaque = (HashPageOpaque) PageGetSpecialPointer(page);
/*
* If _hash_step is called from _hash_first, current will not be valid, so
* we can't dereference it. However, in that case, we presumably want to
* start at the beginning/end of the page...
*/
maxoff = PageGetMaxOffsetNumber(page);
if (ItemPointerIsValid(current))
offnum = ItemPointerGetOffsetNumber(current);
else
offnum = InvalidOffsetNumber;
/*
* 'offnum' now points to the last tuple we examined (if any).
*
* continue to step through tuples until: 1) we get to the end of the
* bucket chain or 2) we find a valid tuple.
*/
do
{
switch (dir)
{
case ForwardScanDirection:
if (offnum != InvalidOffsetNumber)
offnum = OffsetNumberNext(offnum); /* move forward */
else
{
/* new page, locate starting position by binary search */
offnum = _hash_binsearch(page, so->hashso_sk_hash);
}
for (;;)
{
/*
* check if we're still in the range of items with the
* target hash key
*/
if (offnum <= maxoff)
{
Assert(offnum >= FirstOffsetNumber);
itup = (IndexTuple) PageGetItem(page, PageGetItemId(page, offnum));
if (so->hashso_sk_hash == _hash_get_indextuple_hashkey(itup))
break; /* yes, so exit for-loop */
}
/*
* ran off the end of this page, try the next
*/
_hash_readnext(rel, &buf, &page, &opaque);
if (BufferIsValid(buf))
{
maxoff = PageGetMaxOffsetNumber(page);
offnum = _hash_binsearch(page, so->hashso_sk_hash);
}
else
{
/* end of bucket */
itup = NULL;
break; /* exit for-loop */
}
}
break;
case BackwardScanDirection:
if (offnum != InvalidOffsetNumber)
offnum = OffsetNumberPrev(offnum); /* move back */
else
{
/* new page, locate starting position by binary search */
offnum = _hash_binsearch_last(page, so->hashso_sk_hash);
}
for (;;)
{
/*
* check if we're still in the range of items with the
* target hash key
*/
if (offnum >= FirstOffsetNumber)
{
Assert(offnum <= maxoff);
itup = (IndexTuple) PageGetItem(page, PageGetItemId(page, offnum));
if (so->hashso_sk_hash == _hash_get_indextuple_hashkey(itup))
break; /* yes, so exit for-loop */
}
/*
* ran off the end of this page, try the next
*/
_hash_readprev(rel, &buf, &page, &opaque);
if (BufferIsValid(buf))
{
maxoff = PageGetMaxOffsetNumber(page);
offnum = _hash_binsearch_last(page, so->hashso_sk_hash);
}
else
{
/* end of bucket */
itup = NULL;
break; /* exit for-loop */
}
}
break;
default:
/* NoMovementScanDirection */
/* this should not be reached */
itup = NULL;
break;
}
if (itup == NULL)
{
/* we ran off the end of the bucket without finding a match */
*bufP = so->hashso_curbuf = InvalidBuffer;
ItemPointerSetInvalid(current);
return false;
}
/* check the tuple quals, loop around if not met */
} while (!_hash_checkqual(scan, itup));
/* if we made it to here, we've found a valid tuple */
blkno = BufferGetBlockNumber(buf);
*bufP = so->hashso_curbuf = buf;
ItemPointerSet(current, blkno, offnum);
return true;
}
static void
bitmap_xlog_insert_lovitem(bool redo, XLogRecPtr lsn, XLogRecord* record)
{
xl_bm_lovitem *xlrec = (xl_bm_lovitem*) XLogRecGetData(record);
Relation reln;
reln = XLogOpenRelation(xlrec->bm_node);
if (!RelationIsValid(reln))
return;
if (redo)
{
Buffer lovBuffer;
Page lovPage;
#ifdef BM_DEBUG
ereport(LOG, (errcode(LOG),
errmsg("call bitmap_xlog_insert_lovitem: redo=%d, blkno=%d\n",
redo, xlrec->bm_lov_blkno)));
#endif
lovBuffer = XLogReadBuffer(false, reln, xlrec->bm_lov_blkno);
if (!BufferIsValid(lovBuffer))
elog(PANIC, "bm_insert_redo: block unfound: %d",
xlrec->bm_lov_blkno);
lovPage = BufferGetPage(lovBuffer);
if (XLByteLT(PageGetLSN(lovPage), lsn))
{
if(xlrec->bm_isNewItem)
{
OffsetNumber newOffset, itemSize;
newOffset = OffsetNumberNext(PageGetMaxOffsetNumber(lovPage));
if (newOffset != xlrec->bm_lov_offset)
elog(PANIC,
"bm_insert_redo: LOV item is not inserted in pos %d(requested %d)",
newOffset, xlrec->bm_lov_offset);
itemSize = sizeof(BMLOVItemData);
if (itemSize > PageGetFreeSpace(lovPage))
elog(PANIC,
"bm_insert_redo: not enough space in LOV page %d",
xlrec->bm_lov_blkno);
if (PageAddItem(lovPage, (Item)&(xlrec->bm_lovItem), itemSize,
newOffset, LP_USED) == InvalidOffsetNumber)
ereport(ERROR,
(errcode(ERRCODE_INTERNAL_ERROR),
errmsg("failed to add LOV item to \"%s\"",
RelationGetRelationName(reln))));
}
else{
BMLOVItem oldLovItem;
oldLovItem = (BMLOVItem)
PageGetItem(lovPage,
PageGetItemId(lovPage, xlrec->bm_lov_offset));
memcpy(oldLovItem, &(xlrec->bm_lovItem), sizeof(BMLOVItemData));
}
PageSetLSN(lovPage, lsn);
PageSetTLI(lovPage, ThisTimeLineID);
_bitmap_wrtbuf(lovBuffer);
}
else {
_bitmap_relbuf(lovBuffer);
}
}
else
elog(PANIC, "bm_insert_undo: not implemented.");
}
/*
* _bt_killitems - set LP_DEAD state for items an indexscan caller has
* told us were killed
*
* scan->so contains information about the current page and killed tuples
* thereon (generally, this should only be called if so->numKilled > 0).
*
* The caller must have pin on so->currPos.buf, but may or may not have
* read-lock, as indicated by haveLock. Note that we assume read-lock
* is sufficient for setting LP_DEAD status (which is only a hint).
*
* We match items by heap TID before assuming they are the right ones to
* delete. We cope with cases where items have moved right due to insertions.
* If an item has moved off the current page due to a split, we'll fail to
* find it and do nothing (this is not an error case --- we assume the item
* will eventually get marked in a future indexscan). Note that because we
* hold pin on the target page continuously from initially reading the items
* until applying this function, VACUUM cannot have deleted any items from
* the page, and so there is no need to search left from the recorded offset.
* (This observation also guarantees that the item is still the right one
* to delete, which might otherwise be questionable since heap TIDs can get
* recycled.)
*/
void
_bt_killitems(IndexScanDesc scan, bool haveLock)
{
BTScanOpaque so = (BTScanOpaque) scan->opaque;
Page page;
BTPageOpaque opaque;
OffsetNumber minoff;
OffsetNumber maxoff;
int i;
bool killedsomething = false;
Assert(BufferIsValid(so->currPos.buf));
if (!haveLock)
LockBuffer(so->currPos.buf, BT_READ);
page = BufferGetPage(so->currPos.buf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
minoff = P_FIRSTDATAKEY(opaque);
maxoff = PageGetMaxOffsetNumber(page);
for (i = 0; i < so->numKilled; i++)
{
int itemIndex = so->killedItems[i];
BTScanPosItem *kitem = &so->currPos.items[itemIndex];
OffsetNumber offnum = kitem->indexOffset;
Assert(itemIndex >= so->currPos.firstItem &&
itemIndex <= so->currPos.lastItem);
if (offnum < minoff)
continue; /* pure paranoia */
while (offnum <= maxoff)
{
ItemId iid = PageGetItemId(page, offnum);
IndexTuple ituple = (IndexTuple) PageGetItem(page, iid);
if (ItemPointerEquals(&ituple->t_tid, &kitem->heapTid))
{
/* found the item */
ItemIdMarkDead(iid);
killedsomething = true;
break; /* out of inner search loop */
}
offnum = OffsetNumberNext(offnum);
}
}
/*
* Since this can be redone later if needed, it's treated the same as a
* commit-hint-bit status update for heap tuples: we mark the buffer dirty
* but don't make a WAL log entry.
*
* Whenever we mark anything LP_DEAD, we also set the page's
* BTP_HAS_GARBAGE flag, which is likewise just a hint.
*/
if (killedsomething)
{
opaque->btpo_flags |= BTP_HAS_GARBAGE;
SetBufferCommitInfoNeedsSave(so->currPos.buf);
}
if (!haveLock)
LockBuffer(so->currPos.buf, BUFFER_LOCK_UNLOCK);
/*
* Always reset the scan state, so we don't look for same items on other
* pages.
*/
so->numKilled = 0;
}
/*
* _hash_finish_split() -- Finish the previously interrupted split operation
*
* To complete the split operation, we form the hash table of TIDs in new
* bucket which is then used by split operation to skip tuples that are
* already moved before the split operation was previously interrupted.
*
* The caller must hold a pin, but no lock, on the metapage and old bucket's
* primary page buffer. The buffers are returned in the same state. (The
* metapage is only touched if it becomes necessary to add or remove overflow
* pages.)
*/
void
_hash_finish_split(Relation rel, Buffer metabuf, Buffer obuf, Bucket obucket,
uint32 maxbucket, uint32 highmask, uint32 lowmask)
{
HASHCTL hash_ctl;
HTAB *tidhtab;
Buffer bucket_nbuf = InvalidBuffer;
Buffer nbuf;
Page npage;
BlockNumber nblkno;
BlockNumber bucket_nblkno;
HashPageOpaque npageopaque;
Bucket nbucket;
bool found;
/* Initialize hash tables used to track TIDs */
memset(&hash_ctl, 0, sizeof(hash_ctl));
hash_ctl.keysize = sizeof(ItemPointerData);
hash_ctl.entrysize = sizeof(ItemPointerData);
hash_ctl.hcxt = CurrentMemoryContext;
tidhtab =
hash_create("bucket ctids",
256, /* arbitrary initial size */
&hash_ctl,
HASH_ELEM | HASH_BLOBS | HASH_CONTEXT);
bucket_nblkno = nblkno = _hash_get_newblock_from_oldbucket(rel, obucket);
/*
* Scan the new bucket and build hash table of TIDs
*/
for (;;)
{
OffsetNumber noffnum;
OffsetNumber nmaxoffnum;
nbuf = _hash_getbuf(rel, nblkno, HASH_READ,
LH_BUCKET_PAGE | LH_OVERFLOW_PAGE);
/* remember the primary bucket buffer to acquire cleanup lock on it. */
if (nblkno == bucket_nblkno)
bucket_nbuf = nbuf;
npage = BufferGetPage(nbuf);
npageopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
/* Scan each tuple in new page */
nmaxoffnum = PageGetMaxOffsetNumber(npage);
for (noffnum = FirstOffsetNumber;
noffnum <= nmaxoffnum;
noffnum = OffsetNumberNext(noffnum))
{
IndexTuple itup;
/* Fetch the item's TID and insert it in hash table. */
itup = (IndexTuple) PageGetItem(npage,
PageGetItemId(npage, noffnum));
(void) hash_search(tidhtab, &itup->t_tid, HASH_ENTER, &found);
Assert(!found);
}
nblkno = npageopaque->hasho_nextblkno;
/*
* release our write lock without modifying buffer and ensure to
* retain the pin on primary bucket.
*/
if (nbuf == bucket_nbuf)
LockBuffer(nbuf, BUFFER_LOCK_UNLOCK);
else
_hash_relbuf(rel, nbuf);
/* Exit loop if no more overflow pages in new bucket */
if (!BlockNumberIsValid(nblkno))
break;
}
/*
* Conditionally get the cleanup lock on old and new buckets to perform
* the split operation. If we don't get the cleanup locks, silently give
* up and next insertion on old bucket will try again to complete the
* split.
*/
if (!ConditionalLockBufferForCleanup(obuf))
{
hash_destroy(tidhtab);
return;
}
if (!ConditionalLockBufferForCleanup(bucket_nbuf))
{
LockBuffer(obuf, BUFFER_LOCK_UNLOCK);
hash_destroy(tidhtab);
return;
}
npage = BufferGetPage(bucket_nbuf);
npageopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
nbucket = npageopaque->hasho_bucket;
_hash_splitbucket(rel, metabuf, obucket,
nbucket, obuf, bucket_nbuf, tidhtab,
maxbucket, highmask, lowmask);
_hash_dropbuf(rel, bucket_nbuf);
hash_destroy(tidhtab);
}
/*
* Scan all items on the GiST index page identified by *pageItem, and insert
* them into the queue (or directly to output areas)
*
* scan: index scan we are executing
* pageItem: search queue item identifying an index page to scan
* myDistances: distances array associated with pageItem, or NULL at the root
* tbm: if not NULL, gistgetbitmap's output bitmap
* ntids: if not NULL, gistgetbitmap's output tuple counter
*
* If tbm/ntids aren't NULL, we are doing an amgetbitmap scan, and heap
* tuples should be reported directly into the bitmap. If they are NULL,
* we're doing a plain or ordered indexscan. For a plain indexscan, heap
* tuple TIDs are returned into so->pageData[]. For an ordered indexscan,
* heap tuple TIDs are pushed into individual search queue items.
*
* If we detect that the index page has split since we saw its downlink
* in the parent, we push its new right sibling onto the queue so the
* sibling will be processed next.
*/
static void
gistScanPage(IndexScanDesc scan, GISTSearchItem *pageItem, double *myDistances,
TIDBitmap *tbm, int64 *ntids)
{
GISTScanOpaque so = (GISTScanOpaque) scan->opaque;
Buffer buffer;
Page page;
GISTPageOpaque opaque;
OffsetNumber maxoff;
OffsetNumber i;
GISTSearchTreeItem *tmpItem = so->tmpTreeItem;
bool isNew;
MemoryContext oldcxt;
Assert(!GISTSearchItemIsHeap(*pageItem));
buffer = ReadBuffer(scan->indexRelation, pageItem->blkno);
LockBuffer(buffer, GIST_SHARE);
gistcheckpage(scan->indexRelation, buffer);
page = BufferGetPage(buffer);
opaque = GistPageGetOpaque(page);
/*
* Check if we need to follow the rightlink. We need to follow it if the
* page was concurrently split since we visited the parent (in which case
* parentlsn < nsn), or if the the system crashed after a page split but
* before the downlink was inserted into the parent.
*/
if (!XLogRecPtrIsInvalid(pageItem->data.parentlsn) &&
(GistFollowRight(page) ||
XLByteLT(pageItem->data.parentlsn, opaque->nsn)) &&
opaque->rightlink != InvalidBlockNumber /* sanity check */ )
{
/* There was a page split, follow right link to add pages */
GISTSearchItem *item;
/* This can't happen when starting at the root */
Assert(myDistances != NULL);
oldcxt = MemoryContextSwitchTo(so->queueCxt);
/* Create new GISTSearchItem for the right sibling index page */
item = palloc(sizeof(GISTSearchItem));
item->next = NULL;
item->blkno = opaque->rightlink;
item->data.parentlsn = pageItem->data.parentlsn;
/* Insert it into the queue using same distances as for this page */
tmpItem->head = item;
tmpItem->lastHeap = NULL;
memcpy(tmpItem->distances, myDistances,
sizeof(double) * scan->numberOfOrderBys);
(void) rb_insert(so->queue, (RBNode *) tmpItem, &isNew);
MemoryContextSwitchTo(oldcxt);
}
so->nPageData = so->curPageData = 0;
/*
* check all tuples on page
*/
maxoff = PageGetMaxOffsetNumber(page);
for (i = FirstOffsetNumber; i <= maxoff; i = OffsetNumberNext(i))
{
IndexTuple it = (IndexTuple) PageGetItem(page, PageGetItemId(page, i));
bool match;
bool recheck;
/*
* Must call gistindex_keytest in tempCxt, and clean up any leftover
* junk afterward.
*/
oldcxt = MemoryContextSwitchTo(so->giststate->tempCxt);
match = gistindex_keytest(scan, it, page, i, &recheck);
MemoryContextSwitchTo(oldcxt);
MemoryContextReset(so->giststate->tempCxt);
/* Ignore tuple if it doesn't match */
if (!match)
continue;
if (tbm && GistPageIsLeaf(page))
{
/*
* getbitmap scan, so just push heap tuple TIDs into the bitmap
* without worrying about ordering
*/
tbm_add_tuples(tbm, &it->t_tid, 1, recheck);
(*ntids)++;
}
else if (scan->numberOfOrderBys == 0 && GistPageIsLeaf(page))
{
/*
* Non-ordered scan, so report heap tuples in so->pageData[]
*/
so->pageData[so->nPageData].heapPtr = it->t_tid;
so->pageData[so->nPageData].recheck = recheck;
so->nPageData++;
}
else
{
/*
* Must push item into search queue. We get here for any lower
* index page, and also for heap tuples if doing an ordered
* search.
*/
GISTSearchItem *item;
oldcxt = MemoryContextSwitchTo(so->queueCxt);
/* Create new GISTSearchItem for this item */
item = palloc(sizeof(GISTSearchItem));
item->next = NULL;
if (GistPageIsLeaf(page))
{
/* Creating heap-tuple GISTSearchItem */
item->blkno = InvalidBlockNumber;
item->data.heap.heapPtr = it->t_tid;
item->data.heap.recheck = recheck;
}
else
{
/* Creating index-page GISTSearchItem */
item->blkno = ItemPointerGetBlockNumber(&it->t_tid);
/* lsn of current page is lsn of parent page for child */
item->data.parentlsn = PageGetLSN(page);
}
/* Insert it into the queue using new distance data */
tmpItem->head = item;
tmpItem->lastHeap = GISTSearchItemIsHeap(*item) ? item : NULL;
memcpy(tmpItem->distances, so->distances,
sizeof(double) * scan->numberOfOrderBys);
(void) rb_insert(so->queue, (RBNode *) tmpItem, &isNew);
MemoryContextSwitchTo(oldcxt);
}
}
UnlockReleaseBuffer(buffer);
}
/*
* Traverse the tree to find path from root page to specified "child" block.
*
* returns a new insertion stack, starting from the parent of "child", up
* to the root. *downlinkoffnum is set to the offset of the downlink in the
* direct parent of child.
*
* To prevent deadlocks, this should lock only one page at a time.
*/
static GISTInsertStack *
gistFindPath(Relation r, BlockNumber child, OffsetNumber *downlinkoffnum)
{
Page page;
Buffer buffer;
OffsetNumber i,
maxoff;
ItemId iid;
IndexTuple idxtuple;
List *fifo;
GISTInsertStack *top,
*ptr;
BlockNumber blkno;
top = (GISTInsertStack *) palloc0(sizeof(GISTInsertStack));
top->blkno = GIST_ROOT_BLKNO;
top->downlinkoffnum = InvalidOffsetNumber;
fifo = list_make1(top);
while (fifo != NIL)
{
/* Get next page to visit */
top = linitial(fifo);
fifo = list_delete_first(fifo);
buffer = ReadBuffer(r, top->blkno);
LockBuffer(buffer, GIST_SHARE);
gistcheckpage(r, buffer);
page = (Page) BufferGetPage(buffer);
if (GistPageIsLeaf(page))
{
/*
* Because we scan the index top-down, all the rest of the pages
* in the queue must be leaf pages as well.
*/
UnlockReleaseBuffer(buffer);
break;
}
top->lsn = PageGetLSN(page);
/*
* If F_FOLLOW_RIGHT is set, the page to the right doesn't have a
* downlink. This should not normally happen..
*/
if (GistFollowRight(page))
elog(ERROR, "concurrent GiST page split was incomplete");
if (top->parent && top->parent->lsn < GistPageGetNSN(page) &&
GistPageGetOpaque(page)->rightlink != InvalidBlockNumber /* sanity check */ )
{
/*
* Page was split while we looked elsewhere. We didn't see the
* downlink to the right page when we scanned the parent, so add
* it to the queue now.
*
* Put the right page ahead of the queue, so that we visit it
* next. That's important, because if this is the lowest internal
* level, just above leaves, we might already have queued up some
* leaf pages, and we assume that there can't be any non-leaf
* pages behind leaf pages.
*/
ptr = (GISTInsertStack *) palloc0(sizeof(GISTInsertStack));
ptr->blkno = GistPageGetOpaque(page)->rightlink;
ptr->downlinkoffnum = InvalidOffsetNumber;
ptr->parent = top->parent;
fifo = lcons(ptr, fifo);
}
maxoff = PageGetMaxOffsetNumber(page);
for (i = FirstOffsetNumber; i <= maxoff; i = OffsetNumberNext(i))
{
iid = PageGetItemId(page, i);
idxtuple = (IndexTuple) PageGetItem(page, iid);
blkno = ItemPointerGetBlockNumber(&(idxtuple->t_tid));
if (blkno == child)
{
/* Found it! */
UnlockReleaseBuffer(buffer);
*downlinkoffnum = i;
return top;
}
else
{
/* Append this child to the list of pages to visit later */
ptr = (GISTInsertStack *) palloc0(sizeof(GISTInsertStack));
ptr->blkno = blkno;
ptr->downlinkoffnum = i;
ptr->parent = top;
fifo = lappend(fifo, ptr);
}
}
UnlockReleaseBuffer(buffer);
}
elog(ERROR, "failed to re-find parent of a page in index \"%s\", block %u",
RelationGetRelationName(r), child);
return NULL; /* keep compiler quiet */
}
/*
* Prune and repair fragmentation in the specified page.
*
* Caller must have pin and buffer cleanup lock on the page.
*
* OldestXmin is the cutoff XID used to distinguish whether tuples are DEAD
* or RECENTLY_DEAD (see HeapTupleSatisfiesVacuum).
*
* If report_stats is true then we send the number of reclaimed heap-only
* tuples to pgstats. (This must be FALSE during vacuum, since vacuum will
* send its own new total to pgstats, and we don't want this delta applied
* on top of that.)
*
* Returns the number of tuples deleted from the page and sets
* latestRemovedXid.
*/
int
heap_page_prune(Relation relation, Buffer buffer, TransactionId OldestXmin,
bool report_stats, TransactionId *latestRemovedXid)
{
int ndeleted = 0;
Page page = BufferGetPage(buffer);
OffsetNumber offnum,
maxoff;
PruneState prstate;
/*
* Our strategy is to scan the page and make lists of items to change,
* then apply the changes within a critical section. This keeps as much
* logic as possible out of the critical section, and also ensures that
* WAL replay will work the same as the normal case.
*
* First, initialize the new pd_prune_xid value to zero (indicating no
* prunable tuples). If we find any tuples which may soon become
* prunable, we will save the lowest relevant XID in new_prune_xid. Also
* initialize the rest of our working state.
*/
prstate.new_prune_xid = InvalidTransactionId;
prstate.latestRemovedXid = *latestRemovedXid;
prstate.nredirected = prstate.ndead = prstate.nunused = 0;
memset(prstate.marked, 0, sizeof(prstate.marked));
/* Scan the page */
maxoff = PageGetMaxOffsetNumber(page);
for (offnum = FirstOffsetNumber;
offnum <= maxoff;
offnum = OffsetNumberNext(offnum))
{
ItemId itemid;
/* Ignore items already processed as part of an earlier chain */
if (prstate.marked[offnum])
continue;
/* Nothing to do if slot is empty or already dead */
itemid = PageGetItemId(page, offnum);
if (!ItemIdIsUsed(itemid) || ItemIdIsDead(itemid))
continue;
/* Process this item or chain of items */
ndeleted += heap_prune_chain(relation, buffer, offnum,
OldestXmin,
&prstate);
}
/* Any error while applying the changes is critical */
START_CRIT_SECTION();
/* Have we found any prunable items? */
if (prstate.nredirected > 0 || prstate.ndead > 0 || prstate.nunused > 0)
{
/*
* Apply the planned item changes, then repair page fragmentation, and
* update the page's hint bit about whether it has free line pointers.
*/
heap_page_prune_execute(buffer,
prstate.redirected, prstate.nredirected,
prstate.nowdead, prstate.ndead,
prstate.nowunused, prstate.nunused);
/*
* Update the page's pd_prune_xid field to either zero, or the lowest
* XID of any soon-prunable tuple.
*/
((PageHeader) page)->pd_prune_xid = prstate.new_prune_xid;
/*
* Also clear the "page is full" flag, since there's no point in
* repeating the prune/defrag process until something else happens to
* the page.
*/
PageClearFull(page);
MarkBufferDirty(buffer);
/*
* Emit a WAL HEAP_CLEAN record showing what we did
*/
if (RelationNeedsWAL(relation))
{
XLogRecPtr recptr;
recptr = log_heap_clean(relation, buffer,
prstate.redirected, prstate.nredirected,
prstate.nowdead, prstate.ndead,
prstate.nowunused, prstate.nunused,
prstate.latestRemovedXid);
PageSetLSN(BufferGetPage(buffer), recptr);
PageSetTLI(BufferGetPage(buffer), ThisTimeLineID);
}
}
else
{
/*
* If we didn't prune anything, but have found a new value for the
* pd_prune_xid field, update it and mark the buffer dirty. This is
* treated as a non-WAL-logged hint.
*
* Also clear the "page is full" flag if it is set, since there's no
* point in repeating the prune/defrag process until something else
* happens to the page.
*/
if (((PageHeader) page)->pd_prune_xid != prstate.new_prune_xid ||
PageIsFull(page))
{
((PageHeader) page)->pd_prune_xid = prstate.new_prune_xid;
PageClearFull(page);
SetBufferCommitInfoNeedsSave(buffer);
}
}
END_CRIT_SECTION();
/*
* If requested, report the number of tuples reclaimed to pgstats. This is
* ndeleted minus ndead, because we don't want to count a now-DEAD root
* item as a deletion for this purpose.
*/
if (report_stats && ndeleted > prstate.ndead)
pgstat_update_heap_dead_tuples(relation, ndeleted - prstate.ndead);
*latestRemovedXid = prstate.latestRemovedXid;
/*
* XXX Should we update the FSM information of this page ?
*
* There are two schools of thought here. We may not want to update FSM
* information so that the page is not used for unrelated UPDATEs/INSERTs
* and any free space in this page will remain available for further
* UPDATEs in *this* page, thus improving chances for doing HOT updates.
*
* But for a large table and where a page does not receive further UPDATEs
* for a long time, we might waste this space by not updating the FSM
* information. The relation may get extended and fragmented further.
*
* One possibility is to leave "fillfactor" worth of space in this page
* and update FSM with the remaining space.
*
* In any case, the current FSM implementation doesn't accept
* one-page-at-a-time updates, so this is all academic for now.
*/
return ndeleted;
}
/*
* btvacuumpage --- VACUUM one page
*
* This processes a single page for btvacuumscan(). In some cases we
* must go back and re-examine previously-scanned pages; this routine
* recurses when necessary to handle that case.
*
* blkno is the page to process. orig_blkno is the highest block number
* reached by the outer btvacuumscan loop (the same as blkno, unless we
* are recursing to re-examine a previous page).
*/
static void
btvacuumpage(BTVacState *vstate, BlockNumber blkno, BlockNumber orig_blkno)
{
IndexVacuumInfo *info = vstate->info;
IndexBulkDeleteResult *stats = vstate->stats;
IndexBulkDeleteCallback callback = vstate->callback;
void *callback_state = vstate->callback_state;
Relation rel = info->index;
bool delete_now;
BlockNumber recurse_to;
Buffer buf;
Page page;
BTPageOpaque opaque;
restart:
delete_now = false;
recurse_to = P_NONE;
/* call vacuum_delay_point while not holding any buffer lock */
vacuum_delay_point();
/*
* We can't use _bt_getbuf() here because it always applies
* _bt_checkpage(), which will barf on an all-zero page. We want to
* recycle all-zero pages, not fail. Also, we want to use a nondefault
* buffer access strategy.
*/
buf = ReadBufferExtended(rel, MAIN_FORKNUM, blkno, RBM_NORMAL,
info->strategy);
LockBuffer(buf, BT_READ);
page = BufferGetPage(buf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
if (!PageIsNew(page))
_bt_checkpage(rel, buf);
/*
* If we are recursing, the only case we want to do anything with is a
* live leaf page having the current vacuum cycle ID. Any other state
* implies we already saw the page (eg, deleted it as being empty).
*/
if (blkno != orig_blkno)
{
if (_bt_page_recyclable(page) ||
P_IGNORE(opaque) ||
!P_ISLEAF(opaque) ||
opaque->btpo_cycleid != vstate->cycleid)
{
_bt_relbuf(rel, buf);
return;
}
}
/* Page is valid, see what to do with it */
if (_bt_page_recyclable(page))
{
/* Okay to recycle this page */
RecordFreeIndexPage(rel, blkno);
vstate->totFreePages++;
stats->pages_deleted++;
}
else if (P_ISDELETED(opaque))
{
/* Already deleted, but can't recycle yet */
stats->pages_deleted++;
}
else if (P_ISHALFDEAD(opaque))
{
/* Half-dead, try to delete */
delete_now = true;
}
else if (P_ISLEAF(opaque))
{
OffsetNumber deletable[MaxOffsetNumber];
int ndeletable;
OffsetNumber offnum,
minoff,
maxoff;
/*
* Trade in the initial read lock for a super-exclusive write lock on
* this page. We must get such a lock on every leaf page over the
* course of the vacuum scan, whether or not it actually contains any
* deletable tuples --- see nbtree/README.
*/
LockBuffer(buf, BUFFER_LOCK_UNLOCK);
LockBufferForCleanup(buf);
/*
* Remember highest leaf page number we've taken cleanup lock on; see
* notes in btvacuumscan
*/
if (blkno > vstate->lastBlockLocked)
vstate->lastBlockLocked = blkno;
/*
* Check whether we need to recurse back to earlier pages. What we
* are concerned about is a page split that happened since we started
* the vacuum scan. If the split moved some tuples to a lower page
* then we might have missed 'em. If so, set up for tail recursion.
* (Must do this before possibly clearing btpo_cycleid below!)
*/
if (vstate->cycleid != 0 &&
opaque->btpo_cycleid == vstate->cycleid &&
!(opaque->btpo_flags & BTP_SPLIT_END) &&
!P_RIGHTMOST(opaque) &&
opaque->btpo_next < orig_blkno)
recurse_to = opaque->btpo_next;
/*
* Scan over all items to see which ones need deleted according to the
* callback function.
*/
ndeletable = 0;
minoff = P_FIRSTDATAKEY(opaque);
maxoff = PageGetMaxOffsetNumber(page);
if (callback)
{
for (offnum = minoff;
offnum <= maxoff;
offnum = OffsetNumberNext(offnum))
{
IndexTuple itup;
ItemPointer htup;
itup = (IndexTuple) PageGetItem(page,
PageGetItemId(page, offnum));
htup = &(itup->t_tid);
/*
* During Hot Standby we currently assume that
* XLOG_BTREE_VACUUM records do not produce conflicts. That is
* only true as long as the callback function depends only
* upon whether the index tuple refers to heap tuples removed
* in the initial heap scan. When vacuum starts it derives a
* value of OldestXmin. Backends taking later snapshots could
* have a RecentGlobalXmin with a later xid than the vacuum's
* OldestXmin, so it is possible that row versions deleted
* after OldestXmin could be marked as killed by other
* backends. The callback function *could* look at the index
* tuple state in isolation and decide to delete the index
* tuple, though currently it does not. If it ever did, we
* would need to reconsider whether XLOG_BTREE_VACUUM records
* should cause conflicts. If they did cause conflicts they
* would be fairly harsh conflicts, since we haven't yet
* worked out a way to pass a useful value for
* latestRemovedXid on the XLOG_BTREE_VACUUM records. This
* applies to *any* type of index that marks index tuples as
* killed.
*/
if (callback(htup, callback_state))
deletable[ndeletable++] = offnum;
}
}
/*
* Apply any needed deletes. We issue just one _bt_delitems_vacuum()
* call per page, so as to minimize WAL traffic.
*/
if (ndeletable > 0)
{
/*
* Notice that the issued XLOG_BTREE_VACUUM WAL record includes an
* instruction to the replay code to get cleanup lock on all pages
* between the previous lastBlockVacuumed and this page. This
* ensures that WAL replay locks all leaf pages at some point.
*
* Since we can visit leaf pages out-of-order when recursing,
* replay might end up locking such pages an extra time, but it
* doesn't seem worth the amount of bookkeeping it'd take to avoid
* that.
*/
_bt_delitems_vacuum(rel, buf, deletable, ndeletable,
vstate->lastBlockVacuumed);
/*
* Remember highest leaf page number we've issued a
* XLOG_BTREE_VACUUM WAL record for.
*/
if (blkno > vstate->lastBlockVacuumed)
vstate->lastBlockVacuumed = blkno;
stats->tuples_removed += ndeletable;
/* must recompute maxoff */
maxoff = PageGetMaxOffsetNumber(page);
}
else
{
/*
* If the page has been split during this vacuum cycle, it seems
* worth expending a write to clear btpo_cycleid even if we don't
* have any deletions to do. (If we do, _bt_delitems_vacuum takes
* care of this.) This ensures we won't process the page again.
*
* We treat this like a hint-bit update because there's no need to
* WAL-log it.
*/
if (vstate->cycleid != 0 &&
opaque->btpo_cycleid == vstate->cycleid)
{
opaque->btpo_cycleid = 0;
MarkBufferDirtyHint(buf, true);
}
}
/*
* If it's now empty, try to delete; else count the live tuples. We
* don't delete when recursing, though, to avoid putting entries into
* freePages out-of-order (doesn't seem worth any extra code to handle
* the case).
*/
if (minoff > maxoff)
delete_now = (blkno == orig_blkno);
else
stats->num_index_tuples += maxoff - minoff + 1;
}
if (delete_now)
{
MemoryContext oldcontext;
int ndel;
/* Run pagedel in a temp context to avoid memory leakage */
MemoryContextReset(vstate->pagedelcontext);
oldcontext = MemoryContextSwitchTo(vstate->pagedelcontext);
ndel = _bt_pagedel(rel, buf);
/* count only this page, else may double-count parent */
if (ndel)
stats->pages_deleted++;
MemoryContextSwitchTo(oldcontext);
/* pagedel released buffer, so we shouldn't */
}
else
_bt_relbuf(rel, buf);
/*
* This is really tail recursion, but if the compiler is too stupid to
* optimize it as such, we'd eat an uncomfortably large amount of stack
* space per recursion level (due to the deletable[] array). A failure is
* improbable since the number of levels isn't likely to be large ... but
* just in case, let's hand-optimize into a loop.
*/
if (recurse_to != P_NONE)
{
blkno = recurse_to;
goto restart;
}
}
/*
* For all items in this page, find their respective root line pointers.
* If item k is part of a HOT-chain with root at item j, then we set
* root_offsets[k - 1] = j.
*
* The passed-in root_offsets array must have MaxHeapTuplesPerPage entries.
* We zero out all unused entries.
*
* The function must be called with at least share lock on the buffer, to
* prevent concurrent prune operations.
*
* Note: The information collected here is valid only as long as the caller
* holds a pin on the buffer. Once pin is released, a tuple might be pruned
* and reused by a completely unrelated tuple.
*/
void
heap_get_root_tuples(Page page, OffsetNumber *root_offsets)
{
OffsetNumber offnum,
maxoff;
MemSet(root_offsets, 0, MaxHeapTuplesPerPage * sizeof(OffsetNumber));
maxoff = PageGetMaxOffsetNumber(page);
for (offnum = FirstOffsetNumber; offnum <= maxoff; offnum = OffsetNumberNext(offnum))
{
ItemId lp = PageGetItemId(page, offnum);
HeapTupleHeader htup;
OffsetNumber nextoffnum;
TransactionId priorXmax;
/* skip unused and dead items */
if (!ItemIdIsUsed(lp) || ItemIdIsDead(lp))
continue;
if (ItemIdIsNormal(lp))
{
htup = (HeapTupleHeader) PageGetItem(page, lp);
/*
* Check if this tuple is part of a HOT-chain rooted at some other
* tuple. If so, skip it for now; we'll process it when we find
* its root.
*/
if (HeapTupleHeaderIsHeapOnly(htup))
continue;
/*
* This is either a plain tuple or the root of a HOT-chain.
* Remember it in the mapping.
*/
root_offsets[offnum - 1] = offnum;
/* If it's not the start of a HOT-chain, we're done with it */
if (!HeapTupleHeaderIsHotUpdated(htup))
continue;
/* Set up to scan the HOT-chain */
nextoffnum = ItemPointerGetOffsetNumber(&htup->t_ctid);
priorXmax = HeapTupleHeaderGetXmax(htup);
}
else
{
/* Must be a redirect item. We do not set its root_offsets entry */
Assert(ItemIdIsRedirected(lp));
/* Set up to scan the HOT-chain */
nextoffnum = ItemIdGetRedirect(lp);
priorXmax = InvalidTransactionId;
}
/*
* Now follow the HOT-chain and collect other tuples in the chain.
*
* Note: Even though this is a nested loop, the complexity of the
* function is O(N) because a tuple in the page should be visited not
* more than twice, once in the outer loop and once in HOT-chain
* chases.
*/
for (;;)
{
lp = PageGetItemId(page, nextoffnum);
/* Check for broken chains */
if (!ItemIdIsNormal(lp))
break;
htup = (HeapTupleHeader) PageGetItem(page, lp);
if (TransactionIdIsValid(priorXmax) &&
!TransactionIdEquals(priorXmax, HeapTupleHeaderGetXmin(htup)))
break;
/* Remember the root line pointer for this item */
root_offsets[nextoffnum - 1] = offnum;
/* Advance to next chain member, if any */
if (!HeapTupleHeaderIsHotUpdated(htup))
break;
nextoffnum = ItemPointerGetOffsetNumber(&htup->t_ctid);
priorXmax = HeapTupleHeaderGetXmax(htup);
}
}
}
/*
* hashgettuple() -- Get the next tuple in the scan.
*/
bool
hashgettuple(IndexScanDesc scan, ScanDirection dir)
{
HashScanOpaque so = (HashScanOpaque) scan->opaque;
Relation rel = scan->indexRelation;
Buffer buf;
Page page;
OffsetNumber offnum;
ItemPointer current;
bool res;
/* Hash indexes are always lossy since we store only the hash code */
scan->xs_recheck = true;
/*
* We hold pin but not lock on current buffer while outside the hash AM.
* Reacquire the read lock here.
*/
if (BufferIsValid(so->hashso_curbuf))
LockBuffer(so->hashso_curbuf, BUFFER_LOCK_SHARE);
/*
* If we've already initialized this scan, we can just advance it in the
* appropriate direction. If we haven't done so yet, we call a routine to
* get the first item in the scan.
*/
current = &(so->hashso_curpos);
if (ItemPointerIsValid(current))
{
/*
* An insertion into the current index page could have happened while
* we didn't have read lock on it. Re-find our position by looking
* for the TID we previously returned. (Because we hold a pin on the
* primary bucket page, no deletions or splits could have occurred;
* therefore we can expect that the TID still exists in the current
* index page, at an offset >= where we were.)
*/
OffsetNumber maxoffnum;
buf = so->hashso_curbuf;
Assert(BufferIsValid(buf));
page = BufferGetPage(buf);
/*
* We don't need test for old snapshot here as the current buffer is
* pinned, so vacuum can't clean the page.
*/
maxoffnum = PageGetMaxOffsetNumber(page);
for (offnum = ItemPointerGetOffsetNumber(current);
offnum <= maxoffnum;
offnum = OffsetNumberNext(offnum))
{
IndexTuple itup;
itup = (IndexTuple) PageGetItem(page, PageGetItemId(page, offnum));
if (ItemPointerEquals(&(so->hashso_heappos), &(itup->t_tid)))
break;
}
if (offnum > maxoffnum)
elog(ERROR, "failed to re-find scan position within index \"%s\"",
RelationGetRelationName(rel));
ItemPointerSetOffsetNumber(current, offnum);
/*
* Check to see if we should kill the previously-fetched tuple.
*/
if (scan->kill_prior_tuple)
{
/*
* Yes, so remember it for later. (We'll deal with all such
* tuples at once right after leaving the index page or at
* end of scan.) In case if caller reverses the indexscan
* direction it is quite possible that the same item might
* get entered multiple times. But, we don't detect that;
* instead, we just forget any excess entries.
*/
if (so->killedItems == NULL)
so->killedItems = palloc(MaxIndexTuplesPerPage *
sizeof(HashScanPosItem));
if (so->numKilled < MaxIndexTuplesPerPage)
{
so->killedItems[so->numKilled].heapTid = so->hashso_heappos;
so->killedItems[so->numKilled].indexOffset =
ItemPointerGetOffsetNumber(&(so->hashso_curpos));
so->numKilled++;
}
}
/*
* Now continue the scan.
*/
res = _hash_next(scan, dir);
}
else
res = _hash_first(scan, dir);
/*
* Skip killed tuples if asked to.
*/
if (scan->ignore_killed_tuples)
{
while (res)
{
offnum = ItemPointerGetOffsetNumber(current);
page = BufferGetPage(so->hashso_curbuf);
if (!ItemIdIsDead(PageGetItemId(page, offnum)))
break;
res = _hash_next(scan, dir);
}
}
/* Release read lock on current buffer, but keep it pinned */
if (BufferIsValid(so->hashso_curbuf))
LockBuffer(so->hashso_curbuf, BUFFER_LOCK_UNLOCK);
/* Return current heap TID on success */
scan->xs_ctup.t_self = so->hashso_heappos;
return res;
}
Datum
gtsquery_picksplit(PG_FUNCTION_ARGS)
{
GistEntryVector *entryvec = (GistEntryVector *) PG_GETARG_POINTER(0);
GIST_SPLITVEC *v = (GIST_SPLITVEC *) PG_GETARG_POINTER(1);
OffsetNumber maxoff = entryvec->n - 2;
OffsetNumber k,
j;
TSQuerySign datum_l,
datum_r;
int4 size_alpha,
size_beta;
int4 size_waste,
waste = -1;
int4 nbytes;
OffsetNumber seed_1 = 0,
seed_2 = 0;
OffsetNumber *left,
*right;
SPLITCOST *costvector;
nbytes = (maxoff + 2) * sizeof(OffsetNumber);
left = v->spl_left = (OffsetNumber *) palloc(nbytes);
right = v->spl_right = (OffsetNumber *) palloc(nbytes);
v->spl_nleft = v->spl_nright = 0;
for (k = FirstOffsetNumber; k < maxoff; k = OffsetNumberNext(k))
for (j = OffsetNumberNext(k); j <= maxoff; j = OffsetNumberNext(j))
{
size_waste = hemdist(GETENTRY(entryvec, j), GETENTRY(entryvec, k));
if (size_waste > waste)
{
waste = size_waste;
seed_1 = k;
seed_2 = j;
}
}
if (seed_1 == 0 || seed_2 == 0)
{
seed_1 = 1;
seed_2 = 2;
}
datum_l = GETENTRY(entryvec, seed_1);
datum_r = GETENTRY(entryvec, seed_2);
maxoff = OffsetNumberNext(maxoff);
costvector = (SPLITCOST *) palloc(sizeof(SPLITCOST) * maxoff);
for (j = FirstOffsetNumber; j <= maxoff; j = OffsetNumberNext(j))
{
costvector[j - 1].pos = j;
size_alpha = hemdist(GETENTRY(entryvec, seed_1), GETENTRY(entryvec, j));
size_beta = hemdist(GETENTRY(entryvec, seed_2), GETENTRY(entryvec, j));
costvector[j - 1].cost = abs(size_alpha - size_beta);
}
qsort((void *) costvector, maxoff, sizeof(SPLITCOST), comparecost);
for (k = 0; k < maxoff; k++)
{
j = costvector[k].pos;
if (j == seed_1)
{
*left++ = j;
v->spl_nleft++;
continue;
}
else if (j == seed_2)
{
*right++ = j;
v->spl_nright++;
continue;
}
size_alpha = hemdist(datum_l, GETENTRY(entryvec, j));
size_beta = hemdist(datum_r, GETENTRY(entryvec, j));
if (size_alpha < size_beta + WISH_F(v->spl_nleft, v->spl_nright, 0.05))
{
datum_l |= GETENTRY(entryvec, j);
*left++ = j;
v->spl_nleft++;
}
else
{
datum_r |= GETENTRY(entryvec, j);
*right++ = j;
v->spl_nright++;
}
}
*right = *left = FirstOffsetNumber;
v->spl_ldatum = TSQuerySignGetDatum(datum_l);
v->spl_rdatum = TSQuerySignGetDatum(datum_r);
PG_RETURN_POINTER(v);
}
/*
* _hash_splitbucket -- split 'obucket' into 'obucket' and 'nbucket'
*
* We are splitting a bucket that consists of a base bucket page and zero
* or more overflow (bucket chain) pages. We must relocate tuples that
* belong in the new bucket, and compress out any free space in the old
* bucket.
*
* The caller must hold exclusive locks on both buckets to ensure that
* no one else is trying to access them (see README).
*
* The caller must hold a pin, but no lock, on the metapage buffer.
* The buffer is returned in the same state. (The metapage is only
* touched if it becomes necessary to add or remove overflow pages.)
*
* In addition, the caller must have created the new bucket's base page,
* which is passed in buffer nbuf, pinned and write-locked. That lock and
* pin are released here. (The API is set up this way because we must do
* _hash_getnewbuf() before releasing the metapage write lock. So instead of
* passing the new bucket's start block number, we pass an actual buffer.)
*/
static void
_hash_splitbucket(Relation rel,
Buffer metabuf,
Bucket obucket,
Bucket nbucket,
BlockNumber start_oblkno,
Buffer nbuf,
uint32 maxbucket,
uint32 highmask,
uint32 lowmask)
{
Buffer obuf;
Page opage;
Page npage;
HashPageOpaque oopaque;
HashPageOpaque nopaque;
/*
* It should be okay to simultaneously write-lock pages from each bucket,
* since no one else can be trying to acquire buffer lock on pages of
* either bucket.
*/
obuf = _hash_getbuf(rel, start_oblkno, HASH_WRITE, LH_BUCKET_PAGE);
opage = BufferGetPage(obuf);
oopaque = (HashPageOpaque) PageGetSpecialPointer(opage);
npage = BufferGetPage(nbuf);
/* initialize the new bucket's primary page */
nopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
nopaque->hasho_prevblkno = InvalidBlockNumber;
nopaque->hasho_nextblkno = InvalidBlockNumber;
nopaque->hasho_bucket = nbucket;
nopaque->hasho_flag = LH_BUCKET_PAGE;
nopaque->hasho_page_id = HASHO_PAGE_ID;
/*
* Partition the tuples in the old bucket between the old bucket and the
* new bucket, advancing along the old bucket's overflow bucket chain and
* adding overflow pages to the new bucket as needed. Outer loop iterates
* once per page in old bucket.
*/
for (;;)
{
BlockNumber oblkno;
OffsetNumber ooffnum;
OffsetNumber omaxoffnum;
OffsetNumber deletable[MaxOffsetNumber];
int ndeletable = 0;
/* Scan each tuple in old page */
omaxoffnum = PageGetMaxOffsetNumber(opage);
for (ooffnum = FirstOffsetNumber;
ooffnum <= omaxoffnum;
ooffnum = OffsetNumberNext(ooffnum))
{
IndexTuple itup;
Size itemsz;
Bucket bucket;
/* skip dead tuples */
if (ItemIdIsDead(PageGetItemId(opage, ooffnum)))
continue;
/*
* Fetch the item's hash key (conveniently stored in the item) and
* determine which bucket it now belongs in.
*/
itup = (IndexTuple) PageGetItem(opage,
PageGetItemId(opage, ooffnum));
bucket = _hash_hashkey2bucket(_hash_get_indextuple_hashkey(itup),
maxbucket, highmask, lowmask);
if (bucket == nbucket)
{
/*
* insert the tuple into the new bucket. if it doesn't fit on
* the current page in the new bucket, we must allocate a new
* overflow page and place the tuple on that page instead.
*
* XXX we have a problem here if we fail to get space for a
* new overflow page: we'll error out leaving the bucket split
* only partially complete, meaning the index is corrupt,
* since searches may fail to find entries they should find.
*/
itemsz = IndexTupleDSize(*itup);
itemsz = MAXALIGN(itemsz);
if (PageGetFreeSpace(npage) < itemsz)
{
/* write out nbuf and drop lock, but keep pin */
_hash_chgbufaccess(rel, nbuf, HASH_WRITE, HASH_NOLOCK);
/* chain to a new overflow page */
nbuf = _hash_addovflpage(rel, metabuf, nbuf);
npage = BufferGetPage(nbuf);
/* we don't need nopaque within the loop */
}
/*
* Insert tuple on new page, using _hash_pgaddtup to ensure
* correct ordering by hashkey. This is a tad inefficient
* since we may have to shuffle itempointers repeatedly.
* Possible future improvement: accumulate all the items for
* the new page and qsort them before insertion.
*/
(void) _hash_pgaddtup(rel, nbuf, itemsz, itup);
/*
* Mark tuple for deletion from old page.
*/
deletable[ndeletable++] = ooffnum;
}
else
{
/*
* the tuple stays on this page, so nothing to do.
*/
Assert(bucket == obucket);
}
}
oblkno = oopaque->hasho_nextblkno;
/*
* Done scanning this old page. If we moved any tuples, delete them
* from the old page.
*/
if (ndeletable > 0)
{
PageIndexMultiDelete(opage, deletable, ndeletable);
_hash_wrtbuf(rel, obuf);
}
else
_hash_relbuf(rel, obuf);
/* Exit loop if no more overflow pages in old bucket */
if (!BlockNumberIsValid(oblkno))
break;
/* Else, advance to next old page */
obuf = _hash_getbuf(rel, oblkno, HASH_WRITE, LH_OVERFLOW_PAGE);
opage = BufferGetPage(obuf);
oopaque = (HashPageOpaque) PageGetSpecialPointer(opage);
}
/*
* We're at the end of the old bucket chain, so we're done partitioning
* the tuples. Before quitting, call _hash_squeezebucket to ensure the
* tuples remaining in the old bucket (including the overflow pages) are
* packed as tightly as possible. The new bucket is already tight.
*/
_hash_wrtbuf(rel, nbuf);
_hash_squeezebucket(rel, obucket, start_oblkno, NULL);
}
/*
* _bt_pagedel() -- Delete a page from the b-tree, if legal to do so.
*
* This action unlinks the page from the b-tree structure, removing all
* pointers leading to it --- but not touching its own left and right links.
* The page cannot be physically reclaimed right away, since other processes
* may currently be trying to follow links leading to the page; they have to
* be allowed to use its right-link to recover. See nbtree/README.
*
* On entry, the target buffer must be pinned and locked (either read or write
* lock is OK). This lock and pin will be dropped before exiting.
*
* The "stack" argument can be a search stack leading (approximately) to the
* target page, or NULL --- outside callers typically pass NULL since they
* have not done such a search, but internal recursion cases pass the stack
* to avoid duplicated search effort.
*
* Returns the number of pages successfully deleted (zero if page cannot
* be deleted now; could be more than one if parent pages were deleted too).
*
* NOTE: this leaks memory. Rather than trying to clean up everything
* carefully, it's better to run it in a temp context that can be reset
* frequently.
*/
int
_bt_pagedel(Relation rel, Buffer buf, BTStack stack)
{
int result;
BlockNumber target,
leftsib,
rightsib,
parent;
OffsetNumber poffset,
maxoff;
uint32 targetlevel,
ilevel;
ItemId itemid;
IndexTuple targetkey,
itup;
ScanKey itup_scankey;
Buffer lbuf,
rbuf,
pbuf;
bool parent_half_dead;
bool parent_one_child;
bool rightsib_empty;
Buffer metabuf = InvalidBuffer;
Page metapg = NULL;
BTMetaPageData *metad = NULL;
Page page;
BTPageOpaque opaque;
/*
* We can never delete rightmost pages nor root pages. While at it, check
* that page is not already deleted and is empty.
*/
page = BufferGetPage(buf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
if (P_RIGHTMOST(opaque) || P_ISROOT(opaque) || P_ISDELETED(opaque) ||
P_FIRSTDATAKEY(opaque) <= PageGetMaxOffsetNumber(page))
{
/* Should never fail to delete a half-dead page */
Assert(!P_ISHALFDEAD(opaque));
_bt_relbuf(rel, buf);
return 0;
}
/*
* Save info about page, including a copy of its high key (it must have
* one, being non-rightmost).
*/
target = BufferGetBlockNumber(buf);
targetlevel = opaque->btpo.level;
leftsib = opaque->btpo_prev;
itemid = PageGetItemId(page, P_HIKEY);
targetkey = CopyIndexTuple((IndexTuple) PageGetItem(page, itemid));
/*
* To avoid deadlocks, we'd better drop the target page lock before going
* further.
*/
_bt_relbuf(rel, buf);
/*
* We need an approximate pointer to the page's parent page. We use the
* standard search mechanism to search for the page's high key; this will
* give us a link to either the current parent or someplace to its left
* (if there are multiple equal high keys). In recursion cases, the
* caller already generated a search stack and we can just re-use that
* work.
*/
if (stack == NULL)
{
if (!InRecovery)
{
/* we need an insertion scan key to do our search, so build one */
itup_scankey = _bt_mkscankey(rel, targetkey);
/* find the leftmost leaf page containing this key */
stack = _bt_search(rel, rel->rd_rel->relnatts, itup_scankey, false,
&lbuf, BT_READ);
/* don't need a pin on that either */
_bt_relbuf(rel, lbuf);
/*
* If we are trying to delete an interior page, _bt_search did
* more than we needed. Locate the stack item pointing to our
* parent level.
*/
ilevel = 0;
for (;;)
{
if (stack == NULL)
elog(ERROR, "not enough stack items");
if (ilevel == targetlevel)
break;
stack = stack->bts_parent;
ilevel++;
}
}
else
{
/*
* During WAL recovery, we can't use _bt_search (for one reason,
* it might invoke user-defined comparison functions that expect
* facilities not available in recovery mode). Instead, just set
* up a dummy stack pointing to the left end of the parent tree
* level, from which _bt_getstackbuf will walk right to the parent
* page. Painful, but we don't care too much about performance in
* this scenario.
*/
pbuf = _bt_get_endpoint(rel, targetlevel + 1, false);
stack = (BTStack) palloc(sizeof(BTStackData));
stack->bts_blkno = BufferGetBlockNumber(pbuf);
stack->bts_offset = InvalidOffsetNumber;
/* bts_btentry will be initialized below */
stack->bts_parent = NULL;
_bt_relbuf(rel, pbuf);
}
}
/*
* We cannot delete a page that is the rightmost child of its immediate
* parent, unless it is the only child --- in which case the parent has to
* be deleted too, and the same condition applies recursively to it. We
* have to check this condition all the way up before trying to delete. We
* don't need to re-test when deleting a non-leaf page, though.
*/
if (targetlevel == 0 &&
!_bt_parent_deletion_safe(rel, target, stack))
return 0;
/*
* We have to lock the pages we need to modify in the standard order:
* moving right, then up. Else we will deadlock against other writers.
*
* So, we need to find and write-lock the current left sibling of the
* target page. The sibling that was current a moment ago could have
* split, so we may have to move right. This search could fail if either
* the sibling or the target page was deleted by someone else meanwhile;
* if so, give up. (Right now, that should never happen, since page
* deletion is only done in VACUUM and there shouldn't be multiple VACUUMs
* concurrently on the same table.)
*/
if (leftsib != P_NONE)
{
lbuf = _bt_getbuf(rel, leftsib, BT_WRITE);
page = BufferGetPage(lbuf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
while (P_ISDELETED(opaque) || opaque->btpo_next != target)
{
/* step right one page */
leftsib = opaque->btpo_next;
_bt_relbuf(rel, lbuf);
if (leftsib == P_NONE)
{
elog(LOG, "no left sibling (concurrent deletion?) in \"%s\"",
RelationGetRelationName(rel));
return 0;
}
lbuf = _bt_getbuf(rel, leftsib, BT_WRITE);
page = BufferGetPage(lbuf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
}
}
else
lbuf = InvalidBuffer;
/*
* Next write-lock the target page itself. It should be okay to take just
* a write lock not a superexclusive lock, since no scans would stop on an
* empty page.
*/
buf = _bt_getbuf(rel, target, BT_WRITE);
page = BufferGetPage(buf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
/*
* Check page is still empty etc, else abandon deletion. The empty check
* is necessary since someone else might have inserted into it while we
* didn't have it locked; the others are just for paranoia's sake.
*/
if (P_RIGHTMOST(opaque) || P_ISROOT(opaque) || P_ISDELETED(opaque) ||
P_FIRSTDATAKEY(opaque) <= PageGetMaxOffsetNumber(page))
{
_bt_relbuf(rel, buf);
if (BufferIsValid(lbuf))
_bt_relbuf(rel, lbuf);
return 0;
}
if (opaque->btpo_prev != leftsib)
elog(ERROR, "left link changed unexpectedly in block %u of index \"%s\"",
target, RelationGetRelationName(rel));
/*
* And next write-lock the (current) right sibling.
*/
rightsib = opaque->btpo_next;
rbuf = _bt_getbuf(rel, rightsib, BT_WRITE);
page = BufferGetPage(rbuf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
if (opaque->btpo_prev != target)
elog(ERROR, "right sibling's left-link doesn't match: "
"block %u links to %u instead of expected %u in index \"%s\"",
rightsib, opaque->btpo_prev, target,
RelationGetRelationName(rel));
/*
* Next find and write-lock the current parent of the target page. This is
* essentially the same as the corresponding step of splitting.
*/
ItemPointerSet(&(stack->bts_btentry.t_tid), target, P_HIKEY);
pbuf = _bt_getstackbuf(rel, stack, BT_WRITE);
if (pbuf == InvalidBuffer)
elog(ERROR, "failed to re-find parent key in index \"%s\" for deletion target page %u",
RelationGetRelationName(rel), target);
parent = stack->bts_blkno;
poffset = stack->bts_offset;
/*
* If the target is the rightmost child of its parent, then we can't
* delete, unless it's also the only child --- in which case the parent
* changes to half-dead status. The "can't delete" case should have been
* detected by _bt_parent_deletion_safe, so complain if we see it now.
*/
page = BufferGetPage(pbuf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
maxoff = PageGetMaxOffsetNumber(page);
parent_half_dead = false;
parent_one_child = false;
if (poffset >= maxoff)
{
if (poffset == P_FIRSTDATAKEY(opaque))
parent_half_dead = true;
else
elog(ERROR, "failed to delete rightmost child %u of block %u in index \"%s\"",
target, parent, RelationGetRelationName(rel));
}
else
{
/* Will there be exactly one child left in this parent? */
if (OffsetNumberNext(P_FIRSTDATAKEY(opaque)) == maxoff)
parent_one_child = true;
}
/*
* If we are deleting the next-to-last page on the target's level, then
* the rightsib is a candidate to become the new fast root. (In theory, it
* might be possible to push the fast root even further down, but the odds
* of doing so are slim, and the locking considerations daunting.)
*
* We don't support handling this in the case where the parent is becoming
* half-dead, even though it theoretically could occur.
*
* We can safely acquire a lock on the metapage here --- see comments for
* _bt_newroot().
*/
if (leftsib == P_NONE && !parent_half_dead)
{
page = BufferGetPage(rbuf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
Assert(opaque->btpo.level == targetlevel);
if (P_RIGHTMOST(opaque))
{
/* rightsib will be the only one left on the level */
metabuf = _bt_getbuf(rel, BTREE_METAPAGE, BT_WRITE);
metapg = BufferGetPage(metabuf);
metad = BTPageGetMeta(metapg);
/*
* The expected case here is btm_fastlevel == targetlevel+1; if
* the fastlevel is <= targetlevel, something is wrong, and we
* choose to overwrite it to fix it.
*/
if (metad->btm_fastlevel > targetlevel + 1)
{
/* no update wanted */
_bt_relbuf(rel, metabuf);
metabuf = InvalidBuffer;
}
}
}
/*
* Check that the parent-page index items we're about to delete/overwrite
* contain what we expect. This can fail if the index has become
* corrupt for some reason. We want to throw any error before entering
* the critical section --- otherwise it'd be a PANIC.
*
* The test on the target item is just an Assert because _bt_getstackbuf
* should have guaranteed it has the expected contents. The test on the
* next-child downlink is known to sometimes fail in the field, though.
*/
page = BufferGetPage(pbuf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
#ifdef USE_ASSERT_CHECKING
itemid = PageGetItemId(page, poffset);
itup = (IndexTuple) PageGetItem(page, itemid);
Assert(ItemPointerGetBlockNumber(&(itup->t_tid)) == target);
#endif
if (!parent_half_dead)
{
OffsetNumber nextoffset;
nextoffset = OffsetNumberNext(poffset);
itemid = PageGetItemId(page, nextoffset);
itup = (IndexTuple) PageGetItem(page, itemid);
if (ItemPointerGetBlockNumber(&(itup->t_tid)) != rightsib)
elog(ERROR, "right sibling %u of block %u is not next child %u of block %u in index \"%s\"",
rightsib, target, ItemPointerGetBlockNumber(&(itup->t_tid)),
parent, RelationGetRelationName(rel));
}
/*
* Here we begin doing the deletion.
*/
/* No ereport(ERROR) until changes are logged */
START_CRIT_SECTION();
/*
* Update parent. The normal case is a tad tricky because we want to
* delete the target's downlink and the *following* key. Easiest way is
* to copy the right sibling's downlink over the target downlink, and then
* delete the following item.
*/
if (parent_half_dead)
{
PageIndexTupleDelete(page, poffset);
opaque->btpo_flags |= BTP_HALF_DEAD;
}
else
{
OffsetNumber nextoffset;
itemid = PageGetItemId(page, poffset);
itup = (IndexTuple) PageGetItem(page, itemid);
ItemPointerSet(&(itup->t_tid), rightsib, P_HIKEY);
nextoffset = OffsetNumberNext(poffset);
PageIndexTupleDelete(page, nextoffset);
}
/*
* Update siblings' side-links. Note the target page's side-links will
* continue to point to the siblings. Asserts here are just rechecking
* things we already verified above.
*/
if (BufferIsValid(lbuf))
{
page = BufferGetPage(lbuf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
Assert(opaque->btpo_next == target);
opaque->btpo_next = rightsib;
}
page = BufferGetPage(rbuf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
Assert(opaque->btpo_prev == target);
opaque->btpo_prev = leftsib;
rightsib_empty = (P_FIRSTDATAKEY(opaque) > PageGetMaxOffsetNumber(page));
/*
* Mark the page itself deleted. It can be recycled when all current
* transactions are gone.
*/
page = BufferGetPage(buf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
opaque->btpo_flags &= ~BTP_HALF_DEAD;
opaque->btpo_flags |= BTP_DELETED;
opaque->btpo.xact = ReadNewTransactionId();
/* And update the metapage, if needed */
if (BufferIsValid(metabuf))
{
metad->btm_fastroot = rightsib;
metad->btm_fastlevel = targetlevel;
MarkBufferDirty(metabuf);
}
/* Must mark buffers dirty before XLogInsert */
MarkBufferDirty(pbuf);
MarkBufferDirty(rbuf);
MarkBufferDirty(buf);
if (BufferIsValid(lbuf))
MarkBufferDirty(lbuf);
/* XLOG stuff */
if (!rel->rd_istemp)
{
xl_btree_delete_page xlrec;
xl_btree_metadata xlmeta;
uint8 xlinfo;
XLogRecPtr recptr;
XLogRecData rdata[5];
XLogRecData *nextrdata;
xlrec.target.node = rel->rd_node;
ItemPointerSet(&(xlrec.target.tid), parent, poffset);
xlrec.deadblk = target;
xlrec.leftblk = leftsib;
xlrec.rightblk = rightsib;
xlrec.btpo_xact = opaque->btpo.xact;
rdata[0].data = (char *) &xlrec;
rdata[0].len = SizeOfBtreeDeletePage;
rdata[0].buffer = InvalidBuffer;
rdata[0].next = nextrdata = &(rdata[1]);
if (BufferIsValid(metabuf))
{
xlmeta.root = metad->btm_root;
xlmeta.level = metad->btm_level;
xlmeta.fastroot = metad->btm_fastroot;
xlmeta.fastlevel = metad->btm_fastlevel;
nextrdata->data = (char *) &xlmeta;
nextrdata->len = sizeof(xl_btree_metadata);
nextrdata->buffer = InvalidBuffer;
nextrdata->next = nextrdata + 1;
nextrdata++;
xlinfo = XLOG_BTREE_DELETE_PAGE_META;
}
else if (parent_half_dead)
xlinfo = XLOG_BTREE_DELETE_PAGE_HALF;
else
xlinfo = XLOG_BTREE_DELETE_PAGE;
nextrdata->data = NULL;
nextrdata->len = 0;
nextrdata->next = nextrdata + 1;
nextrdata->buffer = pbuf;
nextrdata->buffer_std = true;
nextrdata++;
nextrdata->data = NULL;
nextrdata->len = 0;
nextrdata->buffer = rbuf;
nextrdata->buffer_std = true;
nextrdata->next = NULL;
if (BufferIsValid(lbuf))
{
nextrdata->next = nextrdata + 1;
nextrdata++;
nextrdata->data = NULL;
nextrdata->len = 0;
nextrdata->buffer = lbuf;
nextrdata->buffer_std = true;
nextrdata->next = NULL;
}
recptr = XLogInsert(RM_BTREE_ID, xlinfo, rdata);
if (BufferIsValid(metabuf))
{
PageSetLSN(metapg, recptr);
PageSetTLI(metapg, ThisTimeLineID);
}
page = BufferGetPage(pbuf);
PageSetLSN(page, recptr);
PageSetTLI(page, ThisTimeLineID);
page = BufferGetPage(rbuf);
PageSetLSN(page, recptr);
PageSetTLI(page, ThisTimeLineID);
page = BufferGetPage(buf);
PageSetLSN(page, recptr);
PageSetTLI(page, ThisTimeLineID);
if (BufferIsValid(lbuf))
{
page = BufferGetPage(lbuf);
PageSetLSN(page, recptr);
PageSetTLI(page, ThisTimeLineID);
}
}
END_CRIT_SECTION();
/* release metapage; send out relcache inval if metapage changed */
if (BufferIsValid(metabuf))
{
CacheInvalidateRelcache(rel);
_bt_relbuf(rel, metabuf);
}
/* can always release leftsib immediately */
if (BufferIsValid(lbuf))
_bt_relbuf(rel, lbuf);
/*
* If parent became half dead, recurse to delete it. Otherwise, if right
* sibling is empty and is now the last child of the parent, recurse to
* try to delete it. (These cases cannot apply at the same time, though
* the second case might itself recurse to the first.)
*
* When recursing to parent, we hold the lock on the target page until
* done. This delays any insertions into the keyspace that was just
* effectively reassigned to the parent's right sibling. If we allowed
* that, and there were enough such insertions before we finish deleting
* the parent, page splits within that keyspace could lead to inserting
* out-of-order keys into the grandparent level. It is thought that that
* wouldn't have any serious consequences, but it still seems like a
* pretty bad idea.
*/
if (parent_half_dead)
{
/* recursive call will release pbuf */
_bt_relbuf(rel, rbuf);
result = _bt_pagedel(rel, pbuf, stack->bts_parent) + 1;
_bt_relbuf(rel, buf);
}
else if (parent_one_child && rightsib_empty)
{
_bt_relbuf(rel, pbuf);
_bt_relbuf(rel, buf);
/* recursive call will release rbuf */
result = _bt_pagedel(rel, rbuf, stack) + 1;
}
else
{
_bt_relbuf(rel, pbuf);
_bt_relbuf(rel, buf);
_bt_relbuf(rel, rbuf);
result = 1;
}
return result;
}
