v15-0001-Enhance-nbtree-ScalarArrayOp-execution.patch
application/octet-stream
Filename: v15-0001-Enhance-nbtree-ScalarArrayOp-execution.patch
Type: application/octet-stream
Part: 1
Patch
Same data as JSON:
GET /api/v1/attachments/:id/patch
the parsed metadata as JSON — format, series position, per-file stats; never the diff bytes.
API reference →
Format: format-patch
Series: patch v15-0001
Subject: Enhance nbtree ScalarArrayOp execution.
| File | + | − |
|---|---|---|
| doc/src/sgml/monitoring.sgml | 13 | 0 |
| src/backend/access/nbtree/nbtree.c | 46 | 68 |
| src/backend/access/nbtree/nbtsearch.c | 139 | 53 |
| src/backend/access/nbtree/nbtutils.c | 2474 | 251 |
| src/backend/optimizer/path/indxpath.c | 18 | 72 |
| src/backend/utils/adt/selfuncs.c | 67 | 55 |
| src/include/access/nbtree.h | 33 | 18 |
| src/test/regress/expected/create_index.out | 15 | 18 |
| src/test/regress/expected/join.out | 2 | 3 |
| src/test/regress/sql/create_index.sql | 5 | 7 |
From a2d6c95bf83659cd3e8fa61222eb49945860c057 Mon Sep 17 00:00:00 2001
From: Peter Geoghegan <pg@bowt.ie>
Date: Sat, 17 Jun 2023 17:03:36 -0700
Subject: [PATCH v15] Enhance nbtree ScalarArrayOp execution.
Commit 9e8da0f7 taught nbtree to handle ScalarArrayOpExpr quals
natively. This works by pushing down the full context (the array keys)
to the nbtree index AM, enabling it to execute multiple primitive index
scans that the planner treats as one continuous index scan/index path.
This earlier enhancement enabled nbtree ScalarArrayOp index-only scans.
It also allowed scans with ScalarArrayOp quals to return ordered results
(with some notable restrictions, described further down).
Take this general approach a lot further: teach nbtree SAOP index scans
to determine how best to execute ScalarArrayOp scans (how many primitive
index scans to use under the hood) by applying information about the
physical characteristics of the index at runtime. This approach can be
far more efficient. Many cases that previously required thousands of
index descents now require as few as one single index descent. And, all
SAOP scans reliably avoid duplicative leaf page accesses (just like any
other nbtree index scan).
The scan's arrays now advance using binary searches for the array
element that best matches the next tuple's attribute value. This whole
process makes required scan key arrays (i.e. arrays from scan keys that
can terminate the scan) ratchet forward in lockstep with the index scan.
Non-required arrays (i.e. arrays from scan keys that can only exclude
non-matching tuples) are for the most part advanced via this same search
process. We just can't assume a fixed relationship between the current
element of any non-required array and the progress of the index scan
through the index's key space (that would be wrong).
Naturally, only required SAOP scan keys trigger skipping over leaf pages
(non-required arrays cannot safely end or start primitive index scans).
Consequently, index scans of a composite index with (say) a high-order
inequality scan key (which we'll mark required) and a low-order SAOP
scan key (which we'll mark non-required) will now reliably output rows
in index order. Such scans are always executed as one large index scan
under the hood, which is obviously the most efficient way to do it, for
the usual reason (no more wasting cycles on repeat leaf page accesses).
Generalizing SAOP execution along these lines removes any question of
index scans outputting tuples in any order that isn't the index's order.
This allow us to remove various special cases from the planner -- which
in turn makes the nbtree work more widely applicable and more effective.
Bugfix commit 807a40c5 taught the planner to avoid generating unsafe
path keys: path keys on a multicolumn index path, with a SAOP clause on
any attribute beyond the first/most significant attribute. These cases
are now all safe, so we go back to generating path keys without regard
for the presence of SAOP clauses (just like with any other clause type).
Also undo changes from follow-up bugfix commit a4523c5a, which taught
the planner to produce alternative index paths without any low-order
ScalarArrayOpExpr quals (making the SAOP quals into filter quals).
We'll no longer generate these alternative paths, which can no longer
offer any advantage over the index qual paths that we do still generate.
Affected queries thereby avoid all of the disadvantages that come from
using filter quals within index scan nodes. In particular, they can
avoid the extra heap page accesses previously incurred when using filter
quals to exclude non-matching tuples (index quals can be used instead).
This shift is expected to be fairly common in real world applications,
especially with queries that have multiple SAOPs that can now all be
used as index quals when scanning a composite index. Queries with
low-order SAOPs (especially non-required ones) are also likely to see a
significant reduction in heap page accesses.
There is a theoretical risk that removing restrictions on SAOP index
paths from the planner will break compatibility with any out-of-core
amcanorder-based index AMs that coevolved with nbtree. Such an index AM
could have had similar limitations around SOAP execution, and so could
have come to rely on the planner workarounds removed by this commit.
Although it seems unlikely that such an index AM really exists, it still
warrants a pro forma compatibility item in the release notes.
Author: Peter Geoghegan <pg@bowt.ie>
Author: Matthias van de Meent <boekewurm+postgres@gmail.com>
Reviewed-By: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-By: Matthias van de Meent <boekewurm+postgres@gmail.com>
Reviewed-By: Tomas Vondra <tomas.vondra@enterprisedb.com>
Discussion: https://postgr.es/m/CAH2-Wz=ksvN_sjcnD1+Bt-WtifRA5ok48aDYnq3pkKhxgMQpcw@mail.gmail.com
---
src/include/access/nbtree.h | 51 +-
src/backend/access/nbtree/nbtree.c | 114 +-
src/backend/access/nbtree/nbtsearch.c | 192 +-
src/backend/access/nbtree/nbtutils.c | 2725 ++++++++++++++++++--
src/backend/optimizer/path/indxpath.c | 90 +-
src/backend/utils/adt/selfuncs.c | 122 +-
doc/src/sgml/monitoring.sgml | 13 +
src/test/regress/expected/create_index.out | 33 +-
src/test/regress/expected/join.out | 5 +-
src/test/regress/sql/create_index.sql | 12 +-
10 files changed, 2812 insertions(+), 545 deletions(-)
diff --git a/src/include/access/nbtree.h b/src/include/access/nbtree.h
index 6eb162052..62088f140 100644
--- a/src/include/access/nbtree.h
+++ b/src/include/access/nbtree.h
@@ -960,7 +960,7 @@ typedef struct BTScanPosData
* moreLeft and moreRight track whether we think there may be matching
* index entries to the left and right of the current page, respectively.
* We can clear the appropriate one of these flags when _bt_checkkeys()
- * returns continuescan = false.
+ * sets BTReadPageState.continuescan = false.
*/
bool moreLeft;
bool moreRight;
@@ -1022,9 +1022,8 @@ typedef BTScanPosData *BTScanPos;
/* We need one of these for each equality-type SK_SEARCHARRAY scan key */
typedef struct BTArrayKeyInfo
{
- int scan_key; /* index of associated key in arrayKeyData */
+ int scan_key; /* index of associated key in keyData */
int cur_elem; /* index of current element in elem_values */
- int mark_elem; /* index of marked element in elem_values */
int num_elems; /* number of elems in current array value */
Datum *elem_values; /* array of num_elems Datums */
} BTArrayKeyInfo;
@@ -1037,14 +1036,14 @@ typedef struct BTScanOpaqueData
ScanKey keyData; /* array of preprocessed scan keys */
/* workspace for SK_SEARCHARRAY support */
- ScanKey arrayKeyData; /* modified copy of scan->keyData */
- bool arraysStarted; /* Started array keys, but have yet to "reach
- * past the end" of all arrays? */
- int numArrayKeys; /* number of equality-type array keys (-1 if
- * there are any unsatisfiable array keys) */
- int arrayKeyCount; /* count indicating number of array scan keys
- * processed */
+ int numArrayKeys; /* number of equality-type array keys */
+ ScanDirection primScanDir; /* Scan direction for most recent _bt_first */
+ bool scanBehind; /* Scan might be behind arrays? */
+ bool needPrimScan; /* Need primscan to continue in primScanDir? */
BTArrayKeyInfo *arrayKeys; /* info about each equality-type array key */
+ FmgrInfo *orderProcs; /* ORDER procs for all equality-type keys */
+ int numPrimScans; /* Running tally of # primitive index scans
+ * (used to coordinate parallel workers) */
MemoryContext arrayContext; /* scan-lifespan context for array data */
/* info about killed items if any (killedItems is NULL if never used) */
@@ -1075,6 +1074,26 @@ typedef struct BTScanOpaqueData
typedef BTScanOpaqueData *BTScanOpaque;
+/*
+ * _bt_readpage state used across _bt_checkkeys calls for a page
+ */
+typedef struct BTReadPageState
+{
+ /* Input parameters, set by _bt_readpage for _bt_checkkeys */
+ ScanDirection dir; /* current scan direction */
+ IndexTuple finaltup; /* Needed by scans with array keys */
+
+ /* Output parameter, set by _bt_checkkeys for _bt_readpage */
+ bool continuescan; /* Terminate ongoing (primitive) index scan? */
+
+ /*
+ * Input and output parameters, set and unset by both _bt_readpage and
+ * _bt_checkkeys to manage precheck optimizations
+ */
+ bool prechecked; /* precheck set continuescan? */
+ bool firstmatch; /* at least one match so far? */
+} BTReadPageState;
+
/*
* We use some private sk_flags bits in preprocessed scan keys. We're allowed
* to use bits 16-31 (see skey.h). The uppermost bits are copied from the
@@ -1152,7 +1171,7 @@ extern bool btcanreturn(Relation index, int attno);
extern bool _bt_parallel_seize(IndexScanDesc scan, BlockNumber *pageno);
extern void _bt_parallel_release(IndexScanDesc scan, BlockNumber scan_page);
extern void _bt_parallel_done(IndexScanDesc scan);
-extern void _bt_parallel_advance_array_keys(IndexScanDesc scan);
+extern void _bt_parallel_next_primitive_scan(IndexScanDesc scan);
/*
* prototypes for functions in nbtdedup.c
@@ -1243,15 +1262,11 @@ extern Buffer _bt_get_endpoint(Relation rel, uint32 level, bool rightmost);
*/
extern BTScanInsert _bt_mkscankey(Relation rel, IndexTuple itup);
extern void _bt_freestack(BTStack stack);
-extern void _bt_preprocess_array_keys(IndexScanDesc scan);
+extern bool _bt_start_prim_scan(IndexScanDesc scan, ScanDirection dir);
extern void _bt_start_array_keys(IndexScanDesc scan, ScanDirection dir);
-extern bool _bt_advance_array_keys(IndexScanDesc scan, ScanDirection dir);
-extern void _bt_mark_array_keys(IndexScanDesc scan);
-extern void _bt_restore_array_keys(IndexScanDesc scan);
extern void _bt_preprocess_keys(IndexScanDesc scan);
-extern bool _bt_checkkeys(IndexScanDesc scan, IndexTuple tuple,
- int tupnatts, ScanDirection dir, bool *continuescan,
- bool requiredMatchedByPrecheck, bool haveFirstMatch);
+extern bool _bt_checkkeys(IndexScanDesc scan, BTReadPageState *pstate, bool arrayKeys,
+ IndexTuple tuple, int tupnatts);
extern void _bt_killitems(IndexScanDesc scan);
extern BTCycleId _bt_vacuum_cycleid(Relation rel);
extern BTCycleId _bt_start_vacuum(Relation rel);
diff --git a/src/backend/access/nbtree/nbtree.c b/src/backend/access/nbtree/nbtree.c
index 41df1027d..7a6d18777 100644
--- a/src/backend/access/nbtree/nbtree.c
+++ b/src/backend/access/nbtree/nbtree.c
@@ -46,8 +46,8 @@
* BTPARALLEL_IDLE indicates that no backend is currently advancing the scan
* to a new page; some process can start doing that.
*
- * BTPARALLEL_DONE indicates that the scan is complete (including error exit).
- * We reach this state once for every distinct combination of array keys.
+ * BTPARALLEL_DONE indicates that the primitive index scan is complete
+ * (including error exit). Reached once per primitive index scan.
*/
typedef enum
{
@@ -67,8 +67,8 @@ typedef struct BTParallelScanDescData
BTPS_State btps_pageStatus; /* indicates whether next page is
* available for scan. see above for
* possible states of parallel scan. */
- int btps_arrayKeyCount; /* count indicating number of array scan
- * keys processed by parallel scan */
+ int btps_numPrimScans; /* count indicating number of primitive
+ * index scans (used with array keys) */
slock_t btps_mutex; /* protects above variables */
ConditionVariable btps_cv; /* used to synchronize parallel scan */
} BTParallelScanDescData;
@@ -204,21 +204,7 @@ btgettuple(IndexScanDesc scan, ScanDirection dir)
/* btree indexes are never lossy */
scan->xs_recheck = false;
- /*
- * If we have any array keys, initialize them during first call for a
- * scan. We can't do this in btrescan because we don't know the scan
- * direction at that time.
- */
- if (so->numArrayKeys && !BTScanPosIsValid(so->currPos))
- {
- /* punt if we have any unsatisfiable array keys */
- if (so->numArrayKeys < 0)
- return false;
-
- _bt_start_array_keys(scan, dir);
- }
-
- /* This loop handles advancing to the next array elements, if any */
+ /* Each loop iteration performs another primitive index scan */
do
{
/*
@@ -260,8 +246,8 @@ btgettuple(IndexScanDesc scan, ScanDirection dir)
/* If we have a tuple, return it ... */
if (res)
break;
- /* ... otherwise see if we have more array keys to deal with */
- } while (so->numArrayKeys && _bt_advance_array_keys(scan, dir));
+ /* ... otherwise see if we need another primitive index scan */
+ } while (so->numArrayKeys && _bt_start_prim_scan(scan, dir));
return res;
}
@@ -276,19 +262,7 @@ btgetbitmap(IndexScanDesc scan, TIDBitmap *tbm)
int64 ntids = 0;
ItemPointer heapTid;
- /*
- * If we have any array keys, initialize them.
- */
- if (so->numArrayKeys)
- {
- /* punt if we have any unsatisfiable array keys */
- if (so->numArrayKeys < 0)
- return ntids;
-
- _bt_start_array_keys(scan, ForwardScanDirection);
- }
-
- /* This loop handles advancing to the next array elements, if any */
+ /* Each loop iteration performs another primitive index scan */
do
{
/* Fetch the first page & tuple */
@@ -318,8 +292,8 @@ btgetbitmap(IndexScanDesc scan, TIDBitmap *tbm)
ntids++;
}
}
- /* Now see if we have more array keys to deal with */
- } while (so->numArrayKeys && _bt_advance_array_keys(scan, ForwardScanDirection));
+ /* Now see if we need another primitive index scan */
+ } while (so->numArrayKeys && _bt_start_prim_scan(scan, ForwardScanDirection));
return ntids;
}
@@ -348,10 +322,12 @@ btbeginscan(Relation rel, int nkeys, int norderbys)
else
so->keyData = NULL;
- so->arrayKeyData = NULL; /* assume no array keys for now */
- so->arraysStarted = false;
so->numArrayKeys = 0;
+ so->primScanDir = NoMovementScanDirection;
+ so->scanBehind = false;
+ so->needPrimScan = false;
so->arrayKeys = NULL;
+ so->orderProcs = NULL;
so->arrayContext = NULL;
so->killedItems = NULL; /* until needed */
@@ -391,7 +367,11 @@ btrescan(IndexScanDesc scan, ScanKey scankey, int nscankeys,
}
so->markItemIndex = -1;
- so->arrayKeyCount = 0;
+ so->numArrayKeys = 0;
+ so->primScanDir = NoMovementScanDirection;
+ so->scanBehind = false;
+ so->needPrimScan = false;
+ so->numPrimScans = 0;
BTScanPosUnpinIfPinned(so->markPos);
BTScanPosInvalidate(so->markPos);
@@ -425,9 +405,6 @@ btrescan(IndexScanDesc scan, ScanKey scankey, int nscankeys,
scankey,
scan->numberOfKeys * sizeof(ScanKeyData));
so->numberOfKeys = 0; /* until _bt_preprocess_keys sets it */
-
- /* If any keys are SK_SEARCHARRAY type, set up array-key info */
- _bt_preprocess_array_keys(scan);
}
/*
@@ -455,7 +432,7 @@ btendscan(IndexScanDesc scan)
/* Release storage */
if (so->keyData != NULL)
pfree(so->keyData);
- /* so->arrayKeyData and so->arrayKeys are in arrayContext */
+ /* so->arrayKeys is in arrayContext */
if (so->arrayContext != NULL)
MemoryContextDelete(so->arrayContext);
if (so->killedItems != NULL)
@@ -490,10 +467,6 @@ btmarkpos(IndexScanDesc scan)
BTScanPosInvalidate(so->markPos);
so->markItemIndex = -1;
}
-
- /* Also record the current positions of any array keys */
- if (so->numArrayKeys)
- _bt_mark_array_keys(scan);
}
/*
@@ -504,10 +477,6 @@ btrestrpos(IndexScanDesc scan)
{
BTScanOpaque so = (BTScanOpaque) scan->opaque;
- /* Restore the marked positions of any array keys */
- if (so->numArrayKeys)
- _bt_restore_array_keys(scan);
-
if (so->markItemIndex >= 0)
{
/*
@@ -546,6 +515,9 @@ btrestrpos(IndexScanDesc scan)
if (so->currTuples)
memcpy(so->currTuples, so->markTuples,
so->markPos.nextTupleOffset);
+ /* Reset the scan's array keys (see _bt_steppage for why) */
+ if (so->numArrayKeys)
+ _bt_start_array_keys(scan, so->primScanDir);
}
else
BTScanPosInvalidate(so->currPos);
@@ -572,7 +544,7 @@ btinitparallelscan(void *target)
SpinLockInit(&bt_target->btps_mutex);
bt_target->btps_scanPage = InvalidBlockNumber;
bt_target->btps_pageStatus = BTPARALLEL_NOT_INITIALIZED;
- bt_target->btps_arrayKeyCount = 0;
+ bt_target->btps_numPrimScans = 0;
ConditionVariableInit(&bt_target->btps_cv);
}
@@ -598,7 +570,7 @@ btparallelrescan(IndexScanDesc scan)
SpinLockAcquire(&btscan->btps_mutex);
btscan->btps_scanPage = InvalidBlockNumber;
btscan->btps_pageStatus = BTPARALLEL_NOT_INITIALIZED;
- btscan->btps_arrayKeyCount = 0;
+ btscan->btps_numPrimScans = 0;
SpinLockRelease(&btscan->btps_mutex);
}
@@ -609,7 +581,11 @@ btparallelrescan(IndexScanDesc scan)
*
* The return value is true if we successfully seized the scan and false
* if we did not. The latter case occurs if no pages remain for the current
- * set of scankeys.
+ * primitive index scan.
+ *
+ * When array scan keys are in use, each worker process independently advances
+ * its array keys. It's crucial that each worker process never be allowed to
+ * scan a page from before the current scan position.
*
* If the return value is true, *pageno returns the next or current page
* of the scan (depending on the scan direction). An invalid block number
@@ -640,16 +616,16 @@ _bt_parallel_seize(IndexScanDesc scan, BlockNumber *pageno)
SpinLockAcquire(&btscan->btps_mutex);
pageStatus = btscan->btps_pageStatus;
- if (so->arrayKeyCount < btscan->btps_arrayKeyCount)
+ if (so->numPrimScans < btscan->btps_numPrimScans)
{
- /* Parallel scan has already advanced to a new set of scankeys. */
+ /* Top-level scan already moved on to next primitive index scan */
status = false;
}
else if (pageStatus == BTPARALLEL_DONE)
{
/*
- * We're done with this set of scankeys. This may be the end, or
- * there could be more sets to try.
+ * We're done with this primitive index scan. The top-level index
+ * scan might require additional primitive index scans.
*/
status = false;
}
@@ -681,9 +657,12 @@ _bt_parallel_seize(IndexScanDesc scan, BlockNumber *pageno)
void
_bt_parallel_release(IndexScanDesc scan, BlockNumber scan_page)
{
+ BTScanOpaque so PG_USED_FOR_ASSERTS_ONLY = (BTScanOpaque) scan->opaque;
ParallelIndexScanDesc parallel_scan = scan->parallel_scan;
BTParallelScanDesc btscan;
+ Assert(!so->needPrimScan);
+
btscan = (BTParallelScanDesc) OffsetToPointer((void *) parallel_scan,
parallel_scan->ps_offset);
@@ -717,12 +696,11 @@ _bt_parallel_done(IndexScanDesc scan)
parallel_scan->ps_offset);
/*
- * Mark the parallel scan as done for this combination of scan keys,
- * unless some other process already did so. See also
- * _bt_advance_array_keys.
+ * Mark the primitive index scan as done, unless some other process
+ * already did so. See also _bt_start_prim_scan.
*/
SpinLockAcquire(&btscan->btps_mutex);
- if (so->arrayKeyCount >= btscan->btps_arrayKeyCount &&
+ if (so->numPrimScans >= btscan->btps_numPrimScans &&
btscan->btps_pageStatus != BTPARALLEL_DONE)
{
btscan->btps_pageStatus = BTPARALLEL_DONE;
@@ -736,14 +714,14 @@ _bt_parallel_done(IndexScanDesc scan)
}
/*
- * _bt_parallel_advance_array_keys() -- Advances the parallel scan for array
- * keys.
+ * _bt_parallel_next_primitive_scan() -- Advances parallel primitive scan
+ * counter when array keys are in use.
*
- * Updates the count of array keys processed for both local and parallel
+ * Updates the count of primitive index scans for both local and parallel
* scans.
*/
void
-_bt_parallel_advance_array_keys(IndexScanDesc scan)
+_bt_parallel_next_primitive_scan(IndexScanDesc scan)
{
BTScanOpaque so = (BTScanOpaque) scan->opaque;
ParallelIndexScanDesc parallel_scan = scan->parallel_scan;
@@ -752,13 +730,13 @@ _bt_parallel_advance_array_keys(IndexScanDesc scan)
btscan = (BTParallelScanDesc) OffsetToPointer((void *) parallel_scan,
parallel_scan->ps_offset);
- so->arrayKeyCount++;
+ so->numPrimScans++;
SpinLockAcquire(&btscan->btps_mutex);
if (btscan->btps_pageStatus == BTPARALLEL_DONE)
{
btscan->btps_scanPage = InvalidBlockNumber;
btscan->btps_pageStatus = BTPARALLEL_NOT_INITIALIZED;
- btscan->btps_arrayKeyCount++;
+ btscan->btps_numPrimScans++;
}
SpinLockRelease(&btscan->btps_mutex);
}
diff --git a/src/backend/access/nbtree/nbtsearch.c b/src/backend/access/nbtree/nbtsearch.c
index 23e723a23..be201652c 100644
--- a/src/backend/access/nbtree/nbtsearch.c
+++ b/src/backend/access/nbtree/nbtsearch.c
@@ -907,11 +907,38 @@ _bt_first(IndexScanDesc scan, ScanDirection dir)
*/
if (!so->qual_ok)
{
- /* Notify any other workers that we're done with this scan key. */
+ /* Notify any other workers that this primitive scan is done */
_bt_parallel_done(scan);
return false;
}
+ if (so->numArrayKeys)
+ {
+ if (ScanDirectionIsNoMovement(so->primScanDir))
+ {
+ /*
+ * First primitive index scan (for current btrescan).
+ *
+ * Initialize arrays, and the corresponding scan keys that were
+ * just output by _bt_preprocess_keys.
+ */
+ _bt_start_array_keys(scan, dir);
+ }
+ else
+ {
+ /*
+ * Just stick with the array keys set by _bt_checkkeys at the end
+ * of the previous primitive index scan.
+ *
+ * Note: The initial primitive index scan's _bt_preprocess_keys
+ * call actually outputs new keys. Later calls are just no-ops.
+ * We're just here to build an insertion scan key using values
+ * already set in so->keyData[] by _bt_checkkeys.
+ */
+ }
+ so->primScanDir = dir;
+ }
+
/*
* For parallel scans, get the starting page from shared state. If the
* scan has not started, proceed to find out first leaf page in the usual
@@ -1527,11 +1554,10 @@ _bt_readpage(IndexScanDesc scan, ScanDirection dir, OffsetNumber offnum,
BTPageOpaque opaque;
OffsetNumber minoff;
OffsetNumber maxoff;
- int itemIndex;
- bool continuescan;
- int indnatts;
- bool continuescanPrechecked;
- bool haveFirstMatch = false;
+ BTReadPageState pstate;
+ bool arrayKeys;
+ int itemIndex,
+ indnatts;
/*
* We must have the buffer pinned and locked, but the usual macro can't be
@@ -1551,8 +1577,14 @@ _bt_readpage(IndexScanDesc scan, ScanDirection dir, OffsetNumber offnum,
_bt_parallel_release(scan, BufferGetBlockNumber(so->currPos.buf));
}
- continuescan = true; /* default assumption */
+ pstate.dir = dir;
+ pstate.finaltup = NULL;
+ pstate.continuescan = true; /* default assumption */
+ pstate.prechecked = false;
+ pstate.firstmatch = false;
indnatts = IndexRelationGetNumberOfAttributes(scan->indexRelation);
+ arrayKeys = so->numArrayKeys != 0;
+
minoff = P_FIRSTDATAKEY(opaque);
maxoff = PageGetMaxOffsetNumber(page);
@@ -1598,10 +1630,28 @@ _bt_readpage(IndexScanDesc scan, ScanDirection dir, OffsetNumber offnum,
* corresponding value from the last item on the page. So checking with
* the last item on the page would give a more precise answer.
*
- * We skip this for the first page in the scan to evade the possible
- * slowdown of the point queries.
+ * We skip this for the scan's first page to avoid slowing down point
+ * queries. We also have to avoid applying the optimization in rare cases
+ * where it's not yet clear that the scan is at or ahead of its current
+ * array keys. If we're behind, but not too far behind (the start of
+ * tuples matching the current keys is somewhere before the last item),
+ * then the optimization is unsafe.
+ *
+ * Cases with multiple distinct sets of required array keys for key space
+ * from the same leaf page can _attempt_ to use the precheck optimization,
+ * though. It won't work out, but there's no better way of figuring that
+ * out than just optimistically attempting the precheck.
+ *
+ * The array keys safety issue is related to our reliance on _bt_first
+ * passing us an offnum that's exactly at the beginning of where equal
+ * tuples are to be found. The underlying problem is that we have no
+ * built-in ability to tell the difference between the start of required
+ * equality matches and the end of required equality matches. Array key
+ * advancement within _bt_checkkeys has to act as a "_bt_first surrogate"
+ * whenever the start of tuples matching the next set of array keys is
+ * close to the end of tuples matching the current/last set of array keys.
*/
- if (!firstPage && minoff < maxoff)
+ if (!firstPage && !so->scanBehind && minoff < maxoff)
{
ItemId iid;
IndexTuple itup;
@@ -1610,21 +1660,24 @@ _bt_readpage(IndexScanDesc scan, ScanDirection dir, OffsetNumber offnum,
itup = (IndexTuple) PageGetItem(page, iid);
/*
- * Do the precheck. Note that we pass the pointer to the
- * 'continuescanPrechecked' to the 'continuescan' argument. That will
- * set flag to true if all required keys are satisfied and false
- * otherwise.
+ * Do the precheck, while avoiding advancing the scan's array keys
+ * prematurely
*/
- (void) _bt_checkkeys(scan, itup, indnatts, dir,
- &continuescanPrechecked, false, false);
- }
- else
- {
- continuescanPrechecked = false;
+ _bt_checkkeys(scan, &pstate, false, itup, indnatts);
+ pstate.prechecked = pstate.continuescan;
+ pstate.continuescan = true; /* reset */
}
if (ScanDirectionIsForward(dir))
{
+ /* SK_SEARCHARRAY forward scans must provide high key up front */
+ if (arrayKeys && !P_RIGHTMOST(opaque))
+ {
+ ItemId iid = PageGetItemId(page, P_HIKEY);
+
+ pstate.finaltup = (IndexTuple) PageGetItem(page, iid);
+ }
+
/* load items[] in ascending order */
itemIndex = 0;
@@ -1649,23 +1702,13 @@ _bt_readpage(IndexScanDesc scan, ScanDirection dir, OffsetNumber offnum,
itup = (IndexTuple) PageGetItem(page, iid);
Assert(!BTreeTupleIsPivot(itup));
- passes_quals = _bt_checkkeys(scan, itup, indnatts, dir,
- &continuescan,
- continuescanPrechecked,
- haveFirstMatch);
+ passes_quals = _bt_checkkeys(scan, &pstate, arrayKeys,
+ itup, indnatts);
- /*
- * If the result of prechecking required keys was true, then in
- * assert-enabled builds we also recheck that the _bt_checkkeys()
- * result is the same.
- */
- Assert((!continuescanPrechecked && haveFirstMatch) ||
- passes_quals == _bt_checkkeys(scan, itup, indnatts, dir,
- &continuescan, false, false));
if (passes_quals)
{
/* tuple passes all scan key conditions */
- haveFirstMatch = true;
+ pstate.firstmatch = true;
if (!BTreeTupleIsPosting(itup))
{
/* Remember it */
@@ -1696,7 +1739,7 @@ _bt_readpage(IndexScanDesc scan, ScanDirection dir, OffsetNumber offnum,
}
}
/* When !continuescan, there can't be any more matches, so stop */
- if (!continuescan)
+ if (!pstate.continuescan)
break;
offnum = OffsetNumberNext(offnum);
@@ -1713,17 +1756,18 @@ _bt_readpage(IndexScanDesc scan, ScanDirection dir, OffsetNumber offnum,
* only appear on non-pivot tuples on the right sibling page are
* common.
*/
- if (continuescan && !P_RIGHTMOST(opaque))
+ if (pstate.continuescan && !P_RIGHTMOST(opaque))
{
ItemId iid = PageGetItemId(page, P_HIKEY);
IndexTuple itup = (IndexTuple) PageGetItem(page, iid);
int truncatt;
truncatt = BTreeTupleGetNAtts(itup, scan->indexRelation);
- _bt_checkkeys(scan, itup, truncatt, dir, &continuescan, false, false);
+ pstate.prechecked = false; /* precheck didn't cover HIKEY */
+ _bt_checkkeys(scan, &pstate, arrayKeys, itup, truncatt);
}
- if (!continuescan)
+ if (!pstate.continuescan)
so->currPos.moreRight = false;
Assert(itemIndex <= MaxTIDsPerBTreePage);
@@ -1733,6 +1777,14 @@ _bt_readpage(IndexScanDesc scan, ScanDirection dir, OffsetNumber offnum,
}
else
{
+ /* SK_SEARCHARRAY backward scans must provide final tuple up front */
+ if (arrayKeys && minoff <= maxoff && !P_LEFTMOST(opaque))
+ {
+ ItemId iid = PageGetItemId(page, minoff);
+
+ pstate.finaltup = (IndexTuple) PageGetItem(page, iid);
+ }
+
/* load items[] in descending order */
itemIndex = MaxTIDsPerBTreePage;
@@ -1772,23 +1824,13 @@ _bt_readpage(IndexScanDesc scan, ScanDirection dir, OffsetNumber offnum,
itup = (IndexTuple) PageGetItem(page, iid);
Assert(!BTreeTupleIsPivot(itup));
- passes_quals = _bt_checkkeys(scan, itup, indnatts, dir,
- &continuescan,
- continuescanPrechecked,
- haveFirstMatch);
+ passes_quals = _bt_checkkeys(scan, &pstate, arrayKeys,
+ itup, indnatts);
- /*
- * If the result of prechecking required keys was true, then in
- * assert-enabled builds we also recheck that the _bt_checkkeys()
- * result is the same.
- */
- Assert((!continuescanPrechecked && !haveFirstMatch) ||
- passes_quals == _bt_checkkeys(scan, itup, indnatts, dir,
- &continuescan, false, false));
if (passes_quals && tuple_alive)
{
/* tuple passes all scan key conditions */
- haveFirstMatch = true;
+ pstate.firstmatch = true;
if (!BTreeTupleIsPosting(itup))
{
/* Remember it */
@@ -1824,7 +1866,7 @@ _bt_readpage(IndexScanDesc scan, ScanDirection dir, OffsetNumber offnum,
}
}
}
- if (!continuescan)
+ if (!pstate.continuescan)
{
/* there can't be any more matches, so stop */
so->currPos.moreLeft = false;
@@ -1970,6 +2012,24 @@ _bt_steppage(IndexScanDesc scan, ScanDirection dir)
so->currPos.nextTupleOffset);
so->markPos.itemIndex = so->markItemIndex;
so->markItemIndex = -1;
+
+ /*
+ * When a mark used by a scan with array keys is restored, it must be
+ * forced to visit the next sibling page, even when we expect to skip
+ * over it now.
+ *
+ * This approach allows btrestrpos to just reset the scan's array
+ * keys, and allow the next call to _bt_readpage to advance the array
+ * keys to appropriate values.
+ */
+ Assert(!so->numArrayKeys || dir == so->primScanDir);
+ if (so->needPrimScan)
+ {
+ if (ScanDirectionIsForward(dir))
+ so->markPos.moreRight = true;
+ else
+ so->markPos.moreLeft = true;
+ }
}
if (ScanDirectionIsForward(dir))
@@ -2072,6 +2132,13 @@ _bt_readnextpage(IndexScanDesc scan, BlockNumber blkno, ScanDirection dir)
BTScanPosInvalidate(so->currPos);
return false;
}
+
+ if (so->numArrayKeys)
+ {
+ so->primScanDir = dir;
+ so->needPrimScan = false;
+ }
+
/* check for interrupts while we're not holding any buffer lock */
CHECK_FOR_INTERRUPTS();
/* step right one page */
@@ -2152,6 +2219,13 @@ _bt_readnextpage(IndexScanDesc scan, BlockNumber blkno, ScanDirection dir)
return false;
}
+ if (so->numArrayKeys)
+ {
+ so->primScanDir = dir;
+ so->scanBehind = false; /* Only needed in forward direction */
+ so->needPrimScan = false;
+ }
+
/* Step to next physical page */
so->currPos.buf = _bt_walk_left(rel, so->currPos.buf);
@@ -2530,8 +2604,20 @@ _bt_endpoint(IndexScanDesc scan, ScanDirection dir)
static inline void
_bt_initialize_more_data(BTScanOpaque so, ScanDirection dir)
{
- /* initialize moreLeft/moreRight appropriately for scan direction */
- if (ScanDirectionIsForward(dir))
+ /*
+ * Initialize moreLeft/moreRight for this primitive index scan.
+ *
+ * In general, scans that have array keys might still have matches on
+ * pages in the direction opposite dir, the scan's current scan direction.
+ * When we're called, the top-level scan often won't be at the start of
+ * its first primitive index scan.
+ */
+ if (so->numArrayKeys)
+ {
+ so->currPos.moreLeft = true;
+ so->currPos.moreRight = true;
+ }
+ else if (ScanDirectionIsForward(dir))
{
so->currPos.moreLeft = false;
so->currPos.moreRight = true;
diff --git a/src/backend/access/nbtree/nbtutils.c b/src/backend/access/nbtree/nbtutils.c
index d50317096..b401b3119 100644
--- a/src/backend/access/nbtree/nbtutils.c
+++ b/src/backend/access/nbtree/nbtutils.c
@@ -32,23 +32,65 @@
typedef struct BTSortArrayContext
{
- FmgrInfo flinfo;
+ FmgrInfo *sortproc;
Oid collation;
bool reverse;
} BTSortArrayContext;
+typedef struct BTScanKeyPreproc
+{
+ ScanKey skey;
+ int ikey;
+ int arrayidx;
+} BTScanKeyPreproc;
+
+static void _bt_setup_array_cmp(IndexScanDesc scan, ScanKey skey, Oid elemtype,
+ FmgrInfo *orderproc, FmgrInfo **sortprocp);
static Datum _bt_find_extreme_element(IndexScanDesc scan, ScanKey skey,
- StrategyNumber strat,
+ Oid elemtype, StrategyNumber strat,
Datum *elems, int nelems);
-static int _bt_sort_array_elements(IndexScanDesc scan, ScanKey skey,
- bool reverse,
- Datum *elems, int nelems);
+static int _bt_sort_array_elements(ScanKey skey, FmgrInfo *sortproc,
+ bool reverse, Datum *elems, int nelems);
+static int _bt_merge_arrays(ScanKey skey, FmgrInfo *mergeprocp, bool reverse,
+ Datum *elems_orig, int nelems_orig,
+ Datum *elems_next, int nelems_next);
+static bool _bt_compare_array_scankey_args(IndexScanDesc scan,
+ ScanKey arraysk, ScanKey skey,
+ FmgrInfo *orderproc, BTArrayKeyInfo *array,
+ bool *qual_ok);
+static ScanKey _bt_preprocess_array_keys(IndexScanDesc scan);
+static void _bt_preprocess_array_keys_final(IndexScanDesc scan, int *keyDataMap);
static int _bt_compare_array_elements(const void *a, const void *b, void *arg);
+static inline int32 _bt_compare_array_skey(FmgrInfo *orderproc,
+ Datum tupdatum, bool tupnull,
+ Datum arrdatum, ScanKey cur);
+static int _bt_binsrch_array_skey(FmgrInfo *orderproc,
+ bool cur_elem_start, ScanDirection dir,
+ Datum tupdatum, bool tupnull,
+ BTArrayKeyInfo *array, ScanKey cur,
+ int32 *set_elem_result);
+static bool _bt_advance_array_keys_increment(IndexScanDesc scan, ScanDirection dir);
+static void _bt_rewind_nonrequired_arrays(IndexScanDesc scan, ScanDirection dir);
+static bool _bt_tuple_before_array_skeys(IndexScanDesc scan, ScanDirection dir,
+ IndexTuple tuple, int tupnatts, TupleDesc tupdesc,
+ bool readpagetup, int sktrig, bool *scanBehind);
+static bool _bt_advance_array_keys(IndexScanDesc scan, BTReadPageState *pstate,
+ IndexTuple tuple, int tupnatts, TupleDesc tupdesc,
+ int sktrig);
+#ifdef USE_ASSERT_CHECKING
+static bool _bt_verify_arrays_bt_first(IndexScanDesc scan, ScanDirection dir);
+static bool _bt_verify_keys_with_arraykeys(IndexScanDesc scan);
+#endif
static bool _bt_compare_scankey_args(IndexScanDesc scan, ScanKey op,
ScanKey leftarg, ScanKey rightarg,
+ BTArrayKeyInfo *array, FmgrInfo *orderproc,
bool *result);
static bool _bt_fix_scankey_strategy(ScanKey skey, int16 *indoption);
static void _bt_mark_scankey_required(ScanKey skey);
+static bool _bt_check_compare(ScanDirection dir, BTScanOpaque so,
+ IndexTuple tuple, int tupnatts, TupleDesc tupdesc,
+ bool arrayKeys, bool prechecked, bool firstmatch,
+ bool *continuescan, int *ikey);
static bool _bt_check_rowcompare(ScanKey skey,
IndexTuple tuple, int tupnatts, TupleDesc tupdesc,
ScanDirection dir, bool *continuescan);
@@ -188,29 +230,54 @@ _bt_freestack(BTStack stack)
*
* If there are any SK_SEARCHARRAY scan keys, deconstruct the array(s) and
* set up BTArrayKeyInfo info for each one that is an equality-type key.
- * Prepare modified scan keys in so->arrayKeyData, which will hold the current
- * array elements during each primitive indexscan operation. For inequality
- * array keys, it's sufficient to find the extreme element value and replace
- * the whole array with that scalar value.
+ * Return modified scan keys as input for further, standard preprocessing.
*
- * Note: the reason we need so->arrayKeyData, rather than just scribbling
- * on scan->keyData, is that callers are permitted to call btrescan without
- * supplying a new set of scankey data.
+ * Currently we perform two kinds of preprocessing to deal with redundancies.
+ * For inequality array keys, it's sufficient to find the extreme element
+ * value and replace the whole array with that scalar value. This eliminates
+ * all but one array element as redundant. Similarly, we are capable of
+ * "merging together" multiple equality array keys (from two or more input
+ * scan keys) into a single output scan key containing only the intersecting
+ * array elements. This can eliminate many redundant array elements, as well
+ * as eliminating whole array scan keys as redundant. It can also allow us to
+ * detect contradictory quals.
+ *
+ * It is convenient for _bt_preprocess_keys caller to have to deal with no
+ * more than one equality strategy array scan key per index attribute. We'll
+ * always be able to set things up that way when complete opfamilies are used.
+ * Eliminated array scan keys can be recognized as those that have had their
+ * sk_strategy field set to InvalidStrategy here by us. Caller should avoid
+ * including these in the scan's so->keyData[] output array.
+ *
+ * We set the scan key references from the scan's BTArrayKeyInfo info array to
+ * offsets into the temp modified input array returned to caller. Scans that
+ * have array keys should call _bt_preprocess_array_keys_final when standard
+ * preprocessing steps are complete. This will convert the scan key offset
+ * references into references to the scan's so->keyData[] output scan keys.
+ *
+ * Note: the reason we need to return a temp scan key array, rather than just
+ * scribbling on scan->keyData, is that callers are permitted to call btrescan
+ * without supplying a new set of scankey data.
*/
-void
+static ScanKey
_bt_preprocess_array_keys(IndexScanDesc scan)
{
BTScanOpaque so = (BTScanOpaque) scan->opaque;
+ Relation rel = scan->indexRelation;
int numberOfKeys = scan->numberOfKeys;
- int16 *indoption = scan->indexRelation->rd_indoption;
+ int16 *indoption = rel->rd_indoption;
int numArrayKeys;
+ int prevArrayAtt = -1;
+ Oid prevElemtype = InvalidOid;
ScanKey cur;
- int i;
MemoryContext oldContext;
+ ScanKey arrayKeyData; /* modified copy of scan->keyData */
+
+ Assert(numberOfKeys && ScanDirectionIsNoMovement(so->primScanDir));
/* Quick check to see if there are any array keys */
numArrayKeys = 0;
- for (i = 0; i < numberOfKeys; i++)
+ for (int i = 0; i < numberOfKeys; i++)
{
cur = &scan->keyData[i];
if (cur->sk_flags & SK_SEARCHARRAY)
@@ -220,20 +287,15 @@ _bt_preprocess_array_keys(IndexScanDesc scan)
/* If any arrays are null as a whole, we can quit right now. */
if (cur->sk_flags & SK_ISNULL)
{
- so->numArrayKeys = -1;
- so->arrayKeyData = NULL;
- return;
+ so->qual_ok = false;
+ return NULL;
}
}
}
/* Quit if nothing to do. */
if (numArrayKeys == 0)
- {
- so->numArrayKeys = 0;
- so->arrayKeyData = NULL;
- return;
- }
+ return NULL;
/*
* Make a scan-lifespan context to hold array-associated data, or reset it
@@ -249,18 +311,23 @@ _bt_preprocess_array_keys(IndexScanDesc scan)
oldContext = MemoryContextSwitchTo(so->arrayContext);
/* Create modifiable copy of scan->keyData in the workspace context */
- so->arrayKeyData = (ScanKey) palloc(scan->numberOfKeys * sizeof(ScanKeyData));
- memcpy(so->arrayKeyData,
- scan->keyData,
- scan->numberOfKeys * sizeof(ScanKeyData));
+ arrayKeyData = (ScanKey) palloc(numberOfKeys * sizeof(ScanKeyData));
+ memcpy(arrayKeyData, scan->keyData, numberOfKeys * sizeof(ScanKeyData));
/* Allocate space for per-array data in the workspace context */
- so->arrayKeys = (BTArrayKeyInfo *) palloc0(numArrayKeys * sizeof(BTArrayKeyInfo));
+ so->arrayKeys = (BTArrayKeyInfo *) palloc(numArrayKeys * sizeof(BTArrayKeyInfo));
+
+ /* Allocate space for ORDER procs used to help _bt_checkkeys */
+ so->orderProcs = (FmgrInfo *) palloc(numberOfKeys * sizeof(FmgrInfo));
/* Now process each array key */
numArrayKeys = 0;
- for (i = 0; i < numberOfKeys; i++)
+ for (int i = 0; i < numberOfKeys; i++)
{
+ FmgrInfo sortproc;
+ FmgrInfo *sortprocp = &sortproc;
+ Oid elemtype;
+ bool reverse;
ArrayType *arrayval;
int16 elmlen;
bool elmbyval;
@@ -271,7 +338,7 @@ _bt_preprocess_array_keys(IndexScanDesc scan)
int num_nonnulls;
int j;
- cur = &so->arrayKeyData[i];
+ cur = &arrayKeyData[i];
if (!(cur->sk_flags & SK_SEARCHARRAY))
continue;
@@ -305,10 +372,21 @@ _bt_preprocess_array_keys(IndexScanDesc scan)
/* If there's no non-nulls, the scan qual is unsatisfiable */
if (num_nonnulls == 0)
{
- numArrayKeys = -1;
+ so->qual_ok = false;
break;
}
+ /*
+ * Determine the nominal datatype of the array elements. We have to
+ * support the convention that sk_subtype == InvalidOid means the
+ * opclass input type; this is a hack to simplify life for
+ * ScanKeyInit().
+ */
+ elemtype = cur->sk_subtype;
+ if (elemtype == InvalidOid)
+ elemtype = rel->rd_opcintype[cur->sk_attno - 1];
+ Assert(elemtype == ARR_ELEMTYPE(arrayval));
+
/*
* If the comparison operator is not equality, then the array qual
* degenerates to a simple comparison against the smallest or largest
@@ -319,7 +397,7 @@ _bt_preprocess_array_keys(IndexScanDesc scan)
case BTLessStrategyNumber:
case BTLessEqualStrategyNumber:
cur->sk_argument =
- _bt_find_extreme_element(scan, cur,
+ _bt_find_extreme_element(scan, cur, elemtype,
BTGreaterStrategyNumber,
elem_values, num_nonnulls);
continue;
@@ -329,7 +407,7 @@ _bt_preprocess_array_keys(IndexScanDesc scan)
case BTGreaterEqualStrategyNumber:
case BTGreaterStrategyNumber:
cur->sk_argument =
- _bt_find_extreme_element(scan, cur,
+ _bt_find_extreme_element(scan, cur, elemtype,
BTLessStrategyNumber,
elem_values, num_nonnulls);
continue;
@@ -339,27 +417,392 @@ _bt_preprocess_array_keys(IndexScanDesc scan)
break;
}
+ /*
+ * Attributes with equality-type scan keys will need a 3-way ORDER
+ * proc to perform binary searches for the next matching array
+ * element. Set that up now.
+ *
+ * Array scan keys with cross-type equality operators will require a
+ * separate same-type ORDER proc for sorting their array. Otherwise,
+ * sortproc just points to the same proc used during binary searches.
+ *
+ * Note: we'll also need a 3-way ORDER proc for any non-array equality
+ * strategy scan keys. That doesn't happen until preprocessing
+ * reaches _bt_preprocess_array_keys_final, to avoid looking up ORDER
+ * procs needlessly for scan keys that could be eliminated.
+ */
+ _bt_setup_array_cmp(scan, cur, elemtype,
+ &so->orderProcs[i], &sortprocp);
+
/*
* Sort the non-null elements and eliminate any duplicates. We must
* sort in the same ordering used by the index column, so that the
- * successive primitive indexscans produce data in index order.
+ * arrays can be advanced in lockstep with the scan's progress through
+ * the index's key space.
*/
- num_elems = _bt_sort_array_elements(scan, cur,
- (indoption[cur->sk_attno - 1] & INDOPTION_DESC) != 0,
+ reverse = (indoption[cur->sk_attno - 1] & INDOPTION_DESC) != 0;
+ num_elems = _bt_sort_array_elements(cur, sortprocp, reverse,
elem_values, num_nonnulls);
+ /*
+ * If this scan key is semantically equivalent to a previous equality
+ * operator array scan key, merge the two arrays together to eliminate
+ * redundant non-intersecting elements (and whole scan keys).
+ */
+ if (prevArrayAtt == cur->sk_attno)
+ {
+ FmgrInfo *mergeprocp = NULL;
+ FmgrInfo crosstype_orderproc;
+
+ if (elemtype == prevElemtype)
+ {
+ /*
+ * This array is the same type as an earlier redundant array
+ * against the same index attribute. We can merge the two
+ * arrays together using our existing sort ORDER proc (which
+ * is often also the ORDER proc stored in so->orderProcs[]).
+ *
+ * Note: we do this whenever the array element type matches,
+ * which it may even when a cross-type equality operator was
+ * used (we do need a separate sortproc in cases involving
+ * cross-type operators, though).
+ */
+ mergeprocp = sortprocp;
+ }
+ else
+ {
+ /*
+ * Cross-type merging requires another ORDER proc lookup.
+ *
+ * We need to tolerate it when there is a lack of available
+ * cross-type support.
+ */
+ RegProcedure cmp_proc;
+
+ cmp_proc = get_opfamily_proc(rel->rd_opfamily[cur->sk_attno - 1],
+ prevElemtype, elemtype, BTORDER_PROC);
+ if (RegProcedureIsValid(cmp_proc))
+ {
+ fmgr_info_cxt(cmp_proc, &crosstype_orderproc, so->arrayContext);
+ mergeprocp = &crosstype_orderproc;
+ }
+ }
+
+ if (mergeprocp)
+ {
+ BTArrayKeyInfo *prev = &so->arrayKeys[numArrayKeys - 1];
+
+ Assert(arrayKeyData[prev->scan_key].sk_attno == cur->sk_attno);
+ Assert(arrayKeyData[prev->scan_key].sk_collation ==
+ cur->sk_collation);
+
+ num_elems = _bt_merge_arrays(cur, mergeprocp, reverse,
+ prev->elem_values, prev->num_elems,
+ elem_values, num_elems);
+
+ pfree(elem_values);
+
+ /*
+ * If there are no intersecting elements left from merging
+ * this array into the previous array on the same attribute,
+ * the scan qual is unsatisfiable
+ */
+ if (num_elems == 0)
+ {
+ so->qual_ok = false;
+ break;
+ }
+
+ /*
+ * Lower the number of elements from the previous array. This
+ * scan key/array is redundant. Dealing with that is
+ * finalized within _bt_preprocess_keys.
+ */
+ prev->num_elems = num_elems;
+ cur->sk_strategy = InvalidStrategy; /* for _bt_preprocess_keys */
+ continue;
+ }
+
+ /*
+ * Unable to merge this array with previous array due to a lack of
+ * suitable cross-type opfamily support. Will need to keep both
+ * arrays.
+ */
+ }
+
/*
* And set up the BTArrayKeyInfo data.
+ *
+ * Note: _bt_preprocess_array_keys_final will fix-up each array's
+ * scan_key field later on, after so->keyData[] is finalized.
*/
so->arrayKeys[numArrayKeys].scan_key = i;
so->arrayKeys[numArrayKeys].num_elems = num_elems;
so->arrayKeys[numArrayKeys].elem_values = elem_values;
numArrayKeys++;
+ prevArrayAtt = cur->sk_attno;
+ prevElemtype = elemtype;
}
so->numArrayKeys = numArrayKeys;
MemoryContextSwitchTo(oldContext);
+
+ return arrayKeyData;
+}
+
+/*
+ *
+ * _bt_preprocess_array_keys_final() -- fix up array scan key references
+ *
+ * When _bt_preprocess_array_keys performed initial array preprocessing, it
+ * set each array's array->scan_key to the array's arrayKeys[] entry offset
+ * (that also work as references into the original scan->keyData[] array).
+ * This function handles translation of the scan key references from the
+ * BTArrayKeyInfo info array, from input scan key references (to the keys in
+ * scan->keyData[]), into output references (to the keys in so->keyData[]).
+ * Caller's keyDataMap[] array tells us how to perform this remapping.
+ *
+ * Also reorders so->orderProcs[] entries in-place. This is required for all
+ * remaining equality-type scan keys (not just for those with an array).
+ *
+ * We'll also convert single-element array scan keys into equivalent non-array
+ * equality scan keys, which decrements so->numArrayKeys. It's possible that
+ * this will leave this new btrescan without any arrays at all. This isn't
+ * necessary for correctness; it's just an optimization. Non-array equality
+ * scan keys are slightly faster than equivalent array scan keys at runtime.
+ *
+ * Note: _bt_compare_array_scankey_args always eliminates non-array equality
+ * scan keys that are redundant with some other array equality scan key (just
+ * like it will with any other type of non-array scan key). Our "convert any
+ * single element array to a non-array scan key" optimization is therefore the
+ * only way that preprocessing can leave behind a non-array equality scan key
+ * (for index attributes with a partly-redundant array equality scan key).
+ */
+static void
+_bt_preprocess_array_keys_final(IndexScanDesc scan, int *keyDataMap)
+{
+ BTScanOpaque so = (BTScanOpaque) scan->opaque;
+ Relation rel = scan->indexRelation;
+ int arrayidx = 0;
+ int last_equal_output_ikey PG_USED_FOR_ASSERTS_ONLY = -1;
+
+ Assert(so->qual_ok);
+ Assert(so->numArrayKeys);
+
+ for (int output_ikey = 0; output_ikey < so->numberOfKeys; output_ikey++)
+ {
+ ScanKey outkey = so->keyData + output_ikey;
+ int input_ikey;
+ bool found PG_USED_FOR_ASSERTS_ONLY = false;
+
+ Assert(outkey->sk_strategy != InvalidStrategy);
+
+ if (outkey->sk_strategy != BTEqualStrategyNumber)
+ continue;
+
+ input_ikey = keyDataMap[output_ikey];
+
+ /*
+ * Make it safe for _bt_advance_array_keys to subscript
+ * so->orderProcs[] using simple so->keyData[]-wise offsets.
+ *
+ * It's safe to reorder so->orderProcs[] in-place. Our caller is
+ * required to output equality scan keys in their original input order
+ * (which is also the order that arrays appear in).
+ */
+ Assert(last_equal_output_ikey < output_ikey);
+ Assert(last_equal_output_ikey < input_ikey);
+ last_equal_output_ikey = output_ikey;
+
+ /*
+ * We don't actually have to reorder so->orderProcs[] for non-array
+ * equality scan keys, though -- they don't have a valid entry yet.
+ *
+ * We're lazy about looking up ORDER procs for non-array scan keys,
+ * since not all input scan keys become output scan keys.
+ */
+ if (!(outkey->sk_flags & SK_SEARCHARRAY))
+ {
+ Oid elemtype;
+
+ /* No point in ORDER proc lookup for IS NULL scan keys */
+ if (outkey->sk_flags & SK_SEARCHNULL)
+ continue;
+
+ /*
+ * Lookup this equality key's ORDER proc. _bt_advance_array_keys
+ * will treat it as a degenerate single value array, so we can't
+ * get away with just using the equality operator on its own.
+ */
+ elemtype = outkey->sk_subtype;
+ if (elemtype == InvalidOid)
+ elemtype = rel->rd_opcintype[outkey->sk_attno - 1];
+
+ _bt_setup_array_cmp(scan, outkey, elemtype,
+ &so->orderProcs[output_ikey], NULL);
+ continue;
+ }
+
+ /* Reorder so->orderProcs[] in-place for arrays */
+ so->orderProcs[output_ikey] = so->orderProcs[input_ikey];
+
+ /* Also fix-up array->scan_key references for arrays */
+ for (; arrayidx < so->numArrayKeys; arrayidx++)
+ {
+ BTArrayKeyInfo *array = &so->arrayKeys[arrayidx];
+
+ Assert(array->num_elems > 0);
+
+ if (array->scan_key == input_ikey)
+ {
+ /* found it */
+ array->scan_key = output_ikey;
+ found = true;
+
+ /*
+ * Transform array scan keys that have exactly 1 element
+ * remaining (following all prior preprocessing) into
+ * equivalent non-array scan keys.
+ */
+ if (array->num_elems == 1)
+ {
+ outkey->sk_flags &= ~SK_SEARCHARRAY;
+ outkey->sk_argument = array->elem_values[0];
+ so->numArrayKeys--;
+
+ /* If we're out of array keys, we can quit right away */
+ if (so->numArrayKeys == 0)
+ return;
+
+ /* Shift other arrays forward */
+ memmove(array, array + 1,
+ sizeof(BTArrayKeyInfo) *
+ (so->numArrayKeys - arrayidx));
+
+ /*
+ * Don't increment arrayidx (there was an entry that was
+ * just shifted forward to the offset at arrayidx, which
+ * will still need to be matched)
+ */
+ }
+ else
+ {
+ /* Match found, so done with this array */
+ arrayidx++;
+ }
+
+ break;
+ }
+ }
+
+ Assert(found);
+ }
+}
+
+/*
+ * _bt_setup_array_cmp() -- Set up array comparison functions
+ *
+ * Sets ORDER proc in caller's orderproc argument, which is used during binary
+ * searches of arrays during the index scan. Also sets a same-type ORDER proc
+ * in caller's *sortprocp argument.
+ *
+ * Caller should pass an orderproc pointing to space that'll store the ORDER
+ * proc for the scan, and a *sortprocp pointing to its own separate space.
+ *
+ * In the common case where we don't need to deal with cross-type operators,
+ * only one ORDER proc is actually required by caller. We'll set *sortprocp
+ * to point to the same memory that caller's orderproc continues to point to.
+ * Otherwise, *sortprocp will continue to point to separate memory, which
+ * we'll initialize separately (with an "(elemtype, elemtype)" ORDER proc that
+ * can be used to sort arrays).
+ *
+ * Preprocessing calls here with all equality strategy scan keys, including
+ * those not asociated with any array. See _bt_advance_array_keys for an
+ * explanation of why we need to treat these as degenerate single-value arrays
+ * when the scan advances its arrays.
+ */
+static void
+_bt_setup_array_cmp(IndexScanDesc scan, ScanKey skey, Oid elemtype,
+ FmgrInfo *orderproc, FmgrInfo **sortprocp)
+{
+ BTScanOpaque so = (BTScanOpaque) scan->opaque;
+ Relation rel = scan->indexRelation;
+ RegProcedure cmp_proc;
+ Oid opclasstype = rel->rd_opcintype[skey->sk_attno - 1];
+
+ Assert(skey->sk_strategy == BTEqualStrategyNumber);
+ Assert(OidIsValid(elemtype));
+
+ /*
+ * If scankey operator is not a cross-type comparison, we can use the
+ * cached comparison function; otherwise gotta look it up in the catalogs
+ */
+ if (elemtype == opclasstype)
+ *orderproc = *index_getprocinfo(rel, skey->sk_attno, BTORDER_PROC);
+ else
+ {
+ /*
+ * Have to look up the appropriate comparison function in the opfamily
+ * the hard way.
+ *
+ * Must use the opclass type as the left hand arg type, and the array
+ * element as the right hand arg type (since binary searches search
+ * for the array value that best matches the next index tuple' value).
+ *
+ * Note: it's possible that this would fail, if the opfamily lacks the
+ * required cross-type ORDER proc. But this is no different to the
+ * case where _bt_first fails to find an ORDER proc for its insertion
+ * scan key.
+ */
+ cmp_proc = get_opfamily_proc(rel->rd_opfamily[skey->sk_attno - 1],
+ opclasstype, elemtype, BTORDER_PROC);
+ if (!RegProcedureIsValid(cmp_proc))
+ elog(ERROR, "missing support function %d(%u,%u) for attribute %d of index \"%s\"",
+ BTORDER_PROC, opclasstype, elemtype,
+ skey->sk_attno, RelationGetRelationName(rel));
+
+ /* Set ORDER proc for caller */
+ fmgr_info_cxt(cmp_proc, orderproc, so->arrayContext);
+ }
+
+ if (!sortprocp)
+ {
+ /*
+ * Nothing more to do for caller that doesn't actually have an array
+ * to sort
+ */
+ return;
+ }
+
+ if (elemtype == opclasstype)
+ {
+ /*
+ * A second opfamily support proc lookup can be avoided in the common
+ * case where the ORDER proc used for the scan's binary searches uses
+ * the opclass/on-disk datatype for both its left and right arguments.
+ */
+ *sortprocp = orderproc;
+ return;
+ }
+
+ /*
+ * Look up the appropriate same-type comparison function in the opfamily.
+ *
+ * Note: it's possible that this would fail, if the opfamily is
+ * incomplete, but it seems quite unlikely that an opfamily would omit
+ * non-cross-type support functions for any datatype that it supports at
+ * all.
+ */
+ cmp_proc = get_opfamily_proc(rel->rd_opfamily[skey->sk_attno - 1],
+ elemtype, elemtype, BTORDER_PROC);
+ if (!RegProcedureIsValid(cmp_proc))
+ elog(ERROR, "missing support function %d(%u,%u) for attribute %d of index \"%s\"",
+ BTORDER_PROC, elemtype, elemtype,
+ skey->sk_attno, RelationGetRelationName(rel));
+
+ /* Set same-type ORDER proc for caller */
+ fmgr_info_cxt(cmp_proc, *sortprocp, so->arrayContext);
}
/*
@@ -370,27 +813,17 @@ _bt_preprocess_array_keys(IndexScanDesc scan)
* least element, or BTGreaterStrategyNumber to get the greatest.
*/
static Datum
-_bt_find_extreme_element(IndexScanDesc scan, ScanKey skey,
+_bt_find_extreme_element(IndexScanDesc scan, ScanKey skey, Oid elemtype,
StrategyNumber strat,
Datum *elems, int nelems)
{
Relation rel = scan->indexRelation;
- Oid elemtype,
- cmp_op;
+ Oid cmp_op;
RegProcedure cmp_proc;
FmgrInfo flinfo;
Datum result;
int i;
- /*
- * Determine the nominal datatype of the array elements. We have to
- * support the convention that sk_subtype == InvalidOid means the opclass
- * input type; this is a hack to simplify life for ScanKeyInit().
- */
- elemtype = skey->sk_subtype;
- if (elemtype == InvalidOid)
- elemtype = rel->rd_opcintype[skey->sk_attno - 1];
-
/*
* Look up the appropriate comparison operator in the opfamily.
*
@@ -399,6 +832,8 @@ _bt_find_extreme_element(IndexScanDesc scan, ScanKey skey,
* non-cross-type comparison operators for any datatype that it supports
* at all.
*/
+ Assert(skey->sk_strategy != BTEqualStrategyNumber);
+ Assert(OidIsValid(elemtype));
cmp_op = get_opfamily_member(rel->rd_opfamily[skey->sk_attno - 1],
elemtype,
elemtype,
@@ -433,50 +868,21 @@ _bt_find_extreme_element(IndexScanDesc scan, ScanKey skey,
* The array elements are sorted in-place, and the new number of elements
* after duplicate removal is returned.
*
- * scan and skey identify the index column, whose opfamily determines the
- * comparison semantics. If reverse is true, we sort in descending order.
+ * skey identifies the index column whose opfamily determines the comparison
+ * semantics, and sortproc is a corresponding ORDER proc. If reverse is true,
+ * we sort in descending order.
*/
static int
-_bt_sort_array_elements(IndexScanDesc scan, ScanKey skey,
- bool reverse,
+_bt_sort_array_elements(ScanKey skey, FmgrInfo *sortproc, bool reverse,
Datum *elems, int nelems)
{
- Relation rel = scan->indexRelation;
- Oid elemtype;
- RegProcedure cmp_proc;
BTSortArrayContext cxt;
if (nelems <= 1)
return nelems; /* no work to do */
- /*
- * Determine the nominal datatype of the array elements. We have to
- * support the convention that sk_subtype == InvalidOid means the opclass
- * input type; this is a hack to simplify life for ScanKeyInit().
- */
- elemtype = skey->sk_subtype;
- if (elemtype == InvalidOid)
- elemtype = rel->rd_opcintype[skey->sk_attno - 1];
-
- /*
- * Look up the appropriate comparison function in the opfamily.
- *
- * Note: it's possible that this would fail, if the opfamily is
- * incomplete, but it seems quite unlikely that an opfamily would omit
- * non-cross-type support functions for any datatype that it supports at
- * all.
- */
- cmp_proc = get_opfamily_proc(rel->rd_opfamily[skey->sk_attno - 1],
- elemtype,
- elemtype,
- BTORDER_PROC);
- if (!RegProcedureIsValid(cmp_proc))
- elog(ERROR, "missing support function %d(%u,%u) in opfamily %u",
- BTORDER_PROC, elemtype, elemtype,
- rel->rd_opfamily[skey->sk_attno - 1]);
-
/* Sort the array elements */
- fmgr_info(cmp_proc, &cxt.flinfo);
+ cxt.sortproc = sortproc;
cxt.collation = skey->sk_collation;
cxt.reverse = reverse;
qsort_arg(elems, nelems, sizeof(Datum),
@@ -487,6 +893,196 @@ _bt_sort_array_elements(IndexScanDesc scan, ScanKey skey,
_bt_compare_array_elements, &cxt);
}
+/*
+ * _bt_merge_arrays() -- merge together duplicate array keys
+ *
+ * Both scan keys should have arrays whose elements that have already been
+ * sorted and deduplicated.
+ *
+ * mergeprocp is an ORDER proc whose left hand argument accepts datums of
+ * whatever type the original array's elements are, and whose right hand type
+ * is whatever type the next array's elements are.
+ *
+ * Merging reorganizes caller's original array (the left hand arg) in-place,
+ * while avoiding copying over elements taken from the next array. This is
+ * strictly necessary when the arrays don't share the same element types.
+ */
+static int
+_bt_merge_arrays(ScanKey skey, FmgrInfo *mergeprocp, bool reverse,
+ Datum *elems_orig, int nelems_orig,
+ Datum *elems_next, int nelems_next)
+{
+ BTSortArrayContext cxt;
+ int merged_nelems = 0;
+
+ cxt.sortproc = mergeprocp;
+ cxt.collation = skey->sk_collation;
+ cxt.reverse = reverse;
+ for (int i = 0, j = 0; i < nelems_orig && j < nelems_next;)
+ {
+ Datum *oelem = elems_orig + i;
+ Datum *nelem = elems_next + j;
+ int res;
+
+ res = _bt_compare_array_elements(oelem, nelem, &cxt);
+
+ if (res == 0)
+ {
+ elems_orig[merged_nelems] = *oelem;
+ merged_nelems++;
+ i++;
+ j++;
+ }
+ else if (res < 0)
+ i++;
+ else /* res > 0 */
+ j++;
+ }
+
+ return merged_nelems;
+}
+
+/*
+ * Compare a scalar scankey to an array, excluding redundant array elements.
+ *
+ * Array elements can be eliminated as redundant when excluded by some other
+ * operator against the same attribute. For example, with a qual "WHERE a IN
+ * (1, 2, 3) AND a < 2", all array elements except the value "1" are redundant
+ * (the < scan key can be treated as redundant to the scan, too).
+ *
+ * If the opfamily doesn't supply a complete set of cross-type ORDER procs we
+ * may not be able to determine which elements are redundant. If we have the
+ * required ORDER proc then we'll return true (and we'll set *qual_ok). We
+ * return false if the comparison could not be made (caller should ignore the
+ * value we set *qual_ok to when this happens).
+ *
+ * Note: whenever we return true, we guarantee that skey (which is caller's
+ * non-array scan key) has been eliminated, at a minimum. We may have also
+ * managed to eliminate some subset of the array elements (array is actually
+ * modified for caller when this happens), or all of the array's elements,
+ * which means that the scan has an unsatisfiable qual (in which case we'll
+ * also have set *qual_ok=false for caller).
+ */
+static bool
+_bt_compare_array_scankey_args(IndexScanDesc scan, ScanKey arraysk, ScanKey skey,
+ FmgrInfo *orderproc, BTArrayKeyInfo *array,
+ bool *qual_ok)
+{
+ Relation rel = scan->indexRelation;
+ Oid opclasstype = rel->rd_opcintype[arraysk->sk_attno - 1];
+ int matchelem;
+ int cmpresult = 0;
+ int new_nelems = 0;
+ Oid skey_type;
+ FmgrInfo nonmatch_orderproc;
+ FmgrInfo *usable_orderprocp;
+ int cmpval;
+
+ Assert(arraysk->sk_attno == skey->sk_attno);
+
+ Assert(!(arraysk->sk_flags & (SK_ISNULL | SK_ROW_HEADER | SK_ROW_MEMBER)));
+ Assert((arraysk->sk_flags & SK_SEARCHARRAY) &&
+ arraysk->sk_strategy == BTEqualStrategyNumber);
+
+ Assert(!(skey->sk_flags & (SK_ISNULL | SK_ROW_HEADER | SK_ROW_MEMBER)));
+ Assert(!(skey->sk_flags & SK_SEARCHARRAY) ||
+ skey->sk_strategy != BTEqualStrategyNumber);
+
+ Assert(array->num_elems > 0);
+
+ /*
+ * _bt_binsrch_array_skey is designed to search scan key arrays using
+ * datums of whatever type the relevant rel opclass uses (on-disk type).
+ *
+ * We can reuse the array's ORDER proc whenever the non-array scan key's
+ * type is a match for the corresponding index attribute's opclass type.
+ * Otherwise, we have to do another ORDER proc lookup.
+ *
+ * Note: we do not yet have any usable ORDER procs available for non-array
+ * equality scan keys. _bt_preprocess_array_keys_final will look those up
+ * later on. (We put this work off precisely because it'll waste cycles
+ * when we can determine that the non-array equality key is redundant.)
+ */
+ skey_type = skey->sk_subtype;
+ if (skey_type == InvalidOid)
+ skey_type = rel->rd_opcintype[skey->sk_attno - 1];
+ if (skey_type != opclasstype)
+ {
+ RegProcedure cmp_proc;
+ Oid arraysk_elemtype;
+
+ /*
+ * Need an ORDER proc lookup to detect redundancy/contradictoriness
+ * with this pair of scankeys
+ */
+ usable_orderprocp = &nonmatch_orderproc;
+ arraysk_elemtype = arraysk->sk_subtype;
+ if (arraysk_elemtype == InvalidOid)
+ arraysk_elemtype = rel->rd_opcintype[arraysk->sk_attno - 1];
+ cmp_proc = get_opfamily_proc(rel->rd_opfamily[arraysk->sk_attno - 1],
+ skey_type, arraysk_elemtype,
+ BTORDER_PROC);
+ if (RegProcedureIsValid(cmp_proc))
+ fmgr_info(cmp_proc, usable_orderprocp);
+ else
+ {
+ /* Can't make the comparison */
+ *qual_ok = false; /* suppress compiler warnings */
+ return false;
+ }
+ }
+ else
+ usable_orderprocp = orderproc;
+
+ matchelem = _bt_binsrch_array_skey(usable_orderprocp, false,
+ ForwardScanDirection,
+ skey->sk_argument, false, array,
+ arraysk, &cmpresult);
+
+ if (skey->sk_strategy == BTLessStrategyNumber ||
+ skey->sk_strategy == BTLessEqualStrategyNumber)
+ {
+ /* Resize array downwards as needed */
+ cmpval = skey->sk_strategy == BTLessEqualStrategyNumber ? 0 : 1;
+ if (cmpresult >= cmpval)
+ matchelem++;
+ new_nelems = matchelem;
+ }
+ else if (skey->sk_strategy == BTEqualStrategyNumber)
+ {
+ if (cmpresult != 0)
+ new_nelems = 0;
+ else
+ {
+ /* Move matching element to front of array, then resize to 1 */
+ array->elem_values[0] = array->elem_values[matchelem];
+ new_nelems = 1;
+ }
+ }
+ else if (skey->sk_strategy == BTGreaterEqualStrategyNumber ||
+ skey->sk_strategy == BTGreaterStrategyNumber)
+ {
+ /* Shift all matching elements to the start of the array, resize */
+ cmpval = skey->sk_strategy == BTGreaterStrategyNumber ? 0 : 1;
+ if (cmpresult >= cmpval)
+ matchelem++;
+ new_nelems = array->num_elems - matchelem;
+ memmove(array->elem_values,
+ array->elem_values + matchelem,
+ sizeof(Datum) * new_nelems);
+ }
+ else
+ Assert(false);
+
+ Assert(new_nelems >= 0);
+ Assert(new_nelems <= array->num_elems);
+
+ array->num_elems = new_nelems;
+
+ *qual_ok = new_nelems > 0;
+ return true;
+}
+
/*
* qsort_arg comparator for sorting array elements
*/
@@ -498,7 +1094,7 @@ _bt_compare_array_elements(const void *a, const void *b, void *arg)
BTSortArrayContext *cxt = (BTSortArrayContext *) arg;
int32 compare;
- compare = DatumGetInt32(FunctionCall2Coll(&cxt->flinfo,
+ compare = DatumGetInt32(FunctionCall2Coll(cxt->sortproc,
cxt->collation,
da, db));
if (cxt->reverse)
@@ -506,11 +1102,160 @@ _bt_compare_array_elements(const void *a, const void *b, void *arg)
return compare;
}
+/*
+ * _bt_compare_array_skey() -- apply array comparison function
+ *
+ * Compares caller's tuple attribute value to a scan key/array element.
+ * Helper function used during binary searches of SK_SEARCHARRAY arrays.
+ *
+ * This routine returns:
+ * <0 if tupdatum < arrdatum;
+ * 0 if tupdatum == arrdatum;
+ * >0 if tupdatum > arrdatum.
+ *
+ * This is essentially the same interface as _bt_compare: both functions
+ * compare the value that they're searching for to a binary search pivot.
+ * However, unlike _bt_compare, this function's "tuple argument" comes first,
+ * while its "array/scankey argument" comes second.
+*/
+static inline int32
+_bt_compare_array_skey(FmgrInfo *orderproc,
+ Datum tupdatum, bool tupnull,
+ Datum arrdatum, ScanKey cur)
+{
+ int32 result = 0;
+
+ Assert(cur->sk_strategy == BTEqualStrategyNumber);
+
+ if (tupnull) /* NULL tupdatum */
+ {
+ if (cur->sk_flags & SK_ISNULL)
+ result = 0; /* NULL "=" NULL */
+ else if (cur->sk_flags & SK_BT_NULLS_FIRST)
+ result = -1; /* NULL "<" NOT_NULL */
+ else
+ result = 1; /* NULL ">" NOT_NULL */
+ }
+ else if (cur->sk_flags & SK_ISNULL) /* NOT_NULL tupdatum, NULL arrdatum */
+ {
+ if (cur->sk_flags & SK_BT_NULLS_FIRST)
+ result = 1; /* NOT_NULL ">" NULL */
+ else
+ result = -1; /* NOT_NULL "<" NULL */
+ }
+ else
+ {
+ /*
+ * Like _bt_compare, we need to be careful of cross-type comparisons,
+ * so the left value has to be the value that came from an index tuple
+ */
+ result = DatumGetInt32(FunctionCall2Coll(orderproc, cur->sk_collation,
+ tupdatum, arrdatum));
+
+ /*
+ * We flip the sign by following the obvious rule: flip whenever the
+ * column is a DESC column.
+ *
+ * _bt_compare does it the wrong way around (flip when *ASC*) in order
+ * to compensate for passing its orderproc arguments backwards. We
+ * don't need to play these games because we find it natural to pass
+ * tupdatum as the left value (and arrdatum as the right value).
+ */
+ if (cur->sk_flags & SK_BT_DESC)
+ INVERT_COMPARE_RESULT(result);
+ }
+
+ return result;
+}
+
+/*
+ * _bt_binsrch_array_skey() -- Binary search for next matching array key
+ *
+ * Returns an index to the first array element >= caller's tupdatum argument.
+ * This convention is more natural for forwards scan callers, but that can't
+ * really matter to backwards scan callers. Both callers require handling for
+ * the case where the match we return is < tupdatum, and symmetric handling
+ * for the case where our best match is > tupdatum.
+ *
+ * Also sets *set_elem_result to whatever _bt_compare_array_skey returned when
+ * we compared the returned array element to caller's tupdatum argument. This
+ * helps our caller to determine how advancing its array (to the element we'll
+ * return an offset to) might need to carry to higher order arrays.
+ *
+ * cur_elem_start indicates if the binary search should begin at the array's
+ * current element (or have the current element as an upper bound for backward
+ * scans). It's safe for searches against required scan key arrays to reuse
+ * earlier search bounds like this because such arrays always advance in
+ * lockstep with the index scan's progress through the index's key space.
+ */
+static int
+_bt_binsrch_array_skey(FmgrInfo *orderproc,
+ bool cur_elem_start, ScanDirection dir,
+ Datum tupdatum, bool tupnull,
+ BTArrayKeyInfo *array, ScanKey cur,
+ int32 *set_elem_result)
+{
+ int low_elem = 0,
+ mid_elem = -1,
+ high_elem = array->num_elems - 1,
+ result = 0;
+
+ Assert(cur->sk_flags & SK_SEARCHARRAY);
+ Assert(cur->sk_strategy == BTEqualStrategyNumber);
+
+ if (cur_elem_start)
+ {
+ if (ScanDirectionIsForward(dir))
+ low_elem = array->cur_elem;
+ else
+ high_elem = array->cur_elem;
+ }
+
+ while (high_elem > low_elem)
+ {
+ Datum arrdatum;
+
+ mid_elem = low_elem + ((high_elem - low_elem) / 2);
+ arrdatum = array->elem_values[mid_elem];
+
+ result = _bt_compare_array_skey(orderproc, tupdatum, tupnull,
+ arrdatum, cur);
+
+ if (result == 0)
+ {
+ /*
+ * It's safe to quit as soon as we see an equal array element.
+ * This often saves an extra comparison or two...
+ */
+ low_elem = mid_elem;
+ break;
+ }
+
+ if (result > 0)
+ low_elem = mid_elem + 1;
+ else
+ high_elem = mid_elem;
+ }
+
+ /*
+ * ...but our caller also cares about how its searched-for tuple datum
+ * compares to the low_elem datum. Must always set *set_elem_result with
+ * the result of that comparison specifically.
+ */
+ if (low_elem != mid_elem)
+ result = _bt_compare_array_skey(orderproc, tupdatum, tupnull,
+ array->elem_values[low_elem], cur);
+
+ *set_elem_result = result;
+
+ return low_elem;
+}
+
/*
* _bt_start_array_keys() -- Initialize array keys at start of a scan
*
* Set up the cur_elem counters and fill in the first sk_argument value for
- * each array scankey. We can't do this until we know the scan direction.
+ * each array scankey.
*/
void
_bt_start_array_keys(IndexScanDesc scan, ScanDirection dir)
@@ -518,159 +1263,1158 @@ _bt_start_array_keys(IndexScanDesc scan, ScanDirection dir)
BTScanOpaque so = (BTScanOpaque) scan->opaque;
int i;
+ Assert(so->numArrayKeys);
+ Assert(so->qual_ok);
+ Assert(!ScanDirectionIsNoMovement(dir));
+
for (i = 0; i < so->numArrayKeys; i++)
{
BTArrayKeyInfo *curArrayKey = &so->arrayKeys[i];
- ScanKey skey = &so->arrayKeyData[curArrayKey->scan_key];
+ ScanKey skey = &so->keyData[curArrayKey->scan_key];
Assert(curArrayKey->num_elems > 0);
+ Assert(skey->sk_flags & SK_SEARCHARRAY);
+
if (ScanDirectionIsBackward(dir))
curArrayKey->cur_elem = curArrayKey->num_elems - 1;
else
curArrayKey->cur_elem = 0;
skey->sk_argument = curArrayKey->elem_values[curArrayKey->cur_elem];
}
-
- so->arraysStarted = true;
}
/*
- * _bt_advance_array_keys() -- Advance to next set of array elements
+ * _bt_advance_array_keys_increment() -- Advance to next set of array elements
+ *
+ * Advances the array keys by a single increment in the current scan
+ * direction. When there are multiple array keys this can roll over from the
+ * lowest order array to higher order arrays.
*
* Returns true if there is another set of values to consider, false if not.
* On true result, the scankeys are initialized with the next set of values.
+ * On false result, the scankeys stay the same, and the array keys are not
+ * advanced (every array remains at its final element for scan direction).
*/
-bool
-_bt_advance_array_keys(IndexScanDesc scan, ScanDirection dir)
+static bool
+_bt_advance_array_keys_increment(IndexScanDesc scan, ScanDirection dir)
{
BTScanOpaque so = (BTScanOpaque) scan->opaque;
- bool found = false;
- int i;
/*
* We must advance the last array key most quickly, since it will
* correspond to the lowest-order index column among the available
- * qualifications. This is necessary to ensure correct ordering of output
- * when there are multiple array keys.
+ * qualifications. Rolling over like this is necessary to ensure correct
+ * ordering of output when there are multiple array keys.
+ *
+ * We rely on the assumption that array preprocessing always leaves behind
+ * exactly the arrays needed within the scan's preprocessed so->keyData[]
+ * scan keys (needed for the current btrescan). It is still possible that
+ * there will be more than one array per index attribute, though only when
+ * the attribute's opfamily is incomplete. This is not an exception to
+ * the general rule about so->keyData[] and arrays: there will still be a
+ * 1:1 mapping between each array and each array equality scan key.
*/
- for (i = so->numArrayKeys - 1; i >= 0; i--)
+ for (int i = so->numArrayKeys - 1; i >= 0; i--)
{
BTArrayKeyInfo *curArrayKey = &so->arrayKeys[i];
- ScanKey skey = &so->arrayKeyData[curArrayKey->scan_key];
+ ScanKey skey = &so->keyData[curArrayKey->scan_key];
int cur_elem = curArrayKey->cur_elem;
int num_elems = curArrayKey->num_elems;
+ bool rolled = false;
- if (ScanDirectionIsBackward(dir))
+ if (ScanDirectionIsForward(dir) && ++cur_elem >= num_elems)
{
- if (--cur_elem < 0)
- {
- cur_elem = num_elems - 1;
- found = false; /* need to advance next array key */
- }
- else
- found = true;
+ cur_elem = 0;
+ rolled = true;
}
- else
+ else if (ScanDirectionIsBackward(dir) && --cur_elem < 0)
{
- if (++cur_elem >= num_elems)
- {
- cur_elem = 0;
- found = false; /* need to advance next array key */
- }
- else
- found = true;
+ cur_elem = num_elems - 1;
+ rolled = true;
}
curArrayKey->cur_elem = cur_elem;
skey->sk_argument = curArrayKey->elem_values[cur_elem];
- if (found)
- break;
- }
+ if (!rolled)
+ return true;
- /* advance parallel scan */
- if (scan->parallel_scan != NULL)
- _bt_parallel_advance_array_keys(scan);
+ /* Need to advance next array key, if any */
+ }
/*
- * When no new array keys were found, the scan is "past the end" of the
- * array keys. _bt_start_array_keys can still "restart" the array keys if
- * a rescan is required.
+ * The array keys are now exhausted.
+ *
+ * There isn't actually a distinct state that represents array exhaustion,
+ * since index scans don't always end when btgettuple returns "false". The
+ * scan direction might be reversed, or the scan might yet have its last
+ * saved position restored.
+ *
+ * Restore the array keys to the state they were in immediately before we
+ * were called. This ensures that the arrays can only ever ratchet in the
+ * scan's current direction. Without this, scans would overlook matching
+ * tuples if and when the scan's direction was subsequently reversed.
*/
- if (!found)
- so->arraysStarted = false;
+ _bt_start_array_keys(scan, -dir);
- return found;
+ return false;
}
/*
- * _bt_mark_array_keys() -- Handle array keys during btmarkpos
+ * _bt_rewind_nonrequired_arrays() -- Rewind non-required arrays
*
- * Save the current state of the array keys as the "mark" position.
+ * Called when _bt_advance_array_keys decides to start a new primitive index
+ * scan on the basis of the current scan position being before the position
+ * that _bt_first is capable of repositioning the scan to by applying an
+ * inequality operator required in the opposite-to-scan direction only.
+ *
+ * Although equality strategy scan keys (for both arrays and non-arrays alike)
+ * are either marked required in both directions or in neither direction,
+ * there is a sense in which non-required arrays behave like required arrays.
+ * With a qual such as "WHERE a IN (100, 200) AND b >= 3 AND c IN (5, 6, 7)",
+ * the scan key on "c" is non-required, but nevertheless enables positioning
+ * the scan at the first tuple >= "(100, 3, 5)" on the leaf level during the
+ * first descent of the tree by _bt_first. Later on, there could also be a
+ * second descent, that places the scan right before tuples >= "(200, 3, 5)".
+ * _bt_first must never be allowed to build an insertion scan key whose "c"
+ * entry is set to a value other than 5, the "c" array's first element/value.
+ * (Actually, it's the first in the current scan direction. This example uses
+ * a forward scan.)
+ *
+ * Calling here resets the array scan key elements for the scan's non-required
+ * arrays. This is strictly necessary for correctness in a subset of cases
+ * involving "required in opposite direction"-triggered primitive index scans.
+ * Not all callers are at risk of _bt_first using a non-required array like
+ * this, but advancement always resets the arrays when another primitve scan
+ * is scheduled, just to keep things simple. Array advancement even makes
+ * sure to reset non-required arrays during scans that have no inequalities.
+ * (Advancement still won't call here when there are no inequalities, though
+ * that's just because it's all handled indirectly instead.)
+ *
+ * Note: _bt_verify_arrays_bt_first is called by an assertion to enforce that
+ * everybody got this right.
*/
-void
-_bt_mark_array_keys(IndexScanDesc scan)
+static void
+_bt_rewind_nonrequired_arrays(IndexScanDesc scan, ScanDirection dir)
{
BTScanOpaque so = (BTScanOpaque) scan->opaque;
- int i;
+ int arrayidx = 0;
- for (i = 0; i < so->numArrayKeys; i++)
+ Assert(so->primScanDir == dir);
+
+ for (int ikey = 0; ikey < so->numberOfKeys; ikey++)
{
- BTArrayKeyInfo *curArrayKey = &so->arrayKeys[i];
+ ScanKey cur = so->keyData + ikey;
+ BTArrayKeyInfo *array = NULL;
+ int first_elem_dir;
- curArrayKey->mark_elem = curArrayKey->cur_elem;
+ if (!(cur->sk_flags & SK_SEARCHARRAY) ||
+ cur->sk_strategy != BTEqualStrategyNumber)
+ continue;
+
+ array = &so->arrayKeys[arrayidx++];
+ Assert(array->scan_key == ikey);
+
+ if ((cur->sk_flags & (SK_BT_REQFWD | SK_BT_REQBKWD)))
+ continue;
+
+ if (ScanDirectionIsForward(dir) || !array)
+ first_elem_dir = 0;
+ else
+ first_elem_dir = array->num_elems - 1;
+
+ if (array->cur_elem != first_elem_dir)
+ {
+ array->cur_elem = first_elem_dir;
+ cur->sk_argument = array->elem_values[first_elem_dir];
+ }
}
}
/*
- * _bt_restore_array_keys() -- Handle array keys during btrestrpos
+ * _bt_tuple_before_array_skeys() -- determine if tuple advances array keys
*
- * Restore the array keys to where they were when the mark was set.
+ * We always compare the tuple using the current array keys (which we assume
+ * are already set in so->keyData[]). readpagetup indicates if tuple is the
+ * scan's current _bt_readpage-wise tuple.
+ *
+ * readpagetup callers must only call here when _bt_check_compare already set
+ * continuescan=false. We help these callers deal with _bt_check_compare's
+ * inability to distinguishing between the < and > cases (it uses equality
+ * operator scan keys, whereas we use 3-way ORDER procs).
+ *
+ * Returns true when caller passes a tuple that is < the current set of array
+ * keys for the most significant non-equal column/scan key (or > for backwards
+ * scans). This happens to readpagetup callers when tuple is still before the
+ * start of matches for the scan's current required array keys.
+ *
+ * Returns false when caller's tuple is >= the current array keys (or <=, in
+ * the case of backwards scans). This happens to readpagetup callers when the
+ * scan has reached the point of needing its array keys advanced.
+ *
+ * As an optimization, readpagetup callers pass a _bt_check_compare-set sktrig
+ * value to indicate which scan key triggered _bt_checkkeys to recheck with us
+ * (!readpagetup callers must always pass sktrig=0). This allows us to avoid
+ * wastefully checking earlier scan keys that _bt_check_compare already found
+ * to be satisfied by the current qual/set of array keys. If sktrig indicates
+ * a non-required array that _bt_check_compare just set continuescan=false for
+ * (see _bt_check_compare for an explanation), then we always return false.
+ *
+ * !readpagetup callers optionally pass us *scanBehind, which tracks whether
+ * any missing truncated attributes might have affected array advancement
+ * (compared to what would happen if it was shown the first non-pivot tuple on
+ * the page to the right of caller's finaltup/high key tuple instead). It's
+ * only possible that we'll set *scanBehind to true when caller passes us a
+ * pivot tuple (with truncated attributes) that we return false for.
*/
-void
-_bt_restore_array_keys(IndexScanDesc scan)
+static bool
+_bt_tuple_before_array_skeys(IndexScanDesc scan, ScanDirection dir,
+ IndexTuple tuple, int tupnatts, TupleDesc tupdesc,
+ bool readpagetup, int sktrig, bool *scanBehind)
{
BTScanOpaque so = (BTScanOpaque) scan->opaque;
- bool changed = false;
- int i;
- /* Restore each array key to its position when the mark was set */
- for (i = 0; i < so->numArrayKeys; i++)
+ Assert(so->numArrayKeys);
+ Assert(so->numberOfKeys);
+ Assert(!so->needPrimScan);
+ Assert(sktrig == 0 || readpagetup);
+ Assert(!readpagetup || scanBehind == NULL);
+
+ if (scanBehind)
+ *scanBehind = false;
+
+ for (; sktrig < so->numberOfKeys; sktrig++)
{
- BTArrayKeyInfo *curArrayKey = &so->arrayKeys[i];
- ScanKey skey = &so->arrayKeyData[curArrayKey->scan_key];
- int mark_elem = curArrayKey->mark_elem;
+ ScanKey cur = so->keyData + sktrig;
+ FmgrInfo *orderproc;
+ Datum tupdatum;
+ bool tupnull;
+ int32 result;
- if (curArrayKey->cur_elem != mark_elem)
+ /*
+ * Once we reach a non-required scan key, we're completely done.
+ *
+ * Note: we deliberately don't consider the scan direction here.
+ * _bt_advance_array_keys caller requires that we track *scanBehind
+ * without concern for scan direction.
+ */
+ if ((cur->sk_flags & (SK_BT_REQFWD | SK_BT_REQBKWD)) == 0)
{
- curArrayKey->cur_elem = mark_elem;
- skey->sk_argument = curArrayKey->elem_values[mark_elem];
- changed = true;
+ Assert(!readpagetup || (cur->sk_strategy == BTEqualStrategyNumber &&
+ (cur->sk_flags & SK_SEARCHARRAY)));
+ return false;
+ }
+
+ /* readpagetup calls require one ORDER proc comparison (at most) */
+ Assert(!readpagetup || cur == so->keyData + sktrig);
+
+ if (cur->sk_attno > tupnatts)
+ {
+ Assert(!readpagetup);
+
+ /*
+ * When we reach a high key's truncated attribute, assume that the
+ * tuple attribute's value is >= the scan's equality constraint
+ * scan keys (but set *scanBehind to let interested callers know
+ * that a truncated attribute might have affected our answer).
+ */
+ if (scanBehind)
+ *scanBehind = true;
+
+ return false;
+ }
+
+ /*
+ * Inequality strategy scan keys (that are required in current scan
+ * direction) can only be evaluated by _bt_check_compare
+ */
+ if (cur->sk_strategy != BTEqualStrategyNumber)
+ {
+ /*
+ * Give up right away when _bt_check_compare indicated that a
+ * required inequality scan key wasn't satisfied
+ */
+ if (readpagetup)
+ return false;
+
+ /*
+ * Otherwise we can't give up, since we must check all required
+ * scan keys in order to correctly track *scanBehind for caller
+ */
+ continue;
+ }
+
+ orderproc = &so->orderProcs[sktrig];
+ tupdatum = index_getattr(tuple, cur->sk_attno, tupdesc, &tupnull);
+
+ result = _bt_compare_array_skey(orderproc, tupdatum, tupnull,
+ cur->sk_argument, cur);
+
+ /*
+ * Does this comparison indicate that caller must _not_ advance the
+ * scan's arrays just yet?
+ */
+ if ((ScanDirectionIsForward(dir) && result < 0) ||
+ (ScanDirectionIsBackward(dir) && result > 0))
+ return true;
+
+ /*
+ * Does this comparison indicate that caller should now advance the
+ * scan's arrays? (Must be if we get here during a readpagetup call.)
+ */
+ if (readpagetup || result != 0)
+ {
+ Assert(result != 0);
+ return false;
+ }
+
+ /*
+ * Inconclusive -- need to check later scan keys, too.
+ *
+ * This must be a finaltup precheck, or a call made from an assertion.
+ */
+ Assert(result == 0);
+ Assert(!readpagetup);
+ }
+
+ return false;
+}
+
+/*
+ * _bt_start_prim_scan() -- start scheduled primitive index scan?
+ *
+ * Returns true if _bt_checkkeys scheduled another primitive index scan, just
+ * as the last one ended. Otherwise returns false, indicating that the array
+ * keys are now fully exhausted.
+ *
+ * Only call here during scans with one or more equality type array scan keys,
+ * after _bt_first or _bt_next return false.
+ */
+bool
+_bt_start_prim_scan(IndexScanDesc scan, ScanDirection dir)
+{
+ BTScanOpaque so = (BTScanOpaque) scan->opaque;
+
+ Assert(so->numArrayKeys);
+ Assert(so->primScanDir == dir || !so->qual_ok);
+
+ /*
+ * Array keys are advanced within _bt_checkkeys when the scan reaches the
+ * leaf level (more precisely, they're advanced when the scan reaches the
+ * end of each distinct set of array elements). This process avoids
+ * repeat access to leaf pages (across multiple primitive index scans) by
+ * advancing the scan's array keys when it allows the primitive index scan
+ * to find nearby matching tuples (or when it eliminates ranges of array
+ * key space that can't possibly be satisfied by any index tuple).
+ *
+ * _bt_checkkeys sets a simple flag variable to schedule another primitive
+ * index scan. The flag tells us what to do.
+ *
+ * We cannot rely on _bt_first always reaching _bt_checkkeys. There are
+ * various cases where that won't happen. For example, if the index is
+ * completely empty, then _bt_first won't call _bt_readpage/_bt_checkkeys.
+ * We also don't expect a call to _bt_checkkeys during searches for a
+ * non-existent value that happens to be lower/higher than any existing
+ * value in the index.
+ *
+ * We don't require special handling for these cases -- we don't need to
+ * be explicitly instructed to _not_ perform another primitive index scan.
+ * It's up to code under the control of _bt_first to always set the flag
+ * when another primitive index scan will be required.
+ *
+ * This works correctly, even with the tricky cases listed above, which
+ * all involve access to leaf pages "near the boundaries of the key space"
+ * (whether it's from a leftmost/rightmost page, or an imaginary empty
+ * leaf root page). If _bt_checkkeys cannot be reached by a primitive
+ * index scan for one set of array keys, then it also won't be reached for
+ * any later set ("later" in terms of the direction that we scan the index
+ * and advance the arrays). The array keys won't have advanced in these
+ * cases, but that's the correct behavior (even _bt_advance_array_keys
+ * won't always advance the arrays at the point they become "exhausted").
+ */
+ if (so->needPrimScan)
+ {
+ Assert(_bt_verify_arrays_bt_first(scan, dir));
+
+ /* Flag was set -- must call _bt_first again */
+ so->needPrimScan = false;
+ so->scanBehind = false;
+ if (scan->parallel_scan != NULL)
+ _bt_parallel_next_primitive_scan(scan);
+
+ return true;
+ }
+
+ /* The top-level index scan ran out of tuples in this scan direction */
+ if (scan->parallel_scan != NULL)
+ _bt_parallel_done(scan);
+
+ return false;
+}
+
+/*
+ * _bt_advance_array_keys() -- Advance array elements using a tuple
+ *
+ * The scan always gets a new qual as a consequence of calling here (except
+ * when we determine that the top-level scan has run out of matching tuples).
+ * All later _bt_check_compare calls also use the same new qual that was first
+ * used here (at least until the next call here advances the keys once again).
+ * It's convenient to structure _bt_check_compare rechecks of caller's tuple
+ * (using the new qual) as one the steps of advancing the scan's array keys,
+ * so this function works as a wrapper around _bt_check_compare.
+ *
+ * Like _bt_check_compare, we'll set pstate.continuescan on behalf of the
+ * caller, and return a boolean indicating if caller's tuple satisfies the
+ * scan's new qual. But unlike _bt_check_compare, we set so->needPrimScan
+ * when we set continuescan=false, indicating if a new primitive index scan
+ * has been scheduled (otherwise, the top-level scan has run out of tuples in
+ * the current scan direction).
+ *
+ * Caller must use _bt_tuple_before_array_skeys to determine if the current
+ * place in the scan is >= the current array keys _before_ calling here.
+ * We're responsible for ensuring that caller's tuple is <= the newly advanced
+ * required array keys once we return. We try to find an exact match, but
+ * failing that we'll advance the array keys to whatever set of array elements
+ * comes next in the key space for the current scan direction. Required array
+ * keys "ratchet forwards" (or backwards). They can only advance as the scan
+ * itself advances through the index/key space.
+ *
+ * (The rules are the same for backwards scans, except that the operators are
+ * flipped: just replace the precondition's >= operator with a <=, and the
+ * postcondition's <= operator with with a >=. In other words, just swap the
+ * precondition with the postcondition.)
+ *
+ * We also deal with "advancing" non-required arrays here. Sometimes that'll
+ * be the sole reason for calling here. These calls are the only exception to
+ * the general rule about always advancing the array keys. (That rule only
+ * applies when a required scan key was found to be unsatisfied.)
+ *
+ * Note that we deal with non-array required equality strategy scan keys as
+ * degenerate single element arrays here. Obviously, they can never really
+ * advance in the way that real arrays can, but they must still affect how we
+ * advance real array scan keys (exactly like true array equality scan keys).
+ * We have to keep around a 3-way ORDER proc for these (using the "=" operator
+ * won't do), since in general whether the tuple is < or > _any_ unsatisfied
+ * required equality key influences how the scan's real arrays must advance.
+ *
+ * Note also that we may sometimes need to advance the array keys when the
+ * existing array keys are already an exact match for every corresponding
+ * value from caller's tuple. This is how we deal with inequalities that are
+ * required in the current scan direction. They can advance the array keys
+ * here, even though they don't influence the initial positioning strategy
+ * within _bt_first (only inequalities required in the _opposite_ direction to
+ * the scan influence _bt_first in this way). When sktrig (which is an offset
+ * to the unsatisfied scan key set by _bt_check_compare) is for a required
+ * inequality scan key, we'll perform array key advancement.
+ */
+static bool
+_bt_advance_array_keys(IndexScanDesc scan, BTReadPageState *pstate,
+ IndexTuple tuple, int tupnatts, TupleDesc tupdesc,
+ int sktrig)
+{
+ BTScanOpaque so = (BTScanOpaque) scan->opaque;
+ Relation rel = scan->indexRelation;
+ ScanDirection dir = pstate->dir;
+ int arrayidx = 0;
+ bool arrays_advanced = false,
+ beyond_end_advance = false,
+ sktrig_required = false,
+ has_required_opposite_direction_only = false,
+ oppodir_inequality_sktrig = false,
+ all_required_satisfied = true,
+ all_satisfied = true;
+
+ /*
+ * Precondition array state assertions
+ */
+ Assert(!so->needPrimScan && so->primScanDir == dir);
+ Assert(_bt_verify_keys_with_arraykeys(scan));
+ Assert(!_bt_tuple_before_array_skeys(scan, dir, tuple, tupnatts, tupdesc,
+ false, 0, NULL));
+
+ so->scanBehind = false; /* reset */
+
+ for (int ikey = 0; ikey < so->numberOfKeys; ikey++)
+ {
+ ScanKey cur = so->keyData + ikey;
+ FmgrInfo *orderproc;
+ BTArrayKeyInfo *array = NULL;
+ Datum tupdatum;
+ bool required = false,
+ required_opposite_direction_only = false,
+ tupnull;
+ int32 result;
+ int set_elem = 0;
+
+ if (cur->sk_strategy == BTEqualStrategyNumber)
+ {
+ /* Manage array state */
+ if (cur->sk_flags & SK_SEARCHARRAY)
+ {
+ array = &so->arrayKeys[arrayidx++];
+ Assert(array->scan_key == ikey);
+ }
+ }
+ else
+ {
+ /*
+ * Are any inequalities required in the opposite direction only
+ * present here?
+ */
+ if (((ScanDirectionIsForward(dir) &&
+ (cur->sk_flags & (SK_BT_REQBKWD))) ||
+ (ScanDirectionIsBackward(dir) &&
+ (cur->sk_flags & (SK_BT_REQFWD)))))
+ has_required_opposite_direction_only =
+ required_opposite_direction_only = true;
+ }
+
+ /* Optimization: skip over known-satisfied scan keys */
+ if (ikey < sktrig)
+ continue;
+
+ if (cur->sk_flags & (SK_BT_REQFWD | SK_BT_REQBKWD))
+ {
+ required = true;
+
+ if (ikey == sktrig)
+ {
+ /*
+ * Required scan key wasn't satisfied, so required arrays will
+ * have to advance. Invalidate page-level state that tracks
+ * whether the scan's required-in-opposite-direction-only keys
+ * are known to be satisfied by page's remaining tuples.
+ */
+ sktrig_required = true;
+ pstate->firstmatch = false;
+
+ /* Shouldn't have to invalidate 'prechecked', though */
+ Assert(!pstate->prechecked);
+ }
+
+ if (cur->sk_attno > tupnatts)
+ {
+ /* Set this just like _bt_tuple_before_array_skeys */
+ Assert(sktrig < ikey);
+ so->scanBehind = true;
+ }
+ }
+
+ /*
+ * Handle a required non-array scan key that the initial call to
+ * _bt_check_compare indicated triggered array advancement, if any.
+ *
+ * The non-array scan key's strategy will be <, <=, or = during a
+ * forwards scan (or any one of =, >=, or > during a backwards scan).
+ * It follows that the corresponding tuple attribute's value must now
+ * be either > or >= the scan key value (for backwards scans it must
+ * be either < or <= that value).
+ *
+ * If this is a required equality strategy scan key, this is just an
+ * optimization; _bt_tuple_before_array_skeys already confirmed that
+ * this scan key places us ahead of caller's tuple. There's no need
+ * to repeat that work now. (We only do comparisons of any required
+ * non-array equality scan keys that come after the triggering key.)
+ *
+ * If this is a required inequality strategy scan key, we _must_ rely
+ * on _bt_check_compare like this; we aren't capable of directly
+ * evaluating required inequality strategy scan keys here, on our own.
+ */
+ if (ikey == sktrig && !array)
+ {
+ Assert(required && all_required_satisfied && !arrays_advanced);
+
+ /* Use "beyond end" advancement. See below for an explanation. */
+ beyond_end_advance = true;
+ all_satisfied = all_required_satisfied = false;
+
+ /*
+ * Set a flag that remembers that this was an inequality required
+ * in the opposite scan direction only, that nevertheless
+ * triggered the call here.
+ *
+ * This only happens when an inequality operator (which must be
+ * strict) encounters a group of NULLs that indicate the end of
+ * non-NULL values for tuples in the current scan direction.
+ */
+ if (unlikely(required_opposite_direction_only))
+ oppodir_inequality_sktrig = true;
+
+ continue;
+ }
+
+ /*
+ * Nothing more for us to do with an inequality strategy scan key that
+ * wasn't the one that _bt_check_compare stopped on, though.
+ *
+ * Note: if our later call to _bt_check_compare (to recheck caller's
+ * tuple) sets continuescan=false due to finding this same inequality
+ * unsatisfied (possible when it's required in the scan direction),
+ * we'll deal with it via a recursive "second pass" call.
+ */
+ else if (cur->sk_strategy != BTEqualStrategyNumber)
+ continue;
+
+ /*
+ * Nothing for us to do with an equality strategy scan key that isn't
+ * marked required, either.
+ *
+ * Non-required array scan keys are the only exception. They're a
+ * special case in that _bt_check_compare can set continuescan=false
+ * for them, just as it will given an unsatisfied required scan key.
+ * It's convenient to follow the same convention, since it results in
+ * our getting called to advance non-required arrays in the same way
+ * as required arrays (though we avoid stopping the scan for them).
+ */
+ else if (!required && !array)
+ continue;
+
+ /*
+ * Here we perform steps for all array scan keys after a required
+ * array scan key whose binary search triggered "beyond end of array
+ * element" array advancement due to encountering a tuple attribute
+ * value > the closest matching array key (or < for backwards scans).
+ */
+ if (beyond_end_advance)
+ {
+ int final_elem_dir;
+
+ if (ScanDirectionIsBackward(dir) || !array)
+ final_elem_dir = 0;
+ else
+ final_elem_dir = array->num_elems - 1;
+
+ if (array && array->cur_elem != final_elem_dir)
+ {
+ array->cur_elem = final_elem_dir;
+ cur->sk_argument = array->elem_values[final_elem_dir];
+ arrays_advanced = true;
+ }
+
+ continue;
+ }
+
+ /*
+ * Here we perform steps for all array scan keys after a required
+ * array scan key whose tuple attribute was < the closest matching
+ * array key when we dealt with it (or > for backwards scans).
+ *
+ * This earlier required array key already puts us ahead of caller's
+ * tuple in the key space (for the current scan direction). We must
+ * make sure that subsequent lower-order array keys do not put us too
+ * far ahead (ahead of tuples that have yet to be seen by our caller).
+ * For example, when a tuple "(a, b) = (42, 5)" advances the array
+ * keys on "a" from 40 to 45, we must also set "b" to whatever the
+ * first array element for "b" is. It would be wrong to allow "b" to
+ * be set based on the tuple value.
+ *
+ * Perform the same steps with truncated high key attributes. You can
+ * think of this as a "binary search" for the element closest to the
+ * value -inf. Again, the arrays must never get ahead of the scan.
+ */
+ if (!all_required_satisfied || cur->sk_attno > tupnatts)
+ {
+ int first_elem_dir;
+
+ if (ScanDirectionIsForward(dir) || !array)
+ first_elem_dir = 0;
+ else
+ first_elem_dir = array->num_elems - 1;
+
+ if (array && array->cur_elem != first_elem_dir)
+ {
+ array->cur_elem = first_elem_dir;
+ cur->sk_argument = array->elem_values[first_elem_dir];
+ arrays_advanced = true;
+ }
+
+ continue;
+ }
+
+ /*
+ * Search in scankey's array for the corresponding tuple attribute
+ * value from caller's tuple
+ */
+ orderproc = &so->orderProcs[ikey];
+ tupdatum = index_getattr(tuple, cur->sk_attno, tupdesc, &tupnull);
+
+ if (array)
+ {
+ bool ratchets = (required && !arrays_advanced);
+
+ /*
+ * Binary search for closest match that's available from the array
+ */
+ set_elem = _bt_binsrch_array_skey(orderproc, ratchets, dir,
+ tupdatum, tupnull,
+ array, cur, &result);
+
+ /*
+ * Required arrays only ever ratchet forwards (backwards).
+ *
+ * This condition makes it safe for binary searches to skip over
+ * array elements that the scan must already be ahead of by now.
+ * That is strictly an optimization. Our assertion verifies that
+ * the condition holds, which doesn't depend on the optimization.
+ */
+ Assert(!ratchets ||
+ ((ScanDirectionIsForward(dir) && set_elem >= array->cur_elem) ||
+ (ScanDirectionIsBackward(dir) && set_elem <= array->cur_elem)));
+ Assert(set_elem >= 0 && set_elem < array->num_elems);
+ }
+ else
+ {
+ Assert(required);
+
+ /*
+ * This is a required non-array equality strategy scan key, which
+ * we'll treat as a degenerate single value array.
+ *
+ * This scan key's imaginary "array" can't really advance, but it
+ * can still roll over like any other array. (Actually, this is
+ * no different to real single value arrays, which never advance
+ * without rolling over -- they can never truly advance, either.)
+ */
+ result = _bt_compare_array_skey(orderproc, tupdatum, tupnull,
+ cur->sk_argument, cur);
+ }
+
+ /*
+ * Consider "beyond end of array element" array advancement.
+ *
+ * When the tuple attribute value is > the closest matching array key
+ * (or < in the backwards scan case), we need to ratchet this array
+ * forward (backward) by one increment, so that caller's tuple ends up
+ * being < final array value instead (or > final array value instead).
+ * This process has to work for all of the arrays, not just this one:
+ * it must "carry" to higher-order arrays when the set_elem that we
+ * just found happens to be the final one for the scan's direction.
+ * Incrementing (decrementing) set_elem itself isn't good enough.
+ *
+ * Our approach is to provisionally use set_elem as if it was an exact
+ * match now, then set each later/less significant array to whatever
+ * its final element is. Once outside the loop we'll then "increment
+ * this array's set_elem" by calling _bt_advance_array_keys_increment.
+ * That way the process rolls over to higher order arrays as needed.
+ *
+ * Under this scheme any required arrays only ever ratchet forwards
+ * (or backwards), and always do so to the maximum possible extent
+ * that we can know will be safe without seeing the scan's next tuple.
+ * We don't need any special handling for required scan keys that lack
+ * a real array to advance, nor for redundant scan keys that couldn't
+ * be eliminated by _bt_preprocess_keys. It won't matter if some of
+ * our "true" array scan keys (or even all of them) are non-required.
+ */
+ if (required &&
+ ((ScanDirectionIsForward(dir) && result > 0) ||
+ (ScanDirectionIsBackward(dir) && result < 0)))
+ beyond_end_advance = true;
+
+ Assert(all_required_satisfied && all_satisfied);
+ if (result != 0)
+ {
+ /*
+ * Track whether caller's tuple satisfies our new post-advancement
+ * qual, for required scan keys, as well as for the entire set of
+ * interesting scan keys (all required scan keys plus non-required
+ * array scan keys are considered interesting.)
+ */
+ all_satisfied = false;
+ if (required)
+ all_required_satisfied = false;
+ else
+ {
+ /*
+ * There's no need to advance the arrays using the best
+ * available match for a non-required array. Give up now.
+ * (Though note that sktrig_required calls still have to do
+ * all the usual post-advancement steps, including the recheck
+ * call to _bt_check_compare.)
+ */
+ break;
+ }
+ }
+
+ /* Advance array keys, even when set_elem isn't an exact match */
+ if (array && array->cur_elem != set_elem)
+ {
+ array->cur_elem = set_elem;
+ cur->sk_argument = array->elem_values[set_elem];
+ arrays_advanced = true;
}
}
/*
- * If we changed any keys, we must redo _bt_preprocess_keys. That might
- * sound like overkill, but in cases with multiple keys per index column
- * it seems necessary to do the full set of pushups.
- *
- * Also do this whenever the scan's set of array keys "wrapped around" at
- * the end of the last primitive index scan. There won't have been a call
- * to _bt_preprocess_keys from some other place following wrap around, so
- * we do it for ourselves.
+ * Consider if we need to advance the array keys incrementally to finish
+ * off "beyond end of array element" array advancement. This is the only
+ * way that the array keys can be exhausted, which is how top-level index
+ * scans usually determine that they've run out of tuples to return in the
+ * current scan direction (less often the top-level scan just runs out of
+ * tuples/pages before the scan's array keys are exhausted).
*/
- if (changed || !so->arraysStarted)
+ if (beyond_end_advance)
{
- _bt_preprocess_keys(scan);
- /* The mark should have been set on a consistent set of keys... */
- Assert(so->qual_ok);
- }
-}
+ Assert(sktrig_required && !all_required_satisfied && !all_satisfied);
+ if (_bt_advance_array_keys_increment(scan, dir))
+ arrays_advanced = true;
+ else
+ {
+ /* Arrays are exhausted */
+ goto end_toplevel_scan;
+ }
+ }
+
+ Assert(_bt_verify_keys_with_arraykeys(scan));
+
+ /*
+ * Does tuple now satisfy our new qual? Recheck with _bt_check_compare.
+ *
+ * Calls triggered by an unsatisfied required scan key, whose tuple now
+ * satisfies all required scan keys, but not all nonrequired array keys,
+ * will still require a recheck call to _bt_check_compare. They'll still
+ * need its "second pass" handling of required inequality scan keys.
+ * (Might have missed a still-unsatisfied required inequality scan key
+ * that caller didn't detect as the sktrig scan key during its initial
+ * _bt_check_compare call that used the old/original qual.)
+ *
+ * Calls triggered by an unsatisfied nonrequired array scan key never need
+ * "second pass" handling of required inequalities (nor any other handling
+ * of any required scan key). All that matters is whether caller's tuple
+ * satisfies the new qual, so it's safe to just skip the _bt_check_compare
+ * recheck when we've already determined that it'll just return 'false'.
+ */
+ if ((sktrig_required && all_required_satisfied) ||
+ (!sktrig_required && all_satisfied))
+ {
+ int nsktrig = sktrig + 1;
+
+ Assert(all_required_satisfied);
+
+ /* Recheck _bt_check_compare on behalf of caller */
+ if (_bt_check_compare(dir, so, tuple, tupnatts, tupdesc,
+ false, false, false,
+ &pstate->continuescan, &nsktrig) &&
+ !so->scanBehind)
+ {
+ /* This tuple satisfies the new qual */
+ Assert(all_satisfied);
+ return true;
+ }
+
+ /*
+ * Consider "second pass" handling of required inequalities.
+ *
+ * It's possible that our _bt_check_compare call indicated that the
+ * scan should end due to some unsatisfied inequality that wasn't
+ * initially recognized as such by us. Handle this by calling
+ * ourselves recursively, this time indicating that the trigger is the
+ * inequality that we missed first time around (and using a set of
+ * required array/equality keys that are now exact matches for tuple).
+ *
+ * We make a strong, general guarantee that every _bt_checkkeys call
+ * here will advance the array keys to the maximum possible extent
+ * that we can know to be safe based on caller's tuple alone. If we
+ * didn't perform this step, then that guarantee wouldn't quite hold.
+ */
+ if (unlikely(!pstate->continuescan))
+ {
+ bool satisfied PG_USED_FOR_ASSERTS_ONLY;
+
+ Assert(sktrig_required);
+ Assert(so->keyData[nsktrig].sk_strategy != BTEqualStrategyNumber);
+
+ /*
+ * The tuple must use "beyond end" advancement during the
+ * recursive call, so we cannot possibly end up back here when
+ * recursing. We'll consume a small, fixed amount of stack space.
+ */
+ Assert(!beyond_end_advance);
+
+ /* Advance the array keys a second time using same tuple */
+ satisfied = _bt_advance_array_keys(scan, pstate, tuple, tupnatts,
+ tupdesc, nsktrig);
+
+ /* This tuple doesn't satisfy the inequality */
+ Assert(!satisfied);
+ return false;
+ }
+
+ /*
+ * Some non-required scan key (from new qual) still not satisfied.
+ *
+ * All scan keys required in the current scan direction must still be
+ * satisfied, though, so we can trust all_required_satisfied below.
+ *
+ * Note: it's still too early to tell if the current primitive index
+ * scan can continue (has_required_opposite_direction_only steps might
+ * still start a new primitive index scan instead).
+ */
+ }
+
+ /*
+ * Postcondition array state assertions (for still-unsatisfied tuples).
+ *
+ * Caller's tuple is now < the newly advanced array keys (or > when this
+ * is a backwards scan) when not all required scan keys from the new qual
+ * (including any required inequality keys) were found to be satisfied.
+ */
+ Assert(_bt_tuple_before_array_skeys(scan, dir, tuple, tupnatts, tupdesc,
+ false, 0, NULL) ==
+ !all_required_satisfied);
+
+ /*
+ * When we were called just to deal with "advancing" non-required arrays,
+ * there's no way that we can need to start a new primitive index scan
+ * (and it would be wrong to allow it). Continue ongoing primitive scan.
+ *
+ * Note: this test isn't strictly necessary, but having explicit handling
+ * is still useful. It avoids wasting cycles on considering outcomes that
+ * we already know are impossible (has_required_opposite_direction_only
+ * steps are particularly worth avoiding when they aren't really needed).
+ */
+ if (!sktrig_required)
+ {
+ Assert(all_required_satisfied);
+ goto continue_prim_scan;
+ }
+
+ /*
+ * By here we have established that the scan's required arrays (there must
+ * be at least one that's required) advanced, without becoming exhausted
+ */
+ Assert(sktrig_required && arrays_advanced);
+
+ /*
+ * We generally permit primitive index scans to continue onto the next
+ * sibling page when the page's finaltup satisfies all required scan keys
+ * at the point where we're between pages.
+ *
+ * If caller's tuple is also the page's finaltup, and we see that required
+ * scan keys still aren't satisfied, start a new primitive index scan.
+ */
+ if (!all_required_satisfied && pstate->finaltup == tuple)
+ goto new_prim_scan;
+
+ /*
+ * Proactively check finaltup (don't wait until finaltup is reached by the
+ * scan) when it might well turn out to not be satisfied later on.
+ *
+ * This isn't quite equivalent to looking ahead to check if finaltup will
+ * also be satisfied by all required scan keys, since there isn't any real
+ * handling of inequalities in _bt_tuple_before_array_skeys. It wouldn't
+ * make sense for us to evaluate inequalities when "looking ahead to
+ * finaltup", though. Inequalities that are required in the current scan
+ * direction cannot affect how _bt_first repositions the top-level scan
+ * (unless the scan direction happens to change).
+ *
+ * Note: if so->scanBehind hasn't already been set for finaltup by us,
+ * it'll be set during this call to _bt_tuple_before_array_skeys. Either
+ * way, it'll be set correctly after this point.
+ */
+ if (!all_required_satisfied && pstate->finaltup &&
+ _bt_tuple_before_array_skeys(scan, dir, pstate->finaltup,
+ BTreeTupleGetNAtts(pstate->finaltup, rel),
+ tupdesc, false, 0, &so->scanBehind))
+ goto new_prim_scan;
+
+ /*
+ * When we encounter a truncated finaltup high key attribute, we're
+ * optimistic about the chances of its corresponding required scan key
+ * being satisfied when we go on to check it against tuples from this
+ * page's right sibling leaf page. We consider truncated attributes to be
+ * satisfied by required scan keys, which allows the primitive index scan
+ * to continue to the next leaf page. We must set so->scanBehind to true
+ * to remember that the last page's finaltup had "satisfied" required scan
+ * keys for one or more truncated attribute values (scan keys required in
+ * _either_ scan direction).
+ *
+ * There is a chance that _bt_checkkeys (which checks so->scanBehind) will
+ * find that even the sibling leaf page's finaltup is < the new array
+ * keys. When that happens, our optimistic policy will have incurred a
+ * single extra leaf page access that could have been avoided.
+ *
+ * A pessimistic policy would give backward scans a gratuitous advantage
+ * over forward scans. We'd punish forward scans for applying more
+ * accurate information from the high key, rather than just using the
+ * final non-pivot tuple as finaltup, in the style of backward scans.
+ * Being pessimistic would also give some scans with non-required arrays a
+ * perverse advantage over similar scans that use required arrays instead.
+ *
+ * You can think of this as a speculative bet on what the scan is likely
+ * to find on the next page. It's not much of a gamble, though, since the
+ * untruncated prefix of attributes must strictly satisfy the new qual
+ * (though it's okay if any non-required scan keys fail to be satisfied).
+ */
+ if (so->scanBehind && has_required_opposite_direction_only)
+ {
+ /*
+ * However, we avoid this behavior whenever the scan involves a scan
+ * key required in the opposite direction to the scan only, along with
+ * a finaltup with at least one truncated attribute that's associated
+ * with a scan key marked required (required in either direction).
+ *
+ * _bt_check_compare simply won't stop the scan for a scan key that's
+ * marked required in the opposite scan direction only. That leaves
+ * us without any reliable way of reconsidering any opposite-direction
+ * inequalities if it turns out that starting a new primitive index
+ * scan will allow _bt_first to skip ahead by a great many leaf pages
+ * (see next section for details of how that works).
+ */
+ goto new_prim_scan;
+ }
+
+ /*
+ * Handle inequalities marked required in the opposite scan direction.
+ * They can also signal that we should start a new primitive index scan.
+ *
+ * It's possible that the scan is now positioned where "matching" tuples
+ * begin, and that caller's tuple satisfies all scan keys required in the
+ * current scan direction. But if caller's tuple still doesn't satisfy
+ * other scan keys that are required in the opposite scan direction only
+ * (e.g., a required >= strategy scan key when scan direction is forward),
+ * it's still possible that there are many leaf pages before the page that
+ * _bt_first could skip straight to. Groveling through all those pages
+ * will always give correct answers, but it can be very inefficient. We
+ * must avoid needlessly scanning extra pages.
+ *
+ * Separately, it's possible that _bt_check_compare set continuescan=false
+ * for a scan key that's required in the opposite direction only. This is
+ * a special case, that happens only when _bt_check_compare sees that the
+ * inequality encountered a NULL value. This signals the end of non-NULL
+ * values in the current scan direction, which is reason enough to end the
+ * (primitive) scan. If this happens at the start of a large group of
+ * NULL values, then we shouldn't expect to be called again until after
+ * the scan has already read indefinitely-many leaf pages full of tuples
+ * with NULL suffix values. We need a separate test for this case so that
+ * we don't miss our only opportunity to skip over such a group of pages.
+ *
+ * Apply a test against finaltup to detect and recover from the problem:
+ * if even finaltup doesn't satisfy such an inequality, we just skip by
+ * starting a new primitive index scan. When we skip, we know for sure
+ * that all of the tuples on the current page following caller's tuple are
+ * also before the _bt_first-wise start of tuples for our new qual. That
+ * at least suggests many more skippable pages beyond the current page.
+ */
+ if (has_required_opposite_direction_only && pstate->finaltup &&
+ (all_required_satisfied || oppodir_inequality_sktrig))
+ {
+ int nfinaltupatts = BTreeTupleGetNAtts(pstate->finaltup, rel);
+ ScanDirection flipped;
+ bool continuescanflip;
+ int opsktrig;
+
+ /*
+ * We're checking finaltup (which is usually not caller's tuple), so
+ * cannot reuse work from caller's earlier _bt_check_compare call.
+ *
+ * Flip the scan direction when calling _bt_check_compare this time,
+ * so that it will set continuescanflip=false when it encounters an
+ * inequality required in the opposite scan direction.
+ */
+ Assert(!so->scanBehind);
+ opsktrig = 0;
+ flipped = -dir;
+ _bt_check_compare(flipped, so, pstate->finaltup, nfinaltupatts,
+ tupdesc, false, false, false,
+ &continuescanflip, &opsktrig);
+
+ /*
+ * If we ended up here due to the all_required_satisfied criteria,
+ * test opsktrig in a way that ensures that finaltup contains the same
+ * prefix of key columns as caller's tuple (a prefix that satisfies
+ * earlier required-in-current-direction scan keys).
+ *
+ * If we ended up here due to the oppodir_inequality_sktrig criteria,
+ * test opsktrig in a way that ensures that the same scan key that our
+ * caller found to be unsatisfied (by the scan's tuple) was also the
+ * one unsatisfied just now (by finaltup). That way we'll only start
+ * a new primitive scan when we're sure that both tuples _don't_ share
+ * the same prefix of satisfied equality-constrained attribute values,
+ * and that finaltup has a non-NULL attribute value indicated by the
+ * unsatisfied scan key at offset opsktrig/sktrig. (This depends on
+ * _bt_check_compare not caring about the direction that inequalities
+ * are required in whenever NULL attribute values are unsatisfied. It
+ * only cares about the scan direction, and its relationship to
+ * whether NULLs are stored first or last relative to non-NULLs.)
+ */
+ Assert(all_required_satisfied != oppodir_inequality_sktrig);
+ if (unlikely(!continuescanflip &&
+ ((all_required_satisfied && opsktrig > sktrig) ||
+ (oppodir_inequality_sktrig && opsktrig >= sktrig))))
+ {
+ Assert(so->keyData[opsktrig].sk_strategy != BTEqualStrategyNumber);
+
+ /*
+ * Make sure that any non-required arrays are set to the first
+ * array element for the current scan direction
+ */
+ _bt_rewind_nonrequired_arrays(scan, dir);
+
+ goto new_prim_scan;
+ }
+ }
+
+continue_prim_scan:
+
+ /*
+ * Stick with the ongoing primitive index scan for now.
+ *
+ * It's possible that later tuples will also turn out to have values that
+ * are still < the now-current array keys (or > the current array keys).
+ * Our caller will handle this by performing what amounts to a linear
+ * search of the page, implemented by calling _bt_check_compare and then
+ * _bt_tuple_before_array_skeys for each tuple.
+ *
+ * This approach has various advantages over a binary search of the page.
+ * We expect that our caller will quickly discover the next tuple covered
+ * by the current array keys. Repeated binary searches of the page (one
+ * binary search per array advancement) is unlikely to outperform one
+ * continuous linear search of the whole page.
+ */
+ pstate->continuescan = true; /* Override _bt_check_compare */
+ so->needPrimScan = false; /* _bt_readpage has more tuples to check */
+
+ /* Caller's tuple doesn't match the new qual */
+ return false;
+
+new_prim_scan:
+
+ /*
+ * End this primitive index scan, but schedule another.
+ *
+ * Note: If the scan direction happens to change, this scheduled primitive
+ * index scan won't go ahead after all.
+ */
+ pstate->continuescan = false; /* Tell _bt_readpage we're done... */
+ so->needPrimScan = true; /* ...but call _bt_first again */
+
+ /* Caller's tuple doesn't match the new qual */
+ return false;
+
+end_toplevel_scan:
+
+ /*
+ * End the current primitive index scan, but don't schedule another.
+ *
+ * This ends the entire top-level scan in the current scan direction.
+ *
+ * Note: The scan's arrays (including any non-required arrays) are now in
+ * their final positions for the current scan direction. If the scan
+ * direction happens to change, then the arrays will already be in their
+ * first positions for what will then be the current scan direction.
+ */
+ pstate->continuescan = false; /* Tell _bt_readpage we're done... */
+ so->needPrimScan = false; /* ...don't call _bt_first again, though */
+ so->scanBehind = false;
+
+ /* Caller's tuple doesn't match any qual */
+ return false;
+}
/*
* _bt_preprocess_keys() -- Preprocess scan keys
*
- * The given search-type keys (in scan->keyData[] or so->arrayKeyData[])
+ * The given search-type keys (taken from scan->keyData[])
* are copied to so->keyData[] with possible transformation.
* scan->numberOfKeys is the number of input keys, so->numberOfKeys gets
* the number of output keys (possibly less, never greater).
@@ -691,7 +2435,12 @@ _bt_restore_array_keys(IndexScanDesc scan)
* (but verify) that the input keys are already so sorted --- this is done
* by match_clauses_to_index() in indxpath.c. Some reordering of the keys
* within each attribute may be done as a byproduct of the processing here,
- * but no other code depends on that.
+ * but only array advancement depends on that (when dealing with multiple
+ * arrays against the same attribute). Index scans with array scan keys also
+ * depend on state (maintained here by us) that maps each of our input scan
+ * keys to its corresponding output scan key. This indirection allows index
+ * scans to use an ikey offset-to-output-scankey to look up the cached ORDER
+ * proc for the scankey.
*
* The output keys are marked with flags SK_BT_REQFWD and/or SK_BT_REQBKWD
* if they must be satisfied in order to continue the scan forward or backward
@@ -748,8 +2497,8 @@ _bt_restore_array_keys(IndexScanDesc scan)
*
* Note: the reason we have to copy the preprocessed scan keys into private
* storage is that we are modifying the array based on comparisons of the
- * key argument values, which could change on a rescan or after moving to
- * new elements of array keys. Therefore we can't overwrite the source data.
+ * key argument values, which could change on a rescan. Therefore we can't
+ * overwrite the source data.
*/
void
_bt_preprocess_keys(IndexScanDesc scan)
@@ -762,11 +2511,37 @@ _bt_preprocess_keys(IndexScanDesc scan)
ScanKey inkeys;
ScanKey outkeys;
ScanKey cur;
- ScanKey xform[BTMaxStrategyNumber];
+ BTScanKeyPreproc xform[BTMaxStrategyNumber];
bool test_result;
int i,
j;
AttrNumber attno;
+ ScanKey arrayKeyData;
+ int *keyDataMap = NULL;
+ int arrayidx = 0;
+
+ Assert(!so->needPrimScan);
+
+ /*
+ * We're called at the start of each primitive index scan during top-level
+ * scans that use equality array keys. We can reuse the scan keys that
+ * were output at the start of the scan's first primitive index scan.
+ * There is no need to perform exactly the same work more than once.
+ */
+ if (so->numberOfKeys > 0)
+ {
+ /*
+ * An earlier call to _bt_advance_array_keys already set everything up
+ * for us. Just assert that the scan's existing output scan keys are
+ * consistent with its current array elements.
+ */
+ Assert(so->numArrayKeys && !ScanDirectionIsNoMovement(so->primScanDir));
+ Assert(_bt_verify_keys_with_arraykeys(scan));
+ return;
+ }
+
+ Assert(!BTScanPosIsValid(so->currPos));
+ Assert(!BTScanPosIsValid(so->markPos));
/* initialize result variables */
so->qual_ok = true;
@@ -775,11 +2550,28 @@ _bt_preprocess_keys(IndexScanDesc scan)
if (numberOfKeys < 1)
return; /* done if qual-less scan */
+ /* If any keys are SK_SEARCHARRAY type, set up array-key info */
+ arrayKeyData = _bt_preprocess_array_keys(scan);
+ if (!so->qual_ok)
+ {
+ /* unmatchable array, so give up */
+ so->qual_ok = false;
+ return;
+ }
+
/*
- * Read so->arrayKeyData if array keys are present, else scan->keyData
+ * Treat arrayKeyData[] (a partially preprocessed copy of scan->keyData[])
+ * as our input if _bt_preprocess_array_keys just allocated it, else just
+ * use scan->keyData[]
*/
- if (so->arrayKeyData != NULL)
- inkeys = so->arrayKeyData;
+ if (arrayKeyData)
+ {
+ inkeys = arrayKeyData;
+
+ /* Also maintain keyDataMap for remapping so->orderProc[] later */
+ keyDataMap = MemoryContextAlloc(so->arrayContext,
+ numberOfKeys * sizeof(int));
+ }
else
inkeys = scan->keyData;
@@ -800,6 +2592,18 @@ _bt_preprocess_keys(IndexScanDesc scan)
/* We can mark the qual as required if it's for first index col */
if (cur->sk_attno == 1)
_bt_mark_scankey_required(outkeys);
+ if (arrayKeyData)
+ {
+ /*
+ * Don't call _bt_preprocess_array_keys_final in this fast path
+ * (we'll miss out on the single value array transformation, but
+ * that's not nearly as important when there's only one scan key)
+ */
+ Assert(cur->sk_flags & SK_SEARCHARRAY);
+ Assert(cur->sk_strategy != BTEqualStrategyNumber ||
+ so->arrayKeys[0].scan_key == 0);
+ }
+
return;
}
@@ -859,13 +2663,29 @@ _bt_preprocess_keys(IndexScanDesc scan)
* check, and we've rejected any combination of it with a regular
* equality condition; but not with other types of conditions.
*/
- if (xform[BTEqualStrategyNumber - 1])
+ if (xform[BTEqualStrategyNumber - 1].skey)
{
- ScanKey eq = xform[BTEqualStrategyNumber - 1];
+ ScanKey eq = xform[BTEqualStrategyNumber - 1].skey;
+ BTArrayKeyInfo *array = NULL;
+ FmgrInfo *orderproc = NULL;
+
+ if (arrayKeyData && (eq->sk_flags & SK_SEARCHARRAY))
+ {
+ int eq_in_ikey,
+ eq_arrayidx;
+
+ eq_in_ikey = xform[BTEqualStrategyNumber - 1].ikey;
+ eq_arrayidx = xform[BTEqualStrategyNumber - 1].arrayidx;
+ array = &so->arrayKeys[eq_arrayidx - 1];
+ orderproc = so->orderProcs + eq_in_ikey;
+
+ Assert(array->scan_key == eq_in_ikey);
+ Assert(OidIsValid(orderproc->fn_oid));
+ }
for (j = BTMaxStrategyNumber; --j >= 0;)
{
- ScanKey chk = xform[j];
+ ScanKey chk = xform[j].skey;
if (!chk || j == (BTEqualStrategyNumber - 1))
continue;
@@ -878,6 +2698,7 @@ _bt_preprocess_keys(IndexScanDesc scan)
}
if (_bt_compare_scankey_args(scan, chk, eq, chk,
+ array, orderproc,
&test_result))
{
if (!test_result)
@@ -887,7 +2708,9 @@ _bt_preprocess_keys(IndexScanDesc scan)
return;
}
/* else discard the redundant non-equality key */
- xform[j] = NULL;
+ Assert(!array || array->num_elems > 0);
+ xform[j].skey = NULL;
+ xform[j].ikey = -1;
}
/* else, cannot determine redundancy, keep both keys */
}
@@ -896,36 +2719,36 @@ _bt_preprocess_keys(IndexScanDesc scan)
}
/* try to keep only one of <, <= */
- if (xform[BTLessStrategyNumber - 1]
- && xform[BTLessEqualStrategyNumber - 1])
+ if (xform[BTLessStrategyNumber - 1].skey
+ && xform[BTLessEqualStrategyNumber - 1].skey)
{
- ScanKey lt = xform[BTLessStrategyNumber - 1];
- ScanKey le = xform[BTLessEqualStrategyNumber - 1];
+ ScanKey lt = xform[BTLessStrategyNumber - 1].skey;
+ ScanKey le = xform[BTLessEqualStrategyNumber - 1].skey;
- if (_bt_compare_scankey_args(scan, le, lt, le,
+ if (_bt_compare_scankey_args(scan, le, lt, le, NULL, NULL,
&test_result))
{
if (test_result)
- xform[BTLessEqualStrategyNumber - 1] = NULL;
+ xform[BTLessEqualStrategyNumber - 1].skey = NULL;
else
- xform[BTLessStrategyNumber - 1] = NULL;
+ xform[BTLessStrategyNumber - 1].skey = NULL;
}
}
/* try to keep only one of >, >= */
- if (xform[BTGreaterStrategyNumber - 1]
- && xform[BTGreaterEqualStrategyNumber - 1])
+ if (xform[BTGreaterStrategyNumber - 1].skey
+ && xform[BTGreaterEqualStrategyNumber - 1].skey)
{
- ScanKey gt = xform[BTGreaterStrategyNumber - 1];
- ScanKey ge = xform[BTGreaterEqualStrategyNumber - 1];
+ ScanKey gt = xform[BTGreaterStrategyNumber - 1].skey;
+ ScanKey ge = xform[BTGreaterEqualStrategyNumber - 1].skey;
- if (_bt_compare_scankey_args(scan, ge, gt, ge,
+ if (_bt_compare_scankey_args(scan, ge, gt, ge, NULL, NULL,
&test_result))
{
if (test_result)
- xform[BTGreaterEqualStrategyNumber - 1] = NULL;
+ xform[BTGreaterEqualStrategyNumber - 1].skey = NULL;
else
- xform[BTGreaterStrategyNumber - 1] = NULL;
+ xform[BTGreaterStrategyNumber - 1].skey = NULL;
}
}
@@ -936,11 +2759,13 @@ _bt_preprocess_keys(IndexScanDesc scan)
*/
for (j = BTMaxStrategyNumber; --j >= 0;)
{
- if (xform[j])
+ if (xform[j].skey)
{
ScanKey outkey = &outkeys[new_numberOfKeys++];
- memcpy(outkey, xform[j], sizeof(ScanKeyData));
+ memcpy(outkey, xform[j].skey, sizeof(ScanKeyData));
+ if (arrayKeyData)
+ keyDataMap[new_numberOfKeys - 1] = xform[j].ikey;
if (priorNumberOfEqualCols == attno - 1)
_bt_mark_scankey_required(outkey);
}
@@ -966,6 +2791,8 @@ _bt_preprocess_keys(IndexScanDesc scan)
ScanKey outkey = &outkeys[new_numberOfKeys++];
memcpy(outkey, cur, sizeof(ScanKeyData));
+ if (arrayKeyData)
+ keyDataMap[new_numberOfKeys - 1] = i;
if (numberOfEqualCols == attno - 1)
_bt_mark_scankey_required(outkey);
@@ -977,20 +2804,112 @@ _bt_preprocess_keys(IndexScanDesc scan)
continue;
}
- /* have we seen one of these before? */
- if (xform[j] == NULL)
+ /*
+ * Does this input scan key require further processing as an array?
+ */
+ if (cur->sk_strategy == InvalidStrategy)
{
- /* nope, so remember this scankey */
- xform[j] = cur;
+ /* _bt_preprocess_array_keys marked this array key redundant */
+ Assert(arrayKeyData);
+ Assert(cur->sk_flags & SK_SEARCHARRAY);
+ continue;
+ }
+
+ if (cur->sk_strategy == BTEqualStrategyNumber &&
+ (cur->sk_flags & SK_SEARCHARRAY))
+ {
+ /* _bt_preprocess_array_keys kept this array key */
+ Assert(arrayKeyData);
+ arrayidx++;
+ }
+
+ /*
+ * have we seen a scan key for this same attribute and using this same
+ * operator strategy before now?
+ */
+ if (xform[j].skey == NULL)
+ {
+ /* nope, so this scan key wins by default (at least for now) */
+ xform[j].skey = cur;
+ xform[j].ikey = i;
+ xform[j].arrayidx = arrayidx;
}
else
{
- /* yup, keep only the more restrictive key */
- if (_bt_compare_scankey_args(scan, cur, cur, xform[j],
- &test_result))
+ FmgrInfo *orderproc = NULL;
+ BTArrayKeyInfo *array = NULL;
+
+ /*
+ * Seen one of these before, so keep only the more restrictive key
+ * if possible
+ */
+ if (j == (BTEqualStrategyNumber - 1) && arrayKeyData)
{
+ /*
+ * Have to set up array keys
+ */
+ if ((cur->sk_flags & SK_SEARCHARRAY))
+ {
+ array = &so->arrayKeys[arrayidx - 1];
+ orderproc = so->orderProcs + i;
+
+ Assert(array->scan_key == i);
+ Assert(OidIsValid(orderproc->fn_oid));
+ }
+ else if ((xform[j].skey->sk_flags & SK_SEARCHARRAY))
+ {
+ array = &so->arrayKeys[xform[j].arrayidx - 1];
+ orderproc = so->orderProcs + xform[j].ikey;
+
+ Assert(array->scan_key == xform[j].ikey);
+ Assert(OidIsValid(orderproc->fn_oid));
+ }
+
+ /*
+ * Both scan keys might have arrays, in which case we'll
+ * arbitrarily pass only one of the arrays. That won't
+ * matter, since _bt_compare_scankey_args is aware that two
+ * SEARCHARRAY scan keys mean that _bt_preprocess_array_keys
+ * failed to eliminate redundant arrays through array merging.
+ * _bt_compare_scankey_args just returns false when it sees
+ * this; it won't even try to examine either array.
+ */
+ }
+
+ if (_bt_compare_scankey_args(scan, cur, cur, xform[j].skey,
+ array, orderproc, &test_result))
+ {
+ /* Have all we need to determine redundancy */
if (test_result)
- xform[j] = cur;
+ {
+ Assert(!array || array->num_elems > 0);
+
+ /*
+ * New key is more restrictive, and so replaces old key...
+ */
+ if (j != (BTEqualStrategyNumber - 1) ||
+ !(xform[j].skey->sk_flags & SK_SEARCHARRAY))
+ {
+ Assert(!array || array->scan_key == i);
+ xform[j].skey = cur;
+ xform[j].ikey = i;
+ xform[j].arrayidx = arrayidx;
+ }
+ else
+ {
+ /*
+ * ...unless we have to keep the old key because it's
+ * an array that rendered the new key redundant. We
+ * need to make sure that we don't throw away an array
+ * scan key. _bt_compare_scankey_args expects us to
+ * always keep arrays (and discard non-arrays).
+ */
+ Assert(j == (BTEqualStrategyNumber - 1));
+ Assert(xform[j].skey->sk_flags & SK_SEARCHARRAY);
+ Assert(xform[j].ikey == array->scan_key);
+ Assert(!(cur->sk_flags & SK_SEARCHARRAY));
+ }
+ }
else if (j == (BTEqualStrategyNumber - 1))
{
/* key == a && key == b, but a != b */
@@ -1002,22 +2921,130 @@ _bt_preprocess_keys(IndexScanDesc scan)
else
{
/*
- * We can't determine which key is more restrictive. Keep the
- * previous one in xform[j] and push this one directly to the
- * output array.
+ * We can't determine which key is more restrictive. Push
+ * xform[j] directly to the output array, then set xform[j] to
+ * the new scan key.
+ *
+ * Note: We do things this way around so that our arrays are
+ * always in the same order as their corresponding scan keys,
+ * even with incomplete opfamilies. _bt_advance_array_keys
+ * depends on this.
*/
ScanKey outkey = &outkeys[new_numberOfKeys++];
- memcpy(outkey, cur, sizeof(ScanKeyData));
+ memcpy(outkey, xform[j].skey, sizeof(ScanKeyData));
+ if (arrayKeyData)
+ keyDataMap[new_numberOfKeys - 1] = xform[j].ikey;
if (numberOfEqualCols == attno - 1)
_bt_mark_scankey_required(outkey);
+ xform[j].skey = cur;
+ xform[j].ikey = i;
+ xform[j].arrayidx = arrayidx;
}
}
}
so->numberOfKeys = new_numberOfKeys;
+
+ /*
+ * Now that we've output so->keyData[], and built a temporary mapping from
+ * so->keyData[] (output scan keys) to scan->keyData[] (input scan keys),
+ * fix each array->scan_key reference. (Also consolidates so->orderProc[]
+ * array, so it can be subscripted using so->keyData[]-wise offsets.)
+ */
+ if (arrayKeyData)
+ _bt_preprocess_array_keys_final(scan, keyDataMap);
+
+ /* Could pfree arrayKeyData/keyDataMap now, but not worth the cycles */
}
+#ifdef USE_ASSERT_CHECKING
+/*
+ * Verify that the scan's qual state matches what we expect at the point that
+ * _bt_start_prim_scan is about to start a just-scheduled new primitive scan.
+ *
+ * We enforce a rule against non-required array scan keys: they must start out
+ * with whatever element is the first for the scan's current scan direction.
+ * See _bt_rewind_nonrequired_arrays comments for an explanation.
+ */
+static bool
+_bt_verify_arrays_bt_first(IndexScanDesc scan, ScanDirection dir)
+{
+ BTScanOpaque so = (BTScanOpaque) scan->opaque;
+ int arrayidx = 0;
+
+ for (int ikey = 0; ikey < so->numberOfKeys; ikey++)
+ {
+ ScanKey cur = so->keyData + ikey;
+ BTArrayKeyInfo *array = NULL;
+ int first_elem_dir;
+
+ if (!(cur->sk_flags & SK_SEARCHARRAY) ||
+ cur->sk_strategy != BTEqualStrategyNumber)
+ continue;
+
+ array = &so->arrayKeys[arrayidx++];
+
+ if (((cur->sk_flags & SK_BT_REQFWD) && ScanDirectionIsForward(dir)) ||
+ ((cur->sk_flags & SK_BT_REQBKWD) && ScanDirectionIsBackward(dir)))
+ continue;
+
+ if (ScanDirectionIsForward(dir))
+ first_elem_dir = 0;
+ else
+ first_elem_dir = array->num_elems - 1;
+
+ if (array->cur_elem != first_elem_dir)
+ return false;
+ }
+
+ return _bt_verify_keys_with_arraykeys(scan);
+}
+
+/*
+ * Verify that the scan's "so->keyData[]" scan keys are in agreement with
+ * its array key state
+ */
+static bool
+_bt_verify_keys_with_arraykeys(IndexScanDesc scan)
+{
+ BTScanOpaque so = (BTScanOpaque) scan->opaque;
+ int last_sk_attno = InvalidAttrNumber,
+ arrayidx = 0;
+
+ if (!so->qual_ok)
+ return false;
+
+ for (int ikey = 0; ikey < so->numberOfKeys; ikey++)
+ {
+ ScanKey cur = so->keyData + ikey;
+ BTArrayKeyInfo *array;
+
+ if (cur->sk_strategy != BTEqualStrategyNumber ||
+ !(cur->sk_flags & SK_SEARCHARRAY))
+ continue;
+
+ array = &so->arrayKeys[arrayidx++];
+ if (array->scan_key != ikey)
+ return false;
+
+ if (array->num_elems <= 0)
+ return false;
+
+ if (cur->sk_argument != array->elem_values[array->cur_elem])
+ return false;
+ if (last_sk_attno > cur->sk_attno)
+ return false;
+ last_sk_attno = cur->sk_attno;
+ }
+
+ if (arrayidx != so->numArrayKeys)
+ return false;
+
+ return true;
+}
+#endif
+
/*
* Compare two scankey values using a specified operator.
*
@@ -1033,9 +3060,24 @@ _bt_preprocess_keys(IndexScanDesc scan)
* we store the operator result in *result and return true. We return false
* if the comparison could not be made.
*
+ * If either leftarg or rightarg are an array, we'll apply array-specific
+ * rules to determine which array elements are redundant on behalf of caller.
+ * It is up to our caller to save whichever of the two scan keys is the array,
+ * and discard the non-array scan key (the non-array scan key is guaranteed to
+ * be redundant with any complete opfamily). Caller isn't expected to call
+ * here with a pair of array scan keys provided we're dealing with a complete
+ * opfamily (_bt_preprocess_array_keys will merge array keys together to make
+ * sure of that).
+ *
+ * Note: we'll also shrink caller's array as needed to eliminate redundant
+ * array elements. One reason why caller should prefer to discard non-array
+ * scan keys is so that we'll have the opportunity to shrink the array
+ * multiple times, in multiple calls (for each of several other scan keys on
+ * the same index attribute).
+ *
* Note: op always points at the same ScanKey as either leftarg or rightarg.
- * Since we don't scribble on the scankeys, this aliasing should cause no
- * trouble.
+ * Since we don't scribble on the scankeys themselves, this aliasing should
+ * cause no trouble.
*
* Note: this routine needs to be insensitive to any DESC option applied
* to the index column. For example, "x < 4" is a tighter constraint than
@@ -1044,6 +3086,7 @@ _bt_preprocess_keys(IndexScanDesc scan)
static bool
_bt_compare_scankey_args(IndexScanDesc scan, ScanKey op,
ScanKey leftarg, ScanKey rightarg,
+ BTArrayKeyInfo *array, FmgrInfo *orderproc,
bool *result)
{
Relation rel = scan->indexRelation;
@@ -1117,6 +3160,46 @@ _bt_compare_scankey_args(IndexScanDesc scan, ScanKey op,
*/
Assert(leftarg->sk_attno == rightarg->sk_attno);
+ /*
+ * If either leftarg or rightarg are equality-type array scankeys, we need
+ * specialized handling (since by now we know that IS NULL wasn't used)
+ */
+ if (array)
+ {
+ bool leftarray,
+ rightarray;
+
+ leftarray = ((leftarg->sk_flags & SK_SEARCHARRAY) &&
+ leftarg->sk_strategy == BTEqualStrategyNumber);
+ rightarray = ((rightarg->sk_flags & SK_SEARCHARRAY) &&
+ rightarg->sk_strategy == BTEqualStrategyNumber);
+
+ /*
+ * _bt_preprocess_array_keys is responsible for merging together array
+ * scan keys, and will do so whenever the opfamily has the required
+ * cross-type support. If it failed to do that, we handle it just
+ * like the case where we can't make the comparison ourselves.
+ */
+ if (leftarray && rightarray)
+ {
+ /* Can't make the comparison */
+ *result = false; /* suppress compiler warnings */
+ return false;
+ }
+
+ /*
+ * Otherwise we need to determine if either one of leftarg or rightarg
+ * uses an array, then pass this through to a dedicated helper
+ * function.
+ */
+ if (leftarray)
+ return _bt_compare_array_scankey_args(scan, leftarg, rightarg,
+ orderproc, array, result);
+ else if (rightarray)
+ return _bt_compare_array_scankey_args(scan, rightarg, leftarg,
+ orderproc, array, result);
+ }
+
opcintype = rel->rd_opcintype[leftarg->sk_attno - 1];
/*
@@ -1351,60 +3434,194 @@ _bt_mark_scankey_required(ScanKey skey)
*
* Return true if so, false if not. If the tuple fails to pass the qual,
* we also determine whether there's any need to continue the scan beyond
- * this tuple, and set *continuescan accordingly. See comments for
+ * this tuple, and set pstate.continuescan accordingly. See comments for
* _bt_preprocess_keys(), above, about how this is done.
*
* Forward scan callers can pass a high key tuple in the hopes of having
* us set *continuescan to false, and avoiding an unnecessary visit to
* the page to the right.
*
+ * Advances the scan's array keys when necessary for arrayKeys=true callers.
+ * Caller can avoid all array related side-effects when calling just to do a
+ * page continuescan precheck -- pass arrayKeys=false for that. Scans without
+ * any arrays keys must always pass arrayKeys=false.
+ *
+ * Also stops and starts primitive index scans for arrayKeys=true callers.
+ * Scans with array keys are required to set up page state that helps us with
+ * this. The page's finaltup tuple (the page high key for a forward scan, or
+ * the page's first non-pivot tuple for a backward scan) must be set in
+ * pstate.finaltup ahead of the first call here for the page (or possibly the
+ * first call after an initial continuescan-setting page precheck call). Set
+ * this to NULL for rightmost page (or the leftmost page for backwards scans).
+ *
* scan: index scan descriptor (containing a search-type scankey)
+ * pstate: page level input and output parameters
+ * arrayKeys: should we advance the scan's array keys if necessary?
* tuple: index tuple to test
* tupnatts: number of attributes in tupnatts (high key may be truncated)
- * dir: direction we are scanning in
- * continuescan: output parameter (will be set correctly in all cases)
- * continuescanPrechecked: indicates that *continuescan flag is known to
- * be true for the last item on the page
- * haveFirstMatch: indicates that we already have at least one match
- * in the current page
*/
bool
-_bt_checkkeys(IndexScanDesc scan, IndexTuple tuple, int tupnatts,
- ScanDirection dir, bool *continuescan,
- bool continuescanPrechecked, bool haveFirstMatch)
+_bt_checkkeys(IndexScanDesc scan, BTReadPageState *pstate, bool arrayKeys,
+ IndexTuple tuple, int tupnatts)
{
- TupleDesc tupdesc;
- BTScanOpaque so;
- int keysz;
- int ikey;
- ScanKey key;
+ TupleDesc tupdesc = RelationGetDescr(scan->indexRelation);
+ BTScanOpaque so = (BTScanOpaque) scan->opaque;
+ ScanDirection dir = pstate->dir;
+ int ikey = 0;
+ bool res;
Assert(BTreeTupleGetNAtts(tuple, scan->indexRelation) == tupnatts);
+ Assert(!arrayKeys || (so->primScanDir == dir && so->arrayKeys));
+ Assert(!so->scanBehind || ScanDirectionIsForward(dir));
+ Assert(!so->needPrimScan);
+ res = _bt_check_compare(dir, so, tuple, tupnatts, tupdesc,
+ arrayKeys, pstate->prechecked, pstate->firstmatch,
+ &pstate->continuescan, &ikey);
+
+#ifdef USE_ASSERT_CHECKING
+ if (pstate->prechecked || pstate->firstmatch)
+ {
+ bool dcontinuescan;
+ int dikey = 0;
+
+ Assert(res == _bt_check_compare(dir, so, tuple, tupnatts, tupdesc,
+ arrayKeys, false, false,
+ &dcontinuescan, &dikey));
+ Assert(dcontinuescan == pstate->continuescan && ikey == dikey);
+ }
+#endif
+
+ /*
+ * Only one _bt_check_compare call is required in the common case where
+ * there are no equality strategy array scan keys. Otherwise we can only
+ * accept _bt_check_compare's answer unreservedly when it didn't set
+ * pstate.continuescan=false.
+ */
+ if (!arrayKeys || pstate->continuescan)
+ return res;
+
+ /*
+ * _bt_check_compare call set continuescan=false in the presence of
+ * equality type array keys. This could mean that the tuple is just past
+ * the end of matches for the current array keys.
+ *
+ * It's also possible that the scan is still _before_ the _start_ of
+ * tuples matching the current set of array keys. Check for that first.
+ */
+ if (_bt_tuple_before_array_skeys(scan, dir, tuple, tupnatts, tupdesc,
+ true, ikey, NULL))
+ {
+ /*
+ * Tuple is still before the start of matches according to the scan's
+ * required array keys (according to _all_ of its required equality
+ * strategy keys, actually).
+ *
+ * Note: we will end up here repeatedly given a group of tuples > the
+ * previous array keys and < the now-current keys (though only when
+ * _bt_advance_array_keys determined that key space relevant to the
+ * scan covers some of the page's remaining unscanned tuples).
+ *
+ * _bt_advance_array_keys occasionally sets so->scanBehind to signal
+ * that the scan's current position/tuples might be significantly
+ * behind (multiple pages behind) its current array keys. When this
+ * happens, we check the page finaltup ourselves. We'll start a new
+ * primitive index scan on our own if it turns out that the scan isn't
+ * now on a page that has at least some tuples covered by the key
+ * space of the arrays.
+ *
+ * This scheme allows _bt_advance_array_keys to optimistically assume
+ * that the scan will find array key matches for any truncated
+ * finaltup attributes once the scan reaches the right sibling page
+ * (only the untruncated prefix has to match the scan's array keys).
+ */
+ Assert(!so->scanBehind ||
+ so->keyData[ikey].sk_strategy == BTEqualStrategyNumber);
+ if (unlikely(so->scanBehind) && pstate->finaltup &&
+ _bt_tuple_before_array_skeys(scan, dir, pstate->finaltup,
+ BTreeTupleGetNAtts(pstate->finaltup,
+ scan->indexRelation),
+ tupdesc, false, 0, NULL))
+ {
+ /* Cut our losses -- start a new primitive index scan now */
+ pstate->continuescan = false;
+ so->needPrimScan = true;
+ }
+ else
+ {
+ /* Override _bt_check_compare, continue primitive scan */
+ pstate->continuescan = true;
+ }
+
+ /* This indextuple doesn't match the current qual, in any case */
+ return false;
+ }
+
+ /*
+ * Caller's tuple is >= the current set of array keys and other equality
+ * constraint scan keys (or <= if this is a backwards scan). It's now
+ * clear that we _must_ advance any required array keys in lockstep with
+ * the scan (unless the required array keys become exhausted instead, or
+ * unless the ikey trigger corresponds to a non-required array scan key).
+ *
+ * Note: we might advance the required arrays when all existing keys are
+ * already equal to the values from the tuple at this point. See comments
+ * above _bt_advance_array_keys about inequality driven array advancement.
+ */
+ return _bt_advance_array_keys(scan, pstate, tuple, tupnatts, tupdesc, ikey);
+}
+
+/*
+ * Test whether an indextuple satisfies current scan condition.
+ *
+ * Return true if so, false if not. If not, also sets *continuescan to false
+ * when it's also not possible for any later tuples to pass the current qual
+ * (with the scan's current set of array keys, in the current scan direction),
+ * in addition to setting *ikey to the so->keyData[] subscript/offset for the
+ * unsatisfied scan key (needed when caller must consider advancing the scan's
+ * array keys).
+ *
+ * This is a subroutine for _bt_checkkeys. It is written with the assumption
+ * that reaching the end of each distinct set of array keys ends the ongoing
+ * primitive index scan. It is up to our caller to override that initial
+ * determination when it makes more sense to advance the array keys and
+ * continue with further tuples from the same leaf page.
+ *
+ * Note: we set *continuescan to false for arrayKeys=true callers in the event
+ * of an unsatisfied non-required array equality scan key, despite the fact
+ * that it's never safe to end the current primitive index scan when that
+ * happens. Caller will still need to consider "advancing" the array keys
+ * (which isn't all that different to what happens to truly required arrays).
+ * Caller _must_ unset continuescan once non-required arrays have advanced.
+ * Callers that pass arrayKeys=false won't get this behavior, which is useful
+ * when the focus is on whether the scan's required scan keys are satisfied.
+ */
+static bool
+_bt_check_compare(ScanDirection dir, BTScanOpaque so,
+ IndexTuple tuple, int tupnatts, TupleDesc tupdesc,
+ bool arrayKeys, bool prechecked, bool firstmatch,
+ bool *continuescan, int *ikey)
+{
*continuescan = true; /* default assumption */
- tupdesc = RelationGetDescr(scan->indexRelation);
- so = (BTScanOpaque) scan->opaque;
- keysz = so->numberOfKeys;
-
- for (key = so->keyData, ikey = 0; ikey < keysz; key++, ikey++)
+ for (; *ikey < so->numberOfKeys; (*ikey)++)
{
+ ScanKey key = so->keyData + *ikey;
Datum datum;
bool isNull;
- Datum test;
bool requiredSameDir = false,
- requiredOppositeDir = false;
+ requiredOppositeDirOnly = false;
/*
- * Check if the key is required for ordered scan in the same or
- * opposite direction. Save as flag variables for future usage.
+ * Check if the key is required in the current scan direction, in the
+ * opposite scan direction _only_, or in neither direction
*/
if (((key->sk_flags & SK_BT_REQFWD) && ScanDirectionIsForward(dir)) ||
((key->sk_flags & SK_BT_REQBKWD) && ScanDirectionIsBackward(dir)))
requiredSameDir = true;
else if (((key->sk_flags & SK_BT_REQFWD) && ScanDirectionIsBackward(dir)) ||
((key->sk_flags & SK_BT_REQBKWD) && ScanDirectionIsForward(dir)))
- requiredOppositeDir = true;
+ requiredOppositeDirOnly = true;
/*
* If the caller told us the *continuescan flag is known to be true
@@ -1422,8 +3639,9 @@ _bt_checkkeys(IndexScanDesc scan, IndexTuple tuple, int tupnatts,
* Both cases above work except for the row keys, where NULLs could be
* found in the middle of matching values.
*/
- if ((requiredSameDir || (requiredOppositeDir && haveFirstMatch)) &&
- !(key->sk_flags & SK_ROW_HEADER) && continuescanPrechecked)
+ if (prechecked &&
+ (requiredSameDir || (requiredOppositeDirOnly && firstmatch)) &&
+ !(key->sk_flags & SK_ROW_HEADER))
continue;
if (key->sk_attno > tupnatts)
@@ -1434,7 +3652,6 @@ _bt_checkkeys(IndexScanDesc scan, IndexTuple tuple, int tupnatts,
* right could be any possible value. Assume that truncated
* attribute passes the qual.
*/
- Assert(ScanDirectionIsForward(dir));
Assert(BTreeTupleIsPivot(tuple));
continue;
}
@@ -1495,6 +3712,8 @@ _bt_checkkeys(IndexScanDesc scan, IndexTuple tuple, int tupnatts,
* because it's not possible for any future tuples to pass. On
* a forward scan, however, we must keep going, because we may
* have initially positioned to the start of the index.
+ * (_bt_advance_array_keys also relies on this behavior during
+ * forward scans.)
*/
if ((key->sk_flags & (SK_BT_REQFWD | SK_BT_REQBKWD)) &&
ScanDirectionIsBackward(dir))
@@ -1511,6 +3730,8 @@ _bt_checkkeys(IndexScanDesc scan, IndexTuple tuple, int tupnatts,
* because it's not possible for any future tuples to pass. On
* a backward scan, however, we must keep going, because we
* may have initially positioned to the end of the index.
+ * (_bt_advance_array_keys also relies on this behavior during
+ * backward scans.)
*/
if ((key->sk_flags & (SK_BT_REQFWD | SK_BT_REQBKWD)) &&
ScanDirectionIsForward(dir))
@@ -1524,24 +3745,15 @@ _bt_checkkeys(IndexScanDesc scan, IndexTuple tuple, int tupnatts,
}
/*
- * Apply the key-checking function. When the key is required for the
- * opposite direction scan, it must be already satisfied as soon as
- * there is already match on the page. Except for the NULLs checking,
- * which have already done above.
+ * Apply the key-checking function, though only if we must.
+ *
+ * When a key is required in the opposite-of-scan direction _only_,
+ * then it must already be satisfied if firstmatch=true indicates that
+ * an earlier tuple from this same page satisfied it earlier on.
*/
- if (!(requiredOppositeDir && haveFirstMatch))
- {
- test = FunctionCall2Coll(&key->sk_func, key->sk_collation,
- datum, key->sk_argument);
- }
- else
- {
- test = true;
- Assert(test == FunctionCall2Coll(&key->sk_func, key->sk_collation,
- datum, key->sk_argument));
- }
-
- if (!DatumGetBool(test))
+ if (!(requiredOppositeDirOnly && firstmatch) &&
+ !DatumGetBool(FunctionCall2Coll(&key->sk_func, key->sk_collation,
+ datum, key->sk_argument)))
{
/*
* Tuple fails this qual. If it's a required qual for the current
@@ -1556,6 +3768,14 @@ _bt_checkkeys(IndexScanDesc scan, IndexTuple tuple, int tupnatts,
if (requiredSameDir)
*continuescan = false;
+ /*
+ * Also set continuescan=false for non-required equality-type
+ * array keys that don't pass (during arrayKeys=true calls)
+ */
+ if (arrayKeys && (key->sk_flags & SK_SEARCHARRAY) &&
+ key->sk_strategy == BTEqualStrategyNumber)
+ *continuescan = false;
+
/*
* In any case, this indextuple doesn't match the qual.
*/
@@ -1574,7 +3794,7 @@ _bt_checkkeys(IndexScanDesc scan, IndexTuple tuple, int tupnatts,
* it's not possible for any future tuples in the current scan direction
* to pass the qual.
*
- * This is a subroutine for _bt_checkkeys, which see for more info.
+ * This is a subroutine for _bt_checkkeys/_bt_check_compare.
*/
static bool
_bt_check_rowcompare(ScanKey skey, IndexTuple tuple, int tupnatts,
@@ -1603,7 +3823,6 @@ _bt_check_rowcompare(ScanKey skey, IndexTuple tuple, int tupnatts,
* right could be any possible value. Assume that truncated
* attribute passes the qual.
*/
- Assert(ScanDirectionIsForward(dir));
Assert(BTreeTupleIsPivot(tuple));
cmpresult = 0;
if (subkey->sk_flags & SK_ROW_END)
@@ -1630,6 +3849,8 @@ _bt_check_rowcompare(ScanKey skey, IndexTuple tuple, int tupnatts,
* because it's not possible for any future tuples to pass. On
* a forward scan, however, we must keep going, because we may
* have initially positioned to the start of the index.
+ * (_bt_advance_array_keys also relies on this behavior during
+ * forward scans.)
*/
if ((subkey->sk_flags & (SK_BT_REQFWD | SK_BT_REQBKWD)) &&
ScanDirectionIsBackward(dir))
@@ -1646,6 +3867,8 @@ _bt_check_rowcompare(ScanKey skey, IndexTuple tuple, int tupnatts,
* because it's not possible for any future tuples to pass. On
* a backward scan, however, we must keep going, because we
* may have initially positioned to the end of the index.
+ * (_bt_advance_array_keys also relies on this behavior during
+ * backward scans.)
*/
if ((subkey->sk_flags & (SK_BT_REQFWD | SK_BT_REQBKWD)) &&
ScanDirectionIsForward(dir))
diff --git a/src/backend/optimizer/path/indxpath.c b/src/backend/optimizer/path/indxpath.c
index 32c6a8bbd..2230b1310 100644
--- a/src/backend/optimizer/path/indxpath.c
+++ b/src/backend/optimizer/path/indxpath.c
@@ -106,8 +106,7 @@ static List *build_index_paths(PlannerInfo *root, RelOptInfo *rel,
IndexOptInfo *index, IndexClauseSet *clauses,
bool useful_predicate,
ScanTypeControl scantype,
- bool *skip_nonnative_saop,
- bool *skip_lower_saop);
+ bool *skip_nonnative_saop);
static List *build_paths_for_OR(PlannerInfo *root, RelOptInfo *rel,
List *clauses, List *other_clauses);
static List *generate_bitmap_or_paths(PlannerInfo *root, RelOptInfo *rel,
@@ -706,8 +705,6 @@ eclass_already_used(EquivalenceClass *parent_ec, Relids oldrelids,
* index AM supports them natively, we should just include them in simple
* index paths. If not, we should exclude them while building simple index
* paths, and then make a separate attempt to include them in bitmap paths.
- * Furthermore, we should consider excluding lower-order ScalarArrayOpExpr
- * quals so as to create ordered paths.
*/
static void
get_index_paths(PlannerInfo *root, RelOptInfo *rel,
@@ -716,37 +713,17 @@ get_index_paths(PlannerInfo *root, RelOptInfo *rel,
{
List *indexpaths;
bool skip_nonnative_saop = false;
- bool skip_lower_saop = false;
ListCell *lc;
/*
* Build simple index paths using the clauses. Allow ScalarArrayOpExpr
- * clauses only if the index AM supports them natively, and skip any such
- * clauses for index columns after the first (so that we produce ordered
- * paths if possible).
+ * clauses only if the index AM supports them natively.
*/
indexpaths = build_index_paths(root, rel,
index, clauses,
index->predOK,
ST_ANYSCAN,
- &skip_nonnative_saop,
- &skip_lower_saop);
-
- /*
- * If we skipped any lower-order ScalarArrayOpExprs on an index with an AM
- * that supports them, then try again including those clauses. This will
- * produce paths with more selectivity but no ordering.
- */
- if (skip_lower_saop)
- {
- indexpaths = list_concat(indexpaths,
- build_index_paths(root, rel,
- index, clauses,
- index->predOK,
- ST_ANYSCAN,
- &skip_nonnative_saop,
- NULL));
- }
+ &skip_nonnative_saop);
/*
* Submit all the ones that can form plain IndexScan plans to add_path. (A
@@ -784,7 +761,6 @@ get_index_paths(PlannerInfo *root, RelOptInfo *rel,
index, clauses,
false,
ST_BITMAPSCAN,
- NULL,
NULL);
*bitindexpaths = list_concat(*bitindexpaths, indexpaths);
}
@@ -817,27 +793,19 @@ get_index_paths(PlannerInfo *root, RelOptInfo *rel,
* to true if we found any such clauses (caller must initialize the variable
* to false). If it's NULL, we do not ignore ScalarArrayOpExpr clauses.
*
- * If skip_lower_saop is non-NULL, we ignore ScalarArrayOpExpr clauses for
- * non-first index columns, and we set *skip_lower_saop to true if we found
- * any such clauses (caller must initialize the variable to false). If it's
- * NULL, we do not ignore non-first ScalarArrayOpExpr clauses, but they will
- * result in considering the scan's output to be unordered.
- *
* 'rel' is the index's heap relation
* 'index' is the index for which we want to generate paths
* 'clauses' is the collection of indexable clauses (IndexClause nodes)
* 'useful_predicate' indicates whether the index has a useful predicate
* 'scantype' indicates whether we need plain or bitmap scan support
* 'skip_nonnative_saop' indicates whether to accept SAOP if index AM doesn't
- * 'skip_lower_saop' indicates whether to accept non-first-column SAOP
*/
static List *
build_index_paths(PlannerInfo *root, RelOptInfo *rel,
IndexOptInfo *index, IndexClauseSet *clauses,
bool useful_predicate,
ScanTypeControl scantype,
- bool *skip_nonnative_saop,
- bool *skip_lower_saop)
+ bool *skip_nonnative_saop)
{
List *result = NIL;
IndexPath *ipath;
@@ -848,12 +816,13 @@ build_index_paths(PlannerInfo *root, RelOptInfo *rel,
List *orderbyclausecols;
List *index_pathkeys;
List *useful_pathkeys;
- bool found_lower_saop_clause;
bool pathkeys_possibly_useful;
bool index_is_ordered;
bool index_only_scan;
int indexcol;
+ Assert(skip_nonnative_saop != NULL || scantype == ST_BITMAPSCAN);
+
/*
* Check that index supports the desired scan type(s)
*/
@@ -880,19 +849,11 @@ build_index_paths(PlannerInfo *root, RelOptInfo *rel,
* on by btree and possibly other places.) The list can be empty, if the
* index AM allows that.
*
- * found_lower_saop_clause is set true if we accept a ScalarArrayOpExpr
- * index clause for a non-first index column. This prevents us from
- * assuming that the scan result is ordered. (Actually, the result is
- * still ordered if there are equality constraints for all earlier
- * columns, but it seems too expensive and non-modular for this code to be
- * aware of that refinement.)
- *
* We also build a Relids set showing which outer rels are required by the
* selected clauses. Any lateral_relids are included in that, but not
* otherwise accounted for.
*/
index_clauses = NIL;
- found_lower_saop_clause = false;
outer_relids = bms_copy(rel->lateral_relids);
for (indexcol = 0; indexcol < index->nkeycolumns; indexcol++)
{
@@ -903,30 +864,18 @@ build_index_paths(PlannerInfo *root, RelOptInfo *rel,
IndexClause *iclause = (IndexClause *) lfirst(lc);
RestrictInfo *rinfo = iclause->rinfo;
- /* We might need to omit ScalarArrayOpExpr clauses */
- if (IsA(rinfo->clause, ScalarArrayOpExpr))
+ if (skip_nonnative_saop && !index->amsearcharray &&
+ IsA(rinfo->clause, ScalarArrayOpExpr))
{
- if (!index->amsearcharray)
- {
- if (skip_nonnative_saop)
- {
- /* Ignore because not supported by index */
- *skip_nonnative_saop = true;
- continue;
- }
- /* Caller had better intend this only for bitmap scan */
- Assert(scantype == ST_BITMAPSCAN);
- }
- if (indexcol > 0)
- {
- if (skip_lower_saop)
- {
- /* Caller doesn't want to lose index ordering */
- *skip_lower_saop = true;
- continue;
- }
- found_lower_saop_clause = true;
- }
+ /*
+ * Caller asked us to generate IndexPaths that omit any
+ * ScalarArrayOpExpr clauses when the underlying index AM
+ * lacks native support.
+ *
+ * We must omit this clause (and tell caller about it).
+ */
+ *skip_nonnative_saop = true;
+ continue;
}
/* OK to include this clause */
@@ -956,11 +905,9 @@ build_index_paths(PlannerInfo *root, RelOptInfo *rel,
/*
* 2. Compute pathkeys describing index's ordering, if any, then see how
* many of them are actually useful for this query. This is not relevant
- * if we are only trying to build bitmap indexscans, nor if we have to
- * assume the scan is unordered.
+ * if we are only trying to build bitmap indexscans.
*/
pathkeys_possibly_useful = (scantype != ST_BITMAPSCAN &&
- !found_lower_saop_clause &&
has_useful_pathkeys(root, rel));
index_is_ordered = (index->sortopfamily != NULL);
if (index_is_ordered && pathkeys_possibly_useful)
@@ -1212,7 +1159,6 @@ build_paths_for_OR(PlannerInfo *root, RelOptInfo *rel,
index, &clauseset,
useful_predicate,
ST_BITMAPSCAN,
- NULL,
NULL);
result = list_concat(result, indexpaths);
}
diff --git a/src/backend/utils/adt/selfuncs.c b/src/backend/utils/adt/selfuncs.c
index cea777e9d..772dc664f 100644
--- a/src/backend/utils/adt/selfuncs.c
+++ b/src/backend/utils/adt/selfuncs.c
@@ -6557,8 +6557,6 @@ genericcostestimate(PlannerInfo *root,
double numIndexTuples;
double spc_random_page_cost;
double num_sa_scans;
- double num_outer_scans;
- double num_scans;
double qual_op_cost;
double qual_arg_cost;
List *selectivityQuals;
@@ -6573,7 +6571,7 @@ genericcostestimate(PlannerInfo *root,
/*
* Check for ScalarArrayOpExpr index quals, and estimate the number of
- * index scans that will be performed.
+ * primitive index scans that will be performed.
*/
num_sa_scans = 1;
foreach(l, indexQuals)
@@ -6603,19 +6601,8 @@ genericcostestimate(PlannerInfo *root,
*/
numIndexTuples = costs->numIndexTuples;
if (numIndexTuples <= 0.0)
- {
numIndexTuples = indexSelectivity * index->rel->tuples;
- /*
- * The above calculation counts all the tuples visited across all
- * scans induced by ScalarArrayOpExpr nodes. We want to consider the
- * average per-indexscan number, so adjust. This is a handy place to
- * round to integer, too. (If caller supplied tuple estimate, it's
- * responsible for handling these considerations.)
- */
- numIndexTuples = rint(numIndexTuples / num_sa_scans);
- }
-
/*
* We can bound the number of tuples by the index size in any case. Also,
* always estimate at least one tuple is touched, even when
@@ -6653,27 +6640,31 @@ genericcostestimate(PlannerInfo *root,
*
* The above calculations are all per-index-scan. However, if we are in a
* nestloop inner scan, we can expect the scan to be repeated (with
- * different search keys) for each row of the outer relation. Likewise,
- * ScalarArrayOpExpr quals result in multiple index scans. This creates
- * the potential for cache effects to reduce the number of disk page
- * fetches needed. We want to estimate the average per-scan I/O cost in
- * the presence of caching.
+ * different search keys) for each row of the outer relation. This
+ * creates the potential for cache effects to reduce the number of disk
+ * page fetches needed. We want to estimate the average per-scan I/O cost
+ * in the presence of caching.
*
* We use the Mackert-Lohman formula (see costsize.c for details) to
* estimate the total number of page fetches that occur. While this
* wasn't what it was designed for, it seems a reasonable model anyway.
* Note that we are counting pages not tuples anymore, so we take N = T =
* index size, as if there were one "tuple" per page.
+ *
+ * Note: we assume that there will be no repeat index page fetches across
+ * ScalarArrayOpExpr primitive scans from the same logical index scan.
+ * This is guaranteed to be true for btree indexes, but is very optimistic
+ * with index AMs that cannot natively execute ScalarArrayOpExpr quals.
+ * However, these same index AMs also accept our default pessimistic
+ * approach to counting num_sa_scans (btree caller caps this), so we don't
+ * expect the final indexTotalCost to be wildly over-optimistic.
*/
- num_outer_scans = loop_count;
- num_scans = num_sa_scans * num_outer_scans;
-
- if (num_scans > 1)
+ if (loop_count > 1)
{
double pages_fetched;
/* total page fetches ignoring cache effects */
- pages_fetched = numIndexPages * num_scans;
+ pages_fetched = numIndexPages * loop_count;
/* use Mackert and Lohman formula to adjust for cache effects */
pages_fetched = index_pages_fetched(pages_fetched,
@@ -6683,11 +6674,9 @@ genericcostestimate(PlannerInfo *root,
/*
* Now compute the total disk access cost, and then report a pro-rated
- * share for each outer scan. (Don't pro-rate for ScalarArrayOpExpr,
- * since that's internal to the indexscan.)
+ * share for each outer scan.
*/
- indexTotalCost = (pages_fetched * spc_random_page_cost)
- / num_outer_scans;
+ indexTotalCost = (pages_fetched * spc_random_page_cost) / loop_count;
}
else
{
@@ -6703,10 +6692,8 @@ genericcostestimate(PlannerInfo *root,
* evaluated once at the start of the scan to reduce them to runtime keys
* to pass to the index AM (see nodeIndexscan.c). We model the per-tuple
* CPU costs as cpu_index_tuple_cost plus one cpu_operator_cost per
- * indexqual operator. Because we have numIndexTuples as a per-scan
- * number, we have to multiply by num_sa_scans to get the correct result
- * for ScalarArrayOpExpr cases. Similarly add in costs for any index
- * ORDER BY expressions.
+ * indexqual operator. Similarly add in costs for any index ORDER BY
+ * expressions.
*
* Note: this neglects the possible costs of rechecking lossy operators.
* Detecting that that might be needed seems more expensive than it's
@@ -6719,7 +6706,7 @@ genericcostestimate(PlannerInfo *root,
indexStartupCost = qual_arg_cost;
indexTotalCost += qual_arg_cost;
- indexTotalCost += numIndexTuples * num_sa_scans * (cpu_index_tuple_cost + qual_op_cost);
+ indexTotalCost += numIndexTuples * (cpu_index_tuple_cost + qual_op_cost);
/*
* Generic assumption about index correlation: there isn't any.
@@ -6797,7 +6784,6 @@ btcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
bool eqQualHere;
bool found_saop;
bool found_is_null_op;
- double num_sa_scans;
ListCell *lc;
/*
@@ -6812,17 +6798,12 @@ btcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
*
* For a RowCompareExpr, we consider only the first column, just as
* rowcomparesel() does.
- *
- * If there's a ScalarArrayOpExpr in the quals, we'll actually perform N
- * index scans not one, but the ScalarArrayOpExpr's operator can be
- * considered to act the same as it normally does.
*/
indexBoundQuals = NIL;
indexcol = 0;
eqQualHere = false;
found_saop = false;
found_is_null_op = false;
- num_sa_scans = 1;
foreach(lc, path->indexclauses)
{
IndexClause *iclause = lfirst_node(IndexClause, lc);
@@ -6862,14 +6843,9 @@ btcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
else if (IsA(clause, ScalarArrayOpExpr))
{
ScalarArrayOpExpr *saop = (ScalarArrayOpExpr *) clause;
- Node *other_operand = (Node *) lsecond(saop->args);
- double alength = estimate_array_length(root, other_operand);
clause_op = saop->opno;
found_saop = true;
- /* count number of SA scans induced by indexBoundQuals only */
- if (alength > 1)
- num_sa_scans *= alength;
}
else if (IsA(clause, NullTest))
{
@@ -6929,13 +6905,6 @@ btcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
JOIN_INNER,
NULL);
numIndexTuples = btreeSelectivity * index->rel->tuples;
-
- /*
- * As in genericcostestimate(), we have to adjust for any
- * ScalarArrayOpExpr quals included in indexBoundQuals, and then round
- * to integer.
- */
- numIndexTuples = rint(numIndexTuples / num_sa_scans);
}
/*
@@ -6945,6 +6914,48 @@ btcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
genericcostestimate(root, path, loop_count, &costs);
+ /*
+ * Now compensate for btree's ability to efficiently execute scans with
+ * SAOP clauses.
+ *
+ * btree automatically combines individual ScalarArrayOpExpr primitive
+ * index scans whenever the tuples covered by the next set of array keys
+ * are close to tuples covered by the current set. This makes the final
+ * number of descents particularly difficult to estimate. However, btree
+ * scans never visit any single leaf page more than once. That puts a
+ * natural floor under the worst case number of descents.
+ *
+ * It's particularly important that we not wildly overestimate the number
+ * of descents needed for a clause list with several SAOPs -- the costs
+ * really aren't multiplicative in the way genericcostestimate expects. In
+ * general, most distinct combinations of SAOP keys will tend to not find
+ * any matching tuples. Furthermore, btree scans search for the next set
+ * of array keys using the next tuple in line, and so won't even need a
+ * direct comparison to eliminate most non-matching sets of array keys.
+ *
+ * Clamp the number of descents to the estimated number of leaf page
+ * visits. This is still fairly pessimistic, but tends to result in more
+ * accurate costing of scans with several SAOP clauses -- especially when
+ * each array has more than a few elements. The cost of adding additional
+ * array constants to a low-order SAOP column should saturate past a
+ * certain point (except where selectivity estimates continue to shift).
+ *
+ * Also clamp the number of descents to 1/3 the number of index pages.
+ * This avoids implausibly high estimates with low selectivity paths,
+ * where scans frequently require no more than one or two descents.
+ *
+ * XXX Ideally, we'd also account for the fact that non-boundary SAOP
+ * clause quals (which the B-Tree code uses "non-required" scan keys for)
+ * won't actually contribute to the total number of descents of the index.
+ * This would require pushing down more context into genericcostestimate.
+ */
+ if (costs.num_sa_scans > 1)
+ {
+ costs.num_sa_scans = Min(costs.num_sa_scans, costs.numIndexPages);
+ costs.num_sa_scans = Min(costs.num_sa_scans, index->pages / 3);
+ costs.num_sa_scans = Max(costs.num_sa_scans, 1);
+ }
+
/*
* Add a CPU-cost component to represent the costs of initial btree
* descent. We don't charge any I/O cost for touching upper btree levels,
@@ -6952,9 +6963,9 @@ btcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
* comparisons to descend a btree of N leaf tuples. We charge one
* cpu_operator_cost per comparison.
*
- * If there are ScalarArrayOpExprs, charge this once per SA scan. The
- * ones after the first one are not startup cost so far as the overall
- * plan is concerned, so add them only to "total" cost.
+ * If there are ScalarArrayOpExprs, charge this once per estimated
+ * primitive SA scan. The ones after the first one are not startup cost
+ * so far as the overall plan goes, so just add them to "total" cost.
*/
if (index->tuples > 1) /* avoid computing log(0) */
{
@@ -6971,7 +6982,8 @@ btcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
* in cases where only a single leaf page is expected to be visited. This
* cost is somewhat arbitrarily set at 50x cpu_operator_cost per page
* touched. The number of such pages is btree tree height plus one (ie,
- * we charge for the leaf page too). As above, charge once per SA scan.
+ * we charge for the leaf page too). As above, charge once per estimated
+ * primitive SA scan.
*/
descentCost = (index->tree_height + 1) * DEFAULT_PAGE_CPU_MULTIPLIER * cpu_operator_cost;
costs.indexStartupCost += descentCost;
diff --git a/doc/src/sgml/monitoring.sgml b/doc/src/sgml/monitoring.sgml
index 8736eac28..e49a4c0c1 100644
--- a/doc/src/sgml/monitoring.sgml
+++ b/doc/src/sgml/monitoring.sgml
@@ -4064,6 +4064,19 @@ description | Waiting for a newly initialized WAL file to reach durable storage
</para>
</note>
+ <note>
+ <para>
+ Queries that use certain <acronym>SQL</acronym> constructs to search for
+ rows matching any value out of a list or array of multiple scalar values
+ (see <xref linkend="functions-comparisons"/>) perform multiple
+ <quote>primitive</quote> index scans (up to one primitive scan per scalar
+ value) during query execution. Each internal primitive index scan
+ increments <structname>pg_stat_all_indexes</structname>.<structfield>idx_scan</structfield>,
+ so it's possible for the count of index scans to significantly exceed the
+ total number of index scan executor node executions.
+ </para>
+ </note>
+
</sect2>
<sect2 id="monitoring-pg-statio-all-tables-view">
diff --git a/src/test/regress/expected/create_index.out b/src/test/regress/expected/create_index.out
index 70ab47a92..ef7b84662 100644
--- a/src/test/regress/expected/create_index.out
+++ b/src/test/regress/expected/create_index.out
@@ -1910,7 +1910,7 @@ SELECT count(*) FROM dupindexcols
(1 row)
--
--- Check ordering of =ANY indexqual results (bug in 9.2.0)
+-- Check that index scans with =ANY indexquals return rows in index order
--
explain (costs off)
SELECT unique1 FROM tenk1
@@ -1932,16 +1932,16 @@ ORDER BY unique1;
42
(3 rows)
+-- Non-required array scan key on "tenthous":
explain (costs off)
SELECT thousand, tenthous FROM tenk1
WHERE thousand < 2 AND tenthous IN (1001,3000)
ORDER BY thousand;
- QUERY PLAN
--------------------------------------------------------
+ QUERY PLAN
+--------------------------------------------------------------------------------
Index Only Scan using tenk1_thous_tenthous on tenk1
- Index Cond: (thousand < 2)
- Filter: (tenthous = ANY ('{1001,3000}'::integer[]))
-(3 rows)
+ Index Cond: ((thousand < 2) AND (tenthous = ANY ('{1001,3000}'::integer[])))
+(2 rows)
SELECT thousand, tenthous FROM tenk1
WHERE thousand < 2 AND tenthous IN (1001,3000)
@@ -1952,29 +1952,26 @@ ORDER BY thousand;
1 | 1001
(2 rows)
-SET enable_indexonlyscan = OFF;
+-- Non-required array scan key on "tenthous", backward scan:
explain (costs off)
SELECT thousand, tenthous FROM tenk1
WHERE thousand < 2 AND tenthous IN (1001,3000)
-ORDER BY thousand;
- QUERY PLAN
---------------------------------------------------------------------------------------
- Sort
- Sort Key: thousand
- -> Index Scan using tenk1_thous_tenthous on tenk1
- Index Cond: ((thousand < 2) AND (tenthous = ANY ('{1001,3000}'::integer[])))
-(4 rows)
+ORDER BY thousand DESC, tenthous DESC;
+ QUERY PLAN
+--------------------------------------------------------------------------------
+ Index Only Scan Backward using tenk1_thous_tenthous on tenk1
+ Index Cond: ((thousand < 2) AND (tenthous = ANY ('{1001,3000}'::integer[])))
+(2 rows)
SELECT thousand, tenthous FROM tenk1
WHERE thousand < 2 AND tenthous IN (1001,3000)
-ORDER BY thousand;
+ORDER BY thousand DESC, tenthous DESC;
thousand | tenthous
----------+----------
- 0 | 3000
1 | 1001
+ 0 | 3000
(2 rows)
-RESET enable_indexonlyscan;
--
-- Check elimination of constant-NULL subexpressions
--
diff --git a/src/test/regress/expected/join.out b/src/test/regress/expected/join.out
index 960540002..a031d2341 100644
--- a/src/test/regress/expected/join.out
+++ b/src/test/regress/expected/join.out
@@ -8880,10 +8880,9 @@ where j1.id1 % 1000 = 1 and j2.id1 % 1000 = 1 and j2.id1 >= any (array[1,5]);
Merge Cond: (j1.id1 = j2.id1)
Join Filter: (j2.id2 = j1.id2)
-> Index Scan using j1_id1_idx on j1
- -> Index Only Scan using j2_pkey on j2
+ -> Index Scan using j2_id1_idx on j2
Index Cond: (id1 >= ANY ('{1,5}'::integer[]))
- Filter: ((id1 % 1000) = 1)
-(7 rows)
+(6 rows)
select * from j1
inner join j2 on j1.id1 = j2.id1 and j1.id2 = j2.id2
diff --git a/src/test/regress/sql/create_index.sql b/src/test/regress/sql/create_index.sql
index d49ce9f30..90a33795d 100644
--- a/src/test/regress/sql/create_index.sql
+++ b/src/test/regress/sql/create_index.sql
@@ -753,7 +753,7 @@ SELECT count(*) FROM dupindexcols
WHERE f1 BETWEEN 'WA' AND 'ZZZ' and id < 1000 and f1 ~<~ 'YX';
--
--- Check ordering of =ANY indexqual results (bug in 9.2.0)
+-- Check that index scans with =ANY indexquals return rows in index order
--
explain (costs off)
@@ -765,6 +765,7 @@ SELECT unique1 FROM tenk1
WHERE unique1 IN (1,42,7)
ORDER BY unique1;
+-- Non-required array scan key on "tenthous":
explain (costs off)
SELECT thousand, tenthous FROM tenk1
WHERE thousand < 2 AND tenthous IN (1001,3000)
@@ -774,18 +775,15 @@ SELECT thousand, tenthous FROM tenk1
WHERE thousand < 2 AND tenthous IN (1001,3000)
ORDER BY thousand;
-SET enable_indexonlyscan = OFF;
-
+-- Non-required array scan key on "tenthous", backward scan:
explain (costs off)
SELECT thousand, tenthous FROM tenk1
WHERE thousand < 2 AND tenthous IN (1001,3000)
-ORDER BY thousand;
+ORDER BY thousand DESC, tenthous DESC;
SELECT thousand, tenthous FROM tenk1
WHERE thousand < 2 AND tenthous IN (1001,3000)
-ORDER BY thousand;
-
-RESET enable_indexonlyscan;
+ORDER BY thousand DESC, tenthous DESC;
--
-- Check elimination of constant-NULL subexpressions
--
2.43.0