mirror of
https://github.com/FirebirdSQL/firebird.git
synced 2025-01-22 22:43:03 +01:00
Minor refactoring. Reworked fix for #3218 (Optimizer fails applying stream-local predicates before merging), now it covers both cases mentioned in the ticket.
This commit is contained in:
Dmitry Yemanov
parent
343c1f97df
commit
ae1c85f8f8
@ -778,6 +778,9 @@ RecordSource* Optimizer::compile(BoolExprNodeStack* parentStack)
|
||||
outerStreams.join(localStreams);
|
||||
}
|
||||
|
||||
// Apply local booleans, if any
|
||||
rsb = applyLocalBoolean(rsb, localStreams);
|
||||
|
||||
const auto river = FB_NEW_POOL(getPool()) River(csb, rsb, node, localStreams);
|
||||
river->deactivate(csb);
|
||||
rivers.add(river);
|
||||
@ -858,24 +861,24 @@ RecordSource* Optimizer::compile(BoolExprNodeStack* parentStack)
|
||||
// no merge is made between a new cross river and the
|
||||
// currently active rivers. Where in the new cross river
|
||||
// a stream depends (index) on the active rivers.
|
||||
StreamList dependent_streams, free_streams;
|
||||
findDependentStreams(joinStreams, dependent_streams, free_streams);
|
||||
StreamList dependentStreams, freeStreams;
|
||||
findDependentStreams(joinStreams, dependentStreams, freeStreams);
|
||||
|
||||
// If we have dependent and free streams then we can't rely on
|
||||
// the sort node to be used for index navigation
|
||||
if (dependent_streams.hasData() && free_streams.hasData())
|
||||
if (dependentStreams.hasData() && freeStreams.hasData())
|
||||
{
|
||||
sort = nullptr;
|
||||
sortCanBeUsed = false;
|
||||
}
|
||||
|
||||
if (dependent_streams.hasData())
|
||||
if (dependentStreams.hasData())
|
||||
{
|
||||
// Copy free streams
|
||||
joinStreams.assign(free_streams);
|
||||
joinStreams.assign(freeStreams);
|
||||
|
||||
// Make rivers from the dependent streams
|
||||
generateInnerJoin(dependent_streams, rivers, &sort, rse->rse_plan);
|
||||
generateInnerJoin(dependentStreams, rivers, &sort, rse->rse_plan);
|
||||
|
||||
// Generate one river which holds a cross join rsb between
|
||||
// all currently available rivers
|
||||
@ -885,7 +888,7 @@ RecordSource* Optimizer::compile(BoolExprNodeStack* parentStack)
|
||||
}
|
||||
else
|
||||
{
|
||||
if (free_streams.hasData())
|
||||
if (freeStreams.hasData())
|
||||
{
|
||||
// Deactivate streams from rivers on stack, because
|
||||
// the remaining streams don't have any indexed relationship with them
|
||||
@ -958,7 +961,7 @@ RecordSource* Optimizer::compile(BoolExprNodeStack* parentStack)
|
||||
rsb = FB_NEW_POOL(getPool()) FirstRowsStream(csb, rsb, rse->rse_first);
|
||||
|
||||
if (rse->flags & RseNode::FLAG_SINGULAR)
|
||||
rsb = FB_NEW_POOL(*tdbb->getDefaultPool()) SingularStream(csb, rsb);
|
||||
rsb = FB_NEW_POOL(getPool()) SingularStream(csb, rsb);
|
||||
|
||||
if (rse->flags & RseNode::FLAG_WRITELOCK)
|
||||
{
|
||||
@ -974,11 +977,11 @@ RecordSource* Optimizer::compile(BoolExprNodeStack* parentStack)
|
||||
SCL_update, obj_relations, tail->csb_relation->rel_name);
|
||||
}
|
||||
|
||||
rsb = FB_NEW_POOL(*tdbb->getDefaultPool()) LockedStream(csb, rsb);
|
||||
rsb = FB_NEW_POOL(getPool()) LockedStream(csb, rsb);
|
||||
}
|
||||
|
||||
if (rse->flags & RseNode::FLAG_SCROLLABLE)
|
||||
rsb = FB_NEW_POOL(*tdbb->getDefaultPool()) BufferedStream(csb, rsb);
|
||||
rsb = FB_NEW_POOL(getPool()) BufferedStream(csb, rsb);
|
||||
|
||||
return rsb;
|
||||
}
|
||||
@ -1455,12 +1458,13 @@ SortedStream* Optimizer::generateSort(const StreamList& streams,
|
||||
// Find conjuncts local to the given river and compose an appropriate filter
|
||||
//
|
||||
|
||||
RecordSource* Optimizer::applyLocalBoolean(const River* river)
|
||||
RecordSource* Optimizer::applyLocalBoolean(RecordSource* rsb, const StreamList& streams)
|
||||
{
|
||||
StreamStateHolder stateHolder(csb);
|
||||
stateHolder.deactivate();
|
||||
StreamStateHolder globalHolder(csb);
|
||||
globalHolder.deactivate();
|
||||
|
||||
river->activate(csb);
|
||||
StreamStateHolder localHolder(csb, streams);
|
||||
localHolder.activate(csb);
|
||||
|
||||
BoolExprNode* boolean = nullptr;
|
||||
double selectivity = MAXIMUM_SELECTIVITY;
|
||||
@ -1479,7 +1483,6 @@ RecordSource* Optimizer::applyLocalBoolean(const River* river)
|
||||
}
|
||||
}
|
||||
|
||||
const auto rsb = river->getRecordSource();
|
||||
return boolean ? FB_NEW_POOL(getPool()) FilteredStream(csb, rsb, boolean, selectivity) : rsb;
|
||||
}
|
||||
|
||||
@ -2203,7 +2206,7 @@ void Optimizer::formRivers(const StreamList& streams,
|
||||
// If the whole things is a moby no-op, return false.
|
||||
//
|
||||
|
||||
bool Optimizer::generateEquiJoin(RiverList& org_rivers)
|
||||
bool Optimizer::generateEquiJoin(RiverList& orgRivers)
|
||||
{
|
||||
ULONG selected_rivers[OPT_STREAM_BITS], selected_rivers2[OPT_STREAM_BITS];
|
||||
ValueExprNode** eq_class;
|
||||
@ -2212,13 +2215,13 @@ bool Optimizer::generateEquiJoin(RiverList& org_rivers)
|
||||
// a scratch block large enough to hold values to compute equality
|
||||
// classes.
|
||||
|
||||
const unsigned cnt = (unsigned) org_rivers.getCount();
|
||||
const unsigned orgCount = (unsigned) orgRivers.getCount();
|
||||
|
||||
if (cnt < 2)
|
||||
if (orgCount < 2)
|
||||
return false;
|
||||
|
||||
HalfStaticArray<ValueExprNode*, OPT_STATIC_ITEMS> scratch;
|
||||
scratch.grow(baseConjuncts * cnt);
|
||||
scratch.grow(baseConjuncts * orgCount);
|
||||
ValueExprNode** classes = scratch.begin();
|
||||
|
||||
// Compute equivalence classes among streams. This involves finding groups
|
||||
@ -2231,47 +2234,15 @@ bool Optimizer::generateEquiJoin(RiverList& org_rivers)
|
||||
if (iter & CONJUNCT_USED)
|
||||
continue;
|
||||
|
||||
const auto cmpNode = nodeAs<ComparativeBoolNode>(*iter);
|
||||
NestConst<ValueExprNode> node1;
|
||||
NestConst<ValueExprNode> node2;
|
||||
|
||||
if (!cmpNode || (cmpNode->blrOp != blr_eql && cmpNode->blrOp != blr_equiv))
|
||||
if (!getEquiJoinKeys(*iter, &node1, &node2, true))
|
||||
continue;
|
||||
|
||||
ValueExprNode* node1 = cmpNode->arg1;
|
||||
ValueExprNode* node2 = cmpNode->arg2;
|
||||
|
||||
dsc result, desc1, desc2;
|
||||
node1->getDesc(tdbb, csb, &desc1);
|
||||
node2->getDesc(tdbb, csb, &desc2);
|
||||
|
||||
// Ensure that arguments can be compared in the binary form
|
||||
if (!CVT2_get_binary_comparable_desc(&result, &desc1, &desc2))
|
||||
continue;
|
||||
|
||||
// Cast the arguments, if required
|
||||
if (!DSC_EQUIV(&result, &desc1, true) || !DSC_EQUIV(&result, &desc2, true))
|
||||
for (unsigned i = 0; i < orgRivers.getCount(); i++)
|
||||
{
|
||||
if (!DSC_EQUIV(&result, &desc1, true))
|
||||
{
|
||||
const auto cast = FB_NEW_POOL(getPool()) CastNode(getPool());
|
||||
cast->source = node1;
|
||||
cast->castDesc = result;
|
||||
cast->impureOffset = csb->allocImpure<impure_value>();
|
||||
node1 = cast;
|
||||
}
|
||||
|
||||
if (!DSC_EQUIV(&result, &desc2, true))
|
||||
{
|
||||
const auto cast = FB_NEW_POOL(getPool()) CastNode(getPool());
|
||||
cast->source = node2;
|
||||
cast->castDesc = result;
|
||||
cast->impureOffset = csb->allocImpure<impure_value>();
|
||||
node2 = cast;
|
||||
}
|
||||
}
|
||||
|
||||
for (unsigned i = 0; i < org_rivers.getCount(); i++)
|
||||
{
|
||||
const auto river1 = org_rivers[i];
|
||||
const auto river1 = orgRivers[i];
|
||||
|
||||
if (!river1->isReferenced(node1))
|
||||
{
|
||||
@ -2283,13 +2254,13 @@ bool Optimizer::generateEquiJoin(RiverList& org_rivers)
|
||||
node2 = temp;
|
||||
}
|
||||
|
||||
for (unsigned j = i + 1; j < org_rivers.getCount(); j++)
|
||||
for (unsigned j = i + 1; j < orgRivers.getCount(); j++)
|
||||
{
|
||||
const auto river2 = org_rivers[j];
|
||||
const auto river2 = orgRivers[j];
|
||||
|
||||
if (river2->isReferenced(node2))
|
||||
{
|
||||
for (eq_class = classes; eq_class < last_class; eq_class += cnt)
|
||||
for (eq_class = classes; eq_class < last_class; eq_class += orgCount)
|
||||
{
|
||||
if (fieldEqual(node1, classes[i]) ||
|
||||
fieldEqual(node2, classes[j]))
|
||||
@ -2302,7 +2273,7 @@ bool Optimizer::generateEquiJoin(RiverList& org_rivers)
|
||||
eq_class[j] = node2;
|
||||
|
||||
if (eq_class == last_class)
|
||||
last_class += cnt;
|
||||
last_class += orgCount;
|
||||
}
|
||||
}
|
||||
}
|
||||
@ -2312,23 +2283,23 @@ bool Optimizer::generateEquiJoin(RiverList& org_rivers)
|
||||
// Obviously, if the set of classes is empty, return false
|
||||
// to indicate that nothing could be done.
|
||||
|
||||
unsigned river_cnt = 0;
|
||||
HalfStaticArray<ValueExprNode**, OPT_STATIC_ITEMS> selected_classes(cnt);
|
||||
unsigned riverCount = 0;
|
||||
HalfStaticArray<ValueExprNode**, OPT_STATIC_ITEMS> selected_classes(orgCount);
|
||||
|
||||
for (eq_class = classes; eq_class < last_class; eq_class += cnt)
|
||||
for (eq_class = classes; eq_class < last_class; eq_class += orgCount)
|
||||
{
|
||||
unsigned i = getRiverCount(cnt, eq_class);
|
||||
unsigned i = getRiverCount(orgCount, eq_class);
|
||||
|
||||
if (i > river_cnt)
|
||||
if (i > riverCount)
|
||||
{
|
||||
river_cnt = i;
|
||||
riverCount = i;
|
||||
selected_classes.shrink(0);
|
||||
selected_classes.add(eq_class);
|
||||
classMask(cnt, eq_class, selected_rivers);
|
||||
classMask(orgCount, eq_class, selected_rivers);
|
||||
}
|
||||
else
|
||||
{
|
||||
classMask(cnt, eq_class, selected_rivers2);
|
||||
classMask(orgCount, eq_class, selected_rivers2);
|
||||
|
||||
for (i = 0; i < OPT_STREAM_BITS; i++)
|
||||
{
|
||||
@ -2341,30 +2312,21 @@ bool Optimizer::generateEquiJoin(RiverList& org_rivers)
|
||||
}
|
||||
}
|
||||
|
||||
if (!river_cnt)
|
||||
if (!riverCount)
|
||||
return false;
|
||||
|
||||
// AB: Deactivate currently all streams from every river, because we
|
||||
// need to know which nodes are computable between the rivers used
|
||||
// for the merge.
|
||||
|
||||
StreamStateHolder stateHolder(csb);
|
||||
stateHolder.deactivate();
|
||||
|
||||
HalfStaticArray<RecordSource*, OPT_STATIC_ITEMS> rsbs;
|
||||
HalfStaticArray<NestValueArray*, OPT_STATIC_ITEMS> keys;
|
||||
|
||||
// Unconditionally disable merge joins in favor of hash joins.
|
||||
// This is a temporary debugging measure.
|
||||
bool prefer_merge_over_hash = false;
|
||||
bool useMergeJoin = false;
|
||||
|
||||
// AB: Get the lowest river position from the rivers that are merged
|
||||
|
||||
RiverList rivers_to_merge;
|
||||
unsigned lowest_river_position = MAX_ULONG;
|
||||
unsigned number = 0;
|
||||
StreamList streams;
|
||||
RiverList rivers;
|
||||
unsigned number = 0, lowestPosition = MAX_ULONG;
|
||||
|
||||
for (River** iter = org_rivers.begin(); iter < org_rivers.end(); number++)
|
||||
for (River** iter = orgRivers.begin(); iter < orgRivers.end(); number++)
|
||||
{
|
||||
River* const river = *iter;
|
||||
|
||||
@ -2374,21 +2336,20 @@ bool Optimizer::generateEquiJoin(RiverList& org_rivers)
|
||||
continue;
|
||||
}
|
||||
|
||||
if (number < lowest_river_position)
|
||||
lowest_river_position = number;
|
||||
if (number < lowestPosition)
|
||||
lowestPosition = number;
|
||||
|
||||
rivers_to_merge.add(river);
|
||||
org_rivers.remove(iter);
|
||||
streams.join(river->getStreams());
|
||||
rivers.add(river);
|
||||
orgRivers.remove(iter);
|
||||
|
||||
// Apply local river booleans, if any
|
||||
|
||||
auto rsb = applyLocalBoolean(river);
|
||||
auto rsb = river->getRecordSource();
|
||||
|
||||
// Collect RSBs and keys to join
|
||||
|
||||
const auto key = FB_NEW_POOL(getPool()) SortNode(getPool());
|
||||
|
||||
if (prefer_merge_over_hash)
|
||||
if (useMergeJoin)
|
||||
{
|
||||
ValueExprNode*** selected_class;
|
||||
|
||||
@ -2400,9 +2361,7 @@ bool Optimizer::generateEquiJoin(RiverList& org_rivers)
|
||||
key->expressions.add((*selected_class)[number]);
|
||||
}
|
||||
|
||||
StreamList streams;
|
||||
streams.assign(river->getStreams());
|
||||
rsb = generateSort(streams, nullptr, rsb, key, favorFirstRows(), false);
|
||||
rsb = generateSort(river->getStreams(), nullptr, rsb, key, favorFirstRows(), false);
|
||||
}
|
||||
else
|
||||
{
|
||||
@ -2419,7 +2378,7 @@ bool Optimizer::generateEquiJoin(RiverList& org_rivers)
|
||||
// For a hash join, we need to choose the smallest ones as inner sub-streams,
|
||||
// hence we reverse the order when storing them in the temporary arrays.
|
||||
|
||||
if (prefer_merge_over_hash)
|
||||
if (useMergeJoin)
|
||||
{
|
||||
rsbs.add(rsb);
|
||||
keys.add(&key->expressions);
|
||||
@ -2437,7 +2396,7 @@ bool Optimizer::generateEquiJoin(RiverList& org_rivers)
|
||||
|
||||
RecordSource* rsb = nullptr;
|
||||
|
||||
if (prefer_merge_over_hash)
|
||||
if (useMergeJoin)
|
||||
{
|
||||
rsb = FB_NEW_POOL(getPool())
|
||||
MergeJoin(csb, rsbs.getCount(), (SortedStream**) rsbs.begin(), keys.begin());
|
||||
@ -2448,36 +2407,11 @@ bool Optimizer::generateEquiJoin(RiverList& org_rivers)
|
||||
HashJoin(tdbb, csb, rsbs.getCount(), rsbs.begin(), keys.begin());
|
||||
}
|
||||
|
||||
// Activate streams of all the rivers being merged
|
||||
|
||||
for (const auto river : rivers_to_merge)
|
||||
river->activate(csb);
|
||||
|
||||
// Pick up any boolean that may apply
|
||||
rsb = applyLocalBoolean(rsb, streams);
|
||||
|
||||
BoolExprNode* boolean = nullptr;
|
||||
double selectivity = MAXIMUM_SELECTIVITY;
|
||||
|
||||
for (auto iter = getBaseConjuncts(); iter.hasData(); ++iter)
|
||||
{
|
||||
if (!(iter & CONJUNCT_USED) &&
|
||||
!(iter->nodFlags & ExprNode::FLAG_RESIDUAL) &&
|
||||
iter->computable(csb, INVALID_STREAM, false))
|
||||
{
|
||||
compose(getPool(), &boolean, iter);
|
||||
iter |= CONJUNCT_USED;
|
||||
|
||||
if (!(iter & CONJUNCT_MATCHED))
|
||||
selectivity *= getSelectivity(*iter);
|
||||
}
|
||||
}
|
||||
|
||||
if (boolean)
|
||||
rsb = FB_NEW_POOL(getPool()) FilteredStream(csb, rsb, boolean, selectivity);
|
||||
|
||||
const auto merged_river = FB_NEW_POOL(getPool()) River(csb, rsb, rivers_to_merge);
|
||||
|
||||
org_rivers.insert(lowest_river_position, merged_river);
|
||||
const auto finalRiver = FB_NEW_POOL(getPool()) River(csb, rsb, rivers);
|
||||
orgRivers.insert(lowestPosition, finalRiver);
|
||||
|
||||
return true;
|
||||
}
|
||||
@ -2882,6 +2816,68 @@ RecordSource* Optimizer::generateRetrieval(StreamType stream,
|
||||
}
|
||||
|
||||
|
||||
//
|
||||
// Check whether the given boolean can be involved in a equi-join relationship
|
||||
//
|
||||
|
||||
bool Optimizer::getEquiJoinKeys(BoolExprNode* boolean,
|
||||
NestConst<ValueExprNode>* node1,
|
||||
NestConst<ValueExprNode>* node2,
|
||||
bool needCast)
|
||||
{
|
||||
auto cmpNode = nodeAs<ComparativeBoolNode>(boolean);
|
||||
if (!cmpNode || (cmpNode->blrOp != blr_eql && cmpNode->blrOp != blr_equiv))
|
||||
return false;
|
||||
|
||||
auto arg1 = cmpNode->arg1;
|
||||
auto arg2 = cmpNode->arg2;
|
||||
|
||||
if (!getEquiJoinKeys(arg1, arg2, needCast))
|
||||
return false;
|
||||
|
||||
*node1 = arg1;
|
||||
*node2 = arg2;
|
||||
return true;
|
||||
}
|
||||
|
||||
bool Optimizer::getEquiJoinKeys(NestConst<ValueExprNode>& node1,
|
||||
NestConst<ValueExprNode>& node2,
|
||||
bool needCast)
|
||||
{
|
||||
dsc result, desc1, desc2;
|
||||
node1->getDesc(tdbb, csb, &desc1);
|
||||
node2->getDesc(tdbb, csb, &desc2);
|
||||
|
||||
// Ensure that arguments can be compared in the binary form
|
||||
if (!CVT2_get_binary_comparable_desc(&result, &desc1, &desc2))
|
||||
return false;
|
||||
|
||||
// Cast the arguments to the common data type, if required
|
||||
if (needCast)
|
||||
{
|
||||
if (!DSC_EQUIV(&result, &desc1, true))
|
||||
{
|
||||
const auto cast = FB_NEW_POOL(getPool()) CastNode(getPool());
|
||||
cast->source = node1;
|
||||
cast->castDesc = result;
|
||||
cast->impureOffset = csb->allocImpure<impure_value>();
|
||||
node1 = cast;
|
||||
}
|
||||
|
||||
if (!DSC_EQUIV(&result, &desc2, true))
|
||||
{
|
||||
const auto cast = FB_NEW_POOL(getPool()) CastNode(getPool());
|
||||
cast->source = node2;
|
||||
cast->castDesc = result;
|
||||
cast->impureOffset = csb->allocImpure<impure_value>();
|
||||
node2 = cast;
|
||||
}
|
||||
}
|
||||
|
||||
return true;
|
||||
}
|
||||
|
||||
|
||||
//
|
||||
// Make an alias string suitable for printing as part of the plan.
|
||||
// For views, this means multiple aliases to distinguish the base table.
|
||||
|
||||
@ -156,12 +156,10 @@ class River
|
||||
{
|
||||
public:
|
||||
River(CompilerScratch* csb, RecordSource* rsb, RecordSourceNode* node, const StreamList& streams)
|
||||
: m_rsb(rsb), m_nodes(csb->csb_pool), m_streams(csb->csb_pool)
|
||||
: m_rsb(rsb), m_nodes(csb->csb_pool), m_streams(csb->csb_pool, streams)
|
||||
{
|
||||
if (node)
|
||||
m_nodes.add(node);
|
||||
|
||||
m_streams.assign(streams);
|
||||
}
|
||||
|
||||
River(CompilerScratch* csb, RecordSource* rsb, RiverList& rivers)
|
||||
@ -410,6 +408,14 @@ public:
|
||||
return (rse->flags & RseNode::FLAG_OPT_FIRST_ROWS) != 0;
|
||||
}
|
||||
|
||||
bool getEquiJoinKeys(BoolExprNode* boolean,
|
||||
NestConst<ValueExprNode>* node1,
|
||||
NestConst<ValueExprNode>* node2,
|
||||
bool needCast = false);
|
||||
bool getEquiJoinKeys(NestConst<ValueExprNode>& node1,
|
||||
NestConst<ValueExprNode>& node2,
|
||||
bool needCast = false);
|
||||
|
||||
Firebird::string makeAlias(StreamType stream);
|
||||
void printf(const char* format, ...);
|
||||
|
||||
@ -418,7 +424,7 @@ private:
|
||||
|
||||
RecordSource* compile(BoolExprNodeStack* parentStack);
|
||||
|
||||
RecordSource* applyLocalBoolean(const River* river);
|
||||
RecordSource* applyLocalBoolean(RecordSource* rsb, const StreamList& streams);
|
||||
void checkIndices();
|
||||
void checkSorts();
|
||||
unsigned decompose(BoolExprNode* boolNode, BoolExprNodeStack& stack);
|
||||
|
||||
Loading…
Reference in New Issue
Block a user