database/openGauss-server
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

1、老问题：修复分区剪枝、分区路由未使用正确的collation导致的core问题和剪枝问题

Browse Source 
- 修复range、list分区键比较时未指定字符序的问题。
- 修复hash分区键计算hashvalue未使用collation的场景。
2、修复hash分区剪枝时，const->constcollid与分区键不同导致选错分区的问题：
问题分析：getHashPartitionOid中计算hash使用的时const上的collation，const可能由条件中的常量传入，const的collation与分区键不同，应当使用分区键上的collation
修改方案：getHashPartitionOid中无法方便地获取分区键上的collation，所有需要判断分区的const都会调用getTargetType函数，在此函数中处理collation，使返回的collation都与分区键的相同。getTargetType的src参数可能指向条件上的const，不适合直接修改其const，目前copy后修改。
This commit is contained in:
li-judong
2023-06-26 19:52:41 +08:00
parent 622ffe4cab
commit 97f9f7d20c
15 changed files with 515 additions and 89 deletions
Download Patch File Download Diff File Expand all files Collapse all files

13
src/gausskernel/cbb/utils/partition/partitionkey.cpp
View File

@ -276,6 +276,7 @@ List* untransformPartitionBoundary(Datum options)
return result;
}
/* Use value1's collation first, so the table partition key boundary must be the first parameter! */
int partitonKeyCompare(Const** value1, Const** value2, int len, bool nullEqual)
{
uint8 i = 0;
@ -321,7 +322,7 @@ int partitonKeyCompare(Const** value1, Const** value2, int len, bool nullEqual)
break;
}
constCompare(v1, v2, compare);
constCompare(v1, v2, v1->constcollid, compare);
if (compare != 0) {
break;
}
@ -447,12 +448,12 @@ bool isPartKeyValuesInPartition(RangePartitionMap* partMap, Const** partKeyValue
return greaterThanBottom && lessThanTop;
}
int comparePartitionKey(RangePartitionMap* partMap, Const** values1, Const** values2, int partKeyNum)
int comparePartitionKey(RangePartitionMap* partMap, Const** partkey_value, Const** partkey_bound, int partKeyNum)
{
int compare = 0;
incre_partmap_refcount((PartitionMap*)partMap);
partitonKeyCompareForRouting(values1, values2, (uint32)partKeyNum, compare);
partitonKeyCompareForRouting(partkey_value, partkey_bound, (uint32)partKeyNum, compare);
decre_partmap_refcount((PartitionMap*)partMap);
return compare;
@ -784,7 +785,7 @@ void GetPartitionOidListForRTE(RangeTblEntry *rte, RangeVar *relation)
}
/* function to check whether two partKey are identical */
int ConstCompareWithNull(Const *c1, Const *c2)
int ConstCompareWithNull(Const *c1, Const *c2, Oid collation)
{
if (constIsNull(c1) && constIsNull(c2)) {
return 0;
@ -794,7 +795,7 @@ int ConstCompareWithNull(Const *c1, Const *c2)
}
int compare = -1;
constCompare(c1, c2, compare);
constCompare(c1, c2, collation, compare);
return compare;
}
@ -813,7 +814,7 @@ int ListPartKeyCompare(PartitionKey* k1, PartitionKey* k2)
}
int res;
for (int i = 0; i < k1->count; i++) {
res = ConstCompareWithNull(k1->values[i], k2->values[i]);
res = ConstCompareWithNull(k1->values[i], k2->values[i], k2->values[i]->constcollid);
if (res != 0) {
return res;
}

34
src/gausskernel/cbb/utils/partition/partitionmap.cpp
View File

@ -182,7 +182,7 @@
(compare) = interval_cmp_internal((arg1), (arg2)); \
} while (0)
#define constCompare_baseType(value1, value2, compare) \
#define constCompare_baseType(value1, value2, collation, compare) \
do { \
switch ((value1)->consttype) { \
case INT2OID: \
@ -217,26 +217,23 @@
break; \
case BPCHAROID: \
Assert((value2)->consttype == BPCHAROID); \
Assert((value1)->constcollid == (value2)->constcollid); \
bpchar_cmp_partition(DatumGetBpCharP((value1)->constvalue), \
DatumGetBpCharP((value2)->constvalue), \
(value1)->constcollid, \
(collation), \
(compare)); \
break; \
case VARCHAROID: \
Assert((value2)->consttype == VARCHAROID); \
Assert((value1)->constcollid == (value2)->constcollid); \
text_cmp_partition(DatumGetTextP((value1)->constvalue), \
DatumGetTextP((value2)->constvalue), \
(value1)->constcollid, \
(collation), \
(compare)); \
break; \
case TEXTOID: \
Assert((value2)->consttype == TEXTOID); \
Assert((value1)->constcollid == (value2)->constcollid); \
text_cmp_partition(DatumGetTextP((value1)->constvalue), \
DatumGetTextP((value2)->constvalue), \
(value1)->constcollid, \
(collation), \
(compare)); \
break; \
case DATEOID: \
@ -264,7 +261,7 @@
DatumGetIntervalP((value1)->constvalue), DatumGetIntervalP((value2)->constvalue), (compare)); \
break; \
default: \
(compare) = constCompare_constType((value1), (value2)); \
(compare) = constCompare_constType((value1), (value2), (collation)); \
break; \
} \
} while (0)
@ -275,7 +272,7 @@
* @Param[OUT] value2: right value to compare
* @See also:
*/
void constCompare(Const* value1, Const* value2, int& compare)
void constCompare(Const* value1, Const* value2, Oid collation, int& compare)
{
if (t_thrd.utils_cxt.gValueCompareContext == NULL) {
/*
@ -300,9 +297,9 @@ void constCompare(Const* value1, Const* value2, int& compare)
PG_TRY();
{
if (value1->consttype == value2->consttype) {
constCompare_baseType(value1, value2, compare);
constCompare_baseType(value1, value2, collation, compare);
} else {
compare = constCompare_constType(value1, value2);
compare = constCompare_constType(value1, value2, collation);
}
}
PG_CATCH();
@ -2395,8 +2392,8 @@ Oid getHashPartitionOid(PartitionMap* partMap, Const** partKeyValue, int32* part
decre_partmap_refcount(partMap);
return result;
}
hash_value = hashValueCombination(hash_value, partKeyValue[i]->consttype, partKeyValue[i]->constvalue, false,
LOCATOR_TYPE_HASH);
hash_value = hashValueCombination(hash_value, partKeyValue[i]->consttype, partKeyValue[i]->constvalue,
false, LOCATOR_TYPE_HASH, partKeyValue[i]->constcollid);
i++;
}
@ -2772,7 +2769,7 @@ int partOidGetPartSequence(Relation rel, Oid partOid)
* @return 0: value1==value2 1:value1>vlaue2 -1:value1<value2
*/
int constCompare_constType(Const* value1, Const* value2)
int constCompare_constType(Const* value1, Const* value2, Oid collation)
{
EState* estate = NULL;
ExprContext* econtext = NULL;
@ -2789,13 +2786,12 @@ int constCompare_constType(Const* value1, Const* value2)
eqExpr = (Expr*)makeSimpleA_Expr(AEXPR_OP, "=", (Node*)value1, (Node*)value2, -1);
eqExpr = (Expr*)transformExpr(pstate, (Node*)eqExpr, EXPR_KIND_PARTITION_EXPRESSION);
assign_expr_collations(pstate, (Node*)eqExpr);
((OpExpr*)eqExpr)->inputcollid = collation;
gtExpr = (Expr*)makeSimpleA_Expr(AEXPR_OP, ">", (Node*)value1, (Node*)value2, -1);
gtExpr = (Expr*)transformExpr(pstate, (Node*)gtExpr, EXPR_KIND_PARTITION_EXPRESSION);
((OpExpr*)gtExpr)->inputcollid = value1->constcollid;
assign_expr_collations(pstate, (Node*)gtExpr);
((OpExpr*)gtExpr)->inputcollid = collation;
estate = CreateExecutorState();
econtext = GetPerTupleExprContext(estate);
@ -3094,7 +3090,9 @@ Const **transformConstIntoPartkeyType(FormData_pg_attribute *attrs, int2vector *
for (int i = 0; i < len; i++) {
partKeyPos = partitionKey->values[i];
if (likely(attrs[partKeyPos - 1].atttypid == boundary[i]->consttype) || boundary[i]->ismaxvalue) {
if ((likely(attrs[partKeyPos - 1].atttypid == boundary[i]->consttype) &&
likely(attrs[partKeyPos - 1].attcollation == boundary[i]->constcollid)) ||
boundary[i]->ismaxvalue) {
continue;
}

3
src/gausskernel/dbmind/db4ai/executor/algorithms/bayes/bayes.cpp
View File

@ -53,7 +53,8 @@ ValueInTuple create_value(Datum data, Oid type, bool isnull)
{
uint32_t hashvalue = 0;
if (!isnull) {
hashvalue = DatumGetUInt32(compute_hash(type, data, LOCATOR_TYPE_HASH));
/* As before, used invalid collation. B format collation is not supported yet. */
hashvalue = DatumGetUInt32(compute_hash(type, data, LOCATOR_TYPE_HASH, InvalidOid));
}
ValueInTuple val = {data, type, isnull, hashvalue};
return val;

38
src/gausskernel/optimizer/commands/tablecmds.cpp
View File

@ -22759,6 +22759,30 @@ Node* GetTargetValue(Form_pg_attribute attrs, Const* src, bool isinterval, bool
return NULL;
}
/* convert source const's charset to target partkey's charset */
if (!partkeyIsFunc && DB_IS_CMPT(B_FORMAT) && OidIsValid(attrs->attcollation)) {
assign_expr_collations(NULL, expr);
if (attrs->attcollation != exprCollation(expr)) {
int attcharset = get_valid_charset_by_collation(attrs->attcollation);
expr = coerce_to_target_charset(expr, attcharset, target_oid, target_mod, attrs->attcollation);
Assert(expr != NULL);
if (!IsA(expr, Const)) {
expr = (Node*)evaluate_expr((Expr*)expr, target_oid, target_mod, attrs->attcollation);
} else if (attrs->attcollation != exprCollation(expr)) {
if (expr == (Node*)src) {
/* We are not sure where src comes from, avoid set src->constcollid directly. */
expr = (Node*)copyObject((void*)src);
}
/*
* The expr is used to compute hash or compare it with the partition boundary.
* Set the correct collation to ensure the correctness of the partition pruning and routing.
*/
exprSetCollation(expr, attrs->attcollation);
}
}
}
switch (nodeTag(expr)) {
/* do nothing for Const */
case T_Const:
@ -22942,7 +22966,7 @@ static void sqlcmd_check_list_partition_have_duplicate_values(List** key_values_
ListCell* c2 = NULL;
for (int k = 0; k < bound_idx; ++k) {
forboth (c1, key_values_array[part_idx][bound_idx], c2, key_values_array[part_idx][k]) {
if (ConstCompareWithNull((Const*)lfirst(c1), (Const*)lfirst(c2)) != 0) {
if (ConstCompareWithNull((Const*)lfirst(c1), (Const*)lfirst(c2), ((Const*)lfirst(c2))->constcollid) != 0) {
break;
}
}
@ -22968,7 +22992,7 @@ static void sqlcmd_check_two_list_partition_values_overlapped(List** key_values_
Assert(!(con1->ismaxvalue && con2->ismaxvalue));
break;
}
if (ConstCompareWithNull(con1, con2) != 0) {
if (ConstCompareWithNull(con1, con2, con2->constcollid) != 0) {
break;
}
}
@ -23518,7 +23542,7 @@ static void CheckPartitionValueConflictForAddPartition(Relation rel, Node *partD
RangePartitionMap *partMap = (RangePartitionMap *)rel->partMap;
Const *curBound = (Const *)copyObject(partMap->rangeElements[partNum - 1].boundary[0]);
Const *val = partDef->curStartVal;
if (!curBound->ismaxvalue && val != NULL && partitonKeyCompare(&val, &curBound, 1) != 0) {
if (!curBound->ismaxvalue && val != NULL && partitonKeyCompare(&curBound, &val, 1) != 0) {
ereport(ERROR, (errcode(ERRCODE_INVALID_TABLE_DEFINITION),
errmsg("start value of partition \"%s\" NOT EQUAL up-boundary of last partition.",
partDef->partitionInitName ? partDef->partitionInitName : partDef->partitionName)));
@ -27885,14 +27909,14 @@ static void ATExecSplitPartition(Relation partTableRel, AlterTableCmd* cmd)
// check the first dest partition boundary
if (srcPartIndex != 0) {
if (!partMap->rangeElements[srcPartIndex].isInterval) {
compare = comparePartitionKey(partMap, partMap->rangeElements[srcPartIndex - 1].boundary,
(Const**)lfirst(list_head(destPartBoundaryList)), partKeyNum);
compare = comparePartitionKey(partMap, (Const**)lfirst(list_head(destPartBoundaryList)),
partMap->rangeElements[srcPartIndex - 1].boundary, partKeyNum);
} else {
Const** partKeyValue = (Const**)lfirst(list_head(destPartBoundaryList));
RangeElement& srcPartition = partMap->rangeElements[srcPartIndex];
compare = -ValueCmpLowBoudary(partKeyValue, &srcPartition, partMap->intervalValue);
compare = ValueCmpLowBoudary(partKeyValue, &srcPartition, partMap->intervalValue);
}
if (compare >= 0) {
if (compare <= 0) {
ereport(ERROR,
(errcode(ERRCODE_INVALID_OPERATION),
errmsg("the bound of the first resulting partition is too low")));

8
src/gausskernel/optimizer/util/bucketpruning.cpp
View File

@ -88,7 +88,7 @@ static BucketPruningContext* makePruningContext(
PlannerInfo* root, RelOptInfo* rel, RangeTblEntry* rte, List* restrictInfo);
static Expr* RestrictInfoGetExpr(List* restrictInfo);
static BucketPruningResult* BucketPruningForExpr(BucketPruningContext* bpcxt, Expr* expr);
static int getConstBucketId(Const* val, int bucketmapsize);
static int getConstBucketId(Const* val, Oid collation, int bucketmapsize);
static BucketPruningResult* BucketPruningForBoolExpr(BucketPruningContext* bpcxt, BoolExpr* expr);
static BucketPruningResult* BucketPruningForOpExpr(BucketPruningContext* bpcxt, OpExpr* expr);
static int GetExecBucketId(ExecNodes* exec_nodes, ParamListInfo params);
@ -635,7 +635,7 @@ static BucketPruningResult* BucketPruningForOpExpr(BucketPruningContext* bpcxt,
}
/* time for pruning */
int id = getConstBucketId(constArg, bpcxt->rte->bucketmapsize);
int id = getConstBucketId(constArg, varArg->varcollid, bpcxt->rte->bucketmapsize);
if (pg_strcasecmp(opName, "=") == 0) {
return makePruningResult(id);
@ -653,7 +653,7 @@ static BucketPruningResult* BucketPruningForOpExpr(BucketPruningContext* bpcxt,
* @in Const: the const val which we want the bucketid of it
* @return the bucketid index of the const val
*/
static int getConstBucketId(Const* val, int bucketmapsize)
static int getConstBucketId(Const* val, Oid collation, int bucketmapsize)
{
uint32 hashval = 0;
int bucketid = 0;
@ -662,7 +662,7 @@ static int getConstBucketId(Const* val, int bucketmapsize)
return 0;
}
hashval = compute_hash(val->consttype, val->constvalue, LOCATOR_TYPE_HASH);
hashval = compute_hash(val->consttype, val->constvalue, LOCATOR_TYPE_HASH, collation);
bucketid = compute_modulo((unsigned int)(abs((int)hashval)), bucketmapsize);

38
src/gausskernel/optimizer/util/pruning.cpp
View File

@ -1179,6 +1179,13 @@ static PruningResult* partitionPruningFromScalarArrayOpExpr
return result;
}
/* Do not pruning if collation of operator is different from collation of partkey. */
if (((Var*)larg)->varcollid != arrayExpr->inputcollid) {
result->state = PRUNING_RESULT_FULL;
result->isPbeSinlePartition = false;
return result;
}
if (T_ArrayExpr == nodeTag(rarg)) {
List* eleList = NULL;
ListCell* element = NULL;
@ -1305,6 +1312,13 @@ static PruningResult* partitionPruningFromScalarArrayOpExpr(PartitionType partTy
return result;
}
/* Do not pruning if collation of operator is different from collation of partkey. */
if (((Var*)larg)->varcollid != arrayExpr->inputcollid) {
result->state = PRUNING_RESULT_FULL;
result->isPbeSinlePartition = false;
return result;
}
if (T_ArrayExpr == nodeTag(rarg)) {
List* eleList = NULL;
ListCell* element = NULL;
@ -1540,6 +1554,13 @@ static PruningResult* recordBoundaryFromOpExpr(const OpExpr* expr, PruningContex
}
}
/* Do not pruning if collation of operator is different from collation of partkey. */
if (varArg->varcollid != expr->inputcollid) {
result->state = PRUNING_RESULT_FULL;
result->isPbeSinlePartition = false;
return result;
}
/* Var's column MUST belongs to parition key columns */
partMap = (RangePartitionMap*)(GetPartitionMap(context));
@ -1832,6 +1853,13 @@ static PruningResult* recordEqualFromOpExpr(PartitionType partType, const OpExpr
}
}
/* Do not pruning if collation of operator is different from collation of partkey. */
if (varArg->varcollid != expr->inputcollid) {
result->state = PRUNING_RESULT_FULL;
result->isPbeSinlePartition = false;
return result;
}
PruningResult* res = RecordEqualFromOpExprPart(partType, context, opName, constMax, varArg, attrOffset, result, paramArg, exprPart, constArg, boundary);
return res;
}
@ -2388,8 +2416,8 @@ static void cleanPruningBottom(PruningContext *context, PartitionIdentifier* bot
RangeElement* range = partMap->rangeElements + i;
int compare = 0;
compare = partitonKeyCompare(&value, range->boundary, 1);
if (compare >= 0) {
compare = partitonKeyCompare(range->boundary, &value, 1);
if (compare <= 0) {
continue;
} else {
break;
@ -2421,8 +2449,8 @@ static void cleanPruningTop(PruningContext *context, PartitionIdentifier* topSeq
RangeElement* range = partMap->rangeElements + i;
int compare = 0;
compare = partitonKeyCompare(&value, range->boundary, 1);
if (compare <= 0) {
compare = partitonKeyCompare(range->boundary, &value, 1);
if (compare >= 0) {
continue;
} else {
break;
@ -2708,7 +2736,7 @@ static bool PartialListBoundaryMatched(ListPartElement* part, List* keyPos, Cons
ListCell* keyCell = NULL;
foreach(keyCell, keyPos) {
int id = lfirst_int(keyCell);
if (ConstCompareWithNull(keyValue[id], bound->values[id]) != 0) {
if (ConstCompareWithNull(keyValue[id], bound->values[id], bound->values[id]->constcollid) != 0) {
break;
}
}

12
src/gausskernel/process/stream/streamProducer.cpp
View File

@ -1980,11 +1980,11 @@ static int GetListConsumerNodeIdx(ExecBoundary* enBoundary, Const** values, int
while (maxId >= minId) {
midId = ((uint)minId + (uint)maxId) >> 1;
cmp = partitonKeyCompare(values, enBoundary->eles[midId]->boundary, distLen);
cmp = partitonKeyCompare(enBoundary->eles[midId]->boundary, values, distLen);
if (cmp == 0) {
hit = midId;
break;
} else if (cmp < 0) {
} else if (cmp > 0) {
maxId = midId - 1;
} else {
minId = midId + 1;
@ -2009,18 +2009,18 @@ static int GetRangeConsumerNodeIdx(ExecBoundary* enBoundary, Const** values, int
maxId = enBoundary->count - 1;
minId = 0;
hit = -1;
cmp = partitonKeyCompare(values, enBoundary->eles[maxId]->boundary, distLen);
if (cmp >= 0) {
cmp = partitonKeyCompare(enBoundary->eles[maxId]->boundary, values, distLen);
if (cmp <= 0) {
hit = -1;
} else {
while (maxId > minId) {
midId = ((uint)minId + (uint)maxId) >> 1;
cmp = partitonKeyCompare(values, enBoundary->eles[midId]->boundary, distLen);
cmp = partitonKeyCompare(enBoundary->eles[midId]->boundary, values, distLen);
if (cmp == 0) {
hit = midId + 1;
break;
} else if (cmp > 0) {
} else if (cmp < 0) {
minId = midId + 1;
} else {
maxId = midId;

6
src/gausskernel/runtime/executor/route.cpp
View File

@ -333,6 +333,7 @@ int SendRouter::CountNodeId(struct RouteMsg* attr)
Oid* col_type = (Oid*)palloc0(dist_keynum * sizeof(Oid));
int4* col_typemode = (int4*)palloc0(dist_keynum * sizeof(int4));
Oid* col_conllation = (Oid*)palloc0(dist_keynum * sizeof(Oid));
int i = 0;
int t = 0;
int dist_idx = list_nth_int(rel_loc_info->partAttrNum, t) - 1;
@ -341,6 +342,7 @@ int SendRouter::CountNodeId(struct RouteMsg* attr)
col_type[t] = tupdesc->attrs[i].atttypid;
Type typ = typeidType(col_type[t]);
col_typemode[t] = ((Form_pg_type)GETSTRUCT(typ))->typtypmod;
col_conllation[t] = tupdesc->attrs[i].attcollation;
ReleaseSysCache(typ);
t++;
if (t >= dist_keynum)
@ -358,7 +360,8 @@ int SendRouter::CountNodeId(struct RouteMsg* attr)
col_type[i],
GetDatumFromString(col_type[i], col_typemode[i], (char *)lfirst(cell)),
false,
rel_loc_info->locatorType);
rel_loc_info->locatorType,
col_conllation[i]);
i++;
}
if (unlikely(rel_loc_info->buckets_ptr == NULL)) {
@ -370,6 +373,7 @@ int SendRouter::CountNodeId(struct RouteMsg* attr)
pfree_ext(col_type);
pfree_ext(col_typemode);
pfree_ext(col_conllation);
heap_close(rel, AccessShareLock);
if ((node_id >= list_length(rel_loc_info->nodeList)) || node_id < 0) {

32
src/gausskernel/storage/access/hash/hashfunc.cpp
View File

@ -609,7 +609,7 @@ Datum hash_uint32(uint32 k)
/*
* compute_hash() -- Generic hash function for all datatypes
*/
Datum compute_hash(Oid type, Datum value, char locator)
Datum compute_hash(Oid type, Datum value, char locator, Oid collation)
{
uint8 tmp8;
int16 tmp16;
@ -664,7 +664,7 @@ Datum compute_hash(Oid type, Datum value, char locator)
case NVARCHAR2OID:
case VARCHAROID:
case TEXTOID:
return DirectFunctionCall1(hashtext, value);
return DirectFunctionCall1Coll(hashtext, collation, value);
case OIDVECTOROID:
return DirectFunctionCall1(hashoidvector, value);
@ -687,7 +687,7 @@ Datum compute_hash(Oid type, Datum value, char locator)
return DirectFunctionCall1(hashint8, value);
case BPCHAROID:
return DirectFunctionCall1(hashbpchar, value);
return DirectFunctionCall1Coll(hashbpchar, collation, value);
case RAWOID:
case BYTEAOID:
@ -731,13 +731,13 @@ Datum compute_hash(Oid type, Datum value, char locator)
return (Datum)0;
}
uint32 hashValueCombination(uint32 hashValue, Oid colType, Datum val, bool allIsNull, char locatorType)
uint32 hashValueCombination(uint32 hashValue, Oid colType, Datum val, bool allIsNull, char locatorType, Oid collation)
{
if (!allIsNull) {
hashValue = (hashValue << 1) | ((hashValue & 0x80000000) ? 1 : 0);
hashValue ^= (uint32)compute_hash(colType, val, locatorType);
hashValue ^= (uint32)compute_hash(colType, val, locatorType, collation);
} else {
hashValue = (uint32)compute_hash(colType, val, locatorType);
hashValue = (uint32)compute_hash(colType, val, locatorType, collation);
}
return hashValue;
}
@ -933,11 +933,12 @@ Datum bucketbpchar(PG_FUNCTION_ARGS)
Datum value = PG_GETARG_DATUM(0);
int flag = PG_GETARG_INT32(1);
long hashValue;
int collation = is_b_format_collation(PG_GET_COLLATION()) ? PG_GET_COLLATION() : InvalidOid;
if (flag == 0) { /* hash */
hashValue = (long)compute_hash(BPCHAROID, value, LOCATOR_TYPE_HASH);
hashValue = (long)compute_hash(BPCHAROID, value, LOCATOR_TYPE_HASH, collation);
} else {
hashValue = (long)compute_hash(BPCHAROID, value, LOCATOR_TYPE_MODULO);
hashValue = (long)compute_hash(BPCHAROID, value, LOCATOR_TYPE_MODULO, collation);
}
bucket = compute_modulo(abs(hashValue), BUCKETDATALEN);
result = Int32GetDatum(bucket);
@ -980,11 +981,12 @@ Datum bucketvarchar(PG_FUNCTION_ARGS)
Datum value = PG_GETARG_DATUM(0);
int flag = PG_GETARG_INT32(1);
long hashValue;
int collation = is_b_format_collation(PG_GET_COLLATION()) ? PG_GET_COLLATION() : InvalidOid;
if (flag == 0) { /* hash */
hashValue = (long)compute_hash(VARCHAROID, value, LOCATOR_TYPE_HASH);
hashValue = (long)compute_hash(VARCHAROID, value, LOCATOR_TYPE_HASH, collation);
} else {
hashValue = (long)compute_hash(VARCHAROID, value, LOCATOR_TYPE_MODULO);
hashValue = (long)compute_hash(VARCHAROID, value, LOCATOR_TYPE_MODULO, collation);
}
bucket = compute_modulo(abs(hashValue), BUCKETDATALEN);
result = Int32GetDatum(bucket);
@ -1004,11 +1006,12 @@ Datum bucketnvarchar2(PG_FUNCTION_ARGS)
Datum value = PG_GETARG_DATUM(0);
int flag = PG_GETARG_INT32(1);
long hashValue;
int collation = is_b_format_collation(PG_GET_COLLATION()) ? PG_GET_COLLATION() : InvalidOid;
if (flag == 0) { /* hash */
hashValue = (long)compute_hash(NVARCHAR2OID, value, LOCATOR_TYPE_HASH);
hashValue = (long)compute_hash(NVARCHAR2OID, value, LOCATOR_TYPE_HASH, collation);
} else {
hashValue = (long)compute_hash(NVARCHAR2OID, value, LOCATOR_TYPE_MODULO);
hashValue = (long)compute_hash(NVARCHAR2OID, value, LOCATOR_TYPE_MODULO, collation);
}
bucket = compute_modulo(abs(hashValue), BUCKETDATALEN);
result = Int32GetDatum(bucket);
@ -1436,11 +1439,12 @@ Datum buckettext(PG_FUNCTION_ARGS)
Datum value = PG_GETARG_DATUM(0);
int flag = PG_GETARG_INT32(1);
long hashValue;
int collation = is_b_format_collation(PG_GET_COLLATION()) ? PG_GET_COLLATION() : InvalidOid;
if (flag == 0) { /* hash */
hashValue = (long)compute_hash(TEXTOID, value, LOCATOR_TYPE_HASH);
hashValue = (long)compute_hash(TEXTOID, value, LOCATOR_TYPE_HASH, collation);
} else {
hashValue = (long)compute_hash(TEXTOID, value, LOCATOR_TYPE_MODULO);
hashValue = (long)compute_hash(TEXTOID, value, LOCATOR_TYPE_MODULO, collation);
}
bucket = compute_modulo(abs(hashValue), BUCKETDATALEN);
result = Int32GetDatum(bucket);

4
src/include/access/hash.h
View File

@ -416,8 +416,8 @@ extern void hashbucketcleanup(Relation rel, Bucket cur_bucket,
IndexBulkDeleteCallback callback, void *callback_state);
#ifdef PGXC
extern Datum compute_hash(Oid type, Datum value, char locator);
extern uint32 hashValueCombination(uint32 hashValue, Oid colType, Datum val, bool allIsNull, char locatorType = 'H');
extern Datum compute_hash(Oid type, Datum value, char locator, Oid collation = InvalidOid);
extern uint32 hashValueCombination(uint32 hashValue, Oid colType, Datum val, bool allIsNull, char locatorType = 'H', Oid collation = InvalidOid);
extern char* get_compute_hash_function(Oid type, char locator);
extern Datum getbucket(PG_FUNCTION_ARGS);
extern Datum getbucketbycnt(PG_FUNCTION_ARGS);

33
src/include/utils/partitionkey.h
View File

@ -64,48 +64,49 @@ extern bool GetPartitionOidForRTE(RangeTblEntry *rte, RangeVar *relation, ParseS
extern bool GetSubPartitionOidForRTE(RangeTblEntry *rte, RangeVar *relation, ParseState *pstate, Relation rel);
extern void GetPartitionOidListForRTE(RangeTblEntry *rte, RangeVar *relation);
#define partitonKeyCompareForRouting(value1, value2, len, compare) \
/* the 2nd parameter must be partition boundary */
#define partitonKeyCompareForRouting(partkey_value, partkey_bound, len, compare) \
do { \
uint32 i = 0; \
Const *v1 = NULL; \
Const *v2 = NULL; \
Const *kv = NULL; \
Const *bv = NULL; \
for (; i < (len); i++) { \
v1 = *((value1) + i); \
v2 = *((value2) + i); \
if (v1 == NULL || v2 == NULL) { \
if (v1 == NULL && v2 == NULL) { \
kv = *((partkey_value) + i); \
bv = *((partkey_bound) + i); \
if (kv == NULL || bv == NULL) { \
if (kv == NULL && bv == NULL) { \
ereport(ERROR, \
(errcode(ERRCODE_DATATYPE_MISMATCH), errmsg("NULL can not be compared with NULL"))); \
} else if (v1 == NULL) { \
} else if (kv == NULL) { \
compare = -1; \
} else { \
compare = 1; \
} \
break; \
} \
if (constIsMaxValue(v1) || constIsMaxValue(v2)) { \
if (constIsMaxValue(v1) && constIsMaxValue(v2)) { \
if (constIsMaxValue(kv) || constIsMaxValue(bv)) { \
if (constIsMaxValue(kv) && constIsMaxValue(bv)) { \
compare = 0; \
continue; \
} else if (constIsMaxValue(v1)) { \
} else if (constIsMaxValue(kv)) { \
compare = 1; \
} else { \
compare = -1; \
} \
break; \
} \
if (v1->constisnull || v2->constisnull) { \
if (v1->constisnull && v2->constisnull) { \
if (kv->constisnull || bv->constisnull) { \
if (kv->constisnull && bv->constisnull) { \
ereport(ERROR, (errcode(ERRCODE_UNEXPECTED_NULL_VALUE), \
errmsg("null value can not be compared with null value."))); \
} else if (v1->constisnull) { \
} else if (kv->constisnull) { \
compare = 1; \
} else { \
compare = -1; \
} \
break; \
} \
constCompare(v1, v2, compare); \
constCompare(kv, bv, bv->constcollid, compare); \
if ((compare) != 0) { \
break; \
} \
@ -117,5 +118,5 @@ extern void GetPartitionOidListForRTE(RangeTblEntry *rte, RangeVar *relation);
#define constIsNull(x) ((x)->constisnull)
#define constIsMaxValue(x) ((x)->ismaxvalue)
int ConstCompareWithNull(Const *c1, Const *c2);
int ConstCompareWithNull(Const *c1, Const *c2, Oid collation);
int ListPartKeyCompare(PartitionKey* k1, PartitionKey* k2);

4
src/include/utils/partitionmap.h
View File

@ -136,11 +136,11 @@ extern int GetSubPartitionNumber(Relation rel);
extern bool targetListHasPartitionKey(List* targetList, Oid partitiondtableid);
extern int constCompare_constType(Const* value1, Const* value2);
extern int constCompare_constType(Const* value1, Const* value2, Oid collation);
extern bool partitionHasToast(Oid partOid);
extern void constCompare(Const* value1, Const* value2, int& compare);
extern void constCompare(Const* value1, Const* value2, Oid collation, int& compare);
extern struct ListPartElement* CopyListElements(ListPartElement* src, int elementNum);
extern struct HashPartElement* CopyHashElements(HashPartElement* src, int elementNum, int partkeyNum);

2
src/include/utils/partitionmap_gs.h
View File

@ -508,7 +508,7 @@ extern bool EqualPartitonMap(const PartitionMap* partMap1, const PartitionMap* p
bool isPartKeyValuesInPartition(RangePartitionMap* partMap, Const** partKeyValues, int partkeyColumnNum, int partSeq);
extern int comparePartitionKey(RangePartitionMap* partMap, Const** values1, Const** values2, int partKeyNum);
extern int comparePartitionKey(RangePartitionMap* partMap, Const** partkey_value, Const** partkey_bound, int partKeyNum);
extern int lookupHBucketid(oidvector *buckets, int low, int2 bucket_id);

283
src/test/regress/expected/charset_utf8mb4_b_db.out
View File

@ -4025,8 +4025,140 @@ CREATE TABLE t_multi_charset_partkey (part varchar(32) collate utf8mb4_general_c
partition p3 values less than(MAXVALUE)
); -- ERROR
ERROR: partition bound of partition "p2" is too low
CREATE TABLE t_multi_charset_partkey (part varchar(32) collate utf8mb4_general_ci, a int)
PARTITION BY RANGE(part) (
partition p1 values less than('楂樻柉DB'),
partition p2 values less than(_gbk'高斯db'),
partition p3 values less than(MAXVALUE)
); -- ERROR
ERROR: partition bound of partition "p2" is too low
CREATE TABLE t_multi_charset_partkey (part varchar(32) collate utf8mb4_general_ci, a int)
PARTITION BY LIST(part) (
partition p1 values('高斯DB'),
partition p2 values('高斯db')
); -- ERROR
ERROR: list partition p1 and p2 has overlapped value
CREATE TABLE t_multi_charset_partkey (part varchar(32) collate utf8mb4_general_ci, a int)
PARTITION BY RANGE(part) (
PARTITION pass START('高斯DB') END('高斯db'),
PARTITION excellent START('高斯db') END(MAXVALUE)
); -- unsupported
ERROR: datatype of column "part" is unsupported for partition key in start/end clause.
HINT: Valid datatypes are: smallint, int, bigint, float4/real, float8/double, numeric, date and timestamp [with time zone].
-- -- -- utf8mb4
CREATE TABLE t_multi_charset_partkey (part varchar(32) collate utf8mb4_bin, a int)
CREATE TABLE t_multi_charset_partkey (part text collate utf8mb4_bin, a int)
PARTITION BY HASH(part) (
partition p1,
partition p2,
partition p3,
partition p4
);
-- -- -- insert
INSERT INTO t_multi_charset_partkey VALUES(_gbk'高斯DB', 1);
INSERT INTO t_multi_charset_partkey VALUES(_gbk'高斯db', 2);
INSERT INTO t_multi_charset_partkey VALUES(_utf8mb4'高斯DB1', 3);
INSERT INTO t_multi_charset_partkey VALUES(_utf8mb4'高斯db1', 4);
-- -- -- select
SELECT * FROM t_multi_charset_partkey PARTITION(p1) order by 1,2;
part | a
----------+---
楂樻柉DB | 1
楂樻柉db | 2
(2 rows)
SELECT * FROM t_multi_charset_partkey PARTITION(p2) order by 1,2;
part | a
---------+---
高斯db1 | 4
(1 row)
SELECT * FROM t_multi_charset_partkey PARTITION(p3) order by 1,2;
part | a
------+---
(0 rows)
SELECT * FROM t_multi_charset_partkey PARTITION FOR(_gbk'高斯db') order by 1,2;
part | a
----------+---
楂樻柉DB | 1
楂樻柉db | 2
(2 rows)
SELECT * FROM t_multi_charset_partkey PARTITION FOR(_utf8mb4'高斯DB1') order by 1,2;
part | a
---------+---
高斯DB1 | 3
(1 row)
-- -- -- partition pruning
EXPLAIN (costs off)
SELECT * FROM t_multi_charset_partkey WHERE part=_gbk'高斯DB' order by 1,2;
QUERY PLAN
---------------------------------------------------------------
Sort
Sort Key: a
-> Partitioned Seq Scan on t_multi_charset_partkey
Filter: (part = '楂樻柉DB'::text COLLATE utf8mb4_bin)
Selected Partitions: 1
(5 rows)
SELECT * FROM t_multi_charset_partkey WHERE part=_gbk'高斯DB' order by 1,2;
part | a
----------+---
楂樻柉DB | 1
(1 row)
EXPLAIN (costs off)
SELECT * FROM t_multi_charset_partkey WHERE part=_utf8mb4'高斯db1' order by 1,2;
QUERY PLAN
-------------------------------------------------------
Sort
Sort Key: a
-> Partitioned Seq Scan on t_multi_charset_partkey
Filter: (part = '高斯db1'::text)
Selected Partitions: 2
(5 rows)
SELECT * FROM t_multi_charset_partkey WHERE part=_utf8mb4'高斯db1' order by 1,2;
part | a
---------+---
高斯db1 | 4
(1 row)
EXPLAIN (costs off)
SELECT * FROM t_multi_charset_partkey WHERE part=_gbk'高斯DB' collate gbk_chinese_ci order by 1,2; -- ALL PARTS
QUERY PLAN
-------------------------------------------------------------
Sort
Sort Key: part COLLATE utf8mb4_bin, a
-> Partition Iterator
Iterations: 4
-> Partitioned Seq Scan on t_multi_charset_partkey
Filter: ((part)::text = '楂樻柉DB'::text)
Selected Partitions: 1..4
(7 rows)
SELECT * FROM t_multi_charset_partkey WHERE part=_gbk'高斯DB' collate gbk_chinese_ci order by 1,2;
part | a
----------+---
楂樻柉DB | 1
楂樻柉db | 2
(2 rows)
-- -- -- partiton ddl
ALTER TABLE t_multi_charset_partkey SPLIT PARTITION p1 AT ( '高斯DB' ) INTO ( PARTITION p1, PARTITION p4); -- not support
ERROR: can not split LIST/HASH partition table
ALTER TABLE t_multi_charset_partkey RENAME PARTITION FOR(_gbk'高斯db') TO newp1;
SELECT * FROM t_multi_charset_partkey PARTITION(newp1) order by 1,2;
part | a
----------+---
楂樻柉DB | 1
楂樻柉db | 2
(2 rows)
DROP TABLE t_multi_charset_partkey;
-- -- -- utf8mb4
CREATE TABLE t_multi_charset_partkey (part text collate utf8mb4_bin, a int)
PARTITION BY RANGE(part) (
partition p1 values less than('楂樻柉DB'),
partition p2 values less than('楂樻柉db'),
@ -4069,15 +4201,27 @@ SELECT * FROM t_multi_charset_partkey PARTITION(p5) order by 1,2;
高斯db | 4
(1 row)
SELECT * FROM t_multi_charset_partkey PARTITION FOR(_gbk'高斯db') order by 1,2;
part | a
----------+---
楂樻柉db | 2
(1 row)
SELECT * FROM t_multi_charset_partkey PARTITION FOR(_utf8mb4'高斯DB') order by 1,2;
part | a
--------+---
高斯DB | 3
(1 row)
-- -- -- partition pruning
EXPLAIN (costs off)
SELECT * FROM t_multi_charset_partkey WHERE part=_gbk'高斯DB' order by 1,2;
QUERY PLAN
-----------------------------------------------------------------------
QUERY PLAN
---------------------------------------------------------------
Sort
Sort Key: a
-> Partitioned Seq Scan on t_multi_charset_partkey
Filter: ((part)::text = '楂樻柉DB'::text COLLATE utf8mb4_bin)
Filter: (part = '楂樻柉DB'::text COLLATE utf8mb4_bin)
Selected Partitions: 2
(5 rows)
@ -4094,7 +4238,7 @@ SELECT * FROM t_multi_charset_partkey WHERE part=_utf8mb4'高斯db' order by 1,2
Sort
Sort Key: a
-> Partitioned Seq Scan on t_multi_charset_partkey
Filter: ((part)::text = '高斯db'::text)
Filter: (part = '高斯db'::text)
Selected Partitions: 5
(5 rows)
@ -4124,6 +4268,135 @@ SELECT * FROM t_multi_charset_partkey WHERE part=_gbk'高斯DB' collate gbk_chin
楂樻柉db | 2
(2 rows)
-- -- -- partiton ddl
ALTER TABLE t_multi_charset_partkey SPLIT PARTITION FOR(_gbk'高斯DB') AT (_gbk'高斯DB1 ') INTO (PARTITION p2_1, PARTITION p2_2);
INSERT INTO t_multi_charset_partkey VALUES(_gbk'高斯DB1', 1);
SELECT * FROM t_multi_charset_partkey PARTITION(p2_1) order by 1,2;
part | a
----------+---
楂樻柉DB | 1
(1 row)
SELECT * FROM t_multi_charset_partkey PARTITION(p2_2) order by 1,2;
part | a
-----------+---
楂樻柉DB1 | 1
(1 row)
ALTER TABLE t_multi_charset_partkey RENAME PARTITION FOR(_gbk'高斯db') TO p3_1;
SELECT * FROM t_multi_charset_partkey PARTITION(p3_1) order by 1,2;
part | a
----------+---
楂樻柉db | 2
(1 row)
DROP TABLE t_multi_charset_partkey;
-- -- -- utf8mb4
CREATE TABLE t_multi_charset_partkey (part varchar(32) collate utf8mb4_unicode_ci, part2 varchar(32) collate utf8mb4_general_ci, a int)
PARTITION BY LIST COLUMNS(part, part2) (
partition p1 values in(('楂樻柉DB', '楂樻柉db')),
partition p2 values in(('高斯db', '高斯DB'))
);
-- -- -- insert
INSERT INTO t_multi_charset_partkey VALUES(_gbk'高斯DB', _gbk'高斯DB', 1);
INSERT INTO t_multi_charset_partkey VALUES(_gbk'高斯db', _gbk'高斯db', 2);
INSERT INTO t_multi_charset_partkey VALUES(_utf8mb4'高斯DB', _utf8mb4'高斯DB', 3);
INSERT INTO t_multi_charset_partkey VALUES(_utf8mb4'高斯db', _utf8mb4'高斯db', 4);
-- -- -- select
SELECT * FROM t_multi_charset_partkey PARTITION(p1) order by 1,2;
part | part2 | a
----------+----------+---
楂樻柉DB | 楂樻柉DB | 1
楂樻柉db | 楂樻柉db | 2
(2 rows)
SELECT * FROM t_multi_charset_partkey PARTITION(p2) order by 1,2;
part | part2 | a
--------+--------+---
高斯DB | 高斯DB | 3
高斯db | 高斯db | 4
(2 rows)
SELECT * FROM t_multi_charset_partkey PARTITION FOR(_gbk'高斯DB', _gbk'高斯db') order by 1,2;
part | part2 | a
----------+----------+---
楂樻柉DB | 楂樻柉DB | 1
楂樻柉db | 楂樻柉db | 2
(2 rows)
SELECT * FROM t_multi_charset_partkey PARTITION FOR(_utf8mb4'高斯db', _utf8mb4'高斯db') order by 1,2;
part | part2 | a
--------+--------+---
高斯DB | 高斯DB | 3
高斯db | 高斯db | 4
(2 rows)
-- -- -- partition pruning
EXPLAIN (costs off)
SELECT * FROM t_multi_charset_partkey WHERE part=_gbk'高斯DB' order by 1,2;
QUERY PLAN
------------------------------------------------------------------------------
Sort
Sort Key: part2 COLLATE utf8mb4_general_ci
-> Partitioned Seq Scan on t_multi_charset_partkey
Filter: ((part)::text = '楂樻柉DB'::text COLLATE utf8mb4_unicode_ci)
Selected Partitions: 1
(5 rows)
SELECT * FROM t_multi_charset_partkey WHERE part=_gbk'高斯DB' order by 1,2;
part | part2 | a
----------+----------+---
楂樻柉DB | 楂樻柉DB | 1
楂樻柉db | 楂樻柉db | 2
(2 rows)
EXPLAIN (costs off)
SELECT * FROM t_multi_charset_partkey WHERE part=_utf8mb4'高斯db' order by 1,2;
QUERY PLAN
-------------------------------------------------------
Sort
Sort Key: part2 COLLATE utf8mb4_general_ci
-> Partitioned Seq Scan on t_multi_charset_partkey
Filter: ((part)::text = '高斯db'::text)
Selected Partitions: 2
(5 rows)
SELECT * FROM t_multi_charset_partkey WHERE part=_utf8mb4'高斯db' order by 1,2;
part | part2 | a
--------+--------+---
高斯DB | 高斯DB | 3
高斯db | 高斯db | 4
(2 rows)
EXPLAIN (costs off)
SELECT * FROM t_multi_charset_partkey WHERE part=_gbk'高斯DB' collate gbk_chinese_ci order by 1,2; -- ALL PARTS
QUERY PLAN
-------------------------------------------------------------------------------
Sort
Sort Key: part COLLATE utf8mb4_unicode_ci, part2 COLLATE utf8mb4_general_ci
-> Partition Iterator
Iterations: 2
-> Partitioned Seq Scan on t_multi_charset_partkey
Filter: ((part)::text = '楂樻柉DB'::text)
Selected Partitions: 1..2
(7 rows)
SELECT * FROM t_multi_charset_partkey WHERE part=_gbk'高斯DB' collate gbk_chinese_ci order by 1,2;
part | part2 | a
----------+----------+---
楂樻柉DB | 楂樻柉DB | 1
楂樻柉db | 楂樻柉db | 2
(2 rows)
-- -- -- partiton ddl
ALTER TABLE t_multi_charset_partkey RENAME PARTITION FOR(_gbk'高斯DB', _gbk'高斯db') TO p1_1;
SELECT * FROM t_multi_charset_partkey PARTITION(p1_1) order by 1,2;
part | part2 | a
----------+----------+---
楂樻柉DB | 楂樻柉DB | 1
楂樻柉db | 楂樻柉db | 2
(2 rows)
DROP TABLE t_multi_charset_partkey;
-- -- -- gbk
CREATE TABLE t_multi_charset_partkey (part varchar(32) CHARACTER set gbk collate gbk_bin, a int)

94
src/test/regress/sql/charset_utf8mb4_b_db.sql
View File

@ -1141,9 +1141,61 @@ CREATE TABLE t_multi_charset_partkey (part varchar(32) collate utf8mb4_general_c
partition p2 values less than('高斯db'),
partition p3 values less than(MAXVALUE)
); -- ERROR
CREATE TABLE t_multi_charset_partkey (part varchar(32) collate utf8mb4_general_ci, a int)
PARTITION BY RANGE(part) (
partition p1 values less than('楂樻柉DB'),
partition p2 values less than(_gbk'高斯db'),
partition p3 values less than(MAXVALUE)
); -- ERROR
CREATE TABLE t_multi_charset_partkey (part varchar(32) collate utf8mb4_general_ci, a int)
PARTITION BY LIST(part) (
partition p1 values('高斯DB'),
partition p2 values('高斯db')
); -- ERROR
CREATE TABLE t_multi_charset_partkey (part varchar(32) collate utf8mb4_general_ci, a int)
PARTITION BY RANGE(part) (
PARTITION pass START('高斯DB') END('高斯db'),
PARTITION excellent START('高斯db') END(MAXVALUE)
); -- unsupported
-- -- -- utf8mb4
CREATE TABLE t_multi_charset_partkey (part varchar(32) collate utf8mb4_bin, a int)
CREATE TABLE t_multi_charset_partkey (part text collate utf8mb4_bin, a int)
PARTITION BY HASH(part) (
partition p1,
partition p2,
partition p3,
partition p4
);
-- -- -- insert
INSERT INTO t_multi_charset_partkey VALUES(_gbk'高斯DB', 1);
INSERT INTO t_multi_charset_partkey VALUES(_gbk'高斯db', 2);
INSERT INTO t_multi_charset_partkey VALUES(_utf8mb4'高斯DB1', 3);
INSERT INTO t_multi_charset_partkey VALUES(_utf8mb4'高斯db1', 4);
-- -- -- select
SELECT * FROM t_multi_charset_partkey PARTITION(p1) order by 1,2;
SELECT * FROM t_multi_charset_partkey PARTITION(p2) order by 1,2;
SELECT * FROM t_multi_charset_partkey PARTITION(p3) order by 1,2;
SELECT * FROM t_multi_charset_partkey PARTITION FOR(_gbk'高斯db') order by 1,2;
SELECT * FROM t_multi_charset_partkey PARTITION FOR(_utf8mb4'高斯DB1') order by 1,2;
-- -- -- partition pruning
EXPLAIN (costs off)
SELECT * FROM t_multi_charset_partkey WHERE part=_gbk'高斯DB' order by 1,2;
SELECT * FROM t_multi_charset_partkey WHERE part=_gbk'高斯DB' order by 1,2;
EXPLAIN (costs off)
SELECT * FROM t_multi_charset_partkey WHERE part=_utf8mb4'高斯db1' order by 1,2;
SELECT * FROM t_multi_charset_partkey WHERE part=_utf8mb4'高斯db1' order by 1,2;
EXPLAIN (costs off)
SELECT * FROM t_multi_charset_partkey WHERE part=_gbk'高斯DB' collate gbk_chinese_ci order by 1,2; -- ALL PARTS
SELECT * FROM t_multi_charset_partkey WHERE part=_gbk'高斯DB' collate gbk_chinese_ci order by 1,2;
-- -- -- partiton ddl
ALTER TABLE t_multi_charset_partkey SPLIT PARTITION p1 AT ( '高斯DB' ) INTO ( PARTITION p1, PARTITION p4); -- not support
ALTER TABLE t_multi_charset_partkey RENAME PARTITION FOR(_gbk'高斯db') TO newp1;
SELECT * FROM t_multi_charset_partkey PARTITION(newp1) order by 1,2;
DROP TABLE t_multi_charset_partkey;
-- -- -- utf8mb4
CREATE TABLE t_multi_charset_partkey (part text collate utf8mb4_bin, a int)
PARTITION BY RANGE(part) (
partition p1 values less than('楂樻柉DB'),
partition p2 values less than('楂樻柉db'),
@ -1162,6 +1214,8 @@ SELECT * FROM t_multi_charset_partkey PARTITION(p2) order by 1,2;
SELECT * FROM t_multi_charset_partkey PARTITION(p3) order by 1,2;
SELECT * FROM t_multi_charset_partkey PARTITION(p4) order by 1,2;
SELECT * FROM t_multi_charset_partkey PARTITION(p5) order by 1,2;
SELECT * FROM t_multi_charset_partkey PARTITION FOR(_gbk'高斯db') order by 1,2;
SELECT * FROM t_multi_charset_partkey PARTITION FOR(_utf8mb4'高斯DB') order by 1,2;
-- -- -- partition pruning
EXPLAIN (costs off)
SELECT * FROM t_multi_charset_partkey WHERE part=_gbk'高斯DB' order by 1,2;
@ -1172,6 +1226,44 @@ SELECT * FROM t_multi_charset_partkey WHERE part=_utf8mb4'高斯db' order by 1,2
EXPLAIN (costs off)
SELECT * FROM t_multi_charset_partkey WHERE part=_gbk'高斯DB' collate gbk_chinese_ci order by 1,2; -- ALL PARTS
SELECT * FROM t_multi_charset_partkey WHERE part=_gbk'高斯DB' collate gbk_chinese_ci order by 1,2;
-- -- -- partiton ddl
ALTER TABLE t_multi_charset_partkey SPLIT PARTITION FOR(_gbk'高斯DB') AT (_gbk'高斯DB1 ') INTO (PARTITION p2_1, PARTITION p2_2);
INSERT INTO t_multi_charset_partkey VALUES(_gbk'高斯DB1', 1);
SELECT * FROM t_multi_charset_partkey PARTITION(p2_1) order by 1,2;
SELECT * FROM t_multi_charset_partkey PARTITION(p2_2) order by 1,2;
ALTER TABLE t_multi_charset_partkey RENAME PARTITION FOR(_gbk'高斯db') TO p3_1;
SELECT * FROM t_multi_charset_partkey PARTITION(p3_1) order by 1,2;
DROP TABLE t_multi_charset_partkey;
-- -- -- utf8mb4
CREATE TABLE t_multi_charset_partkey (part varchar(32) collate utf8mb4_unicode_ci, part2 varchar(32) collate utf8mb4_general_ci, a int)
PARTITION BY LIST COLUMNS(part, part2) (
partition p1 values in(('楂樻柉DB', '楂樻柉db')),
partition p2 values in(('高斯db', '高斯DB'))
);
-- -- -- insert
INSERT INTO t_multi_charset_partkey VALUES(_gbk'高斯DB', _gbk'高斯DB', 1);
INSERT INTO t_multi_charset_partkey VALUES(_gbk'高斯db', _gbk'高斯db', 2);
INSERT INTO t_multi_charset_partkey VALUES(_utf8mb4'高斯DB', _utf8mb4'高斯DB', 3);
INSERT INTO t_multi_charset_partkey VALUES(_utf8mb4'高斯db', _utf8mb4'高斯db', 4);
-- -- -- select
SELECT * FROM t_multi_charset_partkey PARTITION(p1) order by 1,2;
SELECT * FROM t_multi_charset_partkey PARTITION(p2) order by 1,2;
SELECT * FROM t_multi_charset_partkey PARTITION FOR(_gbk'高斯DB', _gbk'高斯db') order by 1,2;
SELECT * FROM t_multi_charset_partkey PARTITION FOR(_utf8mb4'高斯db', _utf8mb4'高斯db') order by 1,2;
-- -- -- partition pruning
EXPLAIN (costs off)
SELECT * FROM t_multi_charset_partkey WHERE part=_gbk'高斯DB' order by 1,2;
SELECT * FROM t_multi_charset_partkey WHERE part=_gbk'高斯DB' order by 1,2;
EXPLAIN (costs off)
SELECT * FROM t_multi_charset_partkey WHERE part=_utf8mb4'高斯db' order by 1,2;
SELECT * FROM t_multi_charset_partkey WHERE part=_utf8mb4'高斯db' order by 1,2;
EXPLAIN (costs off)
SELECT * FROM t_multi_charset_partkey WHERE part=_gbk'高斯DB' collate gbk_chinese_ci order by 1,2; -- ALL PARTS
SELECT * FROM t_multi_charset_partkey WHERE part=_gbk'高斯DB' collate gbk_chinese_ci order by 1,2;
-- -- -- partiton ddl
ALTER TABLE t_multi_charset_partkey RENAME PARTITION FOR(_gbk'高斯DB', _gbk'高斯db') TO p1_1;
SELECT * FROM t_multi_charset_partkey PARTITION(p1_1) order by 1,2;
DROP TABLE t_multi_charset_partkey;
-- -- -- gbk