/*
* pg_ndistinct
* output routine for type pg_ndistinct
*
* Produces a human-readable representation of the value.
*/
Datum
pg_ndistinct_out(PG_FUNCTION_ARGS)
{
bytea *data = PG_GETARG_BYTEA_PP(0);
MVNDistinct *ndist = statext_ndistinct_deserialize(data);
int i;
StringInfoData str;
initStringInfo(&str);
appendStringInfoChar(&str, '{');
for (i = 0; i < ndist->nitems; i++)
{
MVNDistinctItem item = ndist->items[i];
int x = -1;
bool first = true;
if (i > 0)
appendStringInfoString(&str, ", ");
while ((x = bms_next_member(item.attrs, x)) >= 0)
{
appendStringInfo(&str, "%s%d", first ? "\"" : ", ", x);
first = false;
}
appendStringInfo(&str, "\": %d", (int) item.ndistinct);
}
appendStringInfoChar(&str, '}');
PG_RETURN_CSTRING(str.data);
}
/*
* fixup_inherited_columns
*
* When user is querying on a table with children, it implicitly accesses
* child tables also. So, we also need to check security label of child
* tables and columns, but here is no guarantee attribute numbers are
* same between the parent ans children.
* It returns a bitmapset which contains attribute number of the child
* table based on the given bitmapset of the parent.
*/
static Bitmapset *
fixup_inherited_columns(Oid parentId, Oid childId, Bitmapset *columns)
{
Bitmapset *result = NULL;
int index;
/*
* obviously, no need to do anything here
*/
if (parentId == childId)
return columns;
index = -1;
while ((index = bms_next_member(columns, index)) >= 0)
{
/* bit numbers are offset by FirstLowInvalidHeapAttributeNumber */
AttrNumber attno = index + FirstLowInvalidHeapAttributeNumber;
char *attname;
/*
* whole-row-reference shall be fixed-up later
*/
if (attno == InvalidAttrNumber)
{
result = bms_add_member(result, index);
continue;
}
attname = get_attname(parentId, attno);
if (!attname)
elog(ERROR, "cache lookup failed for attribute %d of relation %u",
attno, parentId);
attno = get_attnum(childId, attname);
if (attno == InvalidAttrNumber)
elog(ERROR, "cache lookup failed for attribute %s of relation %u",
attname, childId);
result = bms_add_member(result,
attno - FirstLowInvalidHeapAttributeNumber);
pfree(attname);
}
return result;
}
/*
* ExecScanReScan
*
* This must be called within the ReScan function of any plan node type
* that uses ExecScan().
*/
void
ExecScanReScan(ScanState *node)
{
EState *estate = node->ps.state;
/* Rescan EvalPlanQual tuple if we're inside an EvalPlanQual recheck */
if (estate->es_epqScanDone != NULL)
{
Index scanrelid = ((Scan *) node->ps.plan)->scanrelid;
if (scanrelid > 0)
estate->es_epqScanDone[scanrelid - 1] = false;
else
{
Bitmapset *relids;
int rtindex = -1;
/*
* If an FDW or custom scan provider has replaced the join with a
* scan, there are multiple RTIs; reset the epqScanDone flag for
* all of them.
*/
if (IsA(node->ps.plan, ForeignScan))
relids = ((ForeignScan *) node->ps.plan)->fs_relids;
else if (IsA(node->ps.plan, CustomScan))
relids = ((CustomScan *) node->ps.plan)->custom_relids;
else
elog(ERROR, "unexpected scan node: %d",
(int) nodeTag(node->ps.plan));
while ((rtindex = bms_next_member(relids, rtindex)) >= 0)
{
Assert(rtindex > 0);
estate->es_epqScanDone[rtindex - 1] = false;
}
}
}
}
/*
* statext_ndistinct_serialize
* serialize ndistinct to the on-disk bytea format
*/
bytea *
statext_ndistinct_serialize(MVNDistinct *ndistinct)
{
int i;
bytea *output;
char *tmp;
Size len;
Assert(ndistinct->magic == STATS_NDISTINCT_MAGIC);
Assert(ndistinct->type == STATS_NDISTINCT_TYPE_BASIC);
/*
* Base size is size of scalar fields in the struct, plus one base struct
* for each item, including number of items for each.
*/
len = VARHDRSZ + SizeOfMVNDistinct +
ndistinct->nitems * (offsetof(MVNDistinctItem, attrs) + sizeof(int));
/* and also include space for the actual attribute numbers */
for (i = 0; i < ndistinct->nitems; i++)
{
int nmembers;
nmembers = bms_num_members(ndistinct->items[i].attrs);
Assert(nmembers >= 2);
len += sizeof(AttrNumber) * nmembers;
}
output = (bytea *) palloc(len);
SET_VARSIZE(output, len);
tmp = VARDATA(output);
/* Store the base struct values (magic, type, nitems) */
memcpy(tmp, &ndistinct->magic, sizeof(uint32));
tmp += sizeof(uint32);
memcpy(tmp, &ndistinct->type, sizeof(uint32));
tmp += sizeof(uint32);
memcpy(tmp, &ndistinct->nitems, sizeof(uint32));
tmp += sizeof(uint32);
/*
* store number of attributes and attribute numbers for each ndistinct
* entry
*/
for (i = 0; i < ndistinct->nitems; i++)
{
MVNDistinctItem item = ndistinct->items[i];
int nmembers = bms_num_members(item.attrs);
int x;
memcpy(tmp, &item.ndistinct, sizeof(double));
tmp += sizeof(double);
memcpy(tmp, &nmembers, sizeof(int));
tmp += sizeof(int);
x = -1;
while ((x = bms_next_member(item.attrs, x)) >= 0)
{
AttrNumber value = (AttrNumber) x;
memcpy(tmp, &value, sizeof(AttrNumber));
tmp += sizeof(AttrNumber);
}
Assert(tmp <= ((char *) output + len));
}
return output;
}
/* Take care of joins */
void
pathman_join_pathlist_hook(PlannerInfo *root,
RelOptInfo *joinrel,
RelOptInfo *outerrel,
RelOptInfo *innerrel,
JoinType jointype,
JoinPathExtraData *extra)
{
JoinCostWorkspace workspace;
JoinType saved_jointype = jointype;
RangeTblEntry *inner_rte = root->simple_rte_array[innerrel->relid];
const PartRelationInfo *inner_prel;
List *joinclauses,
*otherclauses;
WalkerContext context;
double paramsel;
Node *part_expr;
ListCell *lc;
/* Call hooks set by other extensions */
if (pathman_set_join_pathlist_next)
pathman_set_join_pathlist_next(root, joinrel, outerrel,
innerrel, jointype, extra);
/* Check that both pg_pathman & RuntimeAppend nodes are enabled */
if (!IsPathmanReady() || !pg_pathman_enable_runtimeappend)
return;
/* We should only consider base relations */
if (innerrel->reloptkind != RELOPT_BASEREL)
return;
/* We shouldn't process tables with active children */
if (inner_rte->inh)
return;
/* We can't handle full or right outer joins */
if (jointype == JOIN_FULL || jointype == JOIN_RIGHT)
return;
/* Check that innerrel is a BASEREL with PartRelationInfo */
if (innerrel->reloptkind != RELOPT_BASEREL ||
!(inner_prel = get_pathman_relation_info(inner_rte->relid)))
return;
/*
* Check if query is:
* 1) UPDATE part_table SET = .. FROM part_table.
* 2) DELETE FROM part_table USING part_table.
*
* Either outerrel or innerrel may be a result relation.
*/
if ((root->parse->resultRelation == outerrel->relid ||
root->parse->resultRelation == innerrel->relid) &&
(root->parse->commandType == CMD_UPDATE ||
root->parse->commandType == CMD_DELETE))
{
int rti = -1,
count = 0;
/* Inner relation must be partitioned */
Assert(inner_prel);
/* Check each base rel of outer relation */
while ((rti = bms_next_member(outerrel->relids, rti)) >= 0)
{
Oid outer_baserel = root->simple_rte_array[rti]->relid;
/* Is it partitioned? */
if (get_pathman_relation_info(outer_baserel))
count++;
}
if (count > 0)
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("DELETE and UPDATE queries with a join "
"of partitioned tables are not supported")));
}
/* Skip if inner table is not allowed to act as parent (e.g. FROM ONLY) */
if (PARENTHOOD_DISALLOWED == get_rel_parenthood_status(inner_rte))
return;
/*
* These codes are used internally in the planner, but are not supported
* by the executor (nor, indeed, by most of the planner).
*/
if (jointype == JOIN_UNIQUE_OUTER || jointype == JOIN_UNIQUE_INNER)
jointype = JOIN_INNER; /* replace with a proper value */
/* Extract join clauses which will separate partitions */
if (IS_OUTER_JOIN(extra->sjinfo->jointype))
{
extract_actual_join_clauses_compat(extra->restrictlist,
joinrel->relids,
&joinclauses,
&otherclauses);
}
else
{
/* We can treat all clauses alike for an inner join */
joinclauses = extract_actual_clauses(extra->restrictlist, false);
otherclauses = NIL;
}
/* Make copy of partitioning expression and fix Var's varno attributes */
part_expr = PrelExpressionForRelid(inner_prel, innerrel->relid);
paramsel = 1.0;
foreach (lc, joinclauses)
{
WrapperNode *wrap;
InitWalkerContext(&context, part_expr, inner_prel, NULL);
wrap = walk_expr_tree((Expr *) lfirst(lc), &context);
paramsel *= wrap->paramsel;
}
/*
* detects functional dependencies between groups of columns
*
* Generates all possible subsets of columns (variations) and computes
* the degree of validity for each one. For example when creating statistics
* on three columns (a,b,c) there are 9 possible dependencies
*
* two columns three columns
* ----------- -------------
* (a) -> b (a,b) -> c
* (a) -> c (a,c) -> b
* (b) -> a (b,c) -> a
* (b) -> c
* (c) -> a
* (c) -> b
*/
MVDependencies *
statext_dependencies_build(int numrows, HeapTuple *rows, Bitmapset *attrs,
VacAttrStats **stats)
{
int i,
j,
k;
int numattrs;
int *attnums;
/* result */
MVDependencies *dependencies = NULL;
numattrs = bms_num_members(attrs);
/*
* Transform the bms into an array, to make accessing i-th member easier.
*/
attnums = palloc(sizeof(int) * bms_num_members(attrs));
i = 0;
j = -1;
while ((j = bms_next_member(attrs, j)) >= 0)
attnums[i++] = j;
Assert(numattrs >= 2);
/*
* We'll try build functional dependencies starting from the smallest ones
* covering just 2 columns, to the largest ones, covering all columns
* included in the statistics object. We start from the smallest ones
* because we want to be able to skip already implied ones.
*/
for (k = 2; k <= numattrs; k++)
{
AttrNumber *dependency; /* array with k elements */
/* prepare a DependencyGenerator of variation */
DependencyGenerator DependencyGenerator = DependencyGenerator_init(numattrs, k);
/* generate all possible variations of k values (out of n) */
while ((dependency = DependencyGenerator_next(DependencyGenerator)))
{
double degree;
MVDependency *d;
/* compute how valid the dependency seems */
degree = dependency_degree(numrows, rows, k, dependency, stats, attrs);
/*
* if the dependency seems entirely invalid, don't store it
*/
if (degree == 0.0)
continue;
d = (MVDependency *) palloc0(offsetof(MVDependency, attributes)
+ k * sizeof(AttrNumber));
/* copy the dependency (and keep the indexes into stxkeys) */
d->degree = degree;
d->nattributes = k;
for (i = 0; i < k; i++)
d->attributes[i] = attnums[dependency[i]];
/* initialize the list of dependencies */
if (dependencies == NULL)
{
dependencies
= (MVDependencies *) palloc0(sizeof(MVDependencies));
dependencies->magic = STATS_DEPS_MAGIC;
dependencies->type = STATS_DEPS_TYPE_BASIC;
dependencies->ndeps = 0;
}
dependencies->ndeps++;
dependencies = (MVDependencies *) repalloc(dependencies,
offsetof(MVDependencies, deps)
+ dependencies->ndeps * sizeof(MVDependency));
dependencies->deps[dependencies->ndeps - 1] = d;
}
/*
* we're done with variations of k elements, so free the
* DependencyGenerator
*/
DependencyGenerator_free(DependencyGenerator);
}
return dependencies;
}
/*
* validates functional dependency on the data
*
* An actual work horse of detecting functional dependencies. Given a variation
* of k attributes, it checks that the first (k-1) are sufficient to determine
* the last one.
*/
static double
dependency_degree(int numrows, HeapTuple *rows, int k, AttrNumber *dependency,
VacAttrStats **stats, Bitmapset *attrs)
{
int i,
j;
int nvalues = numrows * k;
MultiSortSupport mss;
SortItem *items;
Datum *values;
bool *isnull;
int *attnums;
/* counters valid within a group */
int group_size = 0;
int n_violations = 0;
/* total number of rows supporting (consistent with) the dependency */
int n_supporting_rows = 0;
/* Make sure we have at least two input attributes. */
Assert(k >= 2);
/* sort info for all attributes columns */
mss = multi_sort_init(k);
/* data for the sort */
items = (SortItem *) palloc(numrows * sizeof(SortItem));
values = (Datum *) palloc(sizeof(Datum) * nvalues);
isnull = (bool *) palloc(sizeof(bool) * nvalues);
/* fix the pointers to values/isnull */
for (i = 0; i < numrows; i++)
{
items[i].values = &values[i * k];
items[i].isnull = &isnull[i * k];
}
/*
* Transform the bms into an array, to make accessing i-th member easier.
*/
attnums = (int *) palloc(sizeof(int) * bms_num_members(attrs));
i = 0;
j = -1;
while ((j = bms_next_member(attrs, j)) >= 0)
attnums[i++] = j;
/*
* Verify the dependency (a,b,...)->z, using a rather simple algorithm:
*
* (a) sort the data lexicographically
*
* (b) split the data into groups by first (k-1) columns
*
* (c) for each group count different values in the last column
*/
/* prepare the sort function for the first dimension, and SortItem array */
for (i = 0; i < k; i++)
{
VacAttrStats *colstat = stats[dependency[i]];
TypeCacheEntry *type;
type = lookup_type_cache(colstat->attrtypid, TYPECACHE_LT_OPR);
if (type->lt_opr == InvalidOid) /* shouldn't happen */
elog(ERROR, "cache lookup failed for ordering operator for type %u",
colstat->attrtypid);
/* prepare the sort function for this dimension */
multi_sort_add_dimension(mss, i, type->lt_opr);
/* accumulate all the data for both columns into an array and sort it */
for (j = 0; j < numrows; j++)
{
items[j].values[i] =
heap_getattr(rows[j], attnums[dependency[i]],
stats[i]->tupDesc, &items[j].isnull[i]);
}
}
/* sort the items so that we can detect the groups */
qsort_arg((void *) items, numrows, sizeof(SortItem),
multi_sort_compare, mss);
/*
* Walk through the sorted array, split it into rows according to the
* first (k-1) columns. If there's a single value in the last column, we
* count the group as 'supporting' the functional dependency. Otherwise we
* count it as contradicting.
*/
/* start with the first row forming a group */
group_size = 1;
/* loop 1 beyond the end of the array so that we count the final group */
for (i = 1; i <= numrows; i++)
{
/*
* Check if the group ended, which may be either because we processed
* all the items (i==numrows), or because the i-th item is not equal
* to the preceding one.
*/
if (i == numrows ||
multi_sort_compare_dims(0, k - 2, &items[i - 1], &items[i], mss) != 0)
{
/*
* If no violations were found in the group then track the rows of
* the group as supporting the functional dependency.
*/
if (n_violations == 0)
n_supporting_rows += group_size;
/* Reset counters for the new group */
n_violations = 0;
group_size = 1;
continue;
}
/* first columns match, but the last one does not (so contradicting) */
else if (multi_sort_compare_dim(k - 1, &items[i - 1], &items[i], mss) != 0)
n_violations++;
group_size++;
}
pfree(items);
pfree(values);
pfree(isnull);
pfree(mss);
/* Compute the 'degree of validity' as (supporting/total). */
return (n_supporting_rows * 1.0 / numrows);
}
/*
* Open the local relation associated with the remote one.
*
* Optionally rebuilds the Relcache mapping if it was invalidated
* by local DDL.
*/
LogicalRepRelMapEntry *
logicalrep_rel_open(LogicalRepRelId remoteid, LOCKMODE lockmode)
{
LogicalRepRelMapEntry *entry;
bool found;
if (LogicalRepRelMap == NULL)
logicalrep_relmap_init();
/* Search for existing entry. */
entry = hash_search(LogicalRepRelMap, (void *) &remoteid,
HASH_FIND, &found);
if (!found)
elog(ERROR, "no relation map entry for remote relation ID %u",
remoteid);
/* Need to update the local cache? */
if (!OidIsValid(entry->localreloid))
{
Oid relid;
int i;
int found;
Bitmapset *idkey;
TupleDesc desc;
LogicalRepRelation *remoterel;
MemoryContext oldctx;
remoterel = &entry->remoterel;
/* Try to find and lock the relation by name. */
relid = RangeVarGetRelid(makeRangeVar(remoterel->nspname,
remoterel->relname, -1),
lockmode, true);
if (!OidIsValid(relid))
ereport(ERROR,
(errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
errmsg("logical replication target relation \"%s.%s\" does not exist",
remoterel->nspname, remoterel->relname)));
entry->localrel = heap_open(relid, NoLock);
/*
* We currently only support writing to regular and partitioned
* tables.
*/
if (entry->localrel->rd_rel->relkind != RELKIND_RELATION)
ereport(ERROR,
(errcode(ERRCODE_WRONG_OBJECT_TYPE),
errmsg("logical replication target relation \"%s.%s\" is not a table",
remoterel->nspname, remoterel->relname)));
/*
* Build the mapping of local attribute numbers to remote attribute
* numbers and validate that we don't miss any replicated columns
* as that would result in potentially unwanted data loss.
*/
desc = RelationGetDescr(entry->localrel);
oldctx = MemoryContextSwitchTo(LogicalRepRelMapContext);
entry->attrmap = palloc(desc->natts * sizeof(int));
MemoryContextSwitchTo(oldctx);
found = 0;
for (i = 0; i < desc->natts; i++)
{
int attnum = logicalrep_rel_att_by_name(remoterel,
NameStr(desc->attrs[i]->attname));
entry->attrmap[i] = attnum;
if (attnum >= 0)
found++;
}
/* TODO, detail message with names of missing columns */
if (found < remoterel->natts)
ereport(ERROR,
(errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
errmsg("logical replication target relation \"%s.%s\" is missing "
"some replicated columns",
remoterel->nspname, remoterel->relname)));
/*
* Check that replica identity matches. We allow for stricter replica
* identity (fewer columns) on subscriber as that will not stop us
* from finding unique tuple. IE, if publisher has identity
* (id,timestamp) and subscriber just (id) this will not be a problem,
* but in the opposite scenario it will.
*
* Don't throw any error here just mark the relation entry as not
* updatable, as replica identity is only for updates and deletes
* but inserts can be replicated even without it.
*/
entry->updatable = true;
idkey = RelationGetIndexAttrBitmap(entry->localrel,
INDEX_ATTR_BITMAP_IDENTITY_KEY);
/* fallback to PK if no replica identity */
if (idkey == NULL)
{
idkey = RelationGetIndexAttrBitmap(entry->localrel,
INDEX_ATTR_BITMAP_PRIMARY_KEY);
/*
* If no replica identity index and no PK, the published table
* must have replica identity FULL.
*/
if (idkey == NULL && remoterel->replident != REPLICA_IDENTITY_FULL)
entry->updatable = false;
}
i = -1;
while ((i = bms_next_member(idkey, i)) >= 0)
{
int attnum = i + FirstLowInvalidHeapAttributeNumber;
if (!AttrNumberIsForUserDefinedAttr(attnum))
ereport(ERROR,
(errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
errmsg("logical replication target relation \"%s.%s\" uses "
"system columns in REPLICA IDENTITY index",
remoterel->nspname, remoterel->relname)));
attnum = AttrNumberGetAttrOffset(attnum);
if (!bms_is_member(entry->attrmap[attnum], remoterel->attkeys))
{
entry->updatable = false;
break;
}
}
entry->localreloid = relid;
}
else
entry->localrel = heap_open(entry->localreloid, lockmode);
return entry;
}