extended_stats_internal.h File Reference

#include "statistics/statistics.h"
#include "utils/sortsupport.h"

Include dependency graph for extended_stats_internal.h:

This graph shows which files directly or indirectly include this file:

Go to the source code of this file.

Data Structures
struct	StdAnalyzeData
struct	ScalarItem
struct	DimensionInfo
struct	MultiSortSupportData
struct	SortItem
struct	StatsBuildData

Typedefs
typedef struct DimensionInfo	DimensionInfo
typedef struct MultiSortSupportData	MultiSortSupportData
typedef MultiSortSupportData *	MultiSortSupport
typedef struct SortItem	SortItem
typedef struct StatsBuildData	StatsBuildData

Functions
MVNDistinct *	statext_ndistinct_build (double totalrows, StatsBuildData *data)
bytea *	statext_ndistinct_serialize (MVNDistinct *ndistinct)
MVNDistinct *	statext_ndistinct_deserialize (bytea *data)
MVDependencies *	statext_dependencies_build (StatsBuildData *data)
bytea *	statext_dependencies_serialize (MVDependencies *dependencies)
MVDependencies *	statext_dependencies_deserialize (bytea *data)
MCVList *	statext_mcv_build (StatsBuildData *data, double totalrows, int stattarget)
bytea *	statext_mcv_serialize (MCVList *mcvlist, VacAttrStats **stats)
MCVList *	statext_mcv_deserialize (bytea *data)
MultiSortSupport	multi_sort_init (int ndims)
void	multi_sort_add_dimension (MultiSortSupport mss, int sortdim, Oid oper, Oid collation)
int	multi_sort_compare (const void *a, const void *b, void *arg)
int	multi_sort_compare_dim (int dim, const SortItem *a, const SortItem *b, MultiSortSupport mss)
int	multi_sort_compare_dims (int start, int end, const SortItem *a, const SortItem *b, MultiSortSupport mss)
int	compare_scalars_simple (const void *a, const void *b, void *arg)
int	compare_datums_simple (Datum a, Datum b, SortSupport ssup)
AttrNumber *	build_attnums_array (Bitmapset *attrs, int nexprs, int *numattrs)
SortItem *	build_sorted_items (StatsBuildData *data, int *nitems, MultiSortSupport mss, int numattrs, AttrNumber *attnums)
bool	examine_opclause_args (List *args, Node **exprp, Const **cstp, bool *expronleftp)
Selectivity	mcv_combine_selectivities (Selectivity simple_sel, Selectivity mcv_sel, Selectivity mcv_basesel, Selectivity mcv_totalsel)
Selectivity	mcv_clauselist_selectivity (PlannerInfo *root, StatisticExtInfo *stat, List *clauses, int varRelid, JoinType jointype, SpecialJoinInfo *sjinfo, RelOptInfo *rel, Selectivity *basesel, Selectivity *totalsel)
Selectivity	mcv_clause_selectivity_or (PlannerInfo *root, StatisticExtInfo *stat, MCVList *mcv, Node *clause, bool **or_matches, Selectivity *basesel, Selectivity *overlap_mcvsel, Selectivity *overlap_basesel, Selectivity *totalsel)

Typedef Documentation

DimensionInfo

typedef struct DimensionInfo DimensionInfo

MultiSortSupport

typedef MultiSortSupportData* MultiSortSupport

Definition at line 51 of file extended_stats_internal.h.

MultiSortSupportData

typedef struct MultiSortSupportData MultiSortSupportData

SortItem

typedef struct SortItem SortItem

StatsBuildData

typedef struct StatsBuildData StatsBuildData

Function Documentation

build_attnums_array()

AttrNumber* build_attnums_array	(	Bitmapset *	attrs,
		int	nexprs,
		int *	numattrs
	)

Definition at line 942 of file extended_stats.c.

{
    int         i,
                j;
    AttrNumber *attnums;
    int         num = bms_num_members(attrs);
 
    if (numattrs)
        *numattrs = num;
 
    /* build attnums from the bitmapset */
    attnums = (AttrNumber *) palloc(sizeof(AttrNumber) * num);
    i = 0;
    j = -1;
    while ((j = bms_next_member(attrs, j)) >= 0)
    {
        int         attnum = (j - nexprs);
 
        /*
         * Make sure the bitmap contains only user-defined attributes. As
         * bitmaps can't contain negative values, this can be violated in two
         * ways. Firstly, the bitmap might contain 0 as a member, and secondly
         * the integer value might be larger than MaxAttrNumber.
         */
        Assert(AttributeNumberIsValid(attnum));
        Assert(attnum <= MaxAttrNumber);
        Assert(attnum >= (-nexprs));
 
        attnums[i++] = (AttrNumber) attnum;
 
        /* protect against overflows */
        Assert(i <= num);
    }
 
    return attnums;
}

References Assert(), attnum, AttributeNumberIsValid, bms_next_member(), bms_num_members(), i, j, MaxAttrNumber, and palloc().

build_sorted_items()

SortItem* build_sorted_items	(	StatsBuildData *	data,
		int *	nitems,
		MultiSortSupport	mss,
		int	numattrs,
		AttrNumber *	attnums
	)

Definition at line 987 of file extended_stats.c.

{
    int         i,
                j,
                len,
                nrows;
    int         nvalues = data->numrows * numattrs;
 
    SortItem   *items;
    Datum      *values;
    bool       *isnull;
    char       *ptr;
    int        *typlen;
 
    /* Compute the total amount of memory we need (both items and values). */
    len = data->numrows * sizeof(SortItem) + nvalues * (sizeof(Datum) + sizeof(bool));
 
    /* Allocate the memory and split it into the pieces. */
    ptr = palloc0(len);
 
    /* items to sort */
    items = (SortItem *) ptr;
    ptr += data->numrows * sizeof(SortItem);
 
    /* values and null flags */
    values = (Datum *) ptr;
    ptr += nvalues * sizeof(Datum);
 
    isnull = (bool *) ptr;
    ptr += nvalues * sizeof(bool);
 
    /* make sure we consumed the whole buffer exactly */
    Assert((ptr - (char *) items) == len);
 
    /* fix the pointers to Datum and bool arrays */
    nrows = 0;
    for (i = 0; i < data->numrows; i++)
    {
        items[nrows].values = &values[nrows * numattrs];
        items[nrows].isnull = &isnull[nrows * numattrs];
 
        nrows++;
    }
 
    /* build a local cache of typlen for all attributes */
    typlen = (int *) palloc(sizeof(int) * data->nattnums);
    for (i = 0; i < data->nattnums; i++)
        typlen[i] = get_typlen(data->stats[i]->attrtypid);
 
    nrows = 0;
    for (i = 0; i < data->numrows; i++)
    {
        bool        toowide = false;
 
        /* load the values/null flags from sample rows */
        for (j = 0; j < numattrs; j++)
        {
            Datum       value;
            bool        isnull;
            int         attlen;
            AttrNumber  attnum = attnums[j];
 
            int         idx;
 
            /* match attnum to the pre-calculated data */
            for (idx = 0; idx < data->nattnums; idx++)
            {
                if (attnum == data->attnums[idx])
                    break;
            }
 
            Assert(idx < data->nattnums);
 
            value = data->values[idx][i];
            isnull = data->nulls[idx][i];
            attlen = typlen[idx];
 
            /*
             * If this is a varlena value, check if it's too wide and if yes
             * then skip the whole item. Otherwise detoast the value.
             *
             * XXX It may happen that we've already detoasted some preceding
             * values for the current item. We don't bother to cleanup those
             * on the assumption that those are small (below WIDTH_THRESHOLD)
             * and will be discarded at the end of analyze.
             */
            if ((!isnull) && (attlen == -1))
            {
                if (toast_raw_datum_size(value) > WIDTH_THRESHOLD)
                {
                    toowide = true;
                    break;
                }
 
                value = PointerGetDatum(PG_DETOAST_DATUM(value));
            }
 
            items[nrows].values[j] = value;
            items[nrows].isnull[j] = isnull;
        }
 
        if (toowide)
            continue;
 
        nrows++;
    }
 
    /* store the actual number of items (ignoring the too-wide ones) */
    *nitems = nrows;
 
    /* all items were too wide */
    if (nrows == 0)
    {
        /* everything is allocated as a single chunk */
        pfree(items);
        return NULL;
    }
 
    /* do the sort, using the multi-sort */
    qsort_interruptible(items, nrows, sizeof(SortItem),
                        multi_sort_compare, mss);
 
    return items;
}

References Assert(), attlen, attnum, data, get_typlen(), i, idx(), SortItem::isnull, j, len, multi_sort_compare(), nitems, palloc(), palloc0(), pfree(), PG_DETOAST_DATUM, PointerGetDatum(), qsort_interruptible(), toast_raw_datum_size(), value, values, SortItem::values, and WIDTH_THRESHOLD.

Referenced by dependency_degree(), and statext_mcv_build().

compare_datums_simple()

int compare_datums_simple	(	Datum	a,
		Datum	b,
		SortSupport	ssup
	)

Definition at line 928 of file extended_stats.c.

{
    return ApplySortComparator(a, false, b, false, ssup);
}

References a, ApplySortComparator(), and b.

Referenced by compare_scalars_simple(), and statext_mcv_serialize().

compare_scalars_simple()

int compare_scalars_simple	(	const void *	a,
		const void *	b,
		void *	arg
	)

Definition at line 920 of file extended_stats.c.

{
    return compare_datums_simple(*(Datum *) a,
                                 *(Datum *) b,
                                 (SortSupport) arg);
}

References a, arg, b, and compare_datums_simple().

Referenced by statext_mcv_serialize().

examine_opclause_args()

bool examine_opclause_args	(	List *	args,
		Node **	exprp,
		Const **	cstp,
		bool *	expronleftp
	)

Definition at line 2056 of file extended_stats.c.

{
    Node       *expr;
    Const      *cst;
    bool        expronleft;
    Node       *leftop,
               *rightop;
 
    /* enforced by statext_is_compatible_clause_internal */
    Assert(list_length(args) == 2);
 
    leftop = linitial(args);
    rightop = lsecond(args);
 
    /* strip RelabelType from either side of the expression */
    if (IsA(leftop, RelabelType))
        leftop = (Node *) ((RelabelType *) leftop)->arg;
 
    if (IsA(rightop, RelabelType))
        rightop = (Node *) ((RelabelType *) rightop)->arg;
 
    if (IsA(rightop, Const))
    {
        expr = (Node *) leftop;
        cst = (Const *) rightop;
        expronleft = true;
    }
    else if (IsA(leftop, Const))
    {
        expr = (Node *) rightop;
        cst = (Const *) leftop;
        expronleft = false;
    }
    else
        return false;
 
    /* return pointers to the extracted parts if requested */
    if (exprp)
        *exprp = expr;
 
    if (cstp)
        *cstp = cst;
 
    if (expronleftp)
        *expronleftp = expronleft;
 
    return true;
}

References arg, generate_unaccent_rules::args, Assert(), IsA, linitial, list_length(), and lsecond.

Referenced by mcv_get_match_bitmap(), and statext_is_compatible_clause_internal().

mcv_clause_selectivity_or()

Selectivity mcv_clause_selectivity_or	(	PlannerInfo *	root,
		StatisticExtInfo *	stat,
		MCVList *	mcv,
		Node *	clause,
		bool **	or_matches,
		Selectivity *	basesel,
		Selectivity *	overlap_mcvsel,
		Selectivity *	overlap_basesel,
		Selectivity *	totalsel
	)

Definition at line 2130 of file mcv.c.

{
    Selectivity s = 0.0;
    bool       *new_matches;
    int         i;
 
    /* build the OR-matches bitmap, if not built already */
    if (*or_matches == NULL)
        *or_matches = palloc0(sizeof(bool) * mcv->nitems);
 
    /* build the match bitmap for the new clause */
    new_matches = mcv_get_match_bitmap(root, list_make1(clause), stat->keys,
                                       stat->exprs, mcv, false);
 
    /*
     * Sum the frequencies for all the MCV items matching this clause and also
     * those matching the overlap between this clause and any of the preceding
     * clauses as described above.
     */
    *basesel = 0.0;
    *overlap_mcvsel = 0.0;
    *overlap_basesel = 0.0;
    *totalsel = 0.0;
    for (i = 0; i < mcv->nitems; i++)
    {
        *totalsel += mcv->items[i].frequency;
 
        if (new_matches[i])
        {
            s += mcv->items[i].frequency;
            *basesel += mcv->items[i].base_frequency;
 
            if ((*or_matches)[i])
            {
                *overlap_mcvsel += mcv->items[i].frequency;
                *overlap_basesel += mcv->items[i].base_frequency;
            }
        }
 
        /* update the OR-matches bitmap for the next clause */
        (*or_matches)[i] = (*or_matches)[i] || new_matches[i];
    }
 
    pfree(new_matches);
 
    return s;
}

References MCVItem::base_frequency, MCVItem::frequency, i, MCVList::items, list_make1, mcv_get_match_bitmap(), MCVList::nitems, palloc0(), and pfree().

Referenced by statext_mcv_clauselist_selectivity().

mcv_clauselist_selectivity()

Selectivity mcv_clauselist_selectivity	(	PlannerInfo *	root,
		StatisticExtInfo *	stat,
		List *	clauses,
		int	varRelid,
		JoinType	jointype,
		SpecialJoinInfo *	sjinfo,
		RelOptInfo *	rel,
		Selectivity *	basesel,
		Selectivity *	totalsel
	)

Definition at line 2052 of file mcv.c.

{
    int         i;
    MCVList    *mcv;
    Selectivity s = 0.0;
    RangeTblEntry *rte = root->simple_rte_array[rel->relid];
 
    /* match/mismatch bitmap for each MCV item */
    bool       *matches = NULL;
 
    /* load the MCV list stored in the statistics object */
    mcv = statext_mcv_load(stat->statOid, rte->inh);
 
    /* build a match bitmap for the clauses */
    matches = mcv_get_match_bitmap(root, clauses, stat->keys, stat->exprs,
                                   mcv, false);
 
    /* sum frequencies for all the matching MCV items */
    *basesel = 0.0;
    *totalsel = 0.0;
    for (i = 0; i < mcv->nitems; i++)
    {
        *totalsel += mcv->items[i].frequency;
 
        if (matches[i] != false)
        {
            *basesel += mcv->items[i].base_frequency;
            s += mcv->items[i].frequency;
        }
    }
 
    return s;
}

References MCVItem::base_frequency, MCVItem::frequency, i, RangeTblEntry::inh, MCVList::items, mcv_get_match_bitmap(), MCVList::nitems, RelOptInfo::relid, and statext_mcv_load().

Referenced by statext_mcv_clauselist_selectivity().

mcv_combine_selectivities()

Selectivity mcv_combine_selectivities	(	Selectivity	simple_sel,
		Selectivity	mcv_sel,
		Selectivity	mcv_basesel,
		Selectivity	mcv_totalsel
	)

Definition at line 2010 of file mcv.c.

{
    Selectivity other_sel;
    Selectivity sel;
 
    /* estimated selectivity of values not covered by MCV matches */
    other_sel = simple_sel - mcv_basesel;
    CLAMP_PROBABILITY(other_sel);
 
    /* this non-MCV selectivity cannot exceed 1 - mcv_totalsel */
    if (other_sel > 1.0 - mcv_totalsel)
        other_sel = 1.0 - mcv_totalsel;
 
    /* overall selectivity is the sum of the MCV and non-MCV parts */
    sel = mcv_sel + other_sel;
    CLAMP_PROBABILITY(sel);
 
    return sel;
}

References CLAMP_PROBABILITY.

Referenced by statext_mcv_clauselist_selectivity().

multi_sort_add_dimension()

void multi_sort_add_dimension	(	MultiSortSupport	mss,
		int	sortdim,
		Oid	oper,
		Oid	collation
	)

Definition at line 852 of file extended_stats.c.

{
    SortSupport ssup = &mss->ssup[sortdim];
 
    ssup->ssup_cxt = CurrentMemoryContext;
    ssup->ssup_collation = collation;
    ssup->ssup_nulls_first = false;
 
    PrepareSortSupportFromOrderingOp(oper, ssup);
}

References CurrentMemoryContext, oper(), PrepareSortSupportFromOrderingOp(), MultiSortSupportData::ssup, SortSupportData::ssup_collation, SortSupportData::ssup_cxt, and SortSupportData::ssup_nulls_first.

Referenced by build_mss(), dependency_degree(), and ndistinct_for_combination().

multi_sort_compare()

int multi_sort_compare	(	const void *	a,
		const void *	b,
		void *	arg
	)

Definition at line 866 of file extended_stats.c.

{
    MultiSortSupport mss = (MultiSortSupport) arg;
    SortItem   *ia = (SortItem *) a;
    SortItem   *ib = (SortItem *) b;
    int         i;
 
    for (i = 0; i < mss->ndims; i++)
    {
        int         compare;
 
        compare = ApplySortComparator(ia->values[i], ia->isnull[i],
                                      ib->values[i], ib->isnull[i],
                                      &mss->ssup[i]);
 
        if (compare != 0)
            return compare;
    }
 
    /* equal by default */
    return 0;
}

References a, ApplySortComparator(), arg, b, compare(), i, SortItem::isnull, MultiSortSupportData::ndims, MultiSortSupportData::ssup, and SortItem::values.

Referenced by build_distinct_groups(), build_sorted_items(), count_distinct_groups(), ndistinct_for_combination(), and statext_mcv_build().

multi_sort_compare_dim()

int multi_sort_compare_dim	(	int	dim,
		const SortItem *	a,
		const SortItem *	b,
		MultiSortSupport	mss
	)

Definition at line 891 of file extended_stats.c.

{
    return ApplySortComparator(a->values[dim], a->isnull[dim],
                               b->values[dim], b->isnull[dim],
                               &mss->ssup[dim]);
}

References a, ApplySortComparator(), b, and MultiSortSupportData::ssup.

Referenced by dependency_degree().

multi_sort_compare_dims()

int multi_sort_compare_dims	(	int	start,
		int	end,
		const SortItem *	a,
		const SortItem *	b,
		MultiSortSupport	mss
	)

Definition at line 900 of file extended_stats.c.

{
    int         dim;
 
    for (dim = start; dim <= end; dim++)
    {
        int         r = ApplySortComparator(a->values[dim], a->isnull[dim],
                                            b->values[dim], b->isnull[dim],
                                            &mss->ssup[dim]);
 
        if (r != 0)
            return r;
    }
 
    return 0;
}

References a, ApplySortComparator(), b, and MultiSortSupportData::ssup.

Referenced by dependency_degree().

multi_sort_init()

MultiSortSupport multi_sort_init

(

int

ndims

)

Definition at line 833 of file extended_stats.c.

{
    MultiSortSupport mss;
 
    Assert(ndims >= 2);
 
    mss = (MultiSortSupport) palloc0(offsetof(MultiSortSupportData, ssup)
                                     + sizeof(SortSupportData) * ndims);
 
    mss->ndims = ndims;
 
    return mss;
}

References Assert(), MultiSortSupportData::ndims, and palloc0().

Referenced by build_mss(), dependency_degree(), and ndistinct_for_combination().

statext_dependencies_build()

MVDependencies* statext_dependencies_build

(

StatsBuildData *

data

)

Definition at line 350 of file dependencies.c.

{
    int         i,
                k;
 
    /* result */
    MVDependencies *dependencies = NULL;
    MemoryContext cxt;
 
    Assert(data->nattnums >= 2);
 
    /* tracks memory allocated by dependency_degree calls */
    cxt = AllocSetContextCreate(CurrentMemoryContext,
                                "dependency_degree cxt",
                                ALLOCSET_DEFAULT_SIZES);
 
    /*
     * 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 <= data->nattnums; k++)
    {
        AttrNumber *dependency; /* array with k elements */
 
        /* prepare a DependencyGenerator of variation */
        DependencyGenerator DependencyGenerator = DependencyGenerator_init(data->nattnums, k);
 
        /* generate all possible variations of k values (out of n) */
        while ((dependency = DependencyGenerator_next(DependencyGenerator)))
        {
            double      degree;
            MVDependency *d;
            MemoryContext oldcxt;
 
            /* release memory used by dependency degree calculation */
            oldcxt = MemoryContextSwitchTo(cxt);
 
            /* compute how valid the dependency seems */
            degree = dependency_degree(data, k, dependency);
 
            MemoryContextSwitchTo(oldcxt);
            MemoryContextReset(cxt);
 
            /*
             * 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] = data->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);
    }
 
    MemoryContextDelete(cxt);
 
    return dependencies;
}

References ALLOCSET_DEFAULT_SIZES, AllocSetContextCreate, Assert(), MVDependency::attributes, CurrentMemoryContext, data, MVDependency::degree, dependency_degree(), DependencyGenerator_free(), DependencyGenerator_init(), DependencyGenerator_next(), MVDependencies::deps, i, if(), MVDependencies::magic, MemoryContextDelete(), MemoryContextReset(), MemoryContextSwitchTo(), MVDependency::nattributes, MVDependencies::ndeps, palloc0(), repalloc(), STATS_DEPS_MAGIC, STATS_DEPS_TYPE_BASIC, and MVDependencies::type.

Referenced by BuildRelationExtStatistics().

statext_dependencies_deserialize()

MVDependencies* statext_dependencies_deserialize

(

bytea *

data

)

Definition at line 501 of file dependencies.c.

{
    int         i;
    Size        min_expected_size;
    MVDependencies *dependencies;
    char       *tmp;
 
    if (data == NULL)
        return NULL;
 
    if (VARSIZE_ANY_EXHDR(data) < SizeOfHeader)
        elog(ERROR, "invalid MVDependencies size %zu (expected at least %zu)",
             VARSIZE_ANY_EXHDR(data), SizeOfHeader);
 
    /* read the MVDependencies header */
    dependencies = (MVDependencies *) palloc0(sizeof(MVDependencies));
 
    /* initialize pointer to the data part (skip the varlena header) */
    tmp = VARDATA_ANY(data);
 
    /* read the header fields and perform basic sanity checks */
    memcpy(&dependencies->magic, tmp, sizeof(uint32));
    tmp += sizeof(uint32);
    memcpy(&dependencies->type, tmp, sizeof(uint32));
    tmp += sizeof(uint32);
    memcpy(&dependencies->ndeps, tmp, sizeof(uint32));
    tmp += sizeof(uint32);
 
    if (dependencies->magic != STATS_DEPS_MAGIC)
        elog(ERROR, "invalid dependency magic %d (expected %d)",
             dependencies->magic, STATS_DEPS_MAGIC);
 
    if (dependencies->type != STATS_DEPS_TYPE_BASIC)
        elog(ERROR, "invalid dependency type %d (expected %d)",
             dependencies->type, STATS_DEPS_TYPE_BASIC);
 
    if (dependencies->ndeps == 0)
        elog(ERROR, "invalid zero-length item array in MVDependencies");
 
    /* what minimum bytea size do we expect for those parameters */
    min_expected_size = SizeOfItem(dependencies->ndeps);
 
    if (VARSIZE_ANY_EXHDR(data) < min_expected_size)
        elog(ERROR, "invalid dependencies size %zu (expected at least %zu)",
             VARSIZE_ANY_EXHDR(data), min_expected_size);
 
    /* allocate space for the MCV items */
    dependencies = repalloc(dependencies, offsetof(MVDependencies, deps)
                            + (dependencies->ndeps * sizeof(MVDependency *)));
 
    for (i = 0; i < dependencies->ndeps; i++)
    {
        double      degree;
        AttrNumber  k;
        MVDependency *d;
 
        /* degree of validity */
        memcpy(&degree, tmp, sizeof(double));
        tmp += sizeof(double);
 
        /* number of attributes */
        memcpy(&k, tmp, sizeof(AttrNumber));
        tmp += sizeof(AttrNumber);
 
        /* is the number of attributes valid? */
        Assert((k >= 2) && (k <= STATS_MAX_DIMENSIONS));
 
        /* now that we know the number of attributes, allocate the dependency */
        d = (MVDependency *) palloc0(offsetof(MVDependency, attributes)
                                     + (k * sizeof(AttrNumber)));
 
        d->degree = degree;
        d->nattributes = k;
 
        /* copy attribute numbers */
        memcpy(d->attributes, tmp, sizeof(AttrNumber) * d->nattributes);
        tmp += sizeof(AttrNumber) * d->nattributes;
 
        dependencies->deps[i] = d;
 
        /* still within the bytea */
        Assert(tmp <= ((char *) data + VARSIZE_ANY(data)));
    }
 
    /* we should have consumed the whole bytea exactly */
    Assert(tmp == ((char *) data + VARSIZE_ANY(data)));
 
    return dependencies;
}

References Assert(), MVDependency::attributes, data, MVDependency::degree, MVDependencies::deps, elog(), ERROR, i, MVDependencies::magic, MVDependency::nattributes, MVDependencies::ndeps, palloc0(), repalloc(), SizeOfHeader, SizeOfItem, STATS_DEPS_MAGIC, STATS_DEPS_TYPE_BASIC, STATS_MAX_DIMENSIONS, MVDependencies::type, VARDATA_ANY, VARSIZE_ANY, and VARSIZE_ANY_EXHDR.

Referenced by pg_dependencies_out(), and statext_dependencies_load().

statext_dependencies_serialize()

bytea* statext_dependencies_serialize

(

MVDependencies *

dependencies

)

Definition at line 446 of file dependencies.c.

{
    int         i;
    bytea      *output;
    char       *tmp;
    Size        len;
 
    /* we need to store ndeps, with a number of attributes for each one */
    len = VARHDRSZ + SizeOfHeader;
 
    /* and also include space for the actual attribute numbers and degrees */
    for (i = 0; i < dependencies->ndeps; i++)
        len += SizeOfItem(dependencies->deps[i]->nattributes);
 
    output = (bytea *) palloc0(len);
    SET_VARSIZE(output, len);
 
    tmp = VARDATA(output);
 
    /* Store the base struct values (magic, type, ndeps) */
    memcpy(tmp, &dependencies->magic, sizeof(uint32));
    tmp += sizeof(uint32);
    memcpy(tmp, &dependencies->type, sizeof(uint32));
    tmp += sizeof(uint32);
    memcpy(tmp, &dependencies->ndeps, sizeof(uint32));
    tmp += sizeof(uint32);
 
    /* store number of attributes and attribute numbers for each dependency */
    for (i = 0; i < dependencies->ndeps; i++)
    {
        MVDependency *d = dependencies->deps[i];
 
        memcpy(tmp, &d->degree, sizeof(double));
        tmp += sizeof(double);
 
        memcpy(tmp, &d->nattributes, sizeof(AttrNumber));
        tmp += sizeof(AttrNumber);
 
        memcpy(tmp, d->attributes, sizeof(AttrNumber) * d->nattributes);
        tmp += sizeof(AttrNumber) * d->nattributes;
 
        /* protect against overflow */
        Assert(tmp <= ((char *) output + len));
    }
 
    /* make sure we've produced exactly the right amount of data */
    Assert(tmp == ((char *) output + len));
 
    return output;
}

References Assert(), MVDependency::attributes, MVDependency::degree, MVDependencies::deps, i, len, MVDependencies::magic, MVDependency::nattributes, MVDependencies::ndeps, output, palloc0(), SET_VARSIZE, SizeOfHeader, SizeOfItem, MVDependencies::type, VARDATA, and VARHDRSZ.

Referenced by statext_store().

statext_mcv_build()

MCVList* statext_mcv_build	(	StatsBuildData *	data,
		double	totalrows,
		int	stattarget
	)

Definition at line 184 of file mcv.c.

{
    int         i,
                numattrs,
                numrows,
                ngroups,
                nitems;
    double      mincount;
    SortItem   *items;
    SortItem   *groups;
    MCVList    *mcvlist = NULL;
    MultiSortSupport mss;
 
    /* comparator for all the columns */
    mss = build_mss(data);
 
    /* sort the rows */
    items = build_sorted_items(data, &nitems, mss,
                               data->nattnums, data->attnums);
 
    if (!items)
        return NULL;
 
    /* for convenience */
    numattrs = data->nattnums;
    numrows = data->numrows;
 
    /* transform the sorted rows into groups (sorted by frequency) */
    groups = build_distinct_groups(nitems, items, mss, &ngroups);
 
    /*
     * The maximum number of MCV items to store, based on the statistics
     * target we computed for the statistics object (from the target set for
     * the object itself, attributes and the system default). In any case, we
     * can't keep more groups than we have available.
     */
    nitems = stattarget;
    if (nitems > ngroups)
        nitems = ngroups;
 
    /*
     * Decide how many items to keep in the MCV list. We can't use the same
     * algorithm as per-column MCV lists, because that only considers the
     * actual group frequency - but we're primarily interested in how the
     * actual frequency differs from the base frequency (product of simple
     * per-column frequencies, as if the columns were independent).
     *
     * Using the same algorithm might exclude items that are close to the
     * "average" frequency of the sample. But that does not say whether the
     * observed frequency is close to the base frequency or not. We also need
     * to consider unexpectedly uncommon items (again, compared to the base
     * frequency), and the single-column algorithm does not have to.
     *
     * We simply decide how many items to keep by computing the minimum count
     * using get_mincount_for_mcv_list() and then keep all items that seem to
     * be more common than that.
     */
    mincount = get_mincount_for_mcv_list(numrows, totalrows);
 
    /*
     * Walk the groups until we find the first group with a count below the
     * mincount threshold (the index of that group is the number of groups we
     * want to keep).
     */
    for (i = 0; i < nitems; i++)
    {
        if (groups[i].count < mincount)
        {
            nitems = i;
            break;
        }
    }
 
    /*
     * At this point, we know the number of items for the MCV list. There
     * might be none (for uniform distribution with many groups), and in that
     * case, there will be no MCV list. Otherwise, construct the MCV list.
     */
    if (nitems > 0)
    {
        int         j;
        SortItem    key;
        MultiSortSupport tmp;
 
        /* frequencies for values in each attribute */
        SortItem  **freqs;
        int        *nfreqs;
 
        /* used to search values */
        tmp = (MultiSortSupport) palloc(offsetof(MultiSortSupportData, ssup)
                                        + sizeof(SortSupportData));
 
        /* compute frequencies for values in each column */
        nfreqs = (int *) palloc0(sizeof(int) * numattrs);
        freqs = build_column_frequencies(groups, ngroups, mss, nfreqs);
 
        /*
         * Allocate the MCV list structure, set the global parameters.
         */
        mcvlist = (MCVList *) palloc0(offsetof(MCVList, items) +
                                      sizeof(MCVItem) * nitems);
 
        mcvlist->magic = STATS_MCV_MAGIC;
        mcvlist->type = STATS_MCV_TYPE_BASIC;
        mcvlist->ndimensions = numattrs;
        mcvlist->nitems = nitems;
 
        /* store info about data type OIDs */
        for (i = 0; i < numattrs; i++)
            mcvlist->types[i] = data->stats[i]->attrtypid;
 
        /* Copy the first chunk of groups into the result. */
        for (i = 0; i < nitems; i++)
        {
            /* just point to the proper place in the list */
            MCVItem    *item = &mcvlist->items[i];
 
            item->values = (Datum *) palloc(sizeof(Datum) * numattrs);
            item->isnull = (bool *) palloc(sizeof(bool) * numattrs);
 
            /* copy values for the group */
            memcpy(item->values, groups[i].values, sizeof(Datum) * numattrs);
            memcpy(item->isnull, groups[i].isnull, sizeof(bool) * numattrs);
 
            /* groups should be sorted by frequency in descending order */
            Assert((i == 0) || (groups[i - 1].count >= groups[i].count));
 
            /* group frequency */
            item->frequency = (double) groups[i].count / numrows;
 
            /* base frequency, if the attributes were independent */
            item->base_frequency = 1.0;
            for (j = 0; j < numattrs; j++)
            {
                SortItem   *freq;
 
                /* single dimension */
                tmp->ndims = 1;
                tmp->ssup[0] = mss->ssup[j];
 
                /* fill search key */
                key.values = &groups[i].values[j];
                key.isnull = &groups[i].isnull[j];
 
                freq = (SortItem *) bsearch_arg(&key, freqs[j], nfreqs[j],
                                                sizeof(SortItem),
                                                multi_sort_compare, tmp);
 
                item->base_frequency *= ((double) freq->count) / numrows;
            }
        }
 
        pfree(nfreqs);
        pfree(freqs);
    }
 
    pfree(items);
    pfree(groups);
 
    return mcvlist;
}

References Assert(), MCVItem::base_frequency, bsearch_arg(), build_column_frequencies(), build_distinct_groups(), build_mss(), build_sorted_items(), SortItem::count, data, MCVItem::frequency, get_mincount_for_mcv_list(), i, SortItem::isnull, MCVItem::isnull, MCVList::items, j, sort-test::key, MCVList::magic, multi_sort_compare(), MCVList::ndimensions, MultiSortSupportData::ndims, MCVList::nitems, nitems, palloc(), palloc0(), pfree(), MultiSortSupportData::ssup, STATS_MCV_MAGIC, STATS_MCV_TYPE_BASIC, MCVList::type, MCVList::types, SortItem::values, and MCVItem::values.

Referenced by BuildRelationExtStatistics().

statext_mcv_deserialize()

MCVList* statext_mcv_deserialize

(

bytea *

data

)

Definition at line 1000 of file mcv.c.

{
    int         dim,
                i;
    Size        expected_size;
    MCVList    *mcvlist;
    char       *raw;
    char       *ptr;
    char       *endptr PG_USED_FOR_ASSERTS_ONLY;
 
    int         ndims,
                nitems;
    DimensionInfo *info = NULL;
 
    /* local allocation buffer (used only for deserialization) */
    Datum     **map = NULL;
 
    /* MCV list */
    Size        mcvlen;
 
    /* buffer used for the result */
    Size        datalen;
    char       *dataptr;
    char       *valuesptr;
    char       *isnullptr;
 
    if (data == NULL)
        return NULL;
 
    /*
     * We can't possibly deserialize a MCV list if there's not even a complete
     * header. We need an explicit formula here, because we serialize the
     * header fields one by one, so we need to ignore struct alignment.
     */
    if (VARSIZE_ANY(data) < MinSizeOfMCVList)
        elog(ERROR, "invalid MCV size %zu (expected at least %zu)",
             VARSIZE_ANY(data), MinSizeOfMCVList);
 
    /* read the MCV list header */
    mcvlist = (MCVList *) palloc0(offsetof(MCVList, items));
 
    /* pointer to the data part (skip the varlena header) */
    raw = (char *) data;
    ptr = VARDATA_ANY(raw);
    endptr = (char *) raw + VARSIZE_ANY(data);
 
    /* get the header and perform further sanity checks */
    memcpy(&mcvlist->magic, ptr, sizeof(uint32));
    ptr += sizeof(uint32);
 
    memcpy(&mcvlist->type, ptr, sizeof(uint32));
    ptr += sizeof(uint32);
 
    memcpy(&mcvlist->nitems, ptr, sizeof(uint32));
    ptr += sizeof(uint32);
 
    memcpy(&mcvlist->ndimensions, ptr, sizeof(AttrNumber));
    ptr += sizeof(AttrNumber);
 
    if (mcvlist->magic != STATS_MCV_MAGIC)
        elog(ERROR, "invalid MCV magic %u (expected %u)",
             mcvlist->magic, STATS_MCV_MAGIC);
 
    if (mcvlist->type != STATS_MCV_TYPE_BASIC)
        elog(ERROR, "invalid MCV type %u (expected %u)",
             mcvlist->type, STATS_MCV_TYPE_BASIC);
 
    if (mcvlist->ndimensions == 0)
        elog(ERROR, "invalid zero-length dimension array in MCVList");
    else if ((mcvlist->ndimensions > STATS_MAX_DIMENSIONS) ||
             (mcvlist->ndimensions < 0))
        elog(ERROR, "invalid length (%d) dimension array in MCVList",
             mcvlist->ndimensions);
 
    if (mcvlist->nitems == 0)
        elog(ERROR, "invalid zero-length item array in MCVList");
    else if (mcvlist->nitems > STATS_MCVLIST_MAX_ITEMS)
        elog(ERROR, "invalid length (%u) item array in MCVList",
             mcvlist->nitems);
 
    nitems = mcvlist->nitems;
    ndims = mcvlist->ndimensions;
 
    /*
     * Check amount of data including DimensionInfo for all dimensions and
     * also the serialized items (including uint16 indexes). Also, walk
     * through the dimension information and add it to the sum.
     */
    expected_size = SizeOfMCVList(ndims, nitems);
 
    /*
     * Check that we have at least the dimension and info records, along with
     * the items. We don't know the size of the serialized values yet. We need
     * to do this check first, before accessing the dimension info.
     */
    if (VARSIZE_ANY(data) < expected_size)
        elog(ERROR, "invalid MCV size %zu (expected %zu)",
             VARSIZE_ANY(data), expected_size);
 
    /* Now copy the array of type Oids. */
    memcpy(mcvlist->types, ptr, sizeof(Oid) * ndims);
    ptr += (sizeof(Oid) * ndims);
 
    /* Now it's safe to access the dimension info. */
    info = palloc(ndims * sizeof(DimensionInfo));
 
    memcpy(info, ptr, ndims * sizeof(DimensionInfo));
    ptr += (ndims * sizeof(DimensionInfo));
 
    /* account for the value arrays */
    for (dim = 0; dim < ndims; dim++)
    {
        /*
         * XXX I wonder if we can/should rely on asserts here. Maybe those
         * checks should be done every time?
         */
        Assert(info[dim].nvalues >= 0);
        Assert(info[dim].nbytes >= 0);
 
        expected_size += info[dim].nbytes;
    }
 
    /*
     * Now we know the total expected MCV size, including all the pieces
     * (header, dimension info. items and deduplicated data). So do the final
     * check on size.
     */
    if (VARSIZE_ANY(data) != expected_size)
        elog(ERROR, "invalid MCV size %zu (expected %zu)",
             VARSIZE_ANY(data), expected_size);
 
    /*
     * We need an array of Datum values for each dimension, so that we can
     * easily translate the uint16 indexes later. We also need a top-level
     * array of pointers to those per-dimension arrays.
     *
     * While allocating the arrays for dimensions, compute how much space we
     * need for a copy of the by-ref data, as we can't simply point to the
     * original values (it might go away).
     */
    datalen = 0;                /* space for by-ref data */
    map = (Datum **) palloc(ndims * sizeof(Datum *));
 
    for (dim = 0; dim < ndims; dim++)
    {
        map[dim] = (Datum *) palloc(sizeof(Datum) * info[dim].nvalues);
 
        /* space needed for a copy of data for by-ref types */
        datalen += info[dim].nbytes_aligned;
    }
 
    /*
     * Now resize the MCV list so that the allocation includes all the data.
     *
     * Allocate space for a copy of the data, as we can't simply reference the
     * serialized data - it's not aligned properly, and it may disappear while
     * we're still using the MCV list, e.g. due to catcache release.
     *
     * We do care about alignment here, because we will allocate all the
     * pieces at once, but then use pointers to different parts.
     */
    mcvlen = MAXALIGN(offsetof(MCVList, items) + (sizeof(MCVItem) * nitems));
 
    /* arrays of values and isnull flags for all MCV items */
    mcvlen += nitems * MAXALIGN(sizeof(Datum) * ndims);
    mcvlen += nitems * MAXALIGN(sizeof(bool) * ndims);
 
    /* we don't quite need to align this, but it makes some asserts easier */
    mcvlen += MAXALIGN(datalen);
 
    /* now resize the deserialized MCV list, and compute pointers to parts */
    mcvlist = repalloc(mcvlist, mcvlen);
 
    /* pointer to the beginning of values/isnull arrays */
    valuesptr = (char *) mcvlist
        + MAXALIGN(offsetof(MCVList, items) + (sizeof(MCVItem) * nitems));
 
    isnullptr = valuesptr + (nitems * MAXALIGN(sizeof(Datum) * ndims));
 
    dataptr = isnullptr + (nitems * MAXALIGN(sizeof(bool) * ndims));
 
    /*
     * Build mapping (index => value) for translating the serialized data into
     * the in-memory representation.
     */
    for (dim = 0; dim < ndims; dim++)
    {
        /* remember start position in the input array */
        char       *start PG_USED_FOR_ASSERTS_ONLY = ptr;
 
        if (info[dim].typbyval)
        {
            /* for by-val types we simply copy data into the mapping */
            for (i = 0; i < info[dim].nvalues; i++)
            {
                Datum       v = 0;
 
                memcpy(&v, ptr, info[dim].typlen);
                ptr += info[dim].typlen;
 
                map[dim][i] = fetch_att(&v, true, info[dim].typlen);
 
                /* no under/overflow of input array */
                Assert(ptr <= (start + info[dim].nbytes));
            }
        }
        else
        {
            /* for by-ref types we need to also make a copy of the data */
 
            /* passed by reference, but fixed length (name, tid, ...) */
            if (info[dim].typlen > 0)
            {
                for (i = 0; i < info[dim].nvalues; i++)
                {
                    memcpy(dataptr, ptr, info[dim].typlen);
                    ptr += info[dim].typlen;
 
                    /* just point into the array */
                    map[dim][i] = PointerGetDatum(dataptr);
                    dataptr += MAXALIGN(info[dim].typlen);
                }
            }
            else if (info[dim].typlen == -1)
            {
                /* varlena */
                for (i = 0; i < info[dim].nvalues; i++)
                {
                    uint32      len;
 
                    /* read the uint32 length */
                    memcpy(&len, ptr, sizeof(uint32));
                    ptr += sizeof(uint32);
 
                    /* the length is data-only */
                    SET_VARSIZE(dataptr, len + VARHDRSZ);
                    memcpy(VARDATA(dataptr), ptr, len);
                    ptr += len;
 
                    /* just point into the array */
                    map[dim][i] = PointerGetDatum(dataptr);
 
                    /* skip to place of the next deserialized value */
                    dataptr += MAXALIGN(len + VARHDRSZ);
                }
            }
            else if (info[dim].typlen == -2)
            {
                /* cstring */
                for (i = 0; i < info[dim].nvalues; i++)
                {
                    uint32      len;
 
                    memcpy(&len, ptr, sizeof(uint32));
                    ptr += sizeof(uint32);
 
                    memcpy(dataptr, ptr, len);
                    ptr += len;
 
                    /* just point into the array */
                    map[dim][i] = PointerGetDatum(dataptr);
                    dataptr += MAXALIGN(len);
                }
            }
 
            /* no under/overflow of input array */
            Assert(ptr <= (start + info[dim].nbytes));
 
            /* no overflow of the output mcv value */
            Assert(dataptr <= ((char *) mcvlist + mcvlen));
        }
 
        /* check we consumed input data for this dimension exactly */
        Assert(ptr == (start + info[dim].nbytes));
    }
 
    /* we should have also filled the MCV list exactly */
    Assert(dataptr == ((char *) mcvlist + mcvlen));
 
    /* deserialize the MCV items and translate the indexes to Datums */
    for (i = 0; i < nitems; i++)
    {
        MCVItem    *item = &mcvlist->items[i];
 
        item->values = (Datum *) valuesptr;
        valuesptr += MAXALIGN(sizeof(Datum) * ndims);
 
        item->isnull = (bool *) isnullptr;
        isnullptr += MAXALIGN(sizeof(bool) * ndims);
 
        memcpy(item->isnull, ptr, sizeof(bool) * ndims);
        ptr += sizeof(bool) * ndims;
 
        memcpy(&item->frequency, ptr, sizeof(double));
        ptr += sizeof(double);
 
        memcpy(&item->base_frequency, ptr, sizeof(double));
        ptr += sizeof(double);
 
        /* finally translate the indexes (for non-NULL only) */
        for (dim = 0; dim < ndims; dim++)
        {
            uint16      index;
 
            memcpy(&index, ptr, sizeof(uint16));
            ptr += sizeof(uint16);
 
            if (item->isnull[dim])
                continue;
 
            item->values[dim] = map[dim][index];
        }
 
        /* check we're not overflowing the input */
        Assert(ptr <= endptr);
    }
 
    /* check that we processed all the data */
    Assert(ptr == endptr);
 
    /* release the buffers used for mapping */
    for (dim = 0; dim < ndims; dim++)
        pfree(map[dim]);
 
    pfree(map);
 
    return mcvlist;
}

References Assert(), MCVItem::base_frequency, data, elog(), ERROR, fetch_att(), MCVItem::frequency, i, MCVItem::isnull, MCVList::items, len, MCVList::magic, MAXALIGN, MinSizeOfMCVList, DimensionInfo::nbytes, DimensionInfo::nbytes_aligned, MCVList::ndimensions, MCVList::nitems, nitems, DimensionInfo::nvalues, palloc(), palloc0(), pfree(), PG_USED_FOR_ASSERTS_ONLY, PointerGetDatum(), repalloc(), SET_VARSIZE, SizeOfMCVList, STATS_MAX_DIMENSIONS, STATS_MCV_MAGIC, STATS_MCV_TYPE_BASIC, STATS_MCVLIST_MAX_ITEMS, MCVList::type, MCVList::types, DimensionInfo::typlen, MCVItem::values, VARDATA, VARDATA_ANY, VARHDRSZ, and VARSIZE_ANY.

Referenced by pg_stats_ext_mcvlist_items(), and statext_mcv_load().

statext_mcv_serialize()

bytea* statext_mcv_serialize	(	MCVList *	mcvlist,
		VacAttrStats **	stats
	)

Definition at line 625 of file mcv.c.

{
    int         i;
    int         dim;
    int         ndims = mcvlist->ndimensions;
 
    SortSupport ssup;
    DimensionInfo *info;
 
    Size        total_length;
 
    /* serialized items (indexes into arrays, etc.) */
    bytea      *raw;
    char       *ptr;
    char       *endptr PG_USED_FOR_ASSERTS_ONLY;
 
    /* values per dimension (and number of non-NULL values) */
    Datum     **values = (Datum **) palloc0(sizeof(Datum *) * ndims);
    int        *counts = (int *) palloc0(sizeof(int) * ndims);
 
    /*
     * We'll include some rudimentary information about the attribute types
     * (length, by-val flag), so that we don't have to look them up while
     * deserializing the MCV list (we already have the type OID in the
     * header).  This is safe because when changing the type of the attribute
     * the statistics gets dropped automatically.  We need to store the info
     * about the arrays of deduplicated values anyway.
     */
    info = (DimensionInfo *) palloc0(sizeof(DimensionInfo) * ndims);
 
    /* sort support data for all attributes included in the MCV list */
    ssup = (SortSupport) palloc0(sizeof(SortSupportData) * ndims);
 
    /* collect and deduplicate values for each dimension (attribute) */
    for (dim = 0; dim < ndims; dim++)
    {
        int         ndistinct;
        TypeCacheEntry *typentry;
 
        /*
         * Lookup the LT operator (can't get it from stats extra_data, as we
         * don't know how to interpret that - scalar vs. array etc.).
         */
        typentry = lookup_type_cache(stats[dim]->attrtypid, TYPECACHE_LT_OPR);
 
        /* copy important info about the data type (length, by-value) */
        info[dim].typlen = stats[dim]->attrtype->typlen;
        info[dim].typbyval = stats[dim]->attrtype->typbyval;
 
        /* allocate space for values in the attribute and collect them */
        values[dim] = (Datum *) palloc0(sizeof(Datum) * mcvlist->nitems);
 
        for (i = 0; i < mcvlist->nitems; i++)
        {
            /* skip NULL values - we don't need to deduplicate those */
            if (mcvlist->items[i].isnull[dim])
                continue;
 
            /* append the value at the end */
            values[dim][counts[dim]] = mcvlist->items[i].values[dim];
            counts[dim] += 1;
        }
 
        /* if there are just NULL values in this dimension, we're done */
        if (counts[dim] == 0)
            continue;
 
        /* sort and deduplicate the data */
        ssup[dim].ssup_cxt = CurrentMemoryContext;
        ssup[dim].ssup_collation = stats[dim]->attrcollid;
        ssup[dim].ssup_nulls_first = false;
 
        PrepareSortSupportFromOrderingOp(typentry->lt_opr, &ssup[dim]);
 
        qsort_interruptible(values[dim], counts[dim], sizeof(Datum),
                            compare_scalars_simple, &ssup[dim]);
 
        /*
         * Walk through the array and eliminate duplicate values, but keep the
         * ordering (so that we can do a binary search later). We know there's
         * at least one item as (counts[dim] != 0), so we can skip the first
         * element.
         */
        ndistinct = 1;          /* number of distinct values */
        for (i = 1; i < counts[dim]; i++)
        {
            /* expect sorted array */
            Assert(compare_datums_simple(values[dim][i - 1], values[dim][i], &ssup[dim]) <= 0);
 
            /* if the value is the same as the previous one, we can skip it */
            if (!compare_datums_simple(values[dim][i - 1], values[dim][i], &ssup[dim]))
                continue;
 
            values[dim][ndistinct] = values[dim][i];
            ndistinct += 1;
        }
 
        /* we must not exceed PG_UINT16_MAX, as we use uint16 indexes */
        Assert(ndistinct <= PG_UINT16_MAX);
 
        /*
         * Store additional info about the attribute - number of deduplicated
         * values, and also size of the serialized data. For fixed-length data
         * types this is trivial to compute, for varwidth types we need to
         * actually walk the array and sum the sizes.
         */
        info[dim].nvalues = ndistinct;
 
        if (info[dim].typbyval) /* by-value data types */
        {
            info[dim].nbytes = info[dim].nvalues * info[dim].typlen;
 
            /*
             * We copy the data into the MCV item during deserialization, so
             * we don't need to allocate any extra space.
             */
            info[dim].nbytes_aligned = 0;
        }
        else if (info[dim].typlen > 0)  /* fixed-length by-ref */
        {
            /*
             * We don't care about alignment in the serialized data, so we
             * pack the data as much as possible. But we also track how much
             * data will be needed after deserialization, and in that case we
             * need to account for alignment of each item.
             *
             * Note: As the items are fixed-length, we could easily compute
             * this during deserialization, but we do it here anyway.
             */
            info[dim].nbytes = info[dim].nvalues * info[dim].typlen;
            info[dim].nbytes_aligned = info[dim].nvalues * MAXALIGN(info[dim].typlen);
        }
        else if (info[dim].typlen == -1)    /* varlena */
        {
            info[dim].nbytes = 0;
            info[dim].nbytes_aligned = 0;
            for (i = 0; i < info[dim].nvalues; i++)
            {
                Size        len;
 
                /*
                 * For varlena values, we detoast the values and store the
                 * length and data separately. We don't bother with alignment
                 * here, which means that during deserialization we need to
                 * copy the fields and only access the copies.
                 */
                values[dim][i] = PointerGetDatum(PG_DETOAST_DATUM(values[dim][i]));
 
                /* serialized length (uint32 length + data) */
                len = VARSIZE_ANY_EXHDR(values[dim][i]);
                info[dim].nbytes += sizeof(uint32); /* length */
                info[dim].nbytes += len;    /* value (no header) */
 
                /*
                 * During deserialization we'll build regular varlena values
                 * with full headers, and we need to align them properly.
                 */
                info[dim].nbytes_aligned += MAXALIGN(VARHDRSZ + len);
            }
        }
        else if (info[dim].typlen == -2)    /* cstring */
        {
            info[dim].nbytes = 0;
            info[dim].nbytes_aligned = 0;
            for (i = 0; i < info[dim].nvalues; i++)
            {
                Size        len;
 
                /*
                 * cstring is handled similar to varlena - first we store the
                 * length as uint32 and then the data. We don't care about
                 * alignment, which means that during deserialization we need
                 * to copy the fields and only access the copies.
                 */
 
                /* c-strings include terminator, so +1 byte */
                len = strlen(DatumGetCString(values[dim][i])) + 1;
                info[dim].nbytes += sizeof(uint32); /* length */
                info[dim].nbytes += len;    /* value */
 
                /* space needed for properly aligned deserialized copies */
                info[dim].nbytes_aligned += MAXALIGN(len);
            }
        }
 
        /* we know (count>0) so there must be some data */
        Assert(info[dim].nbytes > 0);
    }
 
    /*
     * Now we can finally compute how much space we'll actually need for the
     * whole serialized MCV list (varlena header, MCV header, dimension info
     * for each attribute, deduplicated values and items).
     */
    total_length = (3 * sizeof(uint32)) /* magic + type + nitems */
        + sizeof(AttrNumber)    /* ndimensions */
        + (ndims * sizeof(Oid));    /* attribute types */
 
    /* dimension info */
    total_length += ndims * sizeof(DimensionInfo);
 
    /* add space for the arrays of deduplicated values */
    for (i = 0; i < ndims; i++)
        total_length += info[i].nbytes;
 
    /*
     * And finally account for the items (those are fixed-length, thanks to
     * replacing values with uint16 indexes into the deduplicated arrays).
     */
    total_length += mcvlist->nitems * ITEM_SIZE(dim);
 
    /*
     * Allocate space for the whole serialized MCV list (we'll skip bytes, so
     * we set them to zero to make the result more compressible).
     */
    raw = (bytea *) palloc0(VARHDRSZ + total_length);
    SET_VARSIZE(raw, VARHDRSZ + total_length);
 
    ptr = VARDATA(raw);
    endptr = ptr + total_length;
 
    /* copy the MCV list header fields, one by one */
    memcpy(ptr, &mcvlist->magic, sizeof(uint32));
    ptr += sizeof(uint32);
 
    memcpy(ptr, &mcvlist->type, sizeof(uint32));
    ptr += sizeof(uint32);
 
    memcpy(ptr, &mcvlist->nitems, sizeof(uint32));
    ptr += sizeof(uint32);
 
    memcpy(ptr, &mcvlist->ndimensions, sizeof(AttrNumber));
    ptr += sizeof(AttrNumber);
 
    memcpy(ptr, mcvlist->types, sizeof(Oid) * ndims);
    ptr += (sizeof(Oid) * ndims);
 
    /* store information about the attributes (data amounts, ...) */
    memcpy(ptr, info, sizeof(DimensionInfo) * ndims);
    ptr += sizeof(DimensionInfo) * ndims;
 
    /* Copy the deduplicated values for all attributes to the output. */
    for (dim = 0; dim < ndims; dim++)
    {
        /* remember the starting point for Asserts later */
        char       *start PG_USED_FOR_ASSERTS_ONLY = ptr;
 
        for (i = 0; i < info[dim].nvalues; i++)
        {
            Datum       value = values[dim][i];
 
            if (info[dim].typbyval) /* passed by value */
            {
                Datum       tmp;
 
                /*
                 * For byval types, we need to copy just the significant bytes
                 * - we can't use memcpy directly, as that assumes
                 * little-endian behavior.  store_att_byval does almost what
                 * we need, but it requires a properly aligned buffer - the
                 * output buffer does not guarantee that. So we simply use a
                 * local Datum variable (which guarantees proper alignment),
                 * and then copy the value from it.
                 */
                store_att_byval(&tmp, value, info[dim].typlen);
 
                memcpy(ptr, &tmp, info[dim].typlen);
                ptr += info[dim].typlen;
            }
            else if (info[dim].typlen > 0)  /* passed by reference */
            {
                /* no special alignment needed, treated as char array */
                memcpy(ptr, DatumGetPointer(value), info[dim].typlen);
                ptr += info[dim].typlen;
            }
            else if (info[dim].typlen == -1)    /* varlena */
            {
                uint32      len = VARSIZE_ANY_EXHDR(DatumGetPointer(value));
 
                /* copy the length */
                memcpy(ptr, &len, sizeof(uint32));
                ptr += sizeof(uint32);
 
                /* data from the varlena value (without the header) */
                memcpy(ptr, VARDATA_ANY(DatumGetPointer(value)), len);
                ptr += len;
            }
            else if (info[dim].typlen == -2)    /* cstring */
            {
                uint32      len = (uint32) strlen(DatumGetCString(value)) + 1;
 
                /* copy the length */
                memcpy(ptr, &len, sizeof(uint32));
                ptr += sizeof(uint32);
 
                /* value */
                memcpy(ptr, DatumGetCString(value), len);
                ptr += len;
            }
 
            /* no underflows or overflows */
            Assert((ptr > start) && ((ptr - start) <= info[dim].nbytes));
        }
 
        /* we should get exactly nbytes of data for this dimension */
        Assert((ptr - start) == info[dim].nbytes);
    }
 
    /* Serialize the items, with uint16 indexes instead of the values. */
    for (i = 0; i < mcvlist->nitems; i++)
    {
        MCVItem    *mcvitem = &mcvlist->items[i];
 
        /* don't write beyond the allocated space */
        Assert(ptr <= (endptr - ITEM_SIZE(dim)));
 
        /* copy NULL and frequency flags into the serialized MCV */
        memcpy(ptr, mcvitem->isnull, sizeof(bool) * ndims);
        ptr += sizeof(bool) * ndims;
 
        memcpy(ptr, &mcvitem->frequency, sizeof(double));
        ptr += sizeof(double);
 
        memcpy(ptr, &mcvitem->base_frequency, sizeof(double));
        ptr += sizeof(double);
 
        /* store the indexes last */
        for (dim = 0; dim < ndims; dim++)
        {
            uint16      index = 0;
            Datum      *value;
 
            /* do the lookup only for non-NULL values */
            if (!mcvitem->isnull[dim])
            {
                value = (Datum *) bsearch_arg(&mcvitem->values[dim], values[dim],
                                              info[dim].nvalues, sizeof(Datum),
                                              compare_scalars_simple, &ssup[dim]);
 
                Assert(value != NULL);  /* serialization or deduplication
                                         * error */
 
                /* compute index within the deduplicated array */
                index = (uint16) (value - values[dim]);
 
                /* check the index is within expected bounds */
                Assert(index < info[dim].nvalues);
            }
 
            /* copy the index into the serialized MCV */
            memcpy(ptr, &index, sizeof(uint16));
            ptr += sizeof(uint16);
        }
 
        /* make sure we don't overflow the allocated value */
        Assert(ptr <= endptr);
    }
 
    /* at this point we expect to match the total_length exactly */
    Assert(ptr == endptr);
 
    pfree(values);
    pfree(counts);
 
    return raw;
}

References Assert(), VacAttrStats::attrcollid, VacAttrStats::attrtype, MCVItem::base_frequency, bsearch_arg(), compare_datums_simple(), compare_scalars_simple(), CurrentMemoryContext, DatumGetCString(), DatumGetPointer(), MCVItem::frequency, i, MCVItem::isnull, ITEM_SIZE, MCVList::items, len, lookup_type_cache(), TypeCacheEntry::lt_opr, MCVList::magic, MAXALIGN, DimensionInfo::nbytes, DimensionInfo::nbytes_aligned, MCVList::ndimensions, MCVList::nitems, DimensionInfo::nvalues, palloc0(), pfree(), PG_DETOAST_DATUM, PG_UINT16_MAX, PG_USED_FOR_ASSERTS_ONLY, PointerGetDatum(), PrepareSortSupportFromOrderingOp(), qsort_interruptible(), SET_VARSIZE, SortSupportData::ssup_collation, SortSupportData::ssup_cxt, SortSupportData::ssup_nulls_first, store_att_byval(), DimensionInfo::typbyval, MCVList::type, TYPECACHE_LT_OPR, MCVList::types, DimensionInfo::typlen, value, values, MCVItem::values, VARDATA, VARDATA_ANY, VARHDRSZ, and VARSIZE_ANY_EXHDR.

Referenced by statext_store().

statext_ndistinct_build()

MVNDistinct* statext_ndistinct_build	(	double	totalrows,
		StatsBuildData *	data
	)

Definition at line 89 of file mvdistinct.c.

{
    MVNDistinct *result;
    int         k;
    int         itemcnt;
    int         numattrs = data->nattnums;
    int         numcombs = num_combinations(numattrs);
 
    result = palloc(offsetof(MVNDistinct, items) +
                    numcombs * sizeof(MVNDistinctItem));
    result->magic = STATS_NDISTINCT_MAGIC;
    result->type = STATS_NDISTINCT_TYPE_BASIC;
    result->nitems = numcombs;
 
    itemcnt = 0;
    for (k = 2; k <= numattrs; k++)
    {
        int        *combination;
        CombinationGenerator *generator;
 
        /* generate combinations of K out of N elements */
        generator = generator_init(numattrs, k);
 
        while ((combination = generator_next(generator)))
        {
            MVNDistinctItem *item = &result->items[itemcnt];
            int         j;
 
            item->attributes = palloc(sizeof(AttrNumber) * k);
            item->nattributes = k;
 
            /* translate the indexes to attnums */
            for (j = 0; j < k; j++)
            {
                item->attributes[j] = data->attnums[combination[j]];
 
                Assert(AttributeNumberIsValid(item->attributes[j]));
            }
 
            item->ndistinct =
                ndistinct_for_combination(totalrows, data, k, combination);
 
            itemcnt++;
            Assert(itemcnt <= result->nitems);
        }
 
        generator_free(generator);
    }
 
    /* must consume exactly the whole output array */
    Assert(itemcnt == result->nitems);
 
    return result;
}

References Assert(), AttributeNumberIsValid, MVNDistinctItem::attributes, data, generator_free(), generator_init(), generator_next(), MVNDistinct::items, j, MVNDistinct::magic, MVNDistinctItem::nattributes, MVNDistinctItem::ndistinct, ndistinct_for_combination(), MVNDistinct::nitems, nitems, num_combinations(), palloc(), STATS_NDISTINCT_MAGIC, STATS_NDISTINCT_TYPE_BASIC, and MVNDistinct::type.

Referenced by BuildRelationExtStatistics().

statext_ndistinct_deserialize()

MVNDistinct* statext_ndistinct_deserialize

(

bytea *

data

)

Definition at line 251 of file mvdistinct.c.

{
    int         i;
    Size        minimum_size;
    MVNDistinct ndist;
    MVNDistinct *ndistinct;
    char       *tmp;
 
    if (data == NULL)
        return NULL;
 
    /* we expect at least the basic fields of MVNDistinct struct */
    if (VARSIZE_ANY_EXHDR(data) < SizeOfHeader)
        elog(ERROR, "invalid MVNDistinct size %zu (expected at least %zu)",
             VARSIZE_ANY_EXHDR(data), SizeOfHeader);
 
    /* initialize pointer to the data part (skip the varlena header) */
    tmp = VARDATA_ANY(data);
 
    /* read the header fields and perform basic sanity checks */
    memcpy(&ndist.magic, tmp, sizeof(uint32));
    tmp += sizeof(uint32);
    memcpy(&ndist.type, tmp, sizeof(uint32));
    tmp += sizeof(uint32);
    memcpy(&ndist.nitems, tmp, sizeof(uint32));
    tmp += sizeof(uint32);
 
    if (ndist.magic != STATS_NDISTINCT_MAGIC)
        elog(ERROR, "invalid ndistinct magic %08x (expected %08x)",
             ndist.magic, STATS_NDISTINCT_MAGIC);
    if (ndist.type != STATS_NDISTINCT_TYPE_BASIC)
        elog(ERROR, "invalid ndistinct type %d (expected %d)",
             ndist.type, STATS_NDISTINCT_TYPE_BASIC);
    if (ndist.nitems == 0)
        elog(ERROR, "invalid zero-length item array in MVNDistinct");
 
    /* what minimum bytea size do we expect for those parameters */
    minimum_size = MinSizeOfItems(ndist.nitems);
    if (VARSIZE_ANY_EXHDR(data) < minimum_size)
        elog(ERROR, "invalid MVNDistinct size %zu (expected at least %zu)",
             VARSIZE_ANY_EXHDR(data), minimum_size);
 
    /*
     * Allocate space for the ndistinct items (no space for each item's
     * attnos: those live in bitmapsets allocated separately)
     */
    ndistinct = palloc0(MAXALIGN(offsetof(MVNDistinct, items)) +
                        (ndist.nitems * sizeof(MVNDistinctItem)));
    ndistinct->magic = ndist.magic;
    ndistinct->type = ndist.type;
    ndistinct->nitems = ndist.nitems;
 
    for (i = 0; i < ndistinct->nitems; i++)
    {
        MVNDistinctItem *item = &ndistinct->items[i];
 
        /* ndistinct value */
        memcpy(&item->ndistinct, tmp, sizeof(double));
        tmp += sizeof(double);
 
        /* number of attributes */
        memcpy(&item->nattributes, tmp, sizeof(int));
        tmp += sizeof(int);
        Assert((item->nattributes >= 2) && (item->nattributes <= STATS_MAX_DIMENSIONS));
 
        item->attributes
            = (AttrNumber *) palloc(item->nattributes * sizeof(AttrNumber));
 
        memcpy(item->attributes, tmp, sizeof(AttrNumber) * item->nattributes);
        tmp += sizeof(AttrNumber) * item->nattributes;
 
        /* still within the bytea */
        Assert(tmp <= ((char *) data + VARSIZE_ANY(data)));
    }
 
    /* we should have consumed the whole bytea exactly */
    Assert(tmp == ((char *) data + VARSIZE_ANY(data)));
 
    return ndistinct;
}

References Assert(), MVNDistinctItem::attributes, data, elog(), ERROR, i, MVNDistinct::items, MVNDistinct::magic, MAXALIGN, MinSizeOfItems, MVNDistinctItem::nattributes, MVNDistinctItem::ndistinct, MVNDistinct::nitems, palloc(), palloc0(), SizeOfHeader, STATS_MAX_DIMENSIONS, STATS_NDISTINCT_MAGIC, STATS_NDISTINCT_TYPE_BASIC, MVNDistinct::type, VARDATA_ANY, VARSIZE_ANY, and VARSIZE_ANY_EXHDR.

Referenced by pg_ndistinct_out(), and statext_ndistinct_load().

statext_ndistinct_serialize()

bytea* statext_ndistinct_serialize

(

MVNDistinct *

ndistinct

)

Definition at line 180 of file mvdistinct.c.

{
    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 + SizeOfHeader;
 
    /* and also include space for the actual attribute numbers */
    for (i = 0; i < ndistinct->nitems; i++)
    {
        int         nmembers;
 
        nmembers = ndistinct->items[i].nattributes;
        Assert(nmembers >= 2);
 
        len += SizeOfItem(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 entry
     */
    for (i = 0; i < ndistinct->nitems; i++)
    {
        MVNDistinctItem item = ndistinct->items[i];
        int         nmembers = item.nattributes;
 
        memcpy(tmp, &item.ndistinct, sizeof(double));
        tmp += sizeof(double);
        memcpy(tmp, &nmembers, sizeof(int));
        tmp += sizeof(int);
 
        memcpy(tmp, item.attributes, sizeof(AttrNumber) * nmembers);
        tmp += nmembers * sizeof(AttrNumber);
 
        /* protect against overflows */
        Assert(tmp <= ((char *) output + len));
    }
 
    /* check we used exactly the expected space */
    Assert(tmp == ((char *) output + len));
 
    return output;
}

References Assert(), MVNDistinctItem::attributes, i, MVNDistinct::items, len, MVNDistinct::magic, MVNDistinctItem::nattributes, MVNDistinctItem::ndistinct, MVNDistinct::nitems, output, palloc(), SET_VARSIZE, SizeOfHeader, SizeOfItem, STATS_NDISTINCT_MAGIC, STATS_NDISTINCT_TYPE_BASIC, MVNDistinct::type, VARDATA, and VARHDRSZ.

Referenced by statext_store().