4 files changed, 23 insertions, 11 deletions

diff --git a/src/backend/commands/analyze.c b/src/backend/commands/analyze.c
index 5fcedd78554..9ac71220a2a 100644
--- a/src/backend/commands/analyze.c
+++ b/src/backend/commands/analyze.c
@@ -2049,8 +2049,11 @@ compute_distinct_stats(VacAttrStatsP stats,
 
 		if (nmultiple == 0)
 		{
-			/* If we found no repeated values, assume it's a unique column */
-			stats->stadistinct = -1.0;
+			/*
+			 * If we found no repeated non-null values, assume it's a unique
+			 * column; but be sure to discount for any nulls we found.
+			 */
+			stats->stadistinct = -1.0 * (1.0 - stats->stanullfrac);
 		}
 		else if (track_cnt < track_max && toowide_cnt == 0 &&
 				 nmultiple == track_cnt)
@@ -2426,8 +2429,11 @@ compute_scalar_stats(VacAttrStatsP stats,
 
 		if (nmultiple == 0)
 		{
-			/* If we found no repeated values, assume it's a unique column */
-			stats->stadistinct = -1.0;
+			/*
+			 * If we found no repeated non-null values, assume it's a unique
+			 * column; but be sure to discount for any nulls we found.
+			 */
+			stats->stadistinct = -1.0 * (1.0 - stats->stanullfrac);
 		}
 		else if (toowide_cnt == 0 && nmultiple == ndistinct)
 		{
@@ -2753,7 +2759,7 @@ compute_scalar_stats(VacAttrStatsP stats,
 		else
 			stats->stawidth = stats->attrtype->typlen;
 		/* Assume all too-wide values are distinct, so it's a unique column */
-		stats->stadistinct = -1.0;
+		stats->stadistinct = -1.0 * (1.0 - stats->stanullfrac);
 	}
 	else if (null_cnt > 0)
 	{
diff --git a/src/backend/tsearch/ts_typanalyze.c b/src/backend/tsearch/ts_typanalyze.c
index 0f851ead060..817453ce011 100644
--- a/src/backend/tsearch/ts_typanalyze.c
+++ b/src/backend/tsearch/ts_typanalyze.c
@@ -295,7 +295,7 @@ compute_tsvector_stats(VacAttrStats *stats,
 		stats->stawidth = total_width / (double) nonnull_cnt;
 
 		/* Assume it's a unique column (see notes above) */
-		stats->stadistinct = -1.0;
+		stats->stadistinct = -1.0 * (1.0 - stats->stanullfrac);
 
 		/*
 		 * Construct an array of the interesting hashtable items, that is,
diff --git a/src/backend/utils/adt/rangetypes_typanalyze.c b/src/backend/utils/adt/rangetypes_typanalyze.c
index fcb71d3f36e..56504fcf3c2 100644
--- a/src/backend/utils/adt/rangetypes_typanalyze.c
+++ b/src/backend/utils/adt/rangetypes_typanalyze.c
@@ -203,7 +203,9 @@ compute_range_stats(VacAttrStats *stats, AnalyzeAttrFetchFunc fetchfunc,
 		/* Do the simple null-frac and width stats */
 		stats->stanullfrac = (double) null_cnt / (double) samplerows;
 		stats->stawidth = total_width / (double) non_null_cnt;
-		stats->stadistinct = -1.0;
+
+		/* Estimate that non-null values are unique */
+		stats->stadistinct = -1.0 * (1.0 - stats->stanullfrac);
 
 		/* Must copy the target values into anl_context */
 		old_cxt = MemoryContextSwitchTo(stats->anl_context);
diff --git a/src/backend/utils/adt/selfuncs.c b/src/backend/utils/adt/selfuncs.c
index cc2a9a1b6c5..56943f2a87a 100644
--- a/src/backend/utils/adt/selfuncs.c
+++ b/src/backend/utils/adt/selfuncs.c
@@ -4738,6 +4738,7 @@ double
 get_variable_numdistinct(VariableStatData *vardata, bool *isdefault)
 {
 	double		stadistinct;
+	double		stanullfrac = 0.0;
 	double		ntuples;
 
 	*isdefault = false;
@@ -4745,7 +4746,8 @@ get_variable_numdistinct(VariableStatData *vardata, bool *isdefault)
 	/*
 	 * Determine the stadistinct value to use.  There are cases where we can
 	 * get an estimate even without a pg_statistic entry, or can get a better
-	 * value than is in pg_statistic.
+	 * value than is in pg_statistic.  Grab stanullfrac too if we can find it
+	 * (otherwise, assume no nulls, for lack of any better idea).
 	 */
 	if (HeapTupleIsValid(vardata->statsTuple))
 	{
@@ -4754,6 +4756,7 @@ get_variable_numdistinct(VariableStatData *vardata, bool *isdefault)
 
 		stats = (Form_pg_statistic) GETSTRUCT(vardata->statsTuple);
 		stadistinct = stats->stadistinct;
+		stanullfrac = stats->stanullfrac;
 	}
 	else if (vardata->vartype == BOOLOID)
 	{
@@ -4777,7 +4780,7 @@ get_variable_numdistinct(VariableStatData *vardata, bool *isdefault)
 			{
 				case ObjectIdAttributeNumber:
 				case SelfItemPointerAttributeNumber:
-					stadistinct = -1.0; /* unique */
+					stadistinct = -1.0; /* unique (and all non null) */
 					break;
 				case TableOidAttributeNumber:
 					stadistinct = 1.0;	/* only 1 value */
@@ -4799,10 +4802,11 @@ get_variable_numdistinct(VariableStatData *vardata, bool *isdefault)
 	 * If there is a unique index or DISTINCT clause for the variable, assume
 	 * it is unique no matter what pg_statistic says; the statistics could be
 	 * out of date, or we might have found a partial unique index that proves
-	 * the var is unique for this query.
+	 * the var is unique for this query.  However, we'd better still believe
+	 * the null-fraction statistic.
 	 */
 	if (vardata->isunique)
-		stadistinct = -1.0;
+		stadistinct = -1.0 * (1.0 - stanullfrac);
 
 	/*
 	 * If we had an absolute estimate, use that.