workload/schemachange: improve error screening

pkg/sql/rowenc/testutils.go

@@ -188,7 +188,13 @@ func RandDatumWithNullChance(rng *rand.Rand, typ *types.T, nullChance int) tree.
 		return &tree.DBitArray{BitArray: r}
 	case types.StringFamily:
 		// Generate a random ASCII string.
-		p := make([]byte, rng.Intn(10))
+		var length int
+		if typ.Oid() == oid.T_char || typ.Oid() == oid.T_bpchar {
+			length = 1
+		} else {
+			length = rng.Intn(10)
+		}
+		p := make([]byte, length)
 		for i := range p {
 			p[i] = byte(1 + rng.Intn(127))
 		}
@@ -750,6 +756,16 @@ func randInterestingDatum(rng *rand.Rand, typ *types.T) tree.Datum {
 		default:
 			panic(errors.AssertionFailedf("float with an unexpected width %d", typ.Width()))
 		}
+	case types.BitFamily:
+		// A width of 64 is used by all special BitFamily datums in randInterestingDatums.
+		// If the provided bit type, typ, has a width of 0 (representing an arbitrary width) or 64 exactly,
+		// then the special datum will be valid for the provided type. Otherwise, the special type
+		// must be resized to match the width of the provided type.
+		if typ.Width() == 0 || typ.Width() == 64 {
+			return special
+		}
+		return &tree.DBitArray{BitArray: special.(*tree.DBitArray).ToWidth(uint(typ.Width()))}
+
 	default:
 		return special
 	}

pkg/workload/schemachange/error_screening.go

@@ -36,6 +36,15 @@ func viewExists(tx *pgx.Tx, tableName *tree.TableName) (bool, error) {
    )`, tableName.Schema(), tableName.Object())
 }
 
+func sequenceExists(tx *pgx.Tx, seqName *tree.TableName) (bool, error) {
+	return scanBool(tx, `SELECT EXISTS (
+	SELECT sequence_name
+    FROM information_schema.sequences
+   WHERE sequence_schema = $1
+     AND sequence_name = $2
+   )`, seqName.Schema(), seqName.Object())
+}
+
 func columnExistsOnTable(tx *pgx.Tx, tableName *tree.TableName, columnName string) (bool, error) {
 	return scanBool(tx, `SELECT EXISTS (
 	SELECT column_name
@@ -156,3 +165,159 @@ func colIsPrimaryKey(tx *pgx.Tx, tableName *tree.TableName, columnName string) (
        );
 	`, tableName.Schema(), tableName.Object(), columnName)
 }
+
+// valuesViolateUniqueConstraints determines if any unique constraints (including primary constraints)
+// will be violated upon inserting the specified rows into the specified table.
+func violatesUniqueConstraints(
+	tx *pgx.Tx, tableName *tree.TableName, columns []string, rows [][]string,
+) (bool, error) {
+
+	if len(rows) == 0 {
+		return false, fmt.Errorf("violatesUniqueConstraints: no rows provided")
+	}
+
+	// Fetch unique constraints from the database. The format returned is an array of string arrays.
+	// Each string array is a group of column names for which a unique constraint exists.
+	constraints, err := scanStringArrayRows(tx, `
+	 SELECT DISTINCT array_agg(cols.column_name ORDER BY cols.column_name)
+					    FROM (
+					          SELECT d.oid,
+					                 d.table_name,
+					                 d.schema_name,
+					                 conname,
+					                 contype,
+					                 unnest(conkey) AS position
+					            FROM (
+					                  SELECT c.oid AS oid,
+					                         c.relname AS table_name,
+					                         ns.nspname AS schema_name
+					                    FROM pg_catalog.pg_class AS c
+					                    JOIN pg_catalog.pg_namespace AS ns ON
+					                          ns.oid = c.relnamespace
+					                   WHERE ns.nspname = $1
+					                     AND c.relname = $2
+					                 ) AS d
+					            JOIN (
+					                  SELECT conname, conkey, conrelid, contype
+					                    FROM pg_catalog.pg_constraint
+					                   WHERE contype = 'p' OR contype = 'u'
+					                 ) ON conrelid = d.oid
+					         ) AS cons
+					    JOIN (
+					          SELECT table_name,
+					                 table_schema,
+					                 column_name,
+					                 ordinal_position
+					            FROM information_schema.columns
+					         ) AS cols ON cons.schema_name = cols.table_schema
+					                  AND cols.table_name = cons.table_name
+					                  AND cols.ordinal_position = cons.position
+					GROUP BY cons.conname;
+`, tableName.Schema(), tableName.Object())
+	if err != nil {
+		return false, err
+	}
+
+	for _, constraint := range constraints {
+		// previousRows is used to check unique constraints among the values which
+		// will be inserted into the database.
+		previousRows := map[string]bool{}
+		for _, row := range rows {
+			violation, err := violatesUniqueConstraintsHelper(tx, tableName, columns, constraint, row, previousRows)
+			if err != nil {
+				return false, err
+			}
+			if violation {
+				return true, nil
+			}
+		}
+	}
+
+	return false, nil
+}
+
+func violatesUniqueConstraintsHelper(
+	tx *pgx.Tx,
+	tableName *tree.TableName,
+	columns []string,
+	constraint []string,
+	row []string,
+	previousRows map[string]bool,
+) (bool, error) {
+
+	// Put values to be inserted into a column name to value map to simplify lookups.
+	columnsToValues := map[string]string{}
+	for i := 0; i < len(columns); i++ {
+		columnsToValues[columns[i]] = row[i]
+	}
+
+	query := strings.Builder{}
+	query.WriteString(fmt.Sprintf(`SELECT EXISTS (
+			SELECT *
+				FROM %s
+       WHERE
+		`, tableName.String()))
+
+	atLeastOneNonNullValue := false
+	for _, column := range constraint {
+
+		// Null values are not checked because unique constraints do not apply to null values.
+		if columnsToValues[column] != "NULL" {
+			if atLeastOneNonNullValue {
+				query.WriteString(fmt.Sprintf(` AND %s = %s`, column, columnsToValues[column]))
+			} else {
+				query.WriteString(fmt.Sprintf(`%s = %s`, column, columnsToValues[column]))
+			}
+
+			atLeastOneNonNullValue = true
+		}
+	}
+	query.WriteString(")")
+
+	// If there are only null values being inserted for each of the constrained columns,
+	// then checking for uniqueness against other rows is not necessary.
+	if !atLeastOneNonNullValue {
+		return false, nil
+	}
+
+	queryString := query.String()
+
+	// Check for uniqueness against other rows to be inserted. For simplicity, the `SELECT EXISTS`
+	// query used to check for uniqueness against rows in the database can also
+	// be used as a unique key to check for uniqueness among rows to be inserted.
+	if _, duplicateEntry := previousRows[queryString]; duplicateEntry {
+		return true, nil
+	}
+	previousRows[queryString] = true
+
+	// Check for uniqueness against rows in the database.
+	exists, err := scanBool(tx, queryString)
+	if err != nil {
+		return false, err
+	}
+	if exists {
+		return true, nil
+	}
+
+	return false, nil
+}
+
+func scanStringArrayRows(tx *pgx.Tx, query string, args ...interface{}) ([][]string, error) {
+	rows, err := tx.Query(query, args...)
+	if err != nil {
+		return nil, err
+	}
+	defer rows.Close()
+
+	results := [][]string{}
+	for rows.Next() {
+		var columnNames []string
+		err := rows.Scan(&columnNames)
+		if err != nil {
+			return nil, err
+		}
+		results = append(results, columnNames)
+	}
+
+	return results, err
+}