Different SYSDATE behaviors in SQL and PL/SQL

The SYSDATE function is available in both SQL and PL/SQL. They both return the current date-time (down to nearest second) for the database, So it would be reasonable to assume that they "act" the same in both SQL statements and PL/SQL blocks.

That would, however, be a bad assumption to make, because in reality:

In SQL, SYSDATE is called just once for the entire statement.

In PL/SQL, SYSDATE is called every time it is invoked.

Wow. Mind blown. 

Let's take a look.

In the script below, I create a table and insert four rows. Then I create a package that keeps track of distinct dates added to a collection and show those dates. Finally, a function that uses the package.

CREATE TABLE tab (id INT)
/

BEGIN
   INSERT INTO tab VALUES (1);
   INSERT INTO tab VALUES (2);
   INSERT INTO tab VALUES (3);
   INSERT INTO tab VALUES (4);
   COMMIT;
END;
/

CREATE OR REPLACE PACKAGE tracker
   AUTHID DEFINER
IS
   TYPE when_t IS TABLE OF INTEGER INDEX BY VARCHAR2 (100);

   dates   when_t;

   PROCEDURE add_date (date_in IN DATE);

   FUNCTION store_and_return_date (date_in            IN DATE,
                                   call_in_plsql_in   IN VARCHAR2)
      RETURN DATE;

   PROCEDURE show_count;
END;
/

CREATE OR REPLACE PACKAGE BODY tracker
IS
   PROCEDURE add_date (date_in IN DATE)
   IS
   BEGIN
      dates (TO_CHAR (date_in, 'YYYY-MM-DD HH:MI:SS')) := 0;
   END;

   PROCEDURE show_count
   IS
      l_index   VARCHAR2 (100) := dates.FIRST;
   BEGIN
      DBMS_OUTPUT.put_line ('Date Count = ' || dates.COUNT);

      WHILE l_index IS NOT NULL
      LOOP
         DBMS_OUTPUT.put_line (l_index);
         l_index := dates.NEXT (l_index);
      END LOOP;

      dates.delete;
   END;

   FUNCTION store_and_return_date (date_in            IN DATE,
                                   call_in_plsql_in   IN VARCHAR2)
      RETURN DATE
   IS
      d   DATE;
   BEGIN
      /* This is the value for SYSDATE passed in from The SELECT */
      add_date (date_in); 

      IF call_in_plsql_in = 'YES'
      THEN
         /* Now we call SYSDATE inside PL/SQL. Will it be the same
            value as passed in for date_in, or different? The
            count in the collection will tell us */
         add_date (SYSDATE);
      END IF;

      DBMS_SESSION.sleep (1);

      RETURN date_in;
   END;
END;
/

When I call the store_and_return_date, I can specify whether or not I want SYSDATE to be called again from within the PL/SQL function - which will be invoked below inside a SELECT statement.

Notice that I use the date (converted explicitly to a string) as the index in the associative array. That makes it easy to keep track of unique timestamps added.

In the first query below, I do not call SYSDATE inside PL/SQL, only in the call to the function, which happens inside the SQL engine.

In the second query, I call SYDATE in PL/SQL in addition to the invocation in the SELECT statement.

SELECT tracker.store_and_return_date (
          SYSDATE, call_in_plsql_in => 'NO') 
  FROM tab
/

BEGIN
   tracker.show_count;
END;
/

SELECT tracker.store_and_return_date (
          SYSDATE, call_in_plsql_in => 'YES') 
  FROM tab
/

BEGIN
   tracker.show_count;
END;
/

And here's the output:

Date Count = 1
2020-09-29 01:28:58

Date Count = 4
2020-09-29 01:29:02
2020-09-29 01:29:03
2020-09-29 01:29:04
2020-09-29 01:29:05

And there you see the difference between SQL and PL/SQL engines when it comes to executing SYSDATE. There are four rows in the TAB table. But the value for SYSDATE when executed in SQL doesn't change for the duration of the query, so there is just one element in the array.

When I ask for SYSDATE to also be called in PL/SQL, it is executed and the current date-time returned - even when that function is called from within a SQL statement.