I’m using a Polymorphic Table Function (PTF) and adding a new column TIMESTAMP_CALCULATED of type TIMESTAMP with fractional seconds scale 9. In FETCH_ROWS I populate it with a constant value:

2026-01-01 01:01:01.123456789

However, when selecting the PTF output, the value appears rounded to 6 fractional digits, e.g.:

2026-01-01T01:01:01.123457Z (rounded to microseconds)

Additionally, formatting it with:

to_char(TIMESTAMP_CALCULATED,'YYYY-MM-DD HH24:MI:SS.FF9')

produces:

2026-01-01 01:01:01.000000000

This is unexpected: even if the value was rounded to 6 digits, TO_CHAR(...FF9) should not become all zeros.

Reproducible test case (minimal script)

CREATE TABLE TIMESTAMP_INPUT(ID INTEGER, TIMESTAMP_VALUE TIMESTAMP(9));

CREATE OR REPLACE PACKAGE TIMESTAMP_PTF AS

FUNCTION describe (tab IN OUT DBMS_TF.table_t) RETURN DBMS_TF.describe_t;
PROCEDURE fetch_rows;

FUNCTION my_ptf(tab IN TABLE) RETURN TABLE PIPELINED ROW POLYMORPHIC USING TIMESTAMP_PTF;

END;
/

CREATE OR REPLACE PACKAGE BODY TIMESTAMP_PTF AS

FUNCTION describe (tab IN OUT DBMS_TF.table_t) RETURN DBMS_TF.describe_t IS
  new_cols dbms_tf.columns_new_t; 
BEGIN
    new_cols(1) := dbms_tf.column_metadata_t(name => 'TIMESTAMP_CALCULATED', type => dbms_tf.type_timestamp, scale => 9); 
    RETURN dbms_tf.describe_t(new_columns => new_cols);
END describe;

PROCEDURE fetch_rows 
as 
  v_rowset dbms_tf.row_set_t; 
  v_vals_rowset dbms_tf.row_set_t; 
   
  v_row_count binary_integer; 
  v_col_count binary_integer; 
begin 
  dbms_tf.get_row_set(v_rowset, v_row_count, v_col_count); 
  
  for i in 1..v_row_count loop 
    v_vals_rowset(1).tab_timestamp(i) := TO_TIMESTAMP('2026-01-01 01:01:01.123456789','YYYY-MM-DD HH24:MI:SS.FF9');
  end loop;

  dbms_tf.put_row_set(v_vals_rowset); 
end; 

END;
/

DELETE TIMESTAMP_INPUT;
INSERT INTO TIMESTAMP_INPUT(ID, TIMESTAMP_VALUE) 
  VALUES(1, TO_TIMESTAMP('2026-01-01 01:01:01.123456789','YYYY-MM-DD HH24:MI:SS.FF9'));

INSERT INTO TIMESTAMP_INPUT(ID, TIMESTAMP_VALUE) 
  VALUES(2, TO_TIMESTAMP('2026-01-02 02:02:02.123456789','YYYY-MM-DD HH24:MI:SS.FF9'));

INSERT INTO TIMESTAMP_INPUT(ID, TIMESTAMP_VALUE) 
  VALUES(3, TO_TIMESTAMP('2026-01-03 03:03:03.123456789','YYYY-MM-DD HH24:MI:SS.FF9'));  
COMMIT;  

ALTER SESSION SET NLS_TIMESTAMP_FORMAT = 'YYYY-MM-DD HH24:MI:SS.FF9';
ALTER SESSION SET NLS_TIMESTAMP_TZ_FORMAT = 'YYYY-MM-DD HH24:MI:SS.FF9 TZH:TZM';

WITH BASE_DATA AS(
    SELECT ID, TIMESTAMP_VALUE FROM TIMESTAMP_INPUT
)
SELECT id, 
       to_char(TIMESTAMP_VALUE,'YYYY-MM-DD HH24:MI:SS.FF9') TIMESTAMP_VALUE_STR,
       DUMP(TIMESTAMP_VALUE) dump_TIMESTAMP_VALUE, 
       TIMESTAMP_VALUE,
       to_char(TIMESTAMP_CALCULATED,'YYYY-MM-DD HH24:MI:SS.FF9') TIMESTAMP_CALCULATED_STR,
       DUMP(TIMESTAMP_CALCULATED) DUMP_TIMESTAMP_CALCULATED,
       TIMESTAMP_CALCULATED
  FROM TIMESTAMP_PTF.my_ptf(BASE_DATA) pf;

drop table TIMESTAMP_INPUT;
drop package TIMESTAMP_PTF;

Actual results (what I see):
TIMESTAMP_CALCULATED_STR DUMP_TIMESTAMP_CALCULATED TIMESTAMP_CALCULATED
------------------------------- ------------------------------------------------ -----------------------------
2026-01-01 01:01:01.000000000 Typ=180 Len=11: 120,126,1,1,2,2,2,7,91,205,232 2026-01-01T01:01:01.123457Z
2026-01-01 01:01:01.000000000 Typ=180 Len=11: 120,126,1,1,2,2,2,7,91,205,232 2026-01-01T01:01:01.123457Z
2026-01-01 01:01:01.000000000 Typ=180 Len=11: 120,126,1,1,2,2,2,7,91,205,232 2026-01-01T01:01:01.123457Z

Expected results:
TIMESTAMP_CALCULATED should preserve scale 9 (nanoseconds) if the column is defined as TIMESTAMP(9).
TO_CHAR(...FF9) should display the stored fractional seconds, not zeros.

Questions:
How can I force/guarantee TIMESTAMP(9) precision for a new column created in a PTF (and populated via TAB_TIMESTAMP) so that I can reliably display FF9?

Is there a known limitation/bug in PTF/DBMS_TF regarding TIMESTAMP(9) (or metadata propagation) causing rounding to 6 digits and/or TO_CHAR returning zeros?

If this is expected behavior, what is the recommended workaround to preserve nanosecond precision?

SQL> CREATE OR REPLACE PACKAGE BODY TIMESTAMP_PTF AS
  2
  3  FUNCTION describe (tab IN OUT DBMS_TF.table_t) RETURN DBMS_TF.describe_t IS
  4    new_cols dbms_tf.columns_new_t;
  5  BEGIN
  6      new_cols(1) := dbms_tf.column_metadata_t(name => 'TIMESTAMP_CALCULATED', type => dbms_tf.type_varchar2);
  7      RETURN dbms_tf.describe_t(new_columns => new_cols);
  8  END describe;
  9
 10  PROCEDURE fetch_rows
 11  as
 12    v_rowset dbms_tf.row_set_t;
 13    v_vals_rowset dbms_tf.row_set_t;
 14
 15    v_row_count binary_integer;
 16    v_col_count binary_integer;
 17
 18  begin
 19    dbms_tf.get_row_set(v_rowset, v_row_count);
 20
 21    for i in 1..v_row_count loop
 22      v_vals_rowset(1).tab_varchar2(i) := '2026-01-01 01:01:01.123456789';
 23    end loop;
 24
 25    dbms_tf.put_row_set(v_vals_rowset);
 26  end;
 27
 28  END;
 29  /

Package body created.

SQL> SELECT id,
  2         TIMESTAMP_VALUE,
  3         to_timestamp(TIMESTAMP_CALCULATED,'YYYY-MM-DD HH24:MI:SS.FF9') TIMESTAMP_CALCULATED
  4    FROM TIMESTAMP_PTF.my_ptf(t) pf;

        ID TIMESTAMP_VALUE                          TIMESTAMP_CALCULATED
---------- ---------------------------------------- ----------------------------------------
         1 2023-01-01 01:01:01.123456789            2026-01-01 01:01:01.123456789