SQL Server 2016：使用DATETIME加入DATETIME2（3）

Question

I'm getting unexpected results when joining on a DATETIME2(3) and a DATETIME column with PK, in SQL Server 2016.

I have the following table:

CREATE TABLE DATETIME_TEST 
(
    [DATETIME] DATETIME NOT NULL,
    [DATETIME2_3] DATETIME2(3)
);

ALTER TABLE DATETIME_TEST 
    ADD CONSTRAINT PK_DATETIME_TEST PRIMARY KEY ([DATETIME]);

INSERT INTO DATETIME_TEST ([DATETIME], [DATETIME2_3])
VALUES ('20020202 02:02:02.000', '20020202 02:02:02.000'), 
       ('20020202 02:02:02.003', '20020202 02:02:02.003'), 
       ('20020202 02:02:02.007', '20020202 02:02:02.007'),
       ('2019-04-28 07:23:29.447', '2019-04-28 07:23:29.447');

SELECT * 
FROM DATETIME_TEST 
WHERE CONVERT(DATETIME2(3), [DATETIME]) = [DATETIME2_3]

The results :

DATETIME                  DATETIME2_3
-------------------------------------------------
2002-02-02 02:02:02.000   2002-02-02 02:02:02.000
2002-02-02 02:02:02.003   2002-02-02 02:02:02.003
2002-02-02 02:02:02.007   2002-02-02 02:02:02.007
2019-04-28 07:23:29.447   2019-04-28 07:23:29.447

As you can see above, the values are equal.

SELECT      
    a.DATETIME,
    a.DATETIME2_3
FROM
    DATETIME_TEST a
INNER JOIN 
    DATETIME_TEST b ON CONVERT(DATETIME2(3), a.[DATETIME]) = b.[DATETIME2_3]

The results :

DATETIME2_3                  DATETIME
-----------------------------------------------------
2002-02-02 02:02:02.000      2002-02-02 02:02:02.000

Although the values are equal, I only get some of the rows.

But if I remove the PK or change compatibility level to COMPATIBILITY_LEVEL = 120 , then I get all rows as expected

Is it a bug ?

Is there a better way to do this join.

Note: I join to the same table only for the simplicity of the example in real life I join between 2 different tables.

Answer 1

Actually, I have come across this as well. It is definitely a valid question to Microsoft SQL Server Team. Thank you for taking the time to raising this bug.

But as an alternative solution, you should rather try to convert the 'richer' datatype (DATETIME2) to the 'poorer' datatype (DATETIME) as backward compatible way. And then it would give you the results you are looking for:

SELECT      
    a.DATETIME,
    a.DATETIME2_3
FROM
    DATETIME_TEST a
INNER JOIN 
    DATETIME_TEST b ON a.[DATETIME] = CONVERT(DATETIME, b.[DATETIME2_3])