Register localtimestamp as current_timestamp in Oracle9iDialect

[hazzik]

In Oracle current_timestamp and systimestamp return TIMESTAMP WITH TIME ZONE type, which in Oracle ManagedDataAccess v23 turns into DateTimeOffset of expected DateTime, so some tests fail.

Switching it to localtimestamp to fix the problem.

[fredericDelaporte]

So, using current_timestamp for an update or insert with Oracle was resulting into an UTC date stored in database, eventually converted back to the client current local on read if the typing of the table columns was timestamp with time zone too ; if I understand correctly. While on most other db NHibernate handles, it is a local time timestamp which is stored.

So, that is a bug we are fixing here, but one that would require some data migration for the app affected, wouldn't it?

[hazzik]

Not sure on this front. I’ll try to add some verifications/test.

If that change is not acceptable, we would need to add workaround to AbstractDateTimeType to be able to read from DateTimeOffset.

[fredericDelaporte]

As a bug, fixing should always be acceptable. But here, as applications could possibly rely on the effects of this bug to be functional, if that is the case, we need to document it as a possible breaking change.

[hazzik]

Ok, I've checked and here is the behavior:

So, using current_timestamp for an update or insert with Oracle was resulting into an UTC date stored in database

No.

Both localtimestamp and current_timestamp would be converted to same value:

• if storing to TIMESTAMP column it would be stored as date time at session time zone.
• If storing to TIMESTAMP WITH TIME ZONE column it would be stored as date time with session time zone.

So I have following DDL:

create table Entity (
        Id RAW(16) not null,
       Name VARCHAR2(255),
       DT1 TIMESTAMP,
       DT2 TIMESTAMP WITH TIME ZONE,
       primary key (Id)
    )

Then following insert:

INSERT INTO ENTITY (ID, NAME, DT1, DT2)
SELECT SYS_GUID(), 'Bob', localtimestamp, localtimestamp FROM DUAL
UNION ALL
SELECT SYS_GUID(), 'Alice', current_timestamp, current_timestamp FROM DUAL;

And query:

SELECT * FROM ENTITY;

I'm getting following:

I tried to do the same in code with NH, and the behavior is the same.

[hazzik]

To the above: my server is in UTC, and session UTC+10

[hazzik]

So, I would say that this change is pretty safe and should not be breaking. However, we can say a few words in a probably breaking changes section if we want.

In any way, I would like this to be included in 5.6, especially if we want to treat this as a breaking change

[fredericDelaporte]

I thought the union might cause some additional type conversion before inserting and checked with separated insert queries, same results.

[hazzik]

@fredericDelaporte yeah, I was inserting separately. Just combined query when posting here.