Can't get time zone information when using mod_plsql? - oracle

I have written a function to convert a date into a Unix time stamp. The function is written to work no matter what the current DST status is (e.g. EST or EDT). This is the function:
function unix_time_from_date(in_date in date) return number
as
ut number := 0;
tz varchar2(8) := '';
begin
-- Get the local timezone from the passed in date
-- Assuming the date supplied is for the local time zone
select
extract(
timezone_abbr from cast(in_date as timestamp with local time zone)
)
into tz
from dual;
-- Get the Unix timestamp
select
(new_time(in_date, tz, 'GMT') - to_date('01-JAN-1970', 'DD-MM-YYYY')) * (
86400)
into ut
from dual;
return ut;
end unix_time_from_date;
This function works great when I execute it from a client like JDeveloper. From what I gather, this is because the client is supplying time zone information to the first query. However, if I use the function from within a procedure that is called from a mod_plsql page, then I get the error ORA-01857: not a valid time zone. This error is being thrown from the new_time function because tz is set to 'UNK'.
So, I implemented a work-around for this problem like so:
function unix_time_from_date(in_date in date) return number
as
ut number := 0;
tz varchar2(8) := '';
begin
-- Get the local timezone from the passed in date
-- Assuming the date supplied is for the local time zone
select
extract(
timezone_abbr from cast(in_date as timestamp with local time zone)
)
into tz
from dual;
if tz = 'UNK' then
select
extract(
timezone_abbr from cast(sysdate as timestamp with local time zone)
)
into tz
from dual;
end if;
-- Get the Unix timestamp
select
(new_time(in_date, tz, 'GMT') - to_date('01-JAN-1970', 'DD-MM-YYYY')) * (
86400)
into ut
from dual;
return ut;
end unix_time_from_date;
Except, this still fails with tz being set to 'UNK'. Does anyone know what could be happening here? Why can't I get the local time zone abbreviation when the function is called from a Oracle Application Server process?

I guess this does not depends on the date parameter you are passing. It probably depends on the Operating system settings where the database server runs. In JDeveloper it is probably picking up from your Computers (OS) time zone settings. Try doing an ssh on the DB server and run the first two queries in your script (use the actual date in 'DD-MON-YY' format for first query). Both should be returning 'UNK'. The UNK (unknown) is likely because more than one timezone is returned.Examples: In the following examples, suppose that the current time zone is CST (US Central Time).
SELECT NEW_TIME(SYSDATE, 'CST', 'GMT') FROM DUAL --returns the date in London.
SELECT TO_CHAR(NEW_TIME(SYSDATE, 'CST', 'GMT'),'HH24:MI') FROM DUAL --returns the time, based on the 24-hour clock, in London.
SELECT TO_CHAR(NEW_TIME(SYSDATE + (14 / 24), 'PST', 'PST'),'DD-MON-YY HH24:MI') FROM DUAL --returns the date and time in China.
SELECT TO_CHAR(NEW_TIME(SYSDATE + (diff / 24), ‘GMT’, ‘GMT’),’DD-MON-YY HH24:MI’) FROM DUAL; --returns the date and time of your office.

Have you compared the NLS_DATE_FORMAT on your local machine and the server? You may find that a combination of differences in this, and the possibility that an impicit conversion is happening when the date is being passed in may be your problem here.

The function as written does not work when the session calling it does not have the time zone information set. Therefore, you need to explicitly specify the source time zone. The following function solves this problem (and corrects the return type):
function unix_time_from_date
(
in_date in date,
in_src_tz in varchar2 default 'America/New_York'
)
return integer
as
ut integer := 0;
tz varchar2(8) := '';
tz_date timestamp with time zone;
tz_stmt varchar2(255);
begin
-- Get the local time zone abbreviation from the passed in date
tz_stmt := 'select systimestamp at time zone ''' || in_src_tz || ''' from dual';
execute immediate tz_stmt into tz_date;
select
extract(timezone_abbr from tz_date)
into tz
from dual;
-- Get the Unix timestamp
select
(new_time(in_date, tz, 'GMT') - to_date('01-JAN-1970', 'DD-MM-YYYY')) * (86400)
into ut
from dual;
return ut;
end unix_time_from_date;
Note the addition of a second parameter to the function. This parameter, in_src_tz, is used to indicate what time zone the in_date parameter is in. The value of in_src_tz should be one of the timezone listed in the tzname column of the v$timezone_names table.
Also, you cannot simply select the value of the tzabbrev column in the v$timezone_names table due to time zone having multiple abbreviations. By using the extract, you will get the current abbreviation with DST factored in.

Related

Oracle get all records between current day inserted with timestamp

Unable to retrieve records properly with the formatted date values, need select query with right date format to get all records inserted per day
I have a date string like this in my script -  
dateString :='26-MAR-20 05.00.00.00000000 AM';
I want to add 0.313 minutes to this date value. And also I want to
increment it to the next day something like this- '27-MAR-20
05.00.00.00000000 AM'; ​
I tried this 
dateString :='26-MAR-20 05.00.00.00000000 AM';
dateString :=to_char(dateField,'DD-MON-YYYY HH24:MI:SS.FF');
dateField := to_timestamp(dateString, 'DD/MON/YYYY HH24:MI:SS.FF') + 0.313 * INTERVAL '1' MINUTE;
​
I can see the output and inserted these values into the DB but
unable to retrieve records properly with these formatted date values..
I suspect it might be to do with the timestamp fields
When I run this query -
select *from rptallexceptions where exceptiontime between '27-MAR-2020 04.00.00.000000000 AM' and '28-MAR-2020 03.59.00.000000000 AM' order by exceptiontime desc;
​
--- this one gives 3833 records but expected is 4600
it shows only the records on 27th march., records inserted with timestamp containing 28th march are not retrieved..
select *from rptallexceptions where exceptiontime between '28-MAR-2020 04.00.00.000000000 AM' and '29-MAR-2020 03.59.00.000000000 AM' order by exceptiontime desc;
--- this one returns '0' rows
Excerpt from the script:
cnt :=cnt +1;
dateString :=to_char(dateField,'DD-MON-YYYY HH24:MI:SS.FF');
-- add time difference for each exception.
dateField := to_timestamp(dateString, 'DD/MON/YYYY HH24:MI:SS.FF') + 0.313 * INTERVAL '1' MINUTE;
-- after n*4600 exceptions,update date to next date.
IF REMAINDER(cnt,exceptionsPerDay) = 0 THEN
dateField := to_timestamp(dateField + 1,'DD/MON/YYYY HH24:MI:SS.FF');
END IF;
The way I see it, your problem is that you're comparing timestamps to strings.
Sample table:
SQL> create table test as
2 select to_timestamp('27.03.2020 04:00:00:000000', 'dd.mm.yyyy hh24:mi:ss:ff6') datum, 'A' name from dual
3 union all
4 select to_timestamp('28.03.2020 15:23:00:000000', 'dd.mm.yyyy hh24:mi:ss:ff6') datum, 'A' name from dual;
Table created.
Query should use timestamp, not string:
SQL> select *
2 from test
3 where datum between to_timestamp('25.03.2020 12:00:00:000000am', 'dd.mm.yyyy hh:mi:ss:ff6am') --> timestamp
4 and to_timestamp('28.03.2020 12:00:00:000000am', 'dd.mm.yyyy hh:mi:ss:ff6am'); --> timestamp
DATUM N
--------------------------------------------------------------------------- -
27.03.20 04:00:00,000000000 A
SQL>
You were kind of "lucky" (though, one might call it bad luck as your query didn't fail, but it produced wrong results, according to number of rows being returned) for not getting an error as my database raises (due to different NLS settings):
SQL> select *
2 from test
3 where datum between '25.03.2020 12:00:00:000000am' --> string
4 and '28.03.2020 12:00:00:000000am'; --> string
where datum between '25.03.2020 12:00:00:000000am'
*
ERROR at line 3:
ORA-01830: date format picture ends before converting entire input string
SQL>
Strings are handled differently than dates (or timestamps), or numbers. Maybe you'll see the difference clearly in this example:
SQL> with test (col) as
2 (select '1' from dual union all
3 select '9' from dual union all
4 select '20' from dual
5 )
6 select *
7 from test
8 where col < '9';
CO
--
1
20
SQL>
20 < 9? Yes, if those are strings. The same might have happened to you. Try with proper datatype.
When you do:
dateString :=to_char(dateField,'DD-MON-YYYY HH24:MI:SS.FF');
you are converting whatever is the initial value of dateField to a string; but it looks like that has has not been set yet, so you end up with an empty string (which is the same as null). Your original value of dateString is never being used.
If you skip that and just do:
dateString :='26-MAR-20 05.00.00.00000000 AM';
dateField := to_timestamp(dateString, 'DD/MON/YYYY HH24:MI:SS.FF') + 0.313 * INTERVAL '1' MINUTE;
​then you are converting the 2-digit year 20 with a YYYY format element, which will turn it into year 0020, not 2020; but you also don't have AM in the format model, so it will get ORA-01830 anyway, and have HH24 instead of HH. I think your to_char() is attempting to 'correct' the string format, but that's not a great approach.
If you can't control the starting string format then the model has to match that:
dateString :='26-MAR-20 05.00.00.00000000 AM';
dateField := to_timestamp(dateString, 'DD-MON-RR HH:MI:SS.FF AM') + 0.313 * INTERVAL '1' MINUTE;
which gives dateField a value of 2020-03-26 05:00:18.780000.
Later on you do:
dateField := to_timestamp(dateField + 1,'DD/MON/YYYY HH24:MI:SS.FF');
which is also doing an implicit conversion of dateField + 1 - which is converted to a date, incidentally as timestamp + number is a date, not a timestamp - to a string; which will use your NLS_DATE_FORMAT. Presumably that is something like 'DD-MON-YYYY'.
So stepping through that:
dateField + 1 => date '2020-03-27 05:00:18' (losing fractional seconds)
implicit to_char(dateField + 1) => string '27-Mar-2020' (losing time)
to_timestamp(to_char(dateField + 1), '...') => timestamp '2020-03-27 00:00:00.000000' (with time at midnight).
So your between is, at best going to find records between 2020-03-26 05:00:18.780000 and 2020-03-27 00:00:00.0000000 - so it won't pick up any records later than midnight on the 27th.
Don't convert to and from strings when you don't need to; leave data in its native data type (timestamp in this case) and use direct manipulation with intervals.
Using between isn't ideal anyway, because it is inclusive; it would be better to end up with:
exceptiontime >= timestamp '2020-03-26 05:00:18.780000'
and exceptionTime < timestamp '2020-03-27 05:00:18.780000'

Convert epoch to date in Oracle

Was wondering if anyone could help with precision time conversion.
Sample: 1501646399999 which is GMT: Wednesday, August 2, 2017 3:59:59.999 AM
I used the below query, but it always rounds off to 02-AUG-17 04:00:00. Can anyone please guide me
select TO_TIMESTAMP('1970-01-01 00:00:00.000', 'YYYY-MM-DD hh24:mi:SS.FF3') + ((1/86400000) * 1501646399999)
from dual;
The problem is that you're adding a number to your fixed timestamp, which is causing that timestamp to be implicitly converted to a date - which doesn't have sub-second precision.
If you add an interval instead then it stays as a timestamp:
alter session set nls_timestamp_format = 'YYYY-MM-DD HH24:MI:SS.FF3';
select TO_TIMESTAMP('1970-01-01 00:00:00.000', 'YYYY-MM-DD hh24:mi:SS.FF3')
+ numtodsinterval(1501646399999/1000, 'SECOND')
from dual;
TO_TIMESTAMP('1970-01-0
-----------------------
2017-08-02 03:59:59.999
Incidentally, you could slightly simplify your query with a timestamp literal:
select TIMESTAMP '1970-01-01 00:00:00' + numtodsinterval(...)
You may also want to check if you should be declaring that timestamp as being UTC, and converting back to local time zone after adding the epoch value; or leaving it explicitly as UTC but as a timestamp with time zone value. It depends exactly what that number is supposed to represent. (You said it's GMT/UTC, but still...)

Oracle PLSQL get difference between 2 datetime fields ignoring days

I would like to find the simplest way of calculating the difference in hours between 2 dates from Oracle datetime fields whilst ignoring the days, months & years portion. For example:
Datetime 1 (DATE variable) = 10/05/2017 16:00:00
Datetime 2 (DATE variable) = 15/05/2017 19:34:23
Required result (NUMBER output) = 3.576 hours
This is formula will be used in a PLSQL procedure, the output needs to be a number as above. I would imagine some combination of TO_DATE & TRUNC might work. Any help would be most appriciated and apologies if this is a duplicate question.
Use the sssss date mask to get just the time element as the number of seconds since midnight. Then it's just a matter of simple arithmentic:
select (to_number(to_char(datetime2, 'sssss'))
- to_number(to_char(datetime1, 'sssss')) / 3600 as diff_hours
from dual;
PL/SQL version is the same....
declare
Datetime1 DATE := to_date('10/05/2017 16:00:00', 'dd/mm/yyyy hh24:mi:ss');
Datetime2 DATE := to_date('15/05/2017 19:34:23', 'dd/mm/yyyy hh24:mi:ss');
hours_diff number;
begin
hours_diff := (to_number(to_char(datetime2, 'sssss'))
- to_number(to_char(datetime1, 'sssss'))) / 3600 ;
dbms_output.put_line(hours_diff);
end;

Converting a stored UTC time in Oracle 10g to local time

I have a varchar2(20) column with a value like '2015-01-26T20:29:51Z'.
I successfully convert it to a date with to_date(dEnteredDate,'YYYY-MM-DD"T"HH24:MI:SS"Z"')
I wish to convert it to local time by subtracting my SessionTimeZone, currently -07:00 and have a regular date with time, no timezone info.
I do this in MS SQL with;
SET #diff = datediff(hh,GetUTCDate(), GetDate());
SET #dlocal = DATEADD(hh, #diff, #UTCDateTime)
How can I accomplish the same in Oracle 10g?
You can do it like this:
select to_timestamp(dEnteredDate,'YYYY-MM-DD"T"HH24:MI:SS"Z"')- interval '7' hour from test_3;
Here - interval '7' hour will subtract the required time difference and give you the desired result without timezone.
You can do it like this
select cast(to_timestamp_TZ(dEnteredDate||' UTC','YYYY-MM-DD"T"HH24:MI:SS"Z" tzr') at local as date)
from ...

Oracle current_timestamp to seconds conversion

We are using Oracle database.
In our table timestamp is stored as seconds since 1970, how can I convert the time stamp obtained through current_timestamp() function to seconds
This would do it:
select round((cast(current_timestamp as date) - date '1970-01-01')*24*60*60) from dual
Though I wouldn't use current_timestamp if I was only interested in seconds, I would use SYSDATE:
select round((SYSDATE - date '1970-01-01')*24*60*60) from dual
Maybe not completely relevant. I had to resolve other way around problem (e.g. Oracle stores timestamp in V$RMAN_STATUS and V$RMAN_OUTPUT) and I had to convert that to date/timestamp. I was surprised, but the magic date is not 1970-01-01 there, but 1987-07-07. I looked at Oracle's history and the closest date I can think of is when they ported Oracle products to UNIX. Is this right?
Here's my SQL
SELECT /*+ rule */
to_char(min(stamp)/(24*60*60) + date '1987-07-07', 'DD-MON-YYYY HH24:MI:SS') start_tm
, to_char(to_char(max(stamp)/(24*60*60) + date '1987-07-07', 'DD-MON HH24:MI:SS')) end_tm
FROM V$RMAN_STATUS
START WITH (RECID, STAMP) =
(SELECT MAX(session_recid),MAX(session_stamp) FROM V$RMAN_OUTPUT)
CONNECT BY PRIOR RECID = parent_recid ;
I needed to send timestamp to GrayLog via GELF from Oracle DB. I tried different versions and solutions but only one worked correctly.
SQL:
SELECT REPLACE((CAST(dat AS DATE) - TO_DATE('19700101', 'YYYYMMDD')) * 86400 + MOD(EXTRACT(SECOND FROM dat), 1), ',', '.') AS millis
FROM (SELECT SYSTIMESTAMP AT TIME ZONE 'GMT' AS dat FROM dual)
The result for Systmiestamp
2018/12/18 19:47:29,080988 +02:00
will be
1545155249.080988

Resources