I have been using Postgres (version 9.2 and 9.3) with this function: https://wiki.postgresql.org/wiki/Aggregate_Mode for a while. Now recently after upgrading to version 9.4 I run into the following error when using the function:
PG::WrongObjectType: ERROR: WITHIN GROUP is required for ordered-set aggregate mode
LINE 1: SELECT mode(logins_count) AS mode_value FROM "registrations" WHERE "registrations"."cr...
The error occurs when doing:
SELECT mode(logins_count) AS mode_value FROM registrations
WHERE registrations.created_at > '20141105';
I do not understand the error message and I do not get what I have to change?
Postgres 9.4 introduced a new subclass of aggregate functions. Per documentation:
There is a subclass of aggregate functions called ordered-set
aggregates for which an order_by_clause is required, usually because
the aggregate's computation is only sensible in terms of a specific
ordering of its input rows.
One of the new built-in ordered-set aggregate functions is mode()
which happens to collide with the name of your custom aggregate function. (The Postgres Wiki page you are referring to has not been updated since 2013.)
Solution
Use a different name for your custom aggregate function to avoid the collision.
Or better yet: use the new built-in function instead. Read the updated Postgres Wiki for details.
Related
I was creating a function following an example from a database class which included the creation of a temporary variable (base_salary) and using a SELECT INTO to calculate its value later.
However, I did not realize I used a different order for the syntax (SELECT ... FROM ... INTO base_salary) and the function could be used later without any visible issues (values worked as expected).
Is there any difference in using "SELECT ... FROM ... INTO" syntax order? I tried looking about it in the PostgreSQL documentation but found nothing about it. Google search did not provide any meaningful information neither. Only thing I found related to it was from MySQL documentation, which only mentioned about supporting the different order in an older version.
There is no difference. From the docs of pl/pgsql:
The INTO clause can appear almost anywhere in the SQL command.
Customarily it is written either just before or just after the list of
select_expressions in a SELECT command, or at the end of the command for other command types. It is recommended that you follow
this convention in case the PL/pgSQL parser becomes stricter in future
versions.
Notice that in (non-procedural) SQL, there is also a SELECT INTO command which works like CREATE TABLE AS, in this version the INTO must come right after the SELECT clause.
I always use SELECT ... INTO ... FROM , I believe that is the standard supported notation
https://www.w3schools.com/sql/sql_select_into.asp
I would recommend using this, also if there are any updates or if the other version might become unsupported as you mentioned...
i have a DB2 data source and an Oracle 12c target.
The Oracle has a DB link to the DB2 defined which is working in general.
Now i have a huge table in the DB2 which has a timestamp column (lets call it ROW_CHANGED) for row changes. I want to retrieve rows which have changed after a particular time.
Running
SELECT * FROM lib.tbl WHERE ROW_CHANGED >'2016-08-01 10:00:00'
on the DB2 returns exactly 1 row after ca. 90 secs which is fine.
Now i try the same query from the Oracle via the db link:
SELECT * FROM lib.tbl#dblink_name WHERE ROW_CHANGED >TO_TIMESTAMP('2016-08-01 10:00:00')
This runs for hours and ends up in a timeout.
I read some Oracle docs and found distributed query optimization tips but most of them refer to joining a local to a remote table which is not my case.
In my desperation, i have tried the DRIVING_SITE hint, without effect.
Now i wonder when the WHERE part of the query will be evaluated. Since i have to use Oracle syntax and not DB2 syntax for the query, is it possible the Oracle will try to first copy the full table and apply the where clause afterwards? I did some research but did not find anything which would help me in this direction.
The ROW_CHANGED is a hidden column in the DB2, if that matters.
Thx for any hint in advance.
Update
Thanks#all for help. I'll share what did the trick for me.
First of all i have used TO_TIMESTAMP since the DB2 column is also Timestamp (not date) and i had expected to circumvent implicit conversions by this.
Without the explicit conversion i ran into ORA-28534: Heterogeneous Services preprocessing error and i have no hope of touching the DB config within reasonable time.
The explain plan btw did not bring much. It showed a FULL hint and no conversion on the predicates. Indeed it showed the ROW_CHANGED column as Date, i wonder why.
I have tried Justins suggestion to use a bind variable, however i got ORA-28534 again. Next thing i did was to wrap it into a pl/sql block (will run in a SP anyway later).
declare
v_tmstmp TIMESTAMP := 01.08.16 10:00:00;
begin
INSERT INTO ORAUSER.TMP_TBL (SRC_PK,ROW_CHANGED)
SELECT SRC_PK,ROW_CHANGED
FROM lib.tbl#dblink_name
WHERE ROW_CHANGED > v_tmstmp;
end;
This was executing in the same time as in DB2 itself. The date format is DD.MM.YY here since it is the default unfortunately.
When changing the variable assignment to
v_tmstmp TIMESTAMP := TO_TIMESTAMP('01.08.16 10:00:00','DD.MM.YY HH24:MI:SS');
I got the same problem as before.
Meanwhile the DB2 operators have created an index in the ROW_CHANGED column which i requested earlier that day. This has solved the problem in general it seems. Even my original query finishes in no time now.
If you are actually using an Oracle-specific conversion function like to_timestamp, that forces the predicate to be evaluated on the Oracle side. Oracle isn't going to know how to convert a built-in function like to_timestamp into an exactly equivalent function call in DB2.
If you used a bind variable, that would be more likely to get evaluated on the DB2 side. But that may be complicated by the data type mapping between different databases-- there may not be a perfect mapping between one engine's date and another engine's timestamp data type. If this was a numeric column, a bind variable would be almost certain to get pushed. In this case, it probably involves playing around a bit to figure out exactly what data type to use for your variable that works for your framework, Oracle, and DB2.
If using a bind variable doesn't work, you can force the predicate to be evaluated on the remote server using the dbms_hs_passthrough package. That lets you send a query verbatim to the remote server which allows you to do things like use functions defined in your DB2 database. That's a bit of overkill in this situation, hopefully, but it's nice to have the hammer as your backup if the simpler solution doesn't work quickly enough.
I have some data in a postgres table with one column called version (of type varchar). I would like to use my own comparison function to to order/sort on that column, but I am not sure what is the most appropriate answer:
I have an JS implementation of the style comp(left, right) -> -1/0/1, but I don't know how I can use it in a sql order by clause (through plv8)
I could write a C extension, but I am not particularly excited about this (mostly for maintenance reason, as writing the comparison in C would not be too difficult in itself)
others ?
The type of comparisons I am interested are similar to version string ordering used in package managers.
You want:
ORDER BY mycolumn USING operator
See the docs for SELECT. It looks like you may need to define an operator for the function, and a b-tree operator class containing the operator to use it; you can't just write USING myfunc().
(No time to test this and write a demo right now).
I was planning to use the WITH clause with PostgreSQL, but it doesn't seem to support the command. Is there a substitute command?
What I want to do is with one query select several sub-resultsets and use parts of the sub-resultsets to create my final SELECT.
That would have been easy using the WITH clause.
UPDATE:
Opps! I discovered that I misunderstood the error message I got; and pgSQL does support WITH.
PostgreSQL supports common-table expressions (WITH queries) in version 8.4 and above. See common table expressions in the manual.
You should really include your PostgreSQL version, the exact text of the error message, and the exact text of any query you ran in your question. Where practical/relevant also include table definitions, sample data, and expected results.
I have a generic code that is used to retrieve DDL information from a Firebird database (FB2.1). It generates SQL code like
SELECT * FROM MyTable where 'c' <> 'c'
I cannot change this code. Actually, if that matters, it is inside Report Builder 10.
The fact is that some tables from my database are becoming a litle too populated (>1M records) and that query is starting to take too long to execute.
If I try to execute
SELECT * FROM MyTable where SomeIndexedField = SomeImpossibleValue
it will obviously use that index and run very quickly.
Well, it wouldn´t be that hard to the database find out that that is an impossible matcher and make some sort of optimization and avoid testing it against each row.
Is there any way to make my firebird database to optimize that search?
As the filter condition is a negative proposition (and also doesn't refer a column to search, but only a value to compare to another value), Firebird need to do a full table scan (without use any index) to confirm that aren't any record that meet your criteria.
If you can't change you need to wait for the upcoming 3.0 version, that will implement the Boolean data type, and therefore should start to evaluate "constant" fake comparisons in advance (maybe the client library will do this evaluation before send the statement to the server?).