I wrote a query that works like a charm in SQL Server. Unfortunately it needs to be run on an Oracle db. I have been searching the web inside out for a solution on how to convert it, without any success :/
The query looks like this i SQL:
UPDATE tab1 SET budgpost_gr1=
CASE WHEN (budgpost in (1001,1012,50055)) THEN 'BP_GR_A'
WHEN (budgpost in (5,10,98,0)) THEN 'BP_GR_B'
WHEN (budgpost in (11,876,7976,67465))
ELSE 'Missing' END`
My problem is also that the columns budgetpost_gr1
and budgetpost is alphanumeric and Oracle seems to want to see the list as numbers. The list are variables/parameters that is pre-defined as comma separated lists, which is just dumped开发者_如何学C into the query.
You said that budgetpost is alphanumeric. That means it is looking for comparisons against strings. You should try enclosing your parameters in single quotes (and you are missing the final THEN in the Case expression).
UPDATE tab1
SET budgpost_gr1= CASE
WHEN (budgpost in ('1001','1012','50055')) THEN 'BP_GR_A'
WHEN (budgpost in ('5','10','98','0')) THEN 'BP_GR_B'
WHEN (budgpost in ('11','876','7976','67465')) THEN 'What?'
ELSE 'Missing'
END
Use to_number
to convert budgpost
to a number:
when to_number(budgpost,99999) in (1001,1012,50055) THEN 'BP_GR_A'
EDIT: Make sure there are enough 9
's in to_number
to match to largest budget post.
If there are non-numeric budget posts, you could filter them out with a where
clause at then end of the query:
where regexp_like(budgpost, '^-?[[:digit:],.]+$')
Got a solution that runs. Don't know if it is optimal though. What I do is to split the string according to http://blogs.oracle.com/aramamoo/2010/05/how_to_split_comma_separated_string_and_pass_to_in_clause_of_select_statement.html
Using:
select regexp_substr(' 1, 2 , 3 ','[^,]+', 1, level) from dual
connect by regexp_substr('1 , 2 , 3 ', '[^,]+', 1, level) is not null;
So my final code looks like this ($bp_gr1'
are strings like 1,2,3
):
UPDATE TAB1
SET BUDGPOST_GR1 =
CASE
WHEN ( BUDGPOST IN (SELECT REGEXP_SUBSTR ( '$BP_GR1',
'[^,]+',
1,
LEVEL )
FROM DUAL
CONNECT BY REGEXP_SUBSTR ( '$BP_GR1',
'[^,]+',
1,
LEVEL )
IS NOT NULL) )
THEN
'BP_GR1'
WHEN ( BUDGPOST IN (SELECT REGEXP_SUBSTR ( ' $BP_GR2',
'[^,]+',
1,
LEVEL )
FROM DUAL
CONNECT BY REGEXP_SUBSTR ( '$BP_GR2',
'[^,]+',
1,
LEVEL )
IS NOT NULL) )
THEN
'BP_GR2'
WHEN ( BUDGPOST IN (SELECT REGEXP_SUBSTR ( ' $BP_GR3',
'[^,]+',
1,
LEVEL )
FROM DUAL
CONNECT BY REGEXP_SUBSTR ( '$BP_GR3',
'[^,]+',
1,
LEVEL )
IS NOT NULL) )
THEN
'BP_GR3'
WHEN ( BUDGPOST IN (SELECT REGEXP_SUBSTR ( '$BP_GR4',
'[^,]+',
1,
LEVEL )
FROM DUAL
CONNECT BY REGEXP_SUBSTR ( '$BP_GR4',
'[^,]+',
1,
LEVEL )
IS NOT NULL) )
THEN
'BP_GR4'
ELSE
'SAKNAR BUDGETGRUPP'
END;
Is there a way to make it run faster?
"The list are variables/paramaters that is pre-defined as comma separated lists". Do you mean that your query is actually
UPDATE tab1 SET budgpost_gr1=
CASE WHEN (budgpost in ('1001,1012,50055')) THEN 'BP_GR_A'
WHEN (budgpost in ('5,10,98,0')) THEN 'BP_GR_B'
WHEN (budgpost in ('11,876,7976,67465'))
ELSE 'Missing' END`
If so, you need a function to take a string and parse it into a list of numbers.
create type tab_num is table of number;
create or replace function f_str_to_nums (i_str in varchar2) return tab_num is
v_tab_num tab_num := tab_num();
v_start number := 1;
v_end number;
v_delim VARCHAR2(1) := ',';
v_cnt number(1) := 1;
begin
v_end := instr(i_str||v_delim,v_delim,1, v_start);
WHILE v_end > 0 LOOP
v_cnt := v_cnt + 1;
v_tab_num.extend;
v_tab_num(v_tab_num.count) :=
substr(i_str,v_start,v_end-v_start);
v_start := v_end + 1;
v_end := instr(i_str||v_delim,v_delim,v_start);
END LOOP;
RETURN v_tab_num;
end;
/
Then you can use the function like so:
select column_id,
case when column_id in
(select column_value from table(f_str_to_nums('1,2,3,4'))) then 'red'
else 'blue' end
from user_tab_columns
where table_name = 'EMP'
There is another workaround you can use to update using a join. This example below assumes you want to de-normalize a table by including a lookup value (in this case storing a users name in the table). The update includes a join to find the name and the output is evaluated in a CASE statement that supports the name being found or not found. The key to making this work is ensuring all the columns coming out of the join have unique names. In the sample code, notice how b.user_name conflicts with the a.user_name column and must be aliased with the unique name "user_user_name".
UPDATE
(
SELECT a.user_id, a.user_name, b.user_name as user_user_name
FROM some_table a
LEFT OUTER JOIN user_table b ON a.user_id = b.user_id
WHERE a.user_id IS NOT NULL
)
SET user_name = CASE
WHEN user_user_name IS NOT NULL THEN user_user_name
ELSE 'UNKNOWN'
END;
精彩评论