Consider the SELECT statement below:
SELECT 1, 'A'
UNION ALL
SELECT 2, 'B'
UNION ALL
SELECT 3, 'C';
The result is obvious:
1 'A'
2 'B'
3 'C'
I tried to store it as a separate table:
CREATE TABLE tmp AS
SELECT 1, 'A'
UNION ALL
SELECT 2, 'B'
UNION ALL
SELECT 3, 'C';
The contents of the tmp table are surprising:
1 'A'
2 'A'
3 'A'
Ok, that can be fixed providing explicit field names:
CREATE TABLE tmp AS
SELECT 1 AS field1, 'A' AS field2
UNION ALL
SELECT 2, 'B'
UNION ALL
SELECT 3, 'C';
Now I have a question whether the observed behavior is defined and valid. I feel that the statement without explicit field names should end with an error instead of producing a surprising result.
I'm using SQLite.
Related
I have two tables like
create table nodes_tbl as (
select 'a' as nodeid, 'some string' as dummy_string, 0 as subnetid from dual union all
select 'b', 'qwe', 0 from dual union all
select 'c', 'asd', 0 from dual union all
select 'd', 'zxc', 0 from dual union all
select 'e', 'rty', 0 from dual);
And
create table subnets as (
select 'a' as nodeid, 1 as subnetid from dual union all
select 'b', 2 from dual union all
select 'c', 2 from dual union all
select 'd', 3 from dual union all
select 'e', 4 as nodeid from dual);
With several millions of records a join works fast.
select n.NODEID, n.DUMMY_STRING, s.subnetid
from nodes_tbl n, subnets s where s.nodeid=n.nodeid
Writes are fast as well
create table test_tbl as n.NODEID, s.subnetid
from nodes_tbl n, subnets s where s.nodeid=n.nodeid --10M records in 2s.
However, when I try to update table and add values to the column the query is very slow
UPDATE nodes_tbl n
SET subnetid = (SELECT subnetid
FROM subnets s
WHERE s.nodeid = n.nodeid)
WHERE EXISTS (
SELECT subnetid FROM subnets s
WHERE s.nodeid = n.nodeid) --8 minutes for 100K records
Why is insert so much slower than a create table from a select statement?
What is the most efficient way to do this insert?
I know about create view option, but want to avoid it.
Try MERGE instead:
merge into nodes_tbl n
using (select s.subnetid, s.nodeid
from subnets s
) x
on (x.nodeid = n.nodeid)
when matched then update set
n.subnetid = x.subnetid;
Any improvement?
By the way, did you create index on NODEID column in both tables?
I have 2 lists which I need to compare. I need to find if at least one element from List A is found in List B. I know IN doesn't work with 2 lists. What are my other options?
Basically something like this :
SELECT
CASE WHEN ('A','B','C') IN ('A','Z','H') THEN 1 ELSE 0 END "FOUND"
FROM DUAL
Would appreciate any help!
You are probably looking for something like this. The WITH clause is there just to simulate your "lists" (whatever you mean by that); they are not really part of the solution. The query you need is just the last three lines (plus the semicolon at the end).
with
first_list (str) as (
select 'A' from dual union all
select 'B' from dual union all
select 'C' from dual
),
second_list(str) as (
select 'A' from dual union all
select 'Z' from dual union all
select 'H' from dual
)
select case when exists (select * from first_list f join second_list s
on f.str = s.str) then 1 else 0 end as found
from dual
;
FOUND
----------
1
In Oracle you can do:
select
count(*) as total_matches
from table(sys.ODCIVarchar2List('A', 'B', 'C')) x,
table(sys.ODCIVarchar2List('A', 'Z', 'H')) y
where x.column_value = y.column_value;
You need to repeat the conditions:
SELECT (CASE WHEN 'A' IN ('A', 'Z', 'H') OR
'B' IN ('A', 'Z', 'H') OR
'C' IN ('A', 'Z', 'H')
THEN 1 ELSE 0
END) as "FOUND"
FROM DUAL
If you are working with collection of String you can try Multiset Operators.
create type coll_of_varchar2 is table of varchar2(4000);
and:
-- check if exits
select * from dual where cardinality (coll_of_varchar2('A','B','C') multiset intersect coll_of_varchar2('A','Z','H')) > 0;
-- list of maching elments
select * from table(coll_of_varchar2('A','B','C') multiset intersect coll_of_varchar2('A','Z','H'));
Additionally:
-- union of elemtns
select * from table(coll_of_varchar2('A','B','C') multiset union distinct coll_of_varchar2('A','Z','H'));
select * from table(coll_of_varchar2('A','B','C') multiset union all coll_of_varchar2('A','Z','H'));
-- eelemnt from col1 not in col2
select * from table(coll_of_varchar2('A','A','B','C') multiset except all coll_of_varchar2('A','Z','H'));
select * from table(coll_of_varchar2('A','A','B','C') multiset except distinct coll_of_varchar2('A','Z','H'));
-- check if col1 is subset col2
select * from dual where coll_of_varchar2('B','A') submultiset coll_of_varchar2('A','Z','H','B');
I am trying to do something very similar but the first list is another field on the same query created with listagg and containing integer numbers like:
LISTAGG(my_first_list,', ') WITHIN GROUP(
ORDER BY
my_id
) my_first_list
and return this with all the other fields that I am already returning
SELECT
CASE WHEN my_first_list IN ('1,2,3') THEN 1 ELSE 0 END "FOUND"
FROM DUAL
I am using oracle 10g and i need to write a query where in the table that is to be considered for producing the output is based on the user input.
i have written in the following manner, but getting an error.
UNDEFINE CDR
SELECT F.EMPLOYEE_ID FROM
( SELECT DECODE(&&CDR,25,'TABLE 1' ,22,'TABLE 2' ,19,'TABLE 3' ,16,'TABLE 4') FROM DUAL ) F
WHERE F.FLAG='G';
The closest that you can come without dynamic SQL is:
select EMPLOYEE_ID
from table1
where flag = 'G' and &&CDR = 25
union all
select EMPLOYEE_ID
from table2
where flag = 'G' and &&CDR = 19
union all
select EMPLOYEE_ID
from table4
where flag = 'G' and &&CDR = 16
union all
select EMPLOYEE_ID
from table1
where flag = 'G' and &&CDR not in (25, 19, 16)
Is it possible to keep order from a 'IN' conditional clause?
I found this question on SO but in his example the OP have already a sorted 'IN' clause.
My case is different, 'IN' clause is in random order
Something like this :
SELECT SomeField,OtherField
FROM TestResult
WHERE TestResult.SomeField IN (45,2,445,12,789)
I would like to retrieve results in (45,2,445,12,789) order. I'm using an Oracle database. Maybe there is an attribute in SQL I can use with the conditional clause to specify to keep order of the clause.
There will be no reliable ordering unless you use an ORDER BY clause ..
SELECT SomeField,OtherField
FROM TestResult
WHERE TestResult.SomeField IN (45,2,445,12,789)
order by case TestResult.SomeField
when 45 then 1
when 2 then 2
when 445 then 3
...
end
You could split the query into 5 queries union all'd together though ...
SELECT SomeField,OtherField
FROM TestResult
WHERE TestResult.SomeField = 4
union all
SELECT SomeField,OtherField
FROM TestResult
WHERE TestResult.SomeField = 2
union all
...
I'd trust the former method more, and it would probably perform much better.
Decode function comes handy in this case instead of case expressions:
SELECT SomeField,OtherField
FROM TestResult
WHERE TestResult.SomeField IN (45,2,445,12,789)
ORDER BY DECODE(SomeField, 45,1, 2,2, 445,3, 12,4, 789,5)
Note that value,position pairs (e.g. 445,3) are kept together for readability reasons.
Try this:
SELECT T.SomeField,T.OtherField
FROM TestResult T
JOIN
(
SELECT 1 as Id, 45 as Val FROM dual UNION ALL
SELECT 2, 2 FROM dual UNION ALL
SELECT 3, 445 FROM dual UNION ALL
SELECT 4, 12 FROM dual UNION ALL
SELECT 5, 789 FROM dual
) I
ON T.SomeField = I.Val
ORDER BY I.Id
There is an alternative that uses string functions:
with const as (select ',45,2,445,12,789,' as vals)
select tr.*
from TestResult tr cross join const
where instr(const.vals, ','||cast(tr.somefield as varchar(255))||',') > 0
order by instr(const.vals, ','||cast(tr.somefield as varchar(255))||',')
I offer this because you might find it easier to maintain a string of values rather than an intermediate table.
I was able to do this in my application using (using SQL Server 2016)
select ItemID, iName
from Items
where ItemID in (13,11,12,1)
order by CHARINDEX(' ' + Convert("varchar",ItemID) + ' ',' 13 , 11 , 12 , 1 ')
I used a code-side regex to replace \b (word boundary) with a space. Something like...
var mylist = "13,11,12,1";
var spacedlist = replace(mylist,/\b/," ");
Importantly, because I can in my scenario, I cache the result until the next time the related items are updated, so that the query is only run at item creation/modification, rather than with each item viewing, helping to minimize any performance hit.
Pass the values in via a collection (SYS.ODCINUMBERLIST is an example of a built-in collection) and then order the rows by the collection's order:
SELECT t.SomeField,
t.OtherField
FROM TestResult t
INNER JOIN (
SELECT ROWNUM AS rn,
COLUMN_VALUE AS value
FROM TABLE(SYS.ODCINUMBERLIST(45,2,445,12,789))
) i
ON t.somefield = i.value
ORDER BY rn
Then, for the sample data:
CREATE TABLE TestResult ( somefield, otherfield ) AS
SELECT 2, 'A' FROM DUAL UNION ALL
SELECT 5, 'B' FROM DUAL UNION ALL
SELECT 12, 'C' FROM DUAL UNION ALL
SELECT 37, 'D' FROM DUAL UNION ALL
SELECT 45, 'E' FROM DUAL UNION ALL
SELECT 100, 'F' FROM DUAL UNION ALL
SELECT 445, 'G' FROM DUAL UNION ALL
SELECT 789, 'H' FROM DUAL UNION ALL
SELECT 999, 'I' FROM DUAL;
The output is:
SOMEFIELD
OTHERFIELD
45
E
2
A
445
G
12
C
789
H
fiddle
In my SQL Server 2005 database, using an SLQ query, does anyone know the best way to group records together by one field, and get a comma-separated list of the values from another?
So if I have:
UserID Code
1 A
1 C5
1 X
2 V3
3 B
3 D
3 NULL
3 F4
4 NULL
I'd get:
UserID Code
1 A,C5,X
2 V3
3 B,D,F4
4 NULL
Thanks for any help.
WITH Data AS (
SELECT 1 UserId, 'A' Code
UNION ALL
SELECT 1, 'C5'
UNION ALL
SELECT 1, 'X'
UNION ALL
SELECT 2, 'V3'
UNION ALL
SELECT 3, 'B'
UNION ALL
SELECT 3, 'D'
UNION ALL
SELECT 3, NULL
UNION ALL
SELECT 3, 'F4'
UNION ALL
SELECT 4, NULL
)
SELECT U.UserId, STUFF((
SELECT ','+Code FROM Data WHERE Data.UserID = U.UserID FOR XML PATH('')
), 1, 1, '') Code
FROM (SELECT DISTINCT UserID FROM Data) U
Just replace the Data CTE with your table name and you're done.
There it´s a complete review of forms to do that in TSQL
http://www.simple-talk.com/sql/t-sql-programming/concatenating-row-values-in-transact-sql/