this is the original SQL, that works very well.
select a.sessionid
from session a
left
outer
join applicationinfo b
on a.sessionid=b.sessionid
left
outer
join license c
on a.sessionid=c.sessionid
limit 10
;
below is the converted SQL
SELECT s.sessionid
FROM
( SELECT *
FROM session a
LEFT
OUTER
JOIN applicationinfo b
ON (a.sessionid = b.sessionid)
)s
LEFT
OUTER
JOIN license c
ON (s.sessionid = c.sessionid)
limit 10
;
the second SQL will get 10 NULLs. Since those tables are left joined on the sessionid field. Suppose in the second SQL the s.sessionid should not be NULL.
Could someone tell me the reason why those two SQL cannot return the same result. Thanks.
Related
In my query I want total file count and total closed files in the same table.
My first query:
select hi.eksper_id,ef.ad, count(hi.eksper_id) as total_files
from hasar_ihbar as hi
left outer join eksper_firma ef on ef.id=hi.eksper_id
group by hi.eksper_id,ef.ad
My second query:
select ef.id as eksper_id,ef.ad,count(ef.id) closed_files
from hasar_ihbar_rapor hir
left outer join hasar_ihbar hi on hi.id = hir.hasar_ihbar_id
left outer join eksper_firma ef on ef.id = hi.eksper_id
where hir.rapor_tipi = 3 group by ef.id,ef.ad
I want both combined and this is my code:
select ef.id as eksper_id,ef.ad,count(ef.id) closed_files, count(hi.id) AS total_files
from hasar_ihbar_rapor hir
left outer join hasar_ihbar hi on hi.id = hir.hasar_ihbar_id
left outer join eksper_firma ef on ef.id = hi.eksper_id
where hir.rapor_tipi = 3 group by ef.id,ef.ad
I don't know what I did wrong. Thanks for your help.
The double joins effect your counts since there are more rows.
A better way to combine both counts is to create a query that has two subqueries for each count.
SELECT id,
(SELECT COUNT(*)...) total_files,
(SELECT COUNT(*)...) closed_files
FROM ...
I wrote this query against a redshift database
with wt_id(id) as (select cast(11646722011 as bigint))
select
w.id,
nullif(
trim(
regexp_replace(
json_extract_path_text(bn.foo, 'foo-id'), '\\["(.*)"\\]', '$$1'
)
), '') as foo-id
from wt_id w left outer join foo bn on w.id = bn.fid
where bn.day='2019-12-03'
limit 1;
This query returns an empty result set. But how is this possible? because I did a left out join, I should get 1 row where id is 11646722011 and the foo-id can be null if there was no match on the right hand side.
Your where clause is converting the outer join to an inner join because NULL fails the WHERE condition.
Conditions on the second table in a left join should be in the on clause. So you want:
from wt_id w left outer join
foo bn
on w.id = bn.fid and bn.day = '2019-12-03'
Your where clause has made your data loss. This is not the case about left join allows unmatched rows but about no data on that specific day after left join. So, you should replace where with and.
If we see the order of execution the where gets applied after the join as above. Hence, even if the rows gets null it gets removed if the other condition gets true as your day.
I have the following tables with the following attributes:
Op(OpNo, OpName, Date)
OpConvert(OpNo, M_OpNo, Source_ID, Date)
Source(Source_ID, Source_Name, Date)
Fleet(OpNo, S_No, Date)
I have the current multiple JOIN query which gives me the results that I want:
SELECT O.OpNo AS Op_NO, O.OpName, O.Date AS Date_Entered, C.*
FROM Op O
LEFT OUTER JOIN OpConvert C
ON O.OpNo = C.OpNo
LEFT OUTER JOIN Source D
ON C.Source_ID = D.Source_ID
WHERE C.OpNo IS NOT NULL
The problem is this. I need to join the Fleet table on the previous multiple JOIN statement to attach the relevant S_No to the multiple JOIN table. Would I still be able to accomplish this using a LEFT OUTER JOIN or would I have to use a different JOIN statement? Also, which table would I JOIN on?
Please note that I am only familiar with LEFT OUTER JOINS.
Thanks.
I guess in your case you could use INNER JOIN or LEFT JOIN (which is the same thing as LEFT OUTER JOIN in SQL Server.
INNER JOIN means that it will only return records from other tables only if there are corresponding records (based on the join condition) in the Fleet table.
LEFT JOIN means that it will return records from other tables even if there are no corresponding records (based on the join condition) in the Fleet table. All columns from Fleet will return NULL in this case.
As for which table to join, you should really join the table that makes more logical sense based on your data structure.
Yes, you can use all tables mentioned before in your join conditions. Actually, JOINS (no matter of INNER, LEFT OUTER, RIGHT OUTER, CROSS, FULL OUTER or whatever) are left- associative, i. e. they are implicitly evaluated as if they would have been included in parentheses from the left as follows:
FROM ( ( ( Op O
LEFT OUTER JOIN OpConvert C
ON O.OpNo = C.OpNo
)
LEFT OUTER JOIN Source D
ON C.Source_ID = D.Source_ID
)
LEFT OUTER JOIN Fleet
ON ...
)
This is similar to how + or - would implicitly use parentheses, i. e.
2 + 3 - 4 - 5
is evaluated as
(((2 + 3) - 4) - 5)
By the way: If you use C.OpNo IS NOT NULL, then the LEFT OUTER JOIN Source D is treated as if it were an INNER JOIN, as you are explicitly removing all the "OUTER" rows.
I have the following query:
SELECT MS.idReg, MS.dsMotivo, A.contrato FROM MS
INNER JOIN S
ON S.motivoSiniestro = MS.idReg
INNER JOIN C
ON C.n__contrat = S.n__contrat
INNER JOIN A
ON A.n__article = C.n__article
Table MS has only 12 records, the ones I need and others have many more entries.
My problem is that I only want the 12 records from MS and their contrato column but I'm getting much more that that. Have tried many combinations of INNER, OUTER, LEFT and RIGHT joins. Any help?
You get too many records because there are several A.contrato values for each row in the MS table. Sql server does not know which one of all the A.contrato values to take so it returns all of them. First you need to decide which one you want.
If any will do you can simply write your query like this:
SELECT MS.idReg, MS.dsMotivo, MAX(A.contrato)
FROM MS
INNER JOIN S
ON S.motivoSiniestro = MS.idReg
INNER JOIN C
ON C.n__contrat = S.n__contrat
INNER JOIN A
ON A.n__article = C.n__article
GROUP BY MS.idReg, MS.dsMotivo
Try this one -
SELECT MS.idReg, MS.dsMotivo, A.contrato
FROM dbo.MS
OUTER APPLY (
SELECT TOP 1 A.contrato
FROM dbo.S
JOIN dbo.C ON C.n__contrat = S.n__contrat
JOIN dbo.A ON A.n__article = C.n__article
WHERE S.motivoSiniestro = MS.idReg
) s
I wrote a query. this query sum fields from 2 different table. And grouped by main table id field. But second left outer join is not grouped and giving me different results.
SELECT s.*,
f.firma_adi,
sum(sd.fiyat) AS konak,
sum(ss.fiyat) AS sponsor
FROM fuar_sozlesme1 s
INNER JOIN fuar_firma_2012 f
ON ( s.cari = f.cari )
LEFT OUTER JOIN fuar_sozlesme1_detay sd
ON ( sd.sozlesme_id = s.id )
LEFT OUTER JOIN fuar_sozlesme1_sponsor ss
ON ( ss.sozlesme_id = s.id )
GROUP BY s.id
ORDER BY s.id DESC
I know, it is really complicated but I'm stucking on this issue.
My question is: why second left outer join is not correctly sum of field . If I remove second left outer join or first, everything is normal.
The problem is that you have multiple dimensions on your data, and the number of rows is multiplying beyond what you expect. I would suggest that you run the query for one id, without the group by, to see what rows the join is producing.
One way to fix this is by using correlated subqueries:
select s.*, f.firma_adi,
(select SUM(sd.fiyat)
from fuar_sozlesme1_detay fd
where sd.sozlesme_id = s.id
) as konak,
(select SUM(ss.fiyat)
from fuar_sozlesme1_sponsor ss
where (ss.sozlesme_id = s.id)
) as sponsor
from fuar_sozlesme1 s inner join
fuar_firma_2012 f
on (s.cari = f.cari)
order by s.id DESC
By the way, you appear to by using MySQL (because your query is not parsable in any other dialect). You should tag your questions with the version of the database you are using.