How to simulate a N-dimension INTERSECT with generic SQL - mysql

Here is an example of 2 extract from the same table:
SELECT source_id
FROM table_cust_string_value
WHERE cust_string_id=2 AND VALUE LIKE '%TATA%';
SELECT source_id
FROM table_cust_string_value
WHERE cust_string_id=4 AND VALUE LIKE '%TUTU%';
They give 2 sets of source_id.
Right. Now if I need an intersect of those with MySQL (where INTERSECT does not exist) I found this way:
SELECT DISTINCT source_id
FROM (
SELECT source_id
FROM table_cust_string_value
WHERE cust_string_id=2 AND VALUE LIKE '%TATA%'
) t1
INNER JOIN (
SELECT source_id
FROM table_cust_string_value
WHERE cust_string_id=4 AND VALUE LIKE '%TUTU%'
) t2
USING (source_id);
but what if I need to do this from N sets ?
I can't find a solution + I'm worried about the perf. of doing it this way

You can use a grouping approach. Depending on what indexes you have available this might work out better.
SELECT source_id
FROM table_cust_string_value
WHERE cust_string_id IN ( 2, 4 )
GROUP BY source_id
HAVING MAX(CASE WHEN cust_string_id = 2 AND VALUE LIKE '%TATA%' THEN 1 END) = 1
AND MAX(CASE WHEN cust_string_id = 4 AND VALUE LIKE '%TUTU%' THEN 1 END) = 1

Related

How can I get the multiple columns in a single row using MySQL?

I want combine mutiple rows columns as single record in MySQL
For eg: Actual Data
--------------------------------------------------------------
id name Loantype Amount
--------------------------------------------------------------
1 ABC 1 500000
2 ABC 2 3500000
3 XYZ 1 250000
4 XYZ 2 2500000
I tried with the following query
SELECT
id,
(
CASE Loantype
WHEN 1 THEN Amount
ELSE NULL
END
) AS PersonalLoan,
(
CASE Loantype
WHEN 2 THEN Amount
ELSE NULL
END
) AS HomeLoan
FROM
customer
WHERE
name = 'ABC'
but result comes as below
--------------------------------------------------------------
id name PersonalLoan HomeLoan
--------------------------------------------------------------
1 ABC 500000 NULL
1 ABC NULL 2500000
Expected Result set
--------------------------------------------------------------
id name PersonalLoan HomeLoan
--------------------------------------------------------------
1 ABC 500000 3500000
You can self-join the table so that you will be able to combine the 2 kinds of loans into single row:
SELECT t1.id, t1.name, t1.amount AS personal, t2.amount AS home
FROM customer AS t1
LEFT JOIN customer AS t2 ON t1.name = t2.name AND t2.loantype = 2
WHERE t1.loantype = 1
If you are looking for just 1 user - you can speedup the query by limiting the size of the JOIN:
SELECT t1.id, t1.name, t1.amount AS personal, t2.amount AS home
FROM (SELECT * FROM customer WHERE name = "ABC" AND loantype = 1) AS t1
LEFT JOIN customer AS t2 ON t1.name = t2.name AND t2.loantype = 2
Note that generally speaking you shouldn't denormalize data in SQL (e.g. converting rows to columns: SQL is row-oriented, not column-oriented) - I assume this is to simplify display logic - just be careful of using queries like this when you want to pass meaningful data to other parts of the database rather than directly to the user.
You need to GROUP BY first.
You can also simplify your CASE expressions:
ELSE NULL is always implicit and if the CASE WHEN expression is an equality comparison then you can use the simpler switch-style syntax.
So CASE WHEN a = b THEN c ELSE NULL END can be simplified to CASE a WHEN b THEN c END.
I've added COALESCE so the query will return 0 for the SUM aggregates if there are no matching rows instead of NULL. Note that COUNT (unlike SUM) generally doesn't need to be wrapped in a COALESCE (though I forget precisely how MySQL handles this - it also depends on what version of MySQL you're using and what strict-mode and ANSI/ISO-compliance options are enabled).
Note that the database design you posted seems to allow the same customer.name to have multiple loans of the same loantype.
You can avoid this by adding a UNIQUE CONSTRAINT or UNIQUE INDEX or use a Composite Primary Key:
CREATE UNIQUE INDEX UX_name_loantype ON customer ( name, loantype )
To prevent those rows from causing issues in this query, this uses a SUM and a COUNT to make it clear to readers that the data is an aggregate over multiple rows:
SELECT
name,
COUNT( CASE Loantype WHEN 1 THEN 1 END ) AS CountPersonalLoans,
COALESCE( SUM( CASE Loantype WHEN 1 THEN Amount END ), 0 ) AS SumPersonalLoans,
COUNT( CASE Loantype WHEN 2 THEN 1 END ) AS CountHomeLoans,
COALESCE( SUM( CASE Loantype WHEN 2 THEN Amount END ), 0 ) AS SumHomeLoans
FROM
customer
GROUP BY
name
To maximise query code reuse if you want to filter by name then convert this to a VIEW - or if it's a one-off then make it a CTE query, like so:
WITH aggs AS (
SELECT
name,
COUNT( CASE Loantype WHEN 1 THEN 1 END ) AS CountPersonalLoans,
COALESCE( SUM( CASE Loantype WHEN 1 THEN Amount END ), 0 ) AS SumPersonalLoans,
COUNT( CASE Loantype WHEN 2 THEN 1 END ) AS CountHomeLoans,
COALESCE( SUM( CASE Loantype WHEN 2 THEN Amount END ), 0 ) AS SumHomeLoans
FROM
customer
GROUP BY
name
)
SELECT
name,
CountPersonalLoans,
SumPersonalLoans,
CountHomeLoans,
SumHomeLoans
FROM
aggs
WHERE
name = 'ABC'
ORDER BY
name

Join or SubQuery?

I've got a MySQL table that records the addon titles used on various websites, including a version number. For example:
AddonName | Website ID | Version
ZZZ 1 3.3
ZZZ 2 3.4
ZZZ 3 3.4
ZZZ 4 3.1
YYY 1 1.1
YYY 2 1.1
YYY 3 1.1
YYY 4 1.2
I'd like to create a query that lists a distinct list of AddonName, with details of the total count, count of all sites using the latest version, and counts of all sites using out of date versions.
i.e.:
Name | Total Addons | Up to Date | Out of Date
ZZZ 4 2 2
YYY 4 1 3
I can't figure out how to get this type of data returned, even though the information is all there. I tried using JOIN queries, but didn't have any success.
If it helps make things easier, I can add a 'latest' enum field to the table, to mark rows as up-to-date or out-of-date when the are imported.
Assuming max value as latest version.
Try this:
select t1.AddonName,
count(*) as total_Addon,
sum(case when t1.version=t2.version then 1 else 0 end) as up_to_date,
sum(case when t1.version!=t2.version then 1 else 0 end) as out_of_date
from table1 t1
inner join(
select AddonName,max(version) as version
from table1 group by AddonName
)t2 on t1.AddonName=t2.AddonName
group by t1.AddonName
Try:
SELECT your_table.AddonName,
COUNT(`Website ID`),
COUNT(IF(Version = your_table_max.max_version, 1, NULL)) AS `Up to Date`,
COUNT(IF(Version <> your_table_max.max_version, 1, NULL)) AS `Out of Date`
FROM your_table
INNER JOIN (SELECT MAX(Version) as max_version, AddonName
FROM your_table group by AddonName) your_table_max
ON your_table_max.AddonName = your_table.AddonName
GROUP BY your_table.AddonName;
Assuming the latest version is from the last column:
select t.name, count(*) as TotalAddons,
sum(t.version = tt.maxv) as UpToDate,
sum(t.version <> tt.maxv) as OutOfDate
from t join
(select name, max(version) as maxv
from t
group by name
) tt
on t.name = tt.name
group by t.name;
This calculates the maximum version number for each name in a subquery. It then uses that information for the outer aggregation.
This assumes that version is a number. If it is a string (so 1.10 > 1.2), then a similar approach is:
select t.name, count(*) as TotalAddons,
sum(t.version = t.maxv) as UpToDate,
sum(t.version <> t.maxv) as OutOfDate
from (select t.*,
(select version
from t t2
where t2.name = t.name
order by length(version) desc, version desc
limit 1
) as maxv
from t
) t
group by t.name;
Of course, this will also work for numbers as well.
try this this will solve your problem.
select AddonName,count(AddonName) as countAdd,(select count(Version)from test1 as t where t.AddonName = test1.AddonName and t.Version = max(test1.Version)),(select count(Version) from test1 as t where t.AddonName = test1.AddonName and t.Version = min(test1.Version))from test1 GROUP BY AddonName;
Once you add that Latest table, try following (after inserting your table names)
select
AddonName as Name,
count(*) as TotalAddons,
count(case TableName.Version when Latest.LatestVersion
then 1 else null end) as UpToDate,
TotalAddons-UpToDate as OutOfDate
from TableName join Latest
on TableName.AddonName = Latest.AddonName
group by AddonName
WITH cte
AS
(
SELECT t.AddonName, MAX(t.Version) AS latest_version
FROM Table1 t
GROUP BY t.AddonName
)
SELECT t.AddonName, COUNT(t.WebsiteID) AS total_addons,
SUM
(
CASE WHEN t.version = cte.latest_version
THEN 1 ELSE 0 END
) AS up_to_date,
SUM
(
CASE WHEN t.version <> cte.latest_version
THEN 1 ELSE 0 END
) AS out_of_date
FROM Table1 t
JOIN cte ON t.AddonName = cte.AddonName
GROUP BY t.AddonName

How to bypass a reference to an outer table in subquery?

I've been dealing with these two tables:
Document
id company_id etc
=======================
1 2 x
2 2 x
Version
id document_id version date_created date_issued date_accepted
==========================================================================
1 1 1 2013-04-29 2013-04-30 NULL
2 2 1 2013-05-01 NULL NULL
3 1 2 2013-05-01 2013-05-01 2013-05-03
There's a page where I want to list all documents with their attributes.
And I would like to add a single have status from each document.
The status can be derived from the most present date that corresponding Versions have.
It is possible that an older version is being accepted.
The query result I am looking for is like this:
id company_id etc status
==================================
1 2 x accepted
2 2 x created
I started out by making a query which combines all dates and add a status next to it.
It works as expected and when I add the document_id things look alright.
SELECT `status`
FROM (
SELECT max(date_created) as `date`,'created' as `status` FROM version WHERE document_id = 1
UNION
SELECT max(date_issued),'issued' FROM version WHERE document_id = 1
UNION
SELECT max(date_accepted),'accepted' FROM version WHERE document_id = 1
ORDER BY date DESC
LIMIT 1
) as maxi
When I try to incorporate this query as a subquery, I can't make it work.
SELECT *, (
SELECT `status` FROM (
SELECT max(date_created) as `date`,'created' as `status`FROM version WHERE document_id = document.id
UNION
SELECT max(date_issued),'issued' FROM version WHERE document_id = document.id
UNION
SELECT max(date_accepted),'accepted' FROM version WHERE document_id = document.id
ORDER BY date DESC
LIMIT 1
) as maxi
) as `status`
FROM `document`
This will get me the error Unknown column 'document.id' in 'where clause'. So I've read around at SO and figured it simply can't reach the value offer.id since it's a subquery in a subquery. So I tried to take another approach and get all the statuses at once, to avoid the WHERE statement, and JOIN them. I ended up with the next query.
SELECT MAX(`date`),`status`, document_id
FROM (
SELECT datetime_created as `date`, 'created' as `status`,document_id FROM `version`
UNION
SELECT datetime_issued, 'issued',document_id FROM `version`
UNION
SELECT datetime_accepted, 'accepted',document_id FROM `version`
) as dates
GROUP BY offer_id
No error this time but I realized that the status couldn't be the correct one since it got lost during the GROUP BY. I've tried combinations of the two but both flaws keep hindering me. Could any one suggest how to do this in a single query without changing my database? (I know that saving the dates in a separate table would simply things)
I have not tested this, but you can do it like this (you might need to tweak the details)
It is basically looking at it from a completely different angle.
select
d.*,
(CASE GREATEST(ifnull(v.date_created, 0), ifnull(v.date_issued,0), ifnull(v.date_accepted,0) )
WHEN null THEN 'unknown'
WHEN v.date_accepted THEN 'accepted'
WHEN v.date_issued THEN 'issued'
WHEN v.date_created THEN 'created'
END) as status
from document d
left join version v on
v.document_id = d.document_id and
not exists (select 1 from (select * from version) x where x.document_id = v.document_id and x.id <> v.id and x.version > v.version)
Can you normalise your table designs to move the status / dates onto a different table from the Versions?
If no possibly something like this:-
SELECT Document.id, Document.company_id, Document.etc, CASE WHEN Sub1.status = 3 THEN 'accepted' WHEN Sub1.status = 2 THEN 'issued' WHEN Sub1.status = 1 THEN 'created' ELSE NULL END AS status
FROM Document
INNER JOIN (
SELECT document_id, MAX(CASE WHEN date_accepted IS NOT NULL THEN 3 WHEN date_issued IS NOT NULL THEN 2 WHEN date_created IS NOT NULL THEN 1 ELSE NULL END) AS status
FROM Version
GROUP BY document_id
) Sub1
ON Document.id = Sub1.document_id
The subselect gets the highest status for any document from the version table. Each possible versions highest status is returned as a number, and by grouping that on the document id it will get the highest status of any version. This is joined back against the Document table and the number for the version number converted into the text description.
select Doc.document_id,Doc.company_id,Doc.etc,f.status
from Document Doc
inner join
(select Ver.document_id,
case when Ver.date_accepted is not null then 'Accepted'
when Ver.date_issued is not null then 'Issued'
when Ver.date_created is not null then 'Created'
end as status
from version Ver
inner join (
select document_id,MAX(version) VersionId
from version
group by document_id
)t on t.document_id=Ver.document_id
where t.VersionId=Ver.version
)f on Doc.document_id=f.document_id
SQL Fiddle

Don't return any result if the precedent query has at least 1 result?

Basically I have this query:
( SELECT * FROM tbl WHERE type = 'word1' )
UNION
( SELECT * FROM tbl WHERE type = 'word2' ) // Run this query if there are no results with type = 1
Basically I would like to run the second query only if the first hasn't any results. is it possible?
The FIRST "PreCheck" query does a count of how many records ARE of type = 1. After that, if the count is greater than 1, then return 1, otherwise return 2.
Now, THAT answer can be used in the join (which is always a single row via COUNT(*)) which will either have a 1 or 2 value. THAT value will be the second value is the EQUALITY conditon. So, if there IS an entry of 1, the result will be as if
WHERE t1.Type = 1
Thus never allowing any 2 in the test. HOWEVER, if NO entries are found, it will have a value of 2 and thus create a WHERE clause of
WHERE t1.type = 2
select t1.*
from
( select if( count(*) > 0, 1, 2 ) IncludeType
from tbl t2
where t2.type = 1 ) preCheck,
tbl t1
where
t1.type = preCheck.IncludeType
If there is an index on the "type" column, the first query should be almost instantaneous.
You could write
select * from tbl
where type = 1
union
select * from tbl
where type = 2
and not exists( select * from tble where type = 1 )
but this probably won't perform as well as just doing it in your program
It does the trick:
SELECT tbl.* FROM tbl JOIN (SELECT min(type) min_type FROM tbl WHERE type between 1 and 2 ) on min_type = type
First, it selects the lesser of these two types, if any exists, and then oins this one number table to your table. It is actually a simple filter. You can use WHERE instead of JOIN, if you want.
SELECT tbl.* FROM tbl WHERE (SELECT min(type) FROM tbl WHERE type between 1 and 2 ) = type

MySQL - Matching 2 rows (or more!) within a sub-query/table

I have this schema which I need to match 2 rows from
user_data : user_id, field_id, value
A sample output would be:
user_id field_id value
-----------------------------
1 1 Gandalf
1 2 Glamdring
How do I write a query which basically say "Find the user id of the user whose field_id 1 is Gandalf, and field_id 2 is Glamdring?"
SELECT FROM looks at one row at a time. I am stumped on this. I will also need to find a solution that scale gracefully (such as looking at three rows etc.)
You could run a query to get the users that match each of the conditions and intersect the results. Since MySQL doesn't support intersect you can do it with an n-way join:
SELECT T1.user_id
FROM Table1 T1
JOIN Table1 T2 ON T1.user_id = T2.user_id
WHERE T1.field_id = 1 AND T1.value = 'Gandalf'
AND T2.field_id = 2 AND T2.value = 'Glamdring'
I would try the following:
SELECT user_id
FROM user_data
WHERE ( field_id = 1 AND value= 'Gandalf' )
OR ( field_id = 3 AND value = 'Glamdring' )
GROUP BY user_id
HAVING COUNT( field_id ) = 2
It will search for all the rows that match one of your criteria, and use GROUP BY and HAVING afterwards to find the user_id that has the expected count of matches.
select * from user_date where ( field_id= 1 AND value='Gandalf' ) OR ( field_id =2 AND value ='Glamdring' ) ;
The HAVING clause is the key. It turns the query from an "OR" statement into an "AND" statement