I have a table called users, which has the following columns:
id: INT NOT NULL
face: face_type
face_type is an ENUM type that has the following values: 'square', 'round' and 'triangle'.
And I have another table called houses, which has the following columns:
id: INT NOT NULL
user_id: INT NOT NULL
Now, I want to get all the houses grouped by the different type of face types. So, what I have so far is this:
SELECT users.face_type, COUNT(*)
FROM users
LEFT JOIN houses ON houses.user_id = users.id
GROUP BY users.face_type
The problem is that I also want to get rows for face_type which none of the users have, as well as a result for NULL face_type. So, for example, if I have the following data:
users (id, face_type)
1, 'round'
2, 'triangle'
houses (id, user_id)
1, 1
2, 1
3, 2
I would expect the result to be:
face_type, count
'round' 2
'triangle' 1
'square' 0
null 0
I know how to get all the potential values of the face_type ENUM, by doing :
SELECT unnest(enum_range(NULL::face_type)) AS face_types;
But I don't know how to use that to count all potential face types in the aggregate, as well as also calculating for NULL face types.
You can use LEFT JOIN:
SELECT ft.face_type, COUNT(h.user_id)
FROM (SELECT unnest(enum_range(NULL::face_type)) AS face_types
) ft LEFT JOIN
users u
ON u.face_type = ft.face_type LEFT JOIN
houses h
ON h.user_id = u.id
GROUP BY ft.face_type;
To get NULL, just use UNION ALL:
SELECT ft.face_type, COUNT(h.user_id)
FROM (SELECT unnest(enum_range(NULL::face_type)) AS face_types
UNION ALL
SELECT NULL
) ft LEFT JOIN
users u
ON u.face_type = ft.face_type LEFT JOIN
houses h
ON h.user_id = u.id
GROUP BY ft.face_type;
Of course, the = will not every match. If that is possible, then you want to change the JOIN condition to u.face_type is not distinct from ft.face_type.
to COUNT(houses.*)
SELECT face_type.type, COUNT(houses.*)
FROM (SELECT unnest(enum_range(NULL::face_type))) AS face_type(type)
FULL JOIN users ON users.face_type=face_type.type
LEFT JOIN houses ON houses.user_id = users.id
GROUP BY face_type.type
A LEFT JOIN starting from the ENUM and going to users and houses will allow you to recover totals for each enumerated value. To also display the NULL face types, you can use a UNION query.
SELECT
ft.face_type,
COUNT(ho.user_id) as cnt
FROM
(SELECT unnest(enum_range(NULL::face_type)) AS face_types) ft
LEFT JOIN users us ON us.face_type = ft.fact_type
LEFT JOIN houses ho ON ho.user_id = us.id
GROUP BY ft.face_type
UNION
SELECT
null,
COUNT(ho.user_id)
FROM houses ho
INNER JOIN users us ON ho.user_id = us.id AND us.face_type IS NULL
ORDER BY cnt desc
Related
I have 3 tables user, student_data, teacher_data. A user can be either student or a teacher. If it is the teacher I want to join user and teacher_data. And if it is a student then I want to join user with student_data.
How I can do this join with the condition.
I'd combine the two data tables in a sub-query, and then join the users to that.
SELECT
*
FROM
usr u
LEFT JOIN
(
SELECT user_id, datum, xxx, NULL AS yyy FROM student_data
UNION ALL
SELECT user_id, datum, NULL, yyy FROM teacher_data
)
d
ON d.user_id = u.id
https://dbfiddle.uk/?rdbms=oracle_21&fiddle=9b801ea739d42fe50c00ef4e17eaf143
NOTES:
The columns selected from the two data tables must match
Any unmatched columns must either be skipped or filled with NULL
Please don't call a table user, it's a reserved keyword and Oracle won't allow it.
You can write it like this:
select u.user_id,
s.student_id,
t.teacher_id
from usr u
left join student_data s on u.user_id=s.student_id
left join teacher_data t on u.user_id=t.teacher_id
where s.student_id is not null or t.teacher_id is not null
order by u.user_id
For every user_id check if he is a student or teacher, if he is student get his student column values else null, if he is a teacher get his teacher column values else null.
maybe try a union - something like this
select user_id, user_other_stuff
from user, student_data
where user.user_id = student_data.user_id
UNION
select user_id, user_other_stuff
from user, teacher_data
where user.user_id = teacher_data.user_id
I have two tables, activity and contacts in postgresql.
An activity can have multiple contacts in array form, like this
contact Id = {23,54,34}.
I want to delete a activity only if all the contact Ids of that activity are deleted in contacts table and keep the activity if at least one one contact id is still not deleted.
Deleted At is column in contacts table to check for deleted contacts. I don't want to use NOT IN.
Activity table
id contact Id
-------------------
16 {60,61,23}
15 {}
5 {59}
6 {}
You can use simple EXISTS predicate testing contacts table with activity.contacts array:
create table activity (
id int primary key,
contacts int[]
)
create table contacts (
id int primary key,
name varchar(10),
deleted boolean
)
insert into activity
select 16 as id, '{1,2,3}'::int[] as contacts union all
select 15, null union all
select 5, '{4}' union all
select 6, '{6, 5}'
insert into contacts
select 1 as id, 'q' as name, false as deleted union all
select 2, 'w', false union all
select 3, 'e', true union all
select 4, 'r', false union all
select 5, 't', true union all
select 6, 'y', true
delete
from activity a
where not exists (
select null
from contacts c
where not(c.deleted)
and c.id = any(a.contacts)
)
2 rows affected
db<>fiddle here
Use UNNEST to convert the nested-array to rows, then do an anti-join to look for broken references:
(An anti-join is when you perform a x LEFT OUTER JOIN y with a WHERE y.KeyColumn IS NULL - this gives you the opposite of an INNER JOIN with the same join criteria).
WITH unnested AS (
SELECT
Id AS ActivityId,
UNNEST( ContactId ) AS ContactIdFromArray
FROM
crmActivity
)
SELECT
u.ActivityId,
u.ContactIdFromArray AS Missing_ContactId_in_Activity
FROM
unnested AS u
LEFT OUTER JOIN contacts AS c ON
c.ContactId = u.ContactIdFromArray
WHERE
c.ContactId IS NULL
ORDER BY
u.ActivityId
I want to delete a activity only if all the contact Ids of that activity are deleted in contacts table and keep the activity if at least one one contact id is still not deleted.
This can be done with a DELETE FROM using WHERE crmActivity.Id IN with a CTE that generates the correct set of bad crmActivity.Id values, via a GROUP BY with the above query:
WITH unnested AS (
SELECT
Id AS ActivityId,
UNNEST( ContactId ) AS ContactIdFK
FROM
crmActivity
)
WITH brokenContacts AS (
SELECT
u.ActivityId,
u.ContactIdFK,
c.ContactId AS ContactIdPK
FROM
unnested AS u
LEFT OUTER JOIN contacts AS c ON
c.ContactId = u.ContactIdFromArray
)
WITH counts AS (
SELECT
ActivityId,
COUNT(*) AS ContactIdCount,
COUNT( CASE WHEN ContactIdPK IS NULL THEN 1 END ) AS BadContactIdCount
FROM
brokenContacts
GROUP BY
ActivityId
)
WITH badActivityIds AS (
SELECT
ActivityId
FROM
counts
WHERE
BadContactIdCount = ContactIdCount
AND
ContactIdCount > 0
)
DELETE FROM
crmActivity
WHERE
ActivityId IN ( SELECT ActivityId FROM badActivityIds );
The question im trying to answer is : to find the names of guest who visited more than 2 different restaurants on 15-JUNE-20.
There is a:
Guest table with GID,Gname
Visit table with VID, GID, RESTID, VDATE
Restaurant table with RESTID, RNAME
whenever i tried introducing the groupby i would get the error
SELECT GuestN.GID, GuestN.Gname
FROM GuestN
WHERE GuestN.GID IN (
SELECT VisitN.GID
FROM VisitN
WHERE VisitN.Vdate = '15-JUN-20' AND VisitN.restID IN (
SELECT RestaurantN.Restid
FROM RestaurantN having count(*)>2));
The table RestaurantN is not needed since you have restID in the table VisitN and you are not interested in the restaurant'a name, but only on their number.
Join GuestN to VisitN, aggregate and set the condition in the HAVING clause:
SELECT g.GID, g.Gname
FROM GuestN g INNER JOIN VisitN v
ON v.GID = g.GID
WHERE v.Vdate = '15-JUN-20'
GROUP BY g.GID, g.Gname
HAVING COUNT(DISTINCT v.restID) > 2
I have the following tables (with given columns):
houses (id)
users (id, house_id, active)
custom_values (name, house_id, type)
I want to get all the (distinct) houses and the count of associated users that:
have at least 1 associated custom_value which name column contains the string 'red' (case insensitive) AND the custom_value column type value is 'mandatory'.
have at least 100 associated users which status column is 'active'
How can I run this query in PostgreSQL?
Right now I have this query (which was answered in Get records where associated records name contain a string AND associated record count is bigger than threshold), but I don't know how to select the count of users too (:
select h.*
from houses
where
exists (
select 1
from custom_values cv
where cv.house_id = h.house_id and cv.type = 'mandatory' and lower(cv.name) = 'red'
)
and (
select count(*)
from users u
where u.house_id = h.house_id and u.status = 'active'
) >= 100
You can turn the subquery to a lateral join:
select h.*, u.no_users
from houses h
cross join lateral (
select count(*) no_users
from users u
where u.house_id = h.house_id and u.status = 'active'
) u
where
u.cnt >= 100
and exists (
select 1
from custom_values cv
where cv.house_id = h.house_id and cv.type = 'mandatory' and lower(cv.name) = 'red'
)
I have articles and users tables. I also have another table articles_users with FK columns: userId articleId.
I also have userId=1
How to get get all the rows also with extra column that tell me if the userid is linked to this row?
I have try to do this with left join but the problem is articles_users have duplicate entries like:
articleId: 1, userId: 1
articleId: 2, userId: 1
articleId: 3, userId: 2
And it get duplicate articles rows or none.
SELECT * FROM articles LEFT JOIN articles_users ON articles_users.articleid = articles.id
WHERE articles_users.userid = 1
I would recommend case and exists:
SELECT a.*,
(CASE WHEN EXISTS (SELECT 1
FROM articles_users au
WHERE au.articleid = a.id AND
au.userid = 1
)
THEN 1 ELSE 0
END) as flag_1
FROM articles a;
Although you can use JOIN, I wouldn't recommend it. If you followed the same pattern for multiple users, you might end up with multiple rows.
Many databases support boolean types explicitly. In those databases, you can eliminate the CASE:
SELECT a.*,
(SELECT 1
FROM articles_users au
WHERE au.articleid = a.id AND
au.userid = 1
) as flag_1
FROM articles a;