I am trying to optimize a query and I have it down to something like this,
select a.* from
(select id, count(oid) as cnt from stuff1 s1 inner join stuff2 s2 on s1.id=s2.id group by id) as a
right join
(select id,'0' as cnt from stuff2) as b
on a.id = b.id
Basically the goal was to get the count for each oid, where those having 0 count are also included. I had a query previous to this that worked fine but it took 30 seconds to execute. I am looking to optimize the old query with this one, but I am getting NULL values from table b. I need the values from table b to show up with id and 0. Any help would be greatly appreciated.
An example of the data set could be,
Stuff1
| oid | id |
|---- |----|
| 1 | 1 |
| 2 | 1 |
| 3 | 2 |
| 4 | 3 |
Stuff2
| id |
|----|
| 1 |
| 2 |
| 3 |
| 4 |
| 5 |
| 6 |
| 7 |
the query should produce
| id | cnt |
|----|-----|
| 1 | 2 |
| 2 | 1 |
| 3 | 1 |
| 4 | 0 |
| 5 | 0 |
| 6 | 0 |
| 7 | 0 |
Your query is syntactically incorrect (oid may not be defined; id in the select is ambiguous). However, I suspect you want a simple left join:
select s2.id, count(s1.id) as cnt
from stuff2 s2 left join
stuff1 s1
on s1.id = s2.id
group by s2.id;
Related
so what I am trying to do is having 3 tables (pictures, collections, and bridge) with the following columns:
Collections Table:
| id | name |
------------------
| 1 | coll1 |
| 2 | coll2 |
------------------
Pictures Table: (timestamps are unix timestamps)
| id | name | timestamp |
-------------------------
| 5 | Pic5 | 1 |
| 6 | Pic6 | 19 |
| 7 | Pic7 | 3 |
| 8 | Pic8 | 892 |
| 9 | Pic9 | 4 |
-------------------------
Bridge Table:
| id | collection | picture |
-----------------------------
| 1 | 1 | 5 |
| 2 | 1 | 6 |
| 3 | 1 | 7 |
| 4 | 1 | 8 |
| 5 | 2 | 5 |
| 6 | 2 | 9 |
| 7 | 2 | 7 |
-----------------------------
And the result should look like this:
| collection_name | picture_count | newest_picture |
----------------------------------------------------
| coll1 | 4 | 8 |
| coll2 | 3 | 9 |
----------------------------------------------------
newest_picture should always be the picture with the heighest timestamp in that collection and I also want to sort the result by it. picture_count is obviously the count of picture in that collection.
Can this be done in a single statement with table joins and if yes:
how can I do this the best way?
A simple method uses correlated subqueries:
select c.*,
(select count(*)
from bridge b
where b.collection = c.id
) as pic_count,
(select p.id
from bridge b join
pictures p
on b.picture = b.id
where b.collection = c.id
order by p.timestamp desc
limit 1
) as most_recent_picture
from collections c;
A more common approach would use window functions:
select c.id, c.name, count(bp.collection), bp.most_recent_picture
from collections c left join
(select b.*,
first_value(p.id) over (partition by b.collection order by p.timestamp desc) as most_recent_picture
from bridge b join
pictures p
on b.picture = p.id
) bp
on bp.collection = c.id
group by c.id, c.name, bp.most_recent_picture;
I have a number of tables in my database.
Table: ObjectToPerson
For example if I had a number of entries below in the database:
+----+------------+------------+----------+----------+--------------+
| Id | WeekNumber | Date | PersonId | ObjectId | ObjectTypeId |
+----+------------+------------+----------+----------+--------------+
| 1 | 1 | 2015-11-04 | 1 | 1 | 1 |
| 2 | 1 | 2015-11-04 | 1 | 3 | 2 |
| 3 | 1 | 2015-11-04 | 2 | 2 | 1 |
| 4 | 1 | 2015-11-04 | 2 | 4 | 2 |
+----+------------+------------+----------+----------+--------------+
I am wanting to return the results back as two lines as follows:
+------+------------+----------+----------------------------+----------------------------+
| Week | Date | PersonId | ObjectId(ObjectTypeId = 1) | ObjectId(ObjectTypeId = 2) |
+------+------------+----------+----------------------------+----------------------------+
| 1 | 2015-11-04 | 1 | 1 | 3 |
| 1 | 2015-11-04 | 2 | 2 | 4 |
+------+------------+----------+----------------------------+----------------------------+
I am thinking of some sort of Group By query but I just can't seem to get it right.
Select * From ObjectToPerson
Left Join Objects O On O.Id = ObjectToPerson.ObjectId And ObjectToPerson.ObjectTypeId = 1
Left Join Objects O On O.Id = ObjectToPerson.ObjectId And ObjectToPerson.ObjectTypeId = 2
Can someone explain how I would get to this please?
You could use CASE to only select the ObjectId if the type is correct for the column, then use MAX/GROUP BY to group the result into a single row per person/week/date.
SELECT WeekNumber week, date, personid,
MAX(CASE WHEN ObjectTypeId=1 THEN ObjectId END) Type1,
MAX(CASE WHEN ObjectTypeId=2 THEN ObjectId END) Type2
FROM ObjectToPerson
GROUP BY week, date, personid
An SQLfiddle to test with.
You don't want two joins, you want a WHERE clause;
SELECT * FROM ObjectToPerson
LEFT JOIN Objects O ON O.Id = ObjectToPerson.ObjectId
WHERE ObjectToPerson.ObjectTypeId IN(1,2)
I have 3 tables like this
SecretAgents
| id | name |
|----|------|
| 1 | A |
| 2 | B |
Victims
| id | name | agent_id |
|----|------|----------|
| 1 | Z | 1 |
| 2 | Y | 1 |
| 3 | X | 2 |
Data
| id | keys | values | victim_id | form_id |
|----|------|--------|-----------|---------|
| 1 | a1 | x | 1 | 1 |
| 2 | a2 | xx | 1 | 2 |
| 3 | a3 | xxx | 2 | 1 |
| 4 | a5 | xxx | 1 | 1 |
I have to get the count of forms(here victim_id and form_id are composite primary keys) and the count of victims for each agent.
I have tried this for any 2 tables with left joins and group by but I am not able to achieve the same together. If anyone can be generous enough to offer a pointer/solution, that would be super awesome..
EDIT 1: The query
This is definitely not the right query but anyways
SELECT count(DISTINCT v.id) as victimcount, `sa`.`username`, `sa`.`id`, count(DISTINCT d.form_id) as submissions
FROM `SecretAgents` as `sa`
LEFT JOIN `Victims` as `v` ON `v`.`agent_id`=`sa`.`id`
LEFT JOIN `Data` as `d` ON `d`.`victim_id`=`v`.`id`
GROUP BY `v`.`agent_id`
ORDER BY `sa`.`id` ASC
The victimcount is correct but the submissions count becomes wrong. Tried lots of other things too but this is the most relevant...
Thanks
I believe you can count the forms-per-agent like so:
SELECT COUNT(*) as form_count, a.id as id, a.name as agent
FROM Data d
LEFT JOIN Victims v ON v.id = d.victim_id
LEFT JOIN SecretAgents a on v.agent_id = a.id
GROUP BY a.id;
To count the victims, just leave off the Data table.
I was having problems in creating counting rows by grouping based on a given field value.
For example: I have a Table A structure like this:
+------+------------+
| id | Person |
+------+------------+
| 1 | "Sandy" |
| 2 | "Piper" |
| 3 | "Candy" |
| 4 | "Pendy" |
+------------+------+
Also I have a Table B structure like this:
+------+------------+---------+
| id | Person | Point |
+------+------------+---------+
| 1 | "Sandy" | 10 |
| 2 | "Piper" | 20 |
| 3 | "Candy" | 30 |
| 4 | "Sandy" | 10 |
| 5 | "Piper" | 20 |
| 6 | "Zafar" | 30 |
+------------+------+---------+
And needed a result like:
+------+------------+---------+
| id | Person | Point |
+------+------------+---------+
| 1 | "Piper" | 40 |
| 2 | "Candy" | 30 |
| 3 | "Zafar" | 30 |
| 4 | "Sandy" | 20 |
| 5 | "Pendy" | 0 |
+------------+------+---------+
I hope the table examples are itself self-explanatory.
SELECT person
, SUM(point) total
FROM
( SELECT person,point FROM table_b
UNION
ALL
SELECT person,0 FROM table_a
) x
GROUP
BY person
ORDER
BY total DESC;
It is a simple left join with a group by
select tableA.person, sum(tableB.points) from tableA left join tableB on tableA.person = tableB.person group by tableA.person
union
select tableB.person, sum(tableB.points) from tableB left join tableA on tableA.person = tableB.person where tableA.id is null group by tableA.person
I think below sql useful to you.
select a.id, a.Person,b.total_point from (
select id, Person from tablea) as a join
(select Person, sum(Point) as total_point from tableb group by person) as b on a.person =b.person
Thank you
So lets say I have 2 or more tables consisting of dissimilar columns in which a shared key (id) is not necessarily present :
Alpha:
+----+-------+-------+-------+
| id | paula | randy | simon |
+----+-------+-------+-------+
| 1 | 8 | 7 | 2 |
| 2 | 9 | 6 | 2 |
| 3 | 10 | 5 | 2 |
+----+-------+-------+-------+
Beta:
+----+---------+-----+------------+------+
| id | is_nice | sex | dob | gift |
+----+---------+-----+------------+------+
| 2 | 1 | F | 1990-05-25 | iPod |
| 3 | 0 | M | 1990-05-25 | coal |
+----+---------+-----+------------+------+
Gamma:
+----+---------+--------+
| id | is_tall | is_fat |
+----+---------+--------+
| 1 | 1 | 1 |
| 99 | 0 | 1 |
+----+---------+--------+
The desired effect is to mash the tables together on id inserting NULLs where data is not available:
+----+-------+-------+-------+---------+-----+------------+------+---------+--------+
| id | paula | randy | simon | is_nice | sex | dob | gift | is_tall | is_fat |
+----+-------+-------+-------+---------+-----+------------+------+---------+--------+
| 1 | 8 | 7 | 2 | | | | | 1 | 1 |
| 2 | 9 | 6 | 2 | 1 | F | 1990-05-25 | iPod | | |
| 3 | 10 | 5 | 2 | 0 | M | 1990-05-25 | coal | 1 | 1 |
| 99 | | | | | | | | 0 | 0 |
+----+-------+-------+-------+---------+-----+------------+------+---------+--------+
I can use NULL 'dummy' columns and UNION (MySql SELECT union for different columns?) but that seems like a royal pain if the number of tables is great. I'd like to think there is a JOIN method I can use to accomplish this, but I need some help to figure this out.
This works:
SELECT `id`, `paula`, `randy`, ..., NULL AS `is_nice`, ... FROM `Alpha`
UNION SELECT `id`, NULL AS `paula`, ..., FROM `Beta`
UNION SELECT `id`, NULL AS `paula`, ..., `is_fat` FROM `Gamma` ;
but it sure feels like the wrong way to do it. How can I get the same results without having to edit lines and lines of SQL inserting NULL AS whatever all over the place whenever I want to tack on additional tables?
Thanks in advance!
SELECT
allid.id
, a.paula, a.randy a.simon
, b. ...
, c. ...
FROM
( SELECT id
FROM Alpha
UNION
SELECT id
FROM Beta
UNION
SELECT id
FROM Gamma
) AS allid
LEFT JOIN
Alpha AS a
ON a.id = allid.id
LEFT JOIN
Beta AS b
ON b.id = allid.id
LEFT JOIN
Gamma AS g
ON g.id = allid.id
If the tables share no other column except the id, you could use the simple to write (but easier to break):
SELECT
*
FROM
( SELECT id
FROM Alpha
UNION
SELECT id
FROM Beta
UNION
SELECT id
FROM Gamma
) AS allid
NATURAL LEFT JOIN
Alpha
NATURAL LEFT JOIN
Beta
NATURAL LEFT JOIN
Gamma
You want to use LEFT JOINs.
http://dev.mysql.com/doc/refman/5.0/en/join.html
In your example:
SELECT id_t.id, a.paula, a.randy, a.simon, b.is_nice, b.sex, b.dob, b.gift, g.is_tall, g.is_fat
FROM (SELECT DISTINCT id FROM alpha,beta,gamma) as id_t
LEFT JOIN alpha a ON a.id = id_t.id
LEFT JOIN beta b on b.id = id_t.id
LEFT JOIN gamma g on g.id = id_t.id