Left join sum, group trouble - mysql

I have this query
SELECT
`from_id` as user_id,
MAX(`createdon`) as updated_at,
SUM(`unread`) as new,
u.username,
p.sessionid,
s.access
FROM (
SELECT `from_id`, `createdon`, `unread`
FROM `modx_messenger_messages`
WHERE `to_id` = {$id}
UNION
SELECT `to_id`, `createdon`, 0
FROM `modx_messenger_messages`
WHERE `from_id` = {$id}
ORDER BY `createdon` DESC
) as m
LEFT JOIN `modx_users` as u ON (u.id = m.from_id)
LEFT JOIN `modx_user_attributes` as p ON (p.internalKey = m.from_id)
LEFT JOIN `modx_session` as s ON (s.id = p.internalKey)
GROUP BY `from_id`
ORDER BY `new` DESC, `createdon` DESC;
table
id | message | createdon | from_id | to_id | unread
1 | test | NULL | 5 | 6 | 0
2 | test2 | NULL | 6 | 5 | 1
3 | test3 | NULL | 6 | 5 | 1
result new = 28. Why?
If remove joins new = 2, correctly.

Though it depends on the actual database, pure SQL says that a statement using GROUP BY requires all non-aggregated columns to be in the GROUP BY. Without including all columns, weird stuff can happen, which might explain why you get different results. If you know that the other columns are going to be the same within the user_id, you could do MAX(u.username) or something similar (again, depending on your database server). So I'd try and clean up the SQL statement first.

Related

How to count in mysql

I have this query in mySQL where I would like to sum the line product of each doctor but I dont know how to do it.
use avant_medical;
select
sales.doctor_id as DoctorID,
line_products.id as LineProductID,
line_products.name as LineProductName
from `doctors`
inner join `sales` on `doctors`.`id` = `sales`.`doctor_id`
inner join `inventories` on `sales`.`id` = `inventories`.`sale_id`
inner join `products` on `inventories`.`product_id` = `products`.`id`
inner join `line_products` on `products`.`lineProduct_id` = `line_products`.`id`
order by `doctors`.`id` asc;
lPID= lineProductID
|DrID|lPID |
| -- | ----|
| 1 | 7 |
| 1 | 6 |
| 1 | 6 |
| 1 | 7 |
| 1 | 7 |
| 1 | 7 |
| 1 | 6 |
This is how I want:
Doctor 1
lineID | quantity
7 | 4
6 | 3
I try this query only in mySQL
The keyword you are looking for is count, not sum. Summing would add up every lineProductID as if they where regular mathematical values, while counting will add up how many times a given lineProductID is found.
select
sales.doctor_id as DoctorID,
line_products.id as LineProductID,
line_products.name as LineProductName,
-- We count the number of occurrences of each line_product.id
COUNT(line_products.id) as LineProductQty
from `doctors`
inner join `sales` on `doctors`.`id` = `sales`.`doctor_id`
inner join `inventories` on `sales`.`id` = `inventories`.`sale_id`
inner join `products` on `inventories`.`product_id` = `products`.`id`
inner join `line_products` on `products`.`lineProduct_id` = `line_products`.`id`
-- Never forget to properly GROUP your aggregate functions, such as COUNT() or SUM()!
GROUP BY sales.doctor_id, line_products.id, line_products.name
order by `doctors`.`id` asc;
Since you didn't provided full schema to test this, I made a small, very artificial demo, but should be representative of how the query above works.

MySQL Sort users by highest total count from counts on multiple tables

Similar to stackoverflow, I have a database of users who vote, comment, and make other actions. I am trying to return a sorted result of the top 10 users who have made the most actions based on the combined count of all of the actions a user has made, along with the actual count of total actions said user made.
Below is my table structure.
Users Table
Typical users data such as an incrementing id, username, email, etc.
| id | username |
-----------------
| 1 | bob |
| 2 | jane |
Votes Table
Has an incrementing id, user_id fk and type of vote made.
| id | user_id | type |
| 1 | 1 | up_vote |
| 2 | 2 | up_vote |
Comments Table
Same as the votes table, typical stuff here.
| id | user_id | comment |
---------------------------------
| 1 | 1 | hello, world |
| 1 | 1 | goodbye, world |
Intended results:
results needed
| total_actions | user_id | username |
-------------------------------------|
| 3 | 1 | bob |
| 1 | 2 | jane |
What I actually know how to do, albeit probably not the most efficient way...
Users sorted by most votes, along with the count
select `users`.*,
(
select count(*)
from `votes`
where `users`.`id` = `votes`.`user_id`
) as `votes_count`
from `users`
order by `votes_count` desc
limit 10
Users sorted by most comments, along with the count
select `users`.*,
(
select count(*)
from `comments`
where `users`.`id` = `comments`.`user_id`
) as `comments_count`
from `users`
order by `comments_count` desc
limit 10
Any help would be greatly appreciated!
You can left join aggregate queries that compute the total votes and comments per user, and sort in the outer query, like so:
select
coalesce(v.cnt, 0) + coalesce(c.cnt, 0) total_actions,
u.id,
u.username
from users u
left join (select user_id, count(*) cnt from votes group by user_id) v
on v.user_id = u.id
left join (select user_id, count(*) cnt from comments group by user_id) c
on c.user_id = u.id
order by total_actions desc
limit 10
While I prefer GMB's method (using LEFT JOIN with each subquery) I'll show here how to combine your existing queries. Just use both correlated subqueries, and add them together to get the total.
select `users`.*,
(
select count(*)
from `votes`
where `users`.`id` = `votes`.`user_id`
) +
(
select count(*)
from `comments`
where `users`.`id` = `comments`.`user_id`
) as total_actions
from `users`
order by total_actions desc
limit 10

select 1 random row with complex filtering

I've 2 tables:
first table users:
+-------------------------+---------+------+-----+---------+-------+
| Field | Type | Null | Key | Default | Extra |
+-------------------------+---------+------+-----+---------+-------+
| id | int(11) | NO | PRI | NULL | |
| first_name | text | NO | | NULL | |
| age | int(11) | YES | | NULL | |
| settings | text | YES | | NULL | |
+-------------------------+---------+------+-----+---------+-------+
second table proposals:
+---------+---------+------+-----+---------+----------------+
| Field | Type | Null | Key | Default | Extra |
+---------+---------+------+-----+---------+----------------+
| id | int(11) | NO | PRI | NULL | auto_increment |
| from_id | int(11) | NO | | NULL | |
| to_id | int(11) | NO | | NULL | |
| status | int(11) | NO | | NULL | |
+---------+---------+------+-----+---------+----------------+
I need to get 1 random row from users which id is not in to_id in proposals
I'm doing it (without rand) with this sql:
SELECT DISTINCT *
FROM profiles
WHERE
profiles.first_name IS NOT NULL
AND
NOT EXISTS (
SELECT *
FROM proposal
WHERE
proposal.to_id = profiles.id
)
LIMIT 0 , 1
performance is fine: 1 row in set (0.00 sec)
but perfomance is very bad: 1 row in set (1.78 sec) when I add ORDER BY RAND() to the end
I've big holes in users.id and I can't use something like MAX(id)
I'he try set random limit, example:
...
LIMIT 1234 , 1;
Empty set (2.71 sec)
But it takes much time too :(
How to get random 1 user which users.id isn't exists in proposals.to_id with good perfomance?
I think that I need to first get all profiles with a rand() and then filter them, but I do not know how to do it.
I've two problem solutions.
1) With random id, from https://stackoverflow.com/a/4329447/2051938
SELECT *
FROM profiles AS r1
JOIN
(SELECT CEIL(RAND() *
(SELECT MAX(id)
FROM profiles)) AS id)
AS r2
WHERE
r1.id >= r2.id
AND
r1.first_name IS NOT NULL
AND
NOT EXISTS (
SELECT *
FROM proposal
WHERE
proposal.to_id = r1.id
)
LIMIT 0 , 1
2) With ORDER BY RAND()
SELECT *
FROM
(
SELECT *
FROM profiles
WHERE
profiles.first_name IS NOT NULL
ORDER BY RAND()
) AS users
WHERE
NOT EXISTS (
SELECT *
FROM proposal
WHERE
proposal.to_id = users.id
)
LIMIT 0 , 1
First solution is faster but it've problem with "holes in id" and when you got id from the end (users may end earlier than there will be a match)
Second solution is slower but without flaws!
Have you tried switching not exists to left join?
SELECT DISTINCT *
FROM profiles t1
LEFT JOIN
proposal t2
ON t1.id = t2.to_id
WHERE t1.first_name IS NOT NULL AND
t2.to_id IS NULL
ORDER BY RAND()
LIMIT 0 , 1
This will return you all rows of profiles, and to those that are not matched by a row in proposal it will assign NULL values, on which you can filter.
The result should be the same, but the performance may be better.
As RAND() function assigns a random number to every row present in result, performance will be directly proportional to number of records.
If you want to select only one (random) record, you can apply LIMIT <random number from 0 to record count>, 1
e.g.:
SELECT u.id, count(u.id) as `count`
FROM users u
WHERE
first_name IS NOT NULL
AND
NOT EXISTS (
SELECT *
FROM proposal
WHERE
proposal.to_id = u.id
)
LIMIT RAND(0, count-1) , 1
I haven't tried executing this query, however, it MySQL complains about using count in RAND, you can calculate count separately and substitute the value in this query.
First, I don't think the select distinct is necessary. So, try removing that:
SELECT p.*
FROM profiles p
WHERE p.first_name IS NOT NULL AND
NOT EXISTS (SELECT 1
FROM proposal pr
WHERE pr.to_id = p.id
)
ORDER BY rand()
LIMIT 0 , 1;
That might help a bit. Then, a relatively easy way to reduce the time spent is to reduce the data volume. If you know you will always have thousands of rows that meet the conditions, then you can do:
SELECT p.*
FROM profiles
WHERE p.first_name IS NOT NULL AND
NOT EXISTS (SELECT 1
FROM proposal pr
WHERE pr.to_id = p.id
) AND
rand() < 0.01
ORDER BY rand()
LIMIT 0, 1;
The trick is to find the comparison value that ensures that you get at least one row. This is tricky because you have another set of data. Here is one method that uses a subquery:
SELECT p.*
FROM (SELECT p.*, (#rn := #rn + 1) as rn
FROM profiles p CROSS JOIN
(SELECT #rn := 0) params
WHERE p.first_name IS NOT NULL AND
NOT EXISTS (SELECT 1
FROM proposal pr
WHERE pr.to_id = p.id
)
) p
WHERE rand() < 100 / #rn
ORDER BY rand()
LIMIT 0, 1;
This uses a variable to calculate the number of rows and then randomly selects 100 of them for processing. When choosing 100 rows randomly, there is a very, very, very high likelihood that at least one will be chosen.
The downside to this approach is that the subquery needs to be materialized, which adds to the cost of the query. It is, however, cheaper than a sort on the full data.

Joining Max Row where Col

I have this query:
SELECT
`shift`.`uid`,
`shift`.`activity`,
`users`.`fname`,
`users`.`lname`
FROM `shift`, `users`
WHERE `shift`.`uid` = `users`.`id`
It works fine just like that, but I need to add a new column from another table and order by it.
times :
| uid | User | time |
+++++++++++++++++++++
| 3 | bob | 1231 |
| 3 | bob | 1291 |
| 4 | ned | 1651 |
| 5 | ted | 5679 |
| 6 | joe | 7665 |
| 6 | joe | 7864 |
How can I include the maximum time from the time table for each user (WHERE times.uid = shift.uid) and then order by that column?
Trouble is, all the other tables have one row per user but the time table has multiple and I can't figure out the correct combination of joins and group by.
You could join on an aggregate query:
SELECT `shift`.`uid`,
`shift`.`activity`,
`users`.`fname`,
`users`.`lname` ,
t.max_time
FROM `shift`
JOIN `users` ON `shift`.`uid` = `users`.`id`
JOIN (SELECT `uid`, MAX(`time`) AS max_time
FROM `times`
GROUP BY `uid`) t ON shift.uid = t.uid
ORDER BY t.max_time
A pretty simple way to approach this is using a correlated subquery:
SELECT s.`uid`, s.`activity`, u.`fname`, u.`lname`,
(SELECT MAX(tt.time)
FROM timetable tt
WHERE tt.uid = u.id
) as maxtime
FROM `shift` s JOIN
`users` u
ON s.`uid` = u.`id`;
The advantage of this approach is performance. With an index on timetable(uid, time), this should work better than doing an aggregation at the outer level (because the query will take advantage of the index).
SELECT s.uid,
s.activity,
u.fname,
u.lname,
MAX(t.time) as maxtime
FROM shift s,
INNER JOIN users u ON u.id = s.uid
INNER JOIN times t ON t.uid = u.id
GROUP BY s.uid,
s.activity,
u.fname,
u.lname
ORDER BY maxtime

Strange order of results when adding joins

I'm trying to build a commenting system on my website but having issues with ordering the comments correctly. This is a screenshot of what I had before it went wrong:
And this is the query before it went wrong:
SELECT
com.comment_id,
com.parent_id,
com.is_reply,
com.user_id,
com.comment,
com.posted,
usr.username
FROM
blog_comments AS com
LEFT JOIN
users AS usr ON com.user_id = usr.user_id
WHERE
com.article_id = :article_id AND com.moderated = 1 AND com.status = 1
ORDER BY
com.parent_id DESC;
I now want to include each comment's votes from my blog_comment_votes table, using a LEFT OUTER JOIN, and came up with this query, which works, but screws with the order of results:
SELECT
com.comment_id,
com.parent_id,
com.is_reply,
com.user_id,
com.comment,
com.posted,
usr.username,
IFNULL(c.cnt,0) votes
FROM
blog_comments AS com
LEFT JOIN
users AS usr ON com.user_id = usr.user_id
LEFT OUTER JOIN (
SELECT comment_id, COUNT(vote_id) as cnt
FROM blog_comment_votes
GROUP BY comment_id) c
ON com.comment_id = c.comment_id
WHERE
com.article_id = :article_id AND com.moderated = 1 AND com.status = 1
ORDER BY
com.parent_id DESC;
I now get this order, which is bizarre:
I tried adding a GROUP BY clause on com.comment_id but that failed too. I can't understand how adding a simple join can alter the order of results! Can anybody help back on the correct path?
EXAMPLE TABLE DATA AND EXPECTED RESULTS
These are my relevant tables with example data:
[users]
user_id | username
--------|-----------------
1 | PaparazzoKid
[blog_comments]
comment_id | parent_id | is_reply | article_id | user_id | comment
-----------|-----------|----------|------------|---------|---------------------------
1 | 1 | | 1 | 1 | First comment
2 | 2 | 1 | 1 | 20 | Reply to first comment
3 | 3 | | 1 | 391 | Second comment
[blog_comment_votes]
vote_id | comment_id | article_id | user_id
--------|------------|------------|--------------
1 | 2 | 1 | 233
2 | 2 | 1 | 122
So the order should be
First comment
Reply to first comment +2
Second Comment
It's difficult to say without looking at your query results, but my guess is that it's because you are only ordering by parent id and not saying how to order when two records have the same parent id. Try changing your query to look like this:
SELECT
com.comment_id,
com.parent_id,
com.is_reply,
com.user_id,
com.comment,
com.posted,
usr.username,
COUNT(c.votes) votes
FROM
blog_comments AS com
LEFT JOIN
users AS usr ON com.user_id = usr.user_id
LEFT JOIN
blog_comment_votes c ON com.comment_id = c.comment_id
WHERE
com.article_id = :article_id AND com.moderated = 1 AND com.status = 1
GROUP BY
com.comment_id,
com.parent_id,
com.is_reply,
com.user_id,
com.comment,
com.posted,
usr.username
ORDER BY
com.parent_id DESC, com.comment_id;