I am using datatables with a modified ssp.class.php to allow for joins and other custom features. In the example below I only want to return results from computers for the first x rows sorted by their id. Because of this, I list computers with my conditions first then LEFT JOIN users followed by logs (the information I am after).
It works great, BUT because of the left joins I have empty results. For instance, while my result set only contains logs from the correct computers... if logs has no rows for a particular user/computer combination I have a row with empty log data, but with user/computer data which serves me no purpose.
Is my only option to include a WHERE condition to prevent null values in the logs join... WHERE logs.user_id != '' or is there some other logic I can do in the select that I am missing?
SELECT (*see note)
FROM
( SELECT account_id, computer_id, computer_name
FROM computers
ORDER BY computer_id ASC LIMIT 0, ".$_SESSION['user']['licenses']."
) as c
LEFT JOIN users
on users.computer_id = c.computer_id
LEFT JOIN logs
on logs.user_id = users.user_id
You can use just JOIN for the table logs.
You put a LIMIT in the Derived Table accessing the computers table returning nrows. When an Inner Join or a final WHERE-condition filters some rows you will get less than nrows.
If this is not ok for you and you always want nrows, the only way is to move the LIMIT after doing Inner Joins:
SELECT (*see note)
FROM computers as c
JOIN users
on users.computer_id = c.computer_id
JOIN logs
on logs.user_id = users.user_id
ORDER BY computer_id ASC LIMIT 0, ".$_SESSION['user']['licenses']."
But this will probably be [much] slower...
Related
i'm having some problems. with the JEFT JOIN sql statement.
i have 3 tables:
user , products, prod_images.
table.user user_id
table.products user_id item_id
table.prod_images user_id item_id
when i run this query to get data that relation, its work fine.
but only work if the prod_images table are not empty.
when the prod_images empty
the sql right join merge the result and i get null to the products.item_id array
SELECT products.*, prod_images.*, users.*
FROM products
LEFT JOIN prod_images
ON products.item_id=prod_images.item_id
AND prod_images.is_primary = '1'
JOIN users
ON users.user_id=products.seller_id
WHERE products.status = '1'
ORDER BY created DESC
How can i make this query work fine when the table.prod_images
is empty ?
I believe your query is almost ok.
And it returns everything you asked for, but you should avoid of requesting products.*, prod_images.*, users.* because it is easy to confuse ourselves.
So if you change this part at least to SELECT products.item_id PRODUCTS_ITEM_ID, products.*, prod_images.*, users.*
And one another note, your query probably returns no products.item_id not when prod_images has no related records (because it LEFT JOINed) but when users has no related records because it is INNER JOINed. So I would start from changing JOIN users to LEFT JOIN users.
PS Look at #Reno comment. I did not mention that. for sure you should change ON products.item_id=products.item_id to ON products.item_id=prod_images.item_id
my query looks like that:
SELECT
count(users.id)
FROM users
LEFT JOIN mail_sender_jobs_actions ON mail_sender_jobs_actions.userID = users.id
LEFT JOIN table2 ON table2.userID = users.id
LEFT JOIN table3 ON table3.userID = users.id
WHERE {$flexibleWhereClause}
Now, the mail_sender_jobs_actions table CAN (doesnt need to return anything) return multiple entries. I dont want to group the results but still limit the returns of mail_sender_jobs_actions to 1 so I dont get duplicates... Otherwise the count wouldnt work properly.
Scraped the whole web and found nothing working for me as I want to keep the where clause flexible. Any solution?
EDIT
so to explain the situation. We have a table with users (users). We have a table with actions (mail_seder_jobs_actions). We have other tables related to that query which are not relevant (table1, table2, table3)
If a user does an action, an entry is being created in the actions table.
The where clause is flexible, meaning it is possible that somebody wants to only show users with a specific action.
It is also possible that an action is not relevant to the user, so this entry gets ignored.
With where criteria you have there is no point using left join, since the where criteria applies to the table on the right hand side, effectively turning the left join into an inner join.
Apparently yo do not use any columns from the right hand side table, so instead of using joins, I would use an exists subquery.
SELECT
1 as count,
users.email
FROM users
WHERE EXISTS (SELECT 1
FROM mail_sender_jobs_actions
WHERE mail_sender_jobs_actions.userID = users.id
AND mail_sender_jobs_actions.type = '1'
AND mail_sender_jobs_actions.jobID = '106'
AND {$flexibleWhereClause})
However, there is little point in having the count() because it will always return 1. If you want to count how many records each user has in the mail_sender_jobs_actions table, then you have to use left join, group by, and move the where criteria into the join condition:
SELECT
count(mail_sender_jobs_actions.userID),
users.email
FROM users
LEFT JOIN mail_sender_jobs_actions ON mail_sender_jobs_actions.userID = users.id
AND mail_sender_jobs_actions.type = '1'
AND mail_sender_jobs_actions.jobID = '106'
AND {$flexibleWhereClause}
GROUP BY users.email
I have two tables:
users: user_id, user_zip
settings: user_id, pref_ex_loc
I need to find the single most popular 'pref_ex_loc' from the settings table based on a particular user_zip, which will be specified as the variable $userzip.
Here is the query that I have now and obviously it doesn't work.
$popularexloc = "SELECT pref_ex_loc, user_id COUNT(pref_ex_loc) AS countloc
FROM settings FULL OUTER JOIN users ON settings.user_id = users.user_id
WHERE users.user_zip='$userzip'
GROUP BY settings.pref_ex_loc
ORDER BY countloc LIMIT 1";
$popexloc = mysql_query($popularexloc) or die('SQL Error :: '.mysql_error());
$exlocrow = mysql_fetch_array($popexloc);
$mostpopexloc=$exlocrow[0];
echo '<option value="'.$mostpopexloc.'">'.$mostpopexloc.'</option>';
What am I doing wrong here? I'm not getting any kind of error from this either.
Give this a try:
select s.pref_ex_loc from settings s
join users u on (u.user_id = s.user_id)
where user_zip = $userzip
group by s.pref_ex_loc
order by count(*) desc
limit 1
As you said, this will give you the "single most popular 'pref_ex_loc' from the settings table based on a particular user_zip"
Well, for one thing you are missing a comma before the COUNT():
SELECT pref_ex_loc, user_id COUNT(...
You should have a comma between each field in your select-list:
SELECT pref_ex_loc, user_id, COUNT(...
I would recommend using COUNT(*) instead of COUNT(pref_ex_loc). In this case, either should give the right answer, but in MySQL COUNT(*) usually performs slightly better.
You're using outer join, but then in the WHERE clause you're testing one of the columns of users so it's effectively not an outer join anymore. In this query, I believe you simply need an INNER JOIN, unless you need to handle the possibility that none of the users reference any of your pref_ex_loc values. Read A Visual Explanation of SQL Joins.
Also, MySQL does not support FULL OUTER JOIN.
Your user_id in the select-list, when it is neither in the GROUP BY clause nor in an aggregate function, is an ambiguous field, taking its value from one arbitrary row in the group. You should remove user_id from the select-list.
Sort by the countloc DESC to get the greatest value first.
So here's what I see as a better query:
SELECT pref_ex_loc, COUNT(*) AS countloc
FROM settings INNER JOIN users ON settings.user_id = users.user_id
WHERE users.user_zip='$userzip' GROUP BY settings.pref_ex_loc
ORDER BY countloc DESC LIMIT 1
this will allow values (duplicate most popular) with the highest pref_ex_loc to be shown in the list.
It doesn't use LIMIT, because LIMIT forces the maximum number of rows to be shown. Now, here's the question, What if there are two or more rows that ties up with the most popular pref_ex_loc?
SELECT b.pref_ex_loc
FROM users a
INNER JOIN settings b
ON a.user_ID = b.user_ID
WHERE a.user_zip = 1 -- change the value here
GROUP BY b.pref_ex_loc
HAVING COUNT(*) =
(
SELECT MAX(totalCount)
FROM
(
SELECT b.pref_ex_loc, COUNT(*) totalCount
FROM users a
INNER JOIN settings b
ON a.user_ID = b.user_ID
WHERE a.user_zip = 1 -- change the value here
GROUP BY b.pref_ex_loc
) s
)
SQLFiddle Demo
SQLFiddle Demo (with duplicate most popular)
Try with this query:
SELECT user_id, COUNT(pref_ex_loc) AS countloc
FROM users LEFT JOIN settings ON users.user_id = settings.user_id
WHERE users.user_zip='$userzip' GROUP BY user_id ORDER BY countloc LIMIT 1
I am building an online survey system for which I wish to produce statistics. I want query based on the gender of the user. I have the following tables:
survey_question_options
survey_answer
users
I have constructed the following query so that it brings back a null response where there are no answers to the question:
SELECT COUNT(sa.option_id) AS answer , so.option_label
FROM survey_answer sa
RIGHT JOIN survey_question_options so
ON sa.option_id = so.option_id AND
sa.record_date>='2011-09-01' AND
sa.record_date<='2012-08-01'
LEFT JOIN users u
ON (sa.uid = u.uid AND u.gender='F')
WHERE so.question_id=24
GROUP BY so.option_label
ORDER BY so.option_id ASC
My query returns the following results set:
0 Red
1 Yellow
0 Blue
0 Green
However, the gender condition in the LEFT JOIN appears to be ignored in the query. When I change the gender to 'M' the same result is returned. However, the expected result would be 0 for everything.
I am not sure where I am going wrong. Please help.
Thanks in advance.
Well, you are doing a COUNT on a column from the main table, so the gender condition on the LEFT JOIN won't affect the result. You should do the COUNT on a column from the users table. I'm not sure if this is what you want, but you should try:
SELECT COUNT(u.uid) AS answer , so.option_label
FROM survey_answer sa
RIGHT JOIN survey_question_options so
ON sa.option_id = so.option_id AND
sa.record_date>='2011-09-01' AND
sa.record_date<='2012-08-01'
LEFT JOIN users u
ON (sa.uid = u.uid AND u.gender='M')
WHERE so.question_id=24
GROUP BY so.option_label
ORDER BY so.option_id ASC
The left join to the users table is evaluated after the join to the answer table - so although the user record is not returned if the user is the wrong gender, the answer record will be returned (regardless of the user's gender). Try:
SELECT COUNT(sa.option_id) AS answer , so.option_label
FROM (select a.option_id
from survey_answer a
JOIN users u ON a.uid = u.uid AND u.gender='F'
where a.record_date>='2011-09-01' AND
a.record_date<='2012-08-01') sa
RIGHT JOIN survey_question_options so
ON sa.option_id = so.option_id
WHERE so.question_id=24
GROUP BY so.option_label
ORDER BY so.option_id ASC
You're putting your condition in the wrong block. Since you're performing a LEFT JOIN, (which is a left-bound outer join) everything in the left table (the main table) is selected, together with the data from the joined table, where applicable. What you want is to add the data from all users and then restrict the full output of the query. What you've actually done is add the user data from only the female users and then displayed all data.
Sounds technical, but all you have to do is move the AND u.gender='F' into the main WHERE clause instead the ON clause. That will cause SQL to only select the rows for female users after the JOIN has taken place.
this is my query
SELECT U.id AS user_id,C.name AS country,
CASE
WHEN U.facebook_id > 0 THEN CONCAT(F.first_name,' ',F.last_name)
WHEN U.twitter_id > 0 THEN T.name
WHEN U.regular_id > 0 THEN CONCAT(R.first,' ',R.last)
END AS name,
FROM user U LEFT OUTER JOIN regular R
ON U.regular_id = R.id
LEFT OUTER JOIN twitter T
ON U.twitter_id = T.id
LEFT OUTER JOIN facebook F
ON U.facebook_id = F.id
LEFT OUTER JOIN country C
ON U.country_id = C.id
WHERE (CONCAT(F.first_name,' ',F.last_name) LIKE '%' OR T.name LIKE '%' OR CONCAT(R.first,' ',R.last) LIKE '%') AND U.active = 1
LIMIT 100
its realy fast, but in the EXPLAIN it don't show me it uses INDEXES (there is indexes).
but when i add ORDER BY 'name' before the LIMIT its takes long time why? there is a way to solve it?
tables: users 150000, regular 50000, facebook 50000, twitter 50000, country 250 and growing!
It takes a long time because it's a composite column, not a table column. The name column is a result of a case selection, and unlike simple selects with multiple join, MySQL has to use a different sorting algorithm for this kind of data.
I'm talking from ignorance here, but you could store the data in a temporary table and then sort it. It may go faster since you can create indexes for it but it won't be as fast, because of the different storage type.
UPDATE 2011-01-26
CREATE TEMPORARY TABLE `short_select`
SELECT U.id AS user_id,C.name AS country,
CASE
WHEN U.facebook_id > 0 THEN CONCAT(F.first_name,' ',F.last_name)
WHEN U.twitter_id > 0 THEN T.name
WHEN U.regular_id > 0 THEN CONCAT(R.first,' ',R.last)
END AS name,
FROM user U LEFT OUTER JOIN regular R
ON U.regular_id = R.id
LEFT OUTER JOIN twitter T
ON U.twitter_id = T.id
LEFT OUTER JOIN facebook F
ON U.facebook_id = F.id
LEFT OUTER JOIN country C
ON U.country_id = C.id
WHERE (CONCAT(F.first_name,' ',F.last_name) LIKE '%' OR T.name LIKE '%' OR CONCAT(R.first,' ',R.last) LIKE '%') AND U.active = 1
LIMIT 100;
ALTER TABLE `short_select` ADD INDEX(`name`); --add successive columns if you are going to order by them as well.
SELECT * FROM `short_select`
ORDER BY 'name'; -- same as above
Remember temporary tables are dropped upon connection termination, so you don't have to clean them, but you should anyway.
Without actually knowing your DB structure, and assuming you have all of the proper indexes on everything. An Order By statement takes some variable amount of time to sort the elements being returned by a query (index or not). If it is only 10 rows, it will seem almost instant, if you get 2000 rows, it will be a little slower, if you are sorting 15k rows joined across multiple tables, it is going to take some time to sort the returned result. Also make sure your adding indexes to the fields your sorting by. You may want to take the desired result and store everything in a presorted stub table for faster querying later as well (if you query this sorted result set often)
You need to create first 100 records from each name table separately, then union the results, join them with user and country, order and limit the output:
SELECT u.id AS user_id, c.name AS country, n.name
FROM (
SELECT facebook_id AS id, CONCAT(F.first_name, ' ', F.last_name) AS name
FROM facebook
ORDER BY
first_name, last_name
LIMIT 100
UNION ALL
SELECT twitter_id, name
FROM twitter
WHERE twitter_id NOT IN
(
SELECT facebook_id
FROM facebook
)
ORDER BY
name
LIMIT 100
UNION ALL
SELECT regular_id, CONCAT(R.first, ' ', R.last)
FROM regular
WHERE regular_id NOT IN
(
SELECT facebook_id
FROM facebook
)
AND
regular_id NOT IN
(
SELECT twitter_id
FROM twitter
)
ORDER BY
first, last
LIMIT 100
) n
JOIN user u
ON u.id = n.id
JOIN country с
ON c.id = u.country_id
Create the following indexes:
facebook (first_name, last_name)
twitter (name)
regular (first, last)
Note that this query orders slightly differently from your original one: in this query, 'Ronnie James Dio' would be sorted after 'Ronnie Scott'.
The use of functions on the columns prevent indexes from being used.
CONCAT(F.first_name,' ',F.last_name)
The result of the function is not indexed, even though the individual columns may be. Either you have to rewrite the conditions to query the name columns individually, or you have to store and index the result of that function (such as a "full name" column).
The index on [user.active] is unlikely to help you if most of the users are active.
I don't know what your application is all about, but I wonder if it hadn't been easier if you ditched the foreign keys in User table and instead put the UserID as a foreign key in the other tables instead.