I've recently been teaching myself SQL, and have been working on a toy project to do so. Here is a sample schema:
CREATE TABLE user (
id INT PRIMARY KEY NOT NULL AUTO_INCREMENT,
name VARCHAR(50)
);
INSERT INTO user(name) VALUES
("User 1"),
("User 2"),
("User 3"),
("User 4"),
("User 5");
CREATE TABLE friendship (
uid_1 INT,
uid_2 INT,
accepted_time TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
PRIMARY KEY (uid_1, uid_2),
CONSTRAINT fk_uid_1 FOREIGN KEY (uid_1) REFERENCES user (id),
CONSTRAINT fk_uid_2 FOREIGN KEY (uid_2) REFERENCES user (id)
);
INSERT INTO friendship(uid_1, uid_2) VALUES
(1, 2),
(2, 1);
CREATE TABLE event (
id INT PRIMARY KEY NOT NULL AUTO_INCREMENT,
name VARCHAR(50),
owner_id INT,
CONSTRAINT fk_owner_id FOREIGN KEY (owner_id) REFERENCES user (id)
);
INSERT INTO event (name, owner_id) VALUES
("Event 1", 1),
("Event 2", 2),
("Event 3", 3),
("Event 4", 4),
("Event 5", 5),
("Event 6", 1);
CREATE TABLE invite (
event_id INT NOT NULL,
sent_from_id INT NOT NULL,
sent_to_id INT NOT NULL,
PRIMARY KEY (event_id, sent_to_id),
CONSTRAINT fk_event_id FOREIGN KEY (event_id) REFERENCES event (id),
CONSTRAINT fk_sent_from_id FOREIGN KEY (sent_from_id) REFERENCES user (id),
CONSTRAINT fk_sent_to_id FOREIGN KEY (sent_to_id) REFERENCES user (id)
);
INSERT INTO invite(event_id, sent_from_id, sent_to_id) VALUES
(1, 2, 3);
As part of this project, I have a query that gets a list of users, with information populated relative to the currently authenticated user.
A simplified version of the query looks like this:
$select_users_query = "
SELECT
user.id AS id,
user.name AS name,
friendship.accepted_time AS friend_since
FROM user
LEFT JOIN friendship
ON friendship.uid_1 = user.id AND friendship.uid_2 = $relative_to_id
";
Then, at some endpoints, I want to return objects which have one or more users as sub-objects. In order to do this, I've been JOINing tables to the above query as a subquery, but when the returned object has multiple users (e.g., an invite to an event has a sending user, a receiving user, and a user that owns the event in question), the resulting query can end up pretty repetitive:
$select_invites_query = "
SELECT
event.id AS event_id,
event.name AS event_name,
owner.id AS owner_id,
owner.name AS owner_name,
owner.friend_since AS owner_friend_since,
sent_to.id AS sent_to_id,
sent_to.name AS sent_to_name,
sent_to.friend_since AS sent_to_friend_since,
sent_from.id AS sent_from_id,
sent_from.name AS sent_from_name,
sent_from.friend_since AS sent_from_friend_since,
FROM invite
INNER JOIN event
ON event.id = invite.event_id
INNER JOIN ($select_users_query) owner
ON event.owner_id = owner.id
INNER JOIN ($select_users_query) sent_from
ON invite.sent_from_id = sent_from.id
INNER JOIN ($select_users_query) sent_to
ON invite.sent_to_id = sent_to.id
";
My questions are:
Is repeating a subquery like this a performance issue during execution, assuming that the INNER JOINs all match on just a single row?
If not, is the additional parsing required for $select_invites_query a significant concern at all (especially as $select_users_query grows big and complex)?
Would using a variable here be a good idea, or a bad idea? From my inspection of EXPLAIN it seems as though MySQL is able to handle these JOINs pretty efficiently, but would defining a variable force MySQL to pull the unfiltered result set into memory before JOINing?
See SQL Fiddle schema here.
Since you appear to need self joins of same query, consider a Common Table Expression (CTE) (available in MySQL 8.0+) and with PHP parameterization. Below demonstrates with PHP's mysqli API in object-oriented and procedural styles:
$select_invites_query = "
WITH sub AS (
SELECT u.id AS id, u.`name` AS `name`,
f.accepted_time AS friend_since
FROM user
LEFT JOIN friendship f
ON f.uid_1 = u.id AND f.uid_2 = ?
)
SELECT event.id AS event_id, event.`name` AS event_name,
owner.id AS owner_id, owner.`name` AS owner_name,
owner.friend_since AS owner_friend_since,
sent_to.id AS sent_to_id, sent_to.`name` AS sent_to_name,
sent_to.friend_since AS sent_to_friend_since,
sent_from.id AS sent_from_id, sent_from.`name` AS sent_from_name,
sent_from.friend_since AS sent_from_friend_since
FROM invite
INNER JOIN event
ON event.id = invite.event_id
INNER JOIN sub owner
ON event.owner_id = owner.id
INNER JOIN sub sent_from
ON invite.sent_from_id = sent_from.id
INNER JOIN sub sent_to
ON invite.sent_to_id = sent_to.id";
// OBJECT-ORIENTED STYLE
$conn = new mysqli("my_host", "my_user", "my_pwd", "my_db");
$stmt = $conn->prepare($select_invites_query))
$stmt->bind_param("i", $relative_to_id);
$stmt->execute();
...
// PROCEDURAL STYLE
$conn = mysqli_connect("my_host", "my_user", "my_pwd", "my_db");
$stmt = mysqli_prepare($conn, $select_invites_query);
mysqli_stmt_bind_param($stmt, "i", $relative_to_id);
mysqli_stmt_execute($stmt);
...
Related
I want to insert a new dataset into a MySQL table tab with external data, but also with data from another table otherTab using the others' table primary key and another condition. However, it could be that the requested row simply does not exist (anymore) or the result set is empty due to a mismatch in the other supplied data. Then, the original INSERT should fail. All columns are forbidden to be NULL.
My first attempt was:
INSERT INTO tab (id, extid1, extid2, value)
SELECT 1,
(SELECT id FROM otherTab WHERE id = 12 AND data = 'TXT'),
(SELECT id FROM otherTab WHERE id = 34 AND data = 'JPG'),
1234
but the problem with it is that a returned empty result set is cast to the type of the column in tab, leading to a 0 as entry data.
The query shall be efficient and avoid unnecessary querying. This is how I achieve it with four subqueries:
INSERT INTO tab (id, extid1, extid2, value)
SELECT 1,
(SELECT id FROM otherTab WHERE id = 12 AND data = 'TXT'),
(SELECT id FROM otherTab WHERE id = 34 AND data = 'JPG'),
1234
WHERE EXISTS (SELECT id FROM otherTab WHERE id = 12 AND data = 'TXT')
AND EXISTS (SELECT id FROM otherTab WHERE id = 34 AND data = 'JPG')
I tried with other constructs, e.g. (SELECT IFNULL(SELECT id FROM otherTab WHERE id = 12 AND data = 'TXT', NULL)) to enforce NULL or even a string into the target column, but it also gets casted to a 0 or some value instead.
Here is the code for dbFiddle:
code
CREATE TABLE `tab` (
`id` int NOT NULL,
`seUuid4` binary(16) NOT NULL,
`rxUuid4` binary(16) NOT NULL,
`text` varchar(16)
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_0900_ai_ci;
CREATE TABLE `otherTab` (
`uuid4` binary(16) NOT NULL,
`lgUuid4` binary(16) NOT NULL,
`data` varchar(16)
) ENGINE = InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_0900_ai_ci;
ALTER TABLE `otherTab`
ADD PRIMARY KEY(`uuid4`);
ALTER TABLE `tab`
ADD CONSTRAINT `tab_ibfk_1` FOREIGN KEY (`rxUuid4`) REFERENCES `otherTab` (`uuid4`) ON DELETE RESTRICT ON UPDATE RESTRICT,
ADD CONSTRAINT `tab_ibfk_2` FOREIGN KEY (`seUuid4`) REFERENCES `otherTab` (`uuid4`) ON DELETE RESTRICT ON UPDATE RESTRICT;
INSERT INTO `otherTab` (uuid4, lgUuid4, data) VALUES
(UNHEX("22224444aaaa49c782408b2fe8c4dee0"), UNHEX("00001234aaaa4444aaaa432187654321"), "JPG"),
(UNHEX("11113333aaaa49c782408b2fe8c4dee0"), UNHEX("12340000bbbb6666bbbb432187654321"), "TXT");
INSERT INTO tab (id, seUuid4, rxUuid4, text)
SELECT
1,
(SELECT uuid4 FROM otherTab WHERE lgUuid4 = UNHEX('00001234aaaa4444aaaa432187654321') AND data = 'JPK' LIMIT 0,1),
(SELECT uuid4 FROM otherTab WHERE lgUuid4 = UNHEX('12340000bbbb6666bbbb432187654321') AND data = 'TXT' LIMIT 0,1),
'some text'
This interestingly works exactly as expected: Note the JPK instead of JPG. I verified my code and the PDO prepared statement fires out exactly the same command, but it gets inserted as INSERT INTO tab (id, seUuid4, rxUuid4) VALUES (1, 0x00000000000000000000000000000000, 0x00000000000000000000000000000000, 'datatext'); while the SQL client and phpMyadmin deliver the expected cannot insert null error message.
I could not find anything in the PDO options. If it helps, I use PDO with emulated prepared statements, but also tried without with no change.
PS: I posted already at dba.stackexchange.com/posts/276868
You could use a subquery:
INSERT INTO tab (id, extid1, extid2, value)
SELECT
FROM (
SELECT
1 id,
(SELECT id FROM otherTab WHERE id = 12 AND data = 'TXT') extid1,
(SELECT id FROM otherTab WHERE id = 34 AND data = 'JPG') extid2
1234 value
) t
WHERE extid1 IS NOT NULL and extid2 IS NOT NULL
Or, probably better yet, you can CROSS JOIN the two subqueries:
INSERT INTO tab (id, extid1, extid2, value)
SELECT 1, t1.id, t2.id, 1234
FROM (SELECT id FROM otherTab WHERE id = 12 AND data = 'TXT') t1,
CROSS JOIN (SELECT id FROM otherTab WHERE id = 34 AND data = 'JPG') t2
Actually, since you are reurning the same value that you are filtering on, two exists subqueries are probably sufficient:
INSERT INTO tab (id, extid1, extid2, value)
SELECT t.*
FROM (SELECT 1 id, 12 extid1, 34 extid2, 1234 value) t
WHERE EXISTS (SELECT 1 FROM otherTab t1 WHERE t1.id = t.extid1 AND t1.data = 'TXT')
AND EXISTS (SELECT 1 FROM otherTab t1 WHERE t1.id = t.extid2 AND t1.data = 'JPG')
To preface, I'm no DBA or SQL expert. But I've taken on a personal project that requires me to wear all hats in making a social network. (No, I'm not trying to reinvent Facebook. I'm targeting a niche audience.) And yes, I've heard of frameworks such as http://activitystrea.ms/, but I feel like data serialization should be a last resort for my needs.
Anyway, How to implement the activity stream in a social network helped me get the ball rolling, but I have some unanswered questions.
Below is my database schema (some rows have been omitted for simplification):
Action table:
id name
-------------
1 post
2 like
3 follow
4 favorite
5 tag
6 share
Activity table:
id (int)
user_id (int)
action_id (tinyint)
target_id (int)
object_id (tinyint)
date_created (datetime)
The object_id refers to which object type the target_id is. The idea here is to represent (User + Action + Target Object)
User Post(s) Media
User Favorite(s) Scene
User Follow(s) User
Object (type) table:
id name
-------------
1 media
2 scene
3 brand
4 event
5 user
The problem here is that each object has its own separate table. For example:
Media table:
id (int)
type (int)
thumbnail (varchar)
source (varchar)
description (varchar)
Event table:
id (int)
user_id (int)
name (varchar)
city (int)
address (varchar)
starts (time)
ends (time)
about (varchar)
User table:
id (int)
username (varchar)
profile_picture (varchar)
location (int)
What, then, would be the best (i.e., most efficient) way of querying this database?
Obviously I could perform a SELECT statement on the activity table, and then – based on the object_id – use conditional logic in PHP to make a separate query to the appropriate object's table (e.g., media).
Or would it be smarter (and more efficient) to implement some sort of left or inner JOIN on all 5 object tables, as suggested here: MySQL if statement conditional join. I'm not entirely familiar with how JOINS work, and whether SQL is smart enough to only scan the appropriate object table for each activity row, rather than ALL the joined tables.
Of course the first solution means MANY more calls to the database, which is less desirable. However, I'm not sure how else I could retrieve all the relevant columns (e.g., media "source", event "address") in just one query without implementing some conditional logic.
Suppose, you change your activity table a little bit:
Activity table:
id (int)
user_id (int)
action_id (tinyint)
object_id (tinyint)
date_created (datetime)
and your join table for every target type:
activity_id (int)
target_id (int)
and finally your target table (media)
id (int)
type (int)
thumbnail (varchar)
source (varchar)
description (varchar)
and target table (event)
id (int)
user_id (int)
name (varchar)
city (int)
address (varchar)
starts (time)
ends (time)
about (varchar)
now, you can select the data with
SELECT
activity.id,
activity.user_id,
activity.action_id,
action.name,
activity.object_id,
object.name,
media.id as media_id,
media.type,
media.thumbnail,
media.source,
media.description,
event.id as event_id,
event.name,
...
FROM
activity
LEFT JOIN action ON (action.id = activity.action_id)
INNER JOIN mediaToActivity ON (mediaToActivity.activity_id = activity.id)
LEFT JOIN media ON (media.id = mediaToActivity.target_id)
INNER JOIN eventToActivity ON (eventToActivity.activity_id = activity.id)
LEFT JOIN event ON (event.id = eventToActivity.target_id)
with this query you should get all rows in one query (but only the ones which actually exists are filled with data)
Note, I haven't tested this by now...
I pieced together from your discussion what your solution was. Fiddle
create table activity (
id int,
user_id int,
action_id int,
target_id int,
object_id int,
date_created datetime
);
create table action (
id int,
name varchar(80)
);
create table object (
id int,
name varchar(80)
);
create table media (
id int,
type int,
thumbnail varchar(255),
source varchar(255),
description varchar(255)
);
create table event (
id int,
user_id int,
name varchar(255),
city int,
address varchar(255),
starts time,
ends time,
about varchar(255)
);
-- setup
insert into action values (1, "post");
insert into object values (1, "media");
insert into object values (2, "event");
-- new event
insert into event values (1, null, "breakfast", null, "123 main st", null, null, "we will eat");
insert into activity values (1, null, 1, 1, 2, null);
-- new media
insert into media values (1, null, null, null, "new media");
insert into activity values (2, null, 1, 1, 1, null);
SELECT *
FROM
activity
left join event on (event.id = activity.target_id and activity.object_id = 2)
left join media on (media.id = activity.target_id and activity.object_id = 1);
I'm having difficulty trying to find the best way to get my results from a table. I want to get the targeted row from a table by one using the primary key from another using a foreign key.
The tables are would be set similar to this(minus a lot of other attributes for space):
user Table:
user_Id(pk)
name
type
venue_Id(unique/indexed)
venue Table:
venue_Id(fk)
rating
Logic flow is: user_Id is provided by a session variable. Query DB table 'user' to find that user. Go to type of user to identify if user is person or venue. Assuming user is venue, go to DB table 'venue' and query table for rating using foreign key from unique/indexed venue_Id from user table.
The query looks like
SELECT rating FROM `venue` WHERE `user_Id` = '$user_Id' AND `type` = 'venue'
Is this possible, and if so, what is the correct way to go about it?
You have a few ways to retrieve this information.
Using JOIN:
SELECT v.rating
FROM venue v INNER JOIN user u
ON v.venue_id= u.venue_id
AND u.`user_Id` = '$user_Id' AND u.`type` = 'venue'
Using an IN sub-query
SELECT rating
FROM venue
WHERE venue_id IN (SELECT venue_id FROM user
WHERE `user_Id` = '$user_Id' AND `type` = 'venue')
BTW, you should consider protect your code from potential SQL Injections
Its a bit unclear you explained that way.
From what I get, there is 2 table User and Venue.
In User table u have: user_id, venue_id, name, type.
While in Venue table u have: venue_id, rating.
You are expecting to get rating (Venue Table) while you use the WHERE clause in user_id and type which both stored on User Table.
Your Query:
SELECT rating FROM venue WHERE user_Id = '$user_Id' AND type = 'venue'
It is impossible to get it done like above because you are selecting from venue table while user_id and type is not from venue table. So it will make it unidentified even you have chaining the FK. Because FK will only to show and make some constraint to parent child table.
The query should be something like this:
SELECT rating FROM venue v JOIN user u on v.venue_id = u.venue_id WHERE u.user_Id = '$user_Id' AND u.type = 'venue'
Correct me if I am wrong..
Combining rows from two tables based on the tables having columns with equal values is called an equi-join operation, it's the pattern we typically use to "follow" foreign key relationships.
As an example:
$sql = "SELECT v.rating
FROM `venue` v
JOIN `user` s
ON s.venue_Id = v.venue_Id
AND s.type` = 'venue'
WHERE s.user_Id` = '" . mysqli_real_escape_string($con, $user_Id) ."'"
This isn't the only pattern, there are several other query forms that will return an equivalent result.
As an example of using an EXISTS predicate:
$sql = "SELECT v.rating
FROM `venue` v
WHERE EXISTS
( SELECT 1
FROM `user` s
WHERE s.venue_Id = v.venue_Id
AND s.type` = 'venue'
AND s.user_Id` = '"
. mysqli_real_escape_string($con, $user_Id)
."'"
)";
The original query appears to be vulnerable to SQL Injection; the example queries demonstrate the use of the mysqli_real_escape_string function to "escape" unsafe values and make them safe to include in SQL text. (That function would only be appropriate if you are using the mysqli interface. Using prepared statements with bind placeholders is another approach.
I would like to query the name of the friends of person n°4, so the numbers in 'otherPerson_id', but my attempt only query the id or name in 'person_id'. How could I ask to query the informations about the friends of person n°4, and not the information about person n°4 itself?
Here is my attempt :
$q = "SELECT DISTINCT p.idperson, p.name FROM person p INNER JOIN people_friends pf ON p.idperson = pf.person_id AND p.idperson = 4";
$res = $connexion->query($q);
$res->setFetchMode(PDO::FETCH_COLUMN);
while($record = $res->fetch()) {
echo $record[name];
echo "<br/>";
}
the table : (so when I look for the friends of person n°4, I would like it to return 1 and 7 )
EDIT :
CREATE TABLE people_friends
(
person_id integer NOT NULL,
otherperson_id integer NOT NULL,
CONSTRAINT people_friends_pkey PRIMARY KEY (person_id, otherperson_id),
CONSTRAINT people_friends_person_id_fkey FOREIGN KEY (person_id)
REFERENCES person (idperson) MATCH SIMPLE
ON UPDATE NO ACTION ON DELETE NO ACTION,
CONSTRAINT people_friends_person_id_fkey1 FOREIGN KEY (person_id)
REFERENCES person (idperson) MATCH SIMPLE
ON UPDATE NO ACTION ON DELETE NO ACTION
)
CREATE TABLE person
(
idperson integer NOT NULL,
name character varying(50),
map bytea,
CONSTRAINT person_pkey PRIMARY KEY (idperson)
)
Thanks
Try this query.
SELECT DISTINCT
pFriend.idperson, pFriend.name
FROM people_friends pf
INNER JOIN person pFriend on pf.otherperson_id = pFriend.idperson
WHERE
pf.person_id = 4
You just have to add a WHERE clause where you check the pf table.
SELECT DISTINCT p.idperson
,p.name
FROM person p
INNER JOIN people_friends pf
ON p.idperson = pf.person_id
WHERE pf.otherperson_id = 4
I have 3 tables :
wp_users - stores main information about all users,
wp_usermeta - stores additional information about users(first/last name/etc),
wp_friends - stores information about friends from third party services related to a specific user from wp_users
If you are not familiar with WordPress, you can see the structure of both tables at http://codex.wordpress.org/images/9/9e/WP3.0-ERD.png
The structure of my custom table wp_friends is as follows:
CREATE TABLE wp_friends (
id bigint(20) unsigned NOT NULL auto_increment,
uid bigint(20) unsigned NOT NULL default '0',
fr_id VARCHAR (60) NOT NULL default '',
service VARCHAR (20) NOT NULL default '',
name VARCHAR (80) NOT NULL default '',
photo VARCHAR (255) NOT NULL default '',
PRIMARY KEY (id),
KEY uid (uid),
KEY fr_id (fr_id),
KEY service (service)
)`
The uid column is corresponds to the ID column in the wp_users table - this is how I determine which record corresponds to which user.
What I'm trying to do is to create a query that will look in all of the three tables for a match against a keyword. Here is what I've come with so far(the first part was generated by a search function of WordPress):
SELECT
wp_users.ID,wp_users.display_name,wp_users.user_login,
wp_users.user_email,fr.fr_id,fr.name,fr.photo,fr.service
FROM wp_users
INNER JOIN wp_usermeta ON (wp_users.ID = wp_usermeta.user_id)
LEFT JOIN wp_socialaccess_friends AS fr ON fr.uid = 2
WHERE
(
(user_login LIKE '%nik%' OR user_nicename LIKE '%nik%')
AND
(wp_usermeta.meta_key = 'wp_user_level' AND CAST(wp_usermeta.meta_value AS CHAR) != '0')
)
OR ( fr.uid = 2 AND (fr.fr_id LIKE '%nik%' OR fr.name LIKE '%nik%'))
GROUP BY wp_users.ID,fr.fr_id ORDER BY user_login ASC
In the above query, the keyword is "nik"(which also matches a user_login column). The fr.uid part is needed so the returned results are only for the current user. The query fails in the following ways:
It returns all rows from the wp_friends table(because the user_login column is matched as well), that have wp_friends.uid = 2
It returns rows that have wp_friends.uid = 2 but matched with users where wp_users.ID != 2
Is it possible to create a single query, that would return the selected columns, but will also prevent duplicates?
What about joining on a sub-select like:
SELECT
wp_users.ID,wp_users.display_name,wp_users.user_login,
wp_users.user_email
FROM wp_users
INNER JOIN wp_usermeta ON (wp_users.ID = wp_usermeta.user_id)
left join(
select fr_id, uid,name,photo,service from wp_socialaccess_friends where
uid = 2 and
(fr_id LIKE '%nik%' OR name LIKE '%nik%')
) AS fr ON wp_users.ID = fr.uid
WHERE
(
(user_login LIKE '%nik%' OR user_nicename LIKE '%nik%')
AND
(wp_usermeta.meta_key = 'wp_user_level' AND CAST(wp_usermeta.meta_value AS CHAR) != '0')
)
GROUP BY wp_users.ID,fr.fr_id ORDER BY user_login ASC