php - 每次向依赖 FOREIGN KEY 的 MySQL 表插入数据时，是否需要每次执行额外的 SELECT 和 INSERT IGNORE 查询？

Question

我有 4 个 MySQL 表，它们通过FOREIGN KEYs 相互依赖。

请检查以下表结构架构:

CREATE DATABASE IF NOT EXISTS courses
    CHARACTER SET latin1
    COLLATE latin1_bin;

CREATE TABLE IF NOT EXISTS courses.institution
(
    icode INT UNSIGNED NOT NULL AUTO_INCREMENT,
    iname VARCHAR(255) NOT NULL,
    PRIMARY KEY (icode),
    UNIQUE (iname)
)
    ENGINE = InnoDB;

CREATE TABLE IF NOT EXISTS courses.cities
(
    ccode INT UNSIGNED NOT NULL AUTO_INCREMENT,
    cname VARCHAR(255) NOT NULL,
    PRIMARY KEY (ccode),
    UNIQUE (cname)
)
    ENGINE = InnoDB;

CREATE TABLE IF NOT EXISTS courses.skills
(
    scode INT UNSIGNED NOT NULL AUTO_INCREMENT,
    sname VARCHAR(255) NOT NULL,
    PRIMARY KEY (scode),
    UNIQUE (sname)
)
    ENGINE = InnoDB;

CREATE TABLE IF NOT EXISTS courses.relation
(
    icode INT UNSIGNED NOT NULL,
    scode INT UNSIGNED NOT NULL,
    ccode INT UNSIGNED NOT NULL,
    UNIQUE KEY ucols (icode, scode, ccode),
    FOREIGN KEY (icode) REFERENCES courses.institution (icode),
    FOREIGN KEY (scode) REFERENCES courses.skills (scode),
    FOREIGN KEY (ccode) REFERENCES courses.cities (ccode)
)
    ENGINE = InnoDB;

目前我正在执行下面的查询以在relation 表中插入一条记录。

每次只插入一次，需要 4 个 INSERT 查询和 3 个 SELECT 子查询。

INSERT IGNORE INTO institution(iname) VALUES ('ABC Learners');
INSERT IGNORE INTO skills(sname) VALUES ('PHP');
INSERT IGNORE INTO cities(cname) VALUES ('Bangalore');

INSERT IGNORE INTO relation (icode, scode, ccode) VALUES (
    (SELECT icode FROM institution WHERE iname = 'ABC Learners'),
    (SELECT scode FROM skills WHERE sname = 'PHP'),
    (SELECT ccode FROM cities WHERE cname = 'Bangalore')
);

Is it necessary to execute all this queries every time? or is there any better way to perform this action in single or few queries?

检查下面的简单 PHP 代码。在这段用于在 relation 表上插入 7 条记录的代码中，它执行 4 INSERT 查询，每条记录上有 3 SELECT 子查询。

7 条记录总共需要 7 * 7 = 49 次查询。如何解决？

<?php

$db = new mysqli('localhost', 'user', '****', 'courses');

$records = [
    ['ABC Learners', 'CSS', 'Bangalore'],
    ['ABC Learners', 'PHP', 'Bangalore'],
    ['ABC Learners', 'HTML', 'Bangalore'],
    ['ABC Learners', 'PHP', 'Hyderabad'],
    ['XYZ Solutions', 'PHP', 'Hyderabad'],
    ['XYZ Solutions', 'JAVA', 'Hyderabad'],
    ['XYZ Solutions', 'JAVA', 'Bangalore'],
];

foreach ($records as $record) {

    list($institute, $skill, $city) = $record;

    $db->query("INSERT IGNORE INTO institution (iname) VALUES ('{$institute}')");
    $db->query("INSERT IGNORE INTO skills (sname) VALUES ('{$skill}')");
    $db->query("INSERT IGNORE INTO cities (cname) VALUES ('{$city}')");

    $db->query(
        "INSERT IGNORE INTO relation (icode, scode, ccode) VALUES (" .
        "(SELECT icode FROM institution WHERE iname = '{$institute}'), " .
        "(SELECT scode FROM skills WHERE sname = '{$skill}'), " .
        "(SELECT ccode FROM cities WHERE cname = '{$city}'))"
    );
}

$db->close();

注意:以上脚本是示例目的。使用批处理模式 或禁用自动提交 对我没有用。因为很多时候我需要向关系表添加单个新记录(基于用户通过网络面板的请求)

---

更新 1:

经过一些研究和基准测试后，我创建了一个 MySQL 存储函数来加速这个过程，并将性能提高了 250 - 300%

请在这里查看功能:

DELIMITER $$
CREATE FUNCTION courses.record(i_name VARCHAR(255), s_name VARCHAR(255), c_name VARCHAR(255)) RETURNS INT
BEGIN
    DECLARE _icode, _scode, _ccode INT UNSIGNED;

    SELECT icode INTO _icode FROM institution WHERE iname = i_name;
    SELECT scode INTO _scode FROM skills WHERE sname = s_name;
    SELECT ccode INTO _ccode FROM cities WHERE cname = c_name;

    IF _icode IS NULL THEN
        INSERT IGNORE INTO institution (iname) VALUES (i_name);
        SELECT icode INTO _icode FROM institution WHERE iname = i_name;
    END IF;

    IF _scode IS NULL THEN
        INSERT IGNORE INTO skills (sname) VALUES (s_name);
        SELECT scode INTO _scode FROM skills WHERE sname = s_name;
    END IF;

    IF _ccode IS NULL THEN
        INSERT IGNORE INTO cities (cname) VALUES (c_name);
        SELECT ccode INTO _ccode FROM cities WHERE cname = c_name;
    END IF;

    INSERT IGNORE INTO relation (icode, scode, ccode) VALUES (_icode, _scode, _ccode);

    RETURN ROW_COUNT();
END $$
DELIMITER ;

现在，下面这个 PHP 脚本可以通过一次查询在关系表中插入一条记录

<?php

$db = new mysqli('localhost', 'user', '***', 'courses');

$records = [
    ['ABC Learners', 'CSS', 'Bangalore'],
    ['ABC Learners', 'PHP', 'Bangalore'],
    ['ABC Learners', 'HTML', 'Bangalore'],
    ['ABC Learners', 'PHP', 'Hyderabad'],
    ['XYZ Solutions', 'PHP', 'Hyderabad'],
    ['XYZ Solutions', 'JAVA', 'Hyderabad'],
    ['XYZ Solutions', 'JAVA', 'Bangalore'],
];

$query = $db->prepare("SELECT record (?, ?, ?)");
$query->bind_param('sss', $institute, $skill, $city);

foreach ($records as $record) {
    list($institute, $skill, $city) = $record;
    $query->execute();
}

$db->close();

This MySQL stored function increased the performance. But, still I'm using multiple INSERT and SELECT statements in this function. Is it possible to optimize this function with few statements to gain more performance?

Answer 1

<强>1。将 UNIQUE 更改为 PRIMARY key:

在没有明确定义主键的情况下，InnoDB 将为每一行创建自己的隐藏主键。但在您的情况下，在表 relation 中，当前 UNIQUE 键更适合定义为 PRIMARY 键( NOT NULL 并满足唯一性)。因此，请将其更改为 PRIMARY KEY。

<强>2。利用批量插入/选择:

通过使用批量插入和选择，并利用一些应用程序捕获，我们基本上可以将其整体分解为 4 个插入和 3 个选择。检查下面的代码(带注释):

$records = [
    ['ABC Learners', 'CSS', 'Bangalore'],
    ['ABC Learners', 'PHP', 'Bangalore'],
    ['ABC Learners', 'HTML', 'Bangalore'],
    ['ABC Learners', 'PHP', 'Hyderabad'],
    ['XYZ Solutions', 'PHP', 'Hyderabad'],
    ['XYZ Solutions', 'JAVA', 'Hyderabad'],
    ['XYZ Solutions', 'JAVA', 'Bangalore'],
];

// Create a copy of records to avoid changing the original
$records_copy = $records;

// Get unique institutions, skills and cities
$i = array_unique(array_map('array_shift', $records_copy)); 
$s = array_unique(array_map('array_shift', $records_copy)); 
$c = array_unique(array_map('array_shift', $records_copy)); 

// Prepare batch insert and select SQL for institution table
$i_ins_sql = "INSERT IGNORE INTO institution (iname) VALUES ";
$i_sel_sql = "SELECT icode, iname FROM institution WHERE iname IN (";

foreach ($i as $v) {

    $i_ins_sql .= "('" . $db->real_escape_string($v) . "'),";
    $i_sel_sql .= "'" . $db->real_escape_string($v) . "',";
}

// Execute the INSERT query
$db->query( rtrim($i_ins_sql, ',') );

// Execute the SELECT query and fetch the query result and store (cache) it
// Key of the cache array would be name (string) and the value would be 
// the code (integer)
$i_cache = array();
$i_sel_q = $db->query( rtrim($i_sel_sql, ',') . ")" );
while ( $row = $i_sel_q->fetch_assoc() ) {
    $i_cache[$row['iname']] = $row['icode'];
}

/**
 * REPEAT the INSERT and SELECT steps for the skills and cities tables, 
 * using $s and $c arrays respectively, with appropriate field names
 * Eventually, you would have cache arrays, $s_cache and $c_cache.
 **/

// Now let's do a batch INSERT INTO relation table 
$sql = "INSERT IGNORE INTO relation (icode, scode, ccode) VALUES ";

// Loop over original records array
foreach ($records as $record) {

    $sql .= "(" . (int)$i_cache[$record[0]] . ","
                . (int)$i_cache[$record[1]] . "," 
                . (int)$i_cache[$record[2]] . "),";
}

// Execute the Batch INSERT query into relation table
$db->query( rtrim($sql, ',') );

最后但同样重要的是:您的代码对 SQL injection 开放相关攻击。连real_escape_string不能完全保护它。请学会使用Prepared Statements相反。