Mysqli connection design - php

Question

I going to create a website that can get as much as 6,000 alerts daily(those alerts containing up to 140-characters, that's approx. 30 words-tokens-). My question is if i have a function that verify if each token is already in the db, it false, do nothing, if true it get insert, and so on, but for each word it has to loop through the entire db for checking, how should i handle the connection? it is bad to open connections every time i need to check for a word?
function insertTag($tag){
$db = "test";
$user = "Eduardo";
$pass = "weaponx";
$host = "localhost";
$con = new mysqli($host, $user, $pass, $db);
$noInsert = false;
$result = $con->query("select TAG from TAGS");
$num_tags = $result->num_rows;
for($c=0; $c < $num_tags; $c++){
$row = $result->fetch_assoc();
echo "tag ". ($c+1) .": ". $row['TAG'] ."<br/>";
if ($fila['TAG'] === $tag){
echo "$tag: already exist.<br />";
$noInsert = true;
return;
}
else{
$noInsert = false;
}
}
if (!$noInsert){
$result2 = $con->query("insert into TAGS(TAG) values('$tag')");
echo "token $tag: inserted<br />";
}
}
$tags = "danger in detroit";
// insert word in the BD, only if new
for($i=0; $i < count($tags); $i++){
insertTag($tags[$i]);
}
should i use the mysqli version of persistent connection? if so how?

Answer 1

No no, don't do it that way :) Several points:
1) Open mysql connection at the beginning of your script, and do as many queries as you need without opening new connection for every query. (So opening connection need to be outside of your function)
2) You need to use indexes and mark your tag value in database as unique, in that case you can just INSERT every time, and if it is duplicate it will not get inserted. In order to create index for your table you need to use your favorite sql manager, or just do a query, you can read here more:
http://dev.mysql.com/doc/refman/5.0/en/create-index.html
But in short you need to make column that represents your tag name to be unique ('TAG' in 'TAGS' table), and you need to put your con->query statement inside a try catch block, as it will probably throw exception on duplicate and you need to handle that.
3) I think you are inserting wring values as tags, you are using letters instead of words, is it supposed to be like that?
4) your query look incorrect, does it work?
do something like this (obviously after you have added unique index to your table):
function insertTag($con, $tag){
try {
$con->query("INSERT INTO `TAGS` (TAG`) values('$tag')");
echo "token $tag: inserted<br />";
} catch (Exception $e) {
echo "token $tag: NOT inserted<br />";
}
}
$tags = "danger in detroit";
// will create array of words
$tags = explode(' ', $tags);
$db = "test";
$user = "Eduardo";
$pass = "weaponx";
$host = "localhost";
$con = new mysqli($host, $user, $pass, $db);
// insert word in the DB, only if new
for($i=0; $i < count($tags); $i++) {
insertTag($con, $tags[$i]);
}
Also, I think this might be even done with ONE INSERT query for all tags, but I am not 100% sure, need to check.

Answer 2

Connecting and disconnecting from the database is expensive; for performance, you want to avoid connection "churning".
(In our J2EE web servers, we implement connection pools which are a collection of validated database sessions. The applications can "churn" through the connection pool, retrieving and returning connections, but the actual database session stays connected.)
So, to answer your question, it's a bad design to churn through database connections. A better approach is to connect once at the beginning of the process, pass the handle of that database connection to the functions that need it, and then disconnect at the end of the process.
The answer from Avetis Zakharyan provides a good approach. I totally agree with him that there is no need to run a separate SELECT statement. It's much more efficient to have the INSERT statement check whether the value exists or not.
This approach reduces the number of roundtrips to the database, AND it also works across multiple concurrent sessions where two (or more) sessions may be running a SELECT to check for the existence of tag, and both sessions don't find it, and when both sessions insert the same tag. A UNIQUE constraint on the column would avoid duplicates; using INSERT IGNORE avoids throwing an exception, but still...
My preferred approach would be to use a SQL statement that only attempts to insert a row if a matching row doesn't already exist:
INSERT INTO tags (tag)
SELECT v.tagval
FROM (SELECT :tag AS tagval) v
LEFT JOIN tags d ON d.tag = v.tagval
WHERE d.tag IS NULL
This is not dependent on a UNIQUE constraint on the tag columns; and it's not dependent on the existence of any index, but an index with a leading column of tag would be desirable for improved performance.
(The SELECT can be run separately, for testing.)
The query is a classic antijoin... "return rows from v for which there is no match in d".
With PDO, use bind parameters, rather than including values in the SQL text.
$sql = "INSERT INTO tags (tag)
SELECT v.tagval
FROM (SELECT :tag AS tagval) v
LEFT JOIN tags d ON d.tag = v.tagval
WHERE d.tag IS NULL ";
$stmt=$conn->prepare($sql);
$stmt->bindParam(":tag",$tag);
$stmt->execute();
(Note: if the characterset of the client differs from the characterset of the target column, the conversion may need to be explicit. For example, if the client characterset is UTF8 and the column is latin1, then:
FROM (SELECT CONVERT(:tag AS latin1) AS tagval) v