Sunday 22nd of October 2017 04:47:59 AM

Here are some benefits of the structured nature of XML:

  • XML parsers make your application code more reliable and quick to develop by providing validity checking on your XML documents (if you use a DTD).
  • XML allows you to easily generate XML documents (that contain your information), since it is so structured.
  • XML parsers allow you to code faster by giving you a parser for your all your XML documents (with and without DTDs).
MySQL Subqueries
November 15, 2004

Subqueries that can't be rewritten as joins

Standard SQL does not support joins in DELETE statements. If you wanted to delete all records in animal that did not contain an associated food, you would be able to do it quite simply with a subquery, as follows:

mysql> DELETE FROM animal
WHERE id NOT IN (SELECT animal_id
FROM animal_food
WHERE animal_id=animal.id);
Query OK, 2 rows affected (0.00 sec)

Only the two records remain:

mysql> SELECT * FROM animal;
+----+---------------------+
| id | name                |
+----+---------------------+
|  2 | Chacma Baboon       |
|  4 | Small Grey Mongoose |
+----+---------------------+

However, the subquery inside a DELETE statement syntax is not supported in MySQL 4.0. To achieve the same result, you would need to pass the burden to your application. You would first have to SELECT the records you want to delete, then place the results into a variable, and use this variable to create a DELETE statement. Your application logic would be as follows:

SELECT id FROM animal
WHERE NOT EXISTS (SELECT animal_id
FROM animal_food
WHERE animal.id=animal_id);
 

Create a string variable called $results containing the comma-separated results from the former query

 
DELETE FROM animal WHERE ID IN ($results)

However, it is not only DELETE statement subqueries that cannot be rewritten. Some SELECT subqueries are also impossible to rewrite. Here is another example:

mysql> SELECT name FROM animal WHERE id=(SELECT MAX(id) FROM animal_food);
+---------------------+
| name                |
+---------------------+
| Small Grey Mongoose |
+---------------------+

The MAX() function prevents this from being written as a join. MySQL 4.1 is a necessity for doing these sorts of queries.

Can subqueries be better?

Let's look at a more complex example:

mysql> SELECT animal.id,name FROM animal
WHERE animal.id IN (1,2,3) AND EXISTS (SELECT NULL FROM animal_food
WHERE animal_food.id>2 AND animal.id=animal_food.id ) ORDER BY name;
+----+-----------+
| id | name      |
+----+-----------+
|  3 | Porcupine |
+----+-----------+

This can also be rewritten as a join, as follows:

mysql> SELECT DISTINCT animal.id,name FROM animal,animal_food
WHERE animal.id IN (1,2,3) AND animal_food.id > 2 AND animal.id=animal_food.id ORDER BY name;
+----+-----------+
| id | name      |
+----+-----------+
|  3 | Porcupine |
+----+-----------+

Which is better? EXPLAIN shows us the following:


Click for full EXPLAIN SELECT(3) code

With a subquery, MySQL has to examine 4 rows. In addition, it does not use an index. Without the subquery, MySQL needs to examine 3 rows and uses an index. However, it also uses a temporary table, normally an indication of a sub-optimal query. The question of which is better is then quite complex, and highly dependant on the dataset, the tuning parameters (and the DBMS - do not assume MySQL optimizes the same way as other DBMS'). Does the temporary table have to be created on the disk or in memory? How large is the table that is going to be scanned without an index? I suggest doing some thorough benchmarking in your environment to see which suits you best.

Uncorrelated and Correlated subqueries

Let's further complicate matters, and rewrite the correlated subquery as an uncorrelated subquery. The main difference between the two kinds is that, in a correlated subquery, the inner query is dependant upon the outer query, while, with an uncorrelated query, the inner query does not depend upon the outer query. Instead, it runs just once. Whether this is better or not again depends on your data set. If the number of rows returned by the inner part of the uncorrelated query is relatively large, it can counteract the benefit of only running once. The inner part of the correlated query needs to be performed a number of times - again the efficiency of this depends on your situation. The advice of some older texts to always use a correlated query is flawed. Here is the query as an uncorrelated query:

mysql> SELECT animal.id,name FROM animal
WHERE animal.id IN (1,2,3) AND animal.id
IN (SELECT animal_id FROM animal_food WHERE animal_food.id>2 ) ORDER BY name;
+----+---------------+
| id | name          |
+----+---------------+
|  2 | Chacma Baboon |
+----+---------------+

And using EXPLAIN:


Click for full EXPLAIN SELECT (4) code

Sixteen rows need to be examined. Even though there is no temporary table, in our limited data set this is not the most efficient query. Nevertheless, results will vary. There is no easy answer as to which is best. In this case "it depends," "sometimes one, sometimes the other" really is the best answer! So benchmark your results in your environment, and you could see significant improvements. Good luck!

TIP

When it comes right down to it, the extension isn't actually the important part. What matters is the MIME type the server uses when sending a file. However, since the vast majority of web servers use a file's extension to decide which MIME type to use when sending the file, it obviously becomes important to have a friendly server configuration.

If an external style sheet is sent using the wrong MIME type, the inline element with a background color and padding could have the background extend above and below the element. Figure 7-61 gives us some idea of what that might look like. The line height isn't changed, of course, but since padding does extend the background, it should be visible, right?

Here's where the famous phrase returns: "there may be implementation-specific limits." User agents aren't required to support this type of effect.