Go to the first, previous, next, last section, table of contents.


5 How Standards-compatible Is MySQL?

This chapter describes how MySQL relates to the ANSI SQL standards. MySQL has many extensions to the ANSI SQL standards, and here you will find out what they are, and how to use them. You will also find information about functionality missing from MySQL, and how to work around some differences.

5.1 MySQL Extensions to ANSI SQL92

MySQL includes some extensions that you probably will not find in other SQL databases. Be warned that if you use them, your code will not be portable to other SQL servers. In some cases, you can write code that includes MySQL extensions, but is still portable, by using comments of the form /*! ... */. In this case, MySQL will parse and execute the code within the comment as it would any other MySQL statement, but other SQL servers will ignore the extensions. For example:

SELECT /*! STRAIGHT_JOIN */ col_name FROM table1,table2 WHERE ...

If you add a version number after the '!', the syntax will only be executed if the MySQL version is equal to or newer than the used version number:

CREATE /*!32302 TEMPORARY */ TABLE (a int);

The above means that if you have Version 3.23.02 or newer, then MySQL will use the TEMPORARY keyword.

MySQL extensions are listed below:

5.2 Running MySQL in ANSI Mode

If you start mysqld with the --ansi option, the following behavior of MySQL changes:

5.3 MySQL Differences Compared to ANSI SQL92

We try to make MySQL follow the ANSI SQL standard and the ODBC SQL standard, but in some cases MySQL does some things differently:

5.4 Functionality Missing from MySQL

The following functionality is missing in the current version of MySQL. For a prioritized list indicating when new extensions may be added to MySQL, you should consult the online MySQL TODO list. That is the latest version of the TODO list in this manual. See section H MySQL and the future (The TODO).

5.4.1 Sub-selects

The following will not yet work in MySQL:

SELECT * FROM table1 WHERE id IN (SELECT id FROM table2);
SELECT * FROM table1 WHERE id NOT IN (SELECT id FROM table2);
SELECT * FROM table1 WHERE NOT EXISTS (SELECT id FROM table2 where table1.id=table2.id);

However, in many cases you can rewrite the query without a sub-select:

SELECT table1.* FROM table1,table2 WHERE table1.id=table2.id;
SELECT table1.* FROM table1 LEFT JOIN table2 ON table1.id=table2.id where table2.id IS NULL

For more complicated subqueries you can often create temporary tables to hold the subquery. In some cases, however this option will not work. The most frequently encountered of these cases arises with DELETE statements, for which standard SQL does not support joins (except in sub-selects). For this situation there are two options available until subqueries are supported by MySQL.

The first option is to use a procedural programming language (such as Perl or PHP) to submit a SELECT query to obtain the primary keys for the records to be deleted, and then use these values to construct the DELETE statement (DELETE FROM ... WHERE ... IN (key1, key2, ...)).

The second option is to use interactive SQL to contruct a set of DELETE statements automatically, using the MySQL extension CONCAT() (in lieu of the standard || operator). For example:

SELECT CONCAT('DELETE FROM tab1 WHERE pkid = ', tab1.pkid, ';')
  FROM tab1, tab2
 WHERE tab1.col1 = tab2.col2;

You can place this query in a script file and redirect input from it to the mysql command-line interpreter, piping its output back to a second instance of the interpreter:

prompt> mysql --skip-column-names mydb < myscript.sql | mysql mydb

MySQL only supports INSERT ... SELECT ... and REPLACE ... SELECT ... Independent sub-selects will probably be available in Version 4.0. You can now use the function IN() in other contexts, however.

5.4.2 SELECT INTO TABLE

MySQL doesn't yet support the Oracle SQL extension: SELECT ... INTO TABLE .... MySQL supports instead the ANSI SQL syntax INSERT INTO ... SELECT ..., which is basically the same thing. See section 7.21.1 INSERT ... SELECT Syntax.

INSERT INTO tblTemp2 (fldID) SELECT tblTemp1.fldOrder_ID FROM tblTemp1 WHERE
tblTemp1.fldOrder_ID > 100;

Alternatively, you can use SELECT INTO OUTFILE... or CREATE TABLE ... SELECT to solve your problem.

5.4.3 Transactions

As MySQL does nowadays support transactions, the following discussion is only valid if you are only using the non-transaction-safe table types. See section 7.31 BEGIN/COMMIT/ROLLBACK Syntax.

The question is often asked, by the curious and the critical, ``Why is MySQL not a transactional database?'' or ``Why does MySQL not support transactions?''

MySQL has made a conscious decision to support another paradigm for data integrity, ``atomic operations.'' It is our thinking and experience that atomic operations offer equal or even better integrity with much better performance. We, nonetheless, appreciate and understand the transactional database paradigm and plan, within the next few releases, to introduce transaction-safe tables on a per table basis. We will be giving our users the possibility to decide if they need the speed of atomic operations or if they need to use transactional features in their applications.

How does one use the features of MySQL to maintain rigorous integrity and how do these features compare with the transactional paradigm?

First, in the transactional paradigm, if your applications are written in a way that is dependent on the calling of ``rollback'' instead of ``commit'' in critical situations, then transactions are more convenient. Moreover, transactions ensure that unfinished updates or corrupting activities are not committed to the database; the server is given the opportunity to do an automatic rollback and your database is saved.

MySQL, in almost all cases, allows you to solve for potential problems by including simple checks before updates and by running simple scripts that check the databases for inconsistencies and automatically repair or warn if such occurs. Note that just by using the MySQL log or even adding one extra log, one can normally fix tables perfectly with no data integrity loss.

Moreover, fatal transactional updates can be rewritten to be atomic. In fact,we will go so far as to say that all integrity problems that transactions solve can be done with LOCK TABLES or atomic updates, ensuring that you never will get an automatic abort from the database, which is a common problem with transactional databases.

Not even transactions can prevent all loss if the server goes down. In such cases even a transactional system can lose data. The difference between different systems lies in just how small the time-lap is where they could lose data. No system is 100% secure, only ``secure enough.'' Even Oracle, reputed to be the safest of transactional databases, is reported to sometimes lose data in such situations.

To be safe with MySQL, you only need to have backups and have the update logging turned on. With this you can recover from any situation that you could with any transactional database. It is, of course, always good to have backups, independent of which database you use.

The transactional paradigm has its benefits and its drawbacks. Many users and application developers depend on the ease with which they can code around problems where an abort appears to be, or is necessary, and they may have to do a little more work with MySQL to either think differently or write more. If you are new to the atomic operations paradigm, or more familiar or more comfortable with transactions, do not jump to the conclusion that MySQL has not addressed these issues. Reliability and integrity are foremost in our minds. Recent estimates indicate that there are more than 1,000,000 mysqld servers currently running, many of which are in production environments. We hear very, very seldom from our users that they have lost any data, and in almost all of those cases user error is involved. This is, in our opinion, the best proof of MySQL's stability and reliability.

Lastly, in situations where integrity is of highest importance, MySQL's current features allow for transaction-level or better reliability and integrity. If you lock tables with LOCK TABLES, all updates will stall until any integrity checks are made. If you only obtain a read lock (as opposed to a write lock), then reads and inserts are still allowed to happen. The new inserted records will not be seen by any of the clients that have a READ lock until they release their read locks. With INSERT DELAYED you can queue inserts into a local queue, until the locks are released, without having the client wait for the insert to complete. See section 7.21.2 INSERT DELAYED syntax.

``Atomic,'' in the sense that we mean it, is nothing magical. It only means that you can be sure that while each specific update is running, no other user can interfere with it, and there will never be an automatic rollback (which can happen on transaction based systems if you are not very careful). MySQL also guarantees that there will not be any dirty reads. You can find some example of how to write atomic updates in the commit-rollback section. See section 5.6 How to Cope Without COMMIT/ROLLBACK.

We have thought quite a bit about integrity and performance, and we believe that our atomic operations paradigm allows for both high reliability and extremely high performance, on the order of three to five times the speed of the fastest and most optimally tuned of transactional databases. We didn't leave out transactions because they are hard to do. The main reason we went with atomic operations as opposed to transactions is that by doing this we could apply many speed optimizations that would not otherwise have been possible.

Many of our users who have speed foremost in their minds are not at all concerned about transactions. For them transactions are not an issue. For those of our users who are concerned with or have wondered about transactions vis-a-vis MySQL, there is a ``MySQL way'' as we have outlined above. For those where safety is more important than speed, we recommend them to use the BDB, GEMINI or InnoDB tables for all their critical data. See section 8 MySQL Table Types.

One final note: We are currently working on a safe replication schema that we believe to be better than any commercial replication system we know of. This system will work most reliably under the atomic operations, non-transactional, paradigm. Stay tuned.

5.4.4 Stored Procedures and Triggers

A stored procedure is a set of SQL commands that can be compiled and stored in the server. Once this has been done, clients don't need to keep reissuing the entire query but can refer to the stored procedure. This provides better performance because the query has to be parsed only once, and less information needs to be sent between the server and the client. You can also raise the conceptual level by having libraries of functions in the server.

A trigger is a stored procedure that is invoked when a particular event occurs. For example, you can install a stored procedure that is triggered each time a record is deleted from a transaction table and that automatically deletes the corresponding customer from a customer table when all his transactions are deleted.

The planned update language will be able to handle stored procedures, but without triggers. Triggers usually slow down everything, even queries for which they are not needed.

To see when MySQL might get stored procedures, see section H MySQL and the future (The TODO).

5.4.5 Foreign Keys

Note that foreign keys in SQL are not used to join tables, but are used mostly for checking referential integrity (foreign key constraints). If you want to get results from multiple tables from a SELECT statement, you do this by joining tables:

SELECT * from table1,table2 where table1.id = table2.id;

See section 7.20 JOIN Syntax. See section 9.5.6 Using Foreign Keys.

The FOREIGN KEY syntax in MySQL exists only for compatibility with other SQL vendors' CREATE TABLE commands; it doesn't do anything. The FOREIGN KEY syntax without ON DELETE ... is mostly used for documentation purposes. Some ODBC applications may use this to produce automatic WHERE clauses, but this is usually easy to override. FOREIGN KEY is sometimes used as a constraint check, but this check is unnecessary in practice if rows are inserted into the tables in the right order. MySQL only supports these clauses because some applications require them to exist (regardless of whether or not they work).

In MySQL, you can work around the problem of ON DELETE ... not being implemented by adding the appropriate DELETE statement to an application when you delete records from a table that has a foreign key. In practice this is as quick (in some cases quicker) and much more portable than using foreign keys.

In the near future we will extend the FOREIGN KEY implementation so that at least the information will be saved in the table specification file and may be retrieved by mysqldump and ODBC. At a later stage we will implement the foreign key constraints for application that can't easily be coded to avoid them.

5.4.5.1 Reasons NOT to Use Foreign Keys constraints

There are so many problems with foreign key constraints that we don't know where to start:

The only nice aspect of FOREIGN KEY is that it gives ODBC and some other client programs the ability to see how a table is connected and to use this to show connection diagrams and to help in building applications.

MySQL will soon store FOREIGN KEY definitions so that a client can ask for and receive an answer about how the original connection was made. The current `.frm' file format does not have any place for it. At a later stage we will implement the foreign key constraints for application that can't easily be coded to avoid them.

5.4.6 Views

MySQL doesn't yet support views, but we plan to implement these to about 4.1.

Views are mostly useful for letting users access a set of relations as one table (in read-only mode). Many SQL databases don't allow one to update any rows in a view, but you have to do the updates in the separate tables.

As MySQL is mostly used in applications and on web system where the application writer has full control on the database usage, most of our users haven't regarded views to be very important. (At least no one has been interested enough in this to be prepared to finance the implementation of views).

One doesn't need views in MySQL to restrict access to columns as MySQL has a very sophisticated privilege system. See section 6 The MySQL Access Privilege System.

5.4.7 `--' as the Start of a Comment

Some other SQL databases use `--' to start comments. MySQL has `#' as the start comment character, even if the mysql command-line tool removes all lines that start with `--'. You can also use the C comment style /* this is a comment */ with MySQL. See section 7.38 Comment Syntax.

MySQL Version 3.23.3 and above supports the `--' comment style only if the comment is followed by a space. This is because this degenerate comment style has caused many problems with automatically generated SQL queries that have used something like the following code, where we automatically insert the value of the payment for !payment!:

UPDATE tbl_name SET credit=credit-!payment!

What do you think will happen when the value of payment is negative?

Because 1--1 is legal in SQL, we think it is terrible that `--' means start comment.

In MySQL Version 3.23 you can, however, use: 1-- This is a comment

The following discussion only concerns you if you are running a MySQL version earlier than Version 3.23:

If you have a SQL program in a text file that contains `--' comments you should use:

shell> replace " --" " #" < text-file-with-funny-comments.sql \
         | mysql database

instead of the usual:

shell> mysql database < text-file-with-funny-comments.sql

You can also edit the command file ``in place'' to change the `--' comments to `#' comments:

shell> replace " --" " #" -- text-file-with-funny-comments.sql

Change them back with this command:

shell> replace " #" " --" -- text-file-with-funny-comments.sql

5.5 What Standards Does MySQL Follow?

Entry level SQL92. ODBC levels 0-2.

5.6 How to Cope Without COMMIT/ROLLBACK

The following mostly applies only for ISAM, MyISAM, and HEAP tables. If you only use transaction-safe tables (BDB, GEMINI or InnoDB tables) in an an update, you can do COMMIT and ROLLBACK also with MySQL. See section 7.31 BEGIN/COMMIT/ROLLBACK Syntax.

The problem with handling COMMIT-ROLLBACK efficiently with the above table types would require a completely different table layout than MySQL uses today. The table type would also need extra threads that do automatic cleanups on the tables, and the disk usage would be much higher. This would make these table types about 2-4 times slower than they are today.

For the moment, we prefer implementing the SQL server language (something like stored procedures). With this you would very seldom really need COMMIT-ROLLBACK. This would also give much better performance.

Loops that need transactions normally can be coded with the help of LOCK TABLES, and you don't need cursors when you can update records on the fly.

We at TcX had a greater need for a real fast database than a 100% general database. Whenever we find a way to implement these features without any speed loss, we will probably do it. For the moment, there are many more important things to do. Check the TODO for how we prioritize things at the moment. (Customers with higher levels of support can alter this, so things may be reprioritized.)

The current problem is actually ROLLBACK. Without ROLLBACK, you can do any kind of COMMIT action with LOCK TABLES. To support ROLLBACK with the above table types, MySQL would have to be changed to store all old records that were updated and revert everything back to the starting point if ROLLBACK was issued. For simple cases, this isn't that hard to do (the current isamlog could be used for this purpose), but it would be much more difficult to implement ROLLBACK for ALTER/DROP/CREATE TABLE.

To avoid using ROLLBACK, you can use the following strategy:

  1. Use LOCK TABLES ... to lock all the tables you want to access.
  2. Test conditions.
  3. Update if everything is okay.
  4. Use UNLOCK TABLES to release your locks.

This is usually a much faster method than using transactions with possible ROLLBACKs, although not always. The only situation this solution doesn't handle is when someone kills the threads in the middle of an update. In this case, all locks will be released but some of the updates may not have been executed.

You can also use functions to update records in a single operation. You can get a very efficient application by using the following techniques:

For example, when we are doing updates to some customer information, we update only the customer data that has changed and test only that none of the changed data, or data that depend on the changed data, has changed compared to the original row. The test for changed data is done with the WHERE clause in the UPDATE statement. If the record wasn't updated, we give the client a message: "Some of the data you have changed have been changed by another user". Then we show the old row versus the new row in a window, so the user can decide which version of the customer record he should use.

This gives us something that is similar to column locking but is actually even better, because we only update some of the columns, using values that are relative to their current values. This means that typical UPDATE statements look something like these:

UPDATE tablename SET pay_back=pay_back+'relative change';

UPDATE customer
  SET
    customer_date='current_date',
    address='new address',
    phone='new phone',
    money_he_owes_us=money_he_owes_us+'new_money'
  WHERE
    customer_id=id AND address='old address' AND phone='old phone';

As you can see, this is very efficient and works even if another client has changed the values in the pay_back or money_he_owes_us columns.

In many cases, users have wanted ROLLBACK and/or LOCK TABLES for the purpose of managing unique identifiers for some tables. This can be handled much more efficiently by using an AUTO_INCREMENT column and either the SQL function LAST_INSERT_ID() or the C API function mysql_insert_id(). See section 24.1.3.126 mysql_insert_id().

At MySQL AB, we have never had any need for row-level locking because we have always been able to code around it. Some cases really need row locking, but they are very few. If you want row-level locking, you can use a flag column in the table and do something like this:

UPDATE tbl_name SET row_flag=1 WHERE id=ID;

MySQL returns 1 for the number of affected rows if the row was found and row_flag wasn't already 1 in the original row.

You can think of it as MySQL changed the above query to:

UPDATE tbl_name SET row_flag=1 WHERE id=ID and row_flag <> 1;


Go to the first, previous, next, last section, table of contents.