My sql technical reference manual

MySQL Technical Reference Manual.
What is MySQL?
MySQL is a SQL (Structured Query Language) database server. SQL is the most popular
database language in the world. MySQL is a client server implementation that consists of a server
daemon mysqld and many different client programs/libraries.
The main goals of MySQL are speed and robustness.
The base upon which MySQL is built is a set of routines that have been used in a highly
demanding production environment for many years.

The main features in MySQL
• Multi-threaded
• C, C++, JAVA, Perl, Python and TCL API’s.
• Lots of column types like: signed/unsigned integers 1,2,3,4,8 bytes long, FLOAT,
CHAR,VARCHAR, TEXT, BLOB, DATE, SET and ENUM types.
• Join optimizer with one-sweep multi-join (all joins made in one pass).
• Full function support in the SELECT and WHERE parts.
Example: select column1+column2 from table where column1/column2 > 0
• SQL functions are implemented through a very optimized class library and should be as
fast as they can get! Usually there shouldn’t be any memory allocation at all after the
query initialization.
• Full support for SQL GROUP BY and ORDER BY. Support for group functions (SUM,
MAX and MIN).
• A privilege and password system with is very flexible and secure. Allows host based
verification.
• All password traffic on the net is encrypted.
• Very fast B-tree disk tables with key compression.
• Fixed and variable length records.
• 16 indexes/table. Each index may consist of 1 to 15 columns/parts of columns. Max key
length is 127 bytes. A key may be a prefix of a CHAR field.
• ODBC Open-DataBase-Connectivity for Windows95 (with source). All ODBC 2.5
functions and lots of others.
• In memory hash tables always used as temporary tables.
• Can handle big databases (we are using MySQL with some databases that contain
50,000,000 records).
• All columns have default values. One can always use INSERT on any subset of columns.
• Uses GNU autoconf for portability.
• Written in C and C++. Tested with gcc 2.7.2.
• A thread based memory allocation system (very fast and no memory trashing).
• No memory leaks. Tested with a commercial memory leakage detector (purify).
• A very fast table check and repair utility (isamchk).
• All data saved in ISO8859_1 format. All comparisons for normal string columns are case
insensitive.
1

• Full ISO8859_1 (Latin1) support. For example Scandinavian åäö is allowed in table and
column names.
• Sorts by ISO8859_1 Latin1 (the Swedish way at the moment). It is possible to change
this in the source by adding new sort order arrays.
• Alias on tables and columns as in the SQL92 standard. avg & count).
• INSERT,UPDATE and DELETE returns how many rows were affected.
• Function names do not clash with table or column names. For example ABS is a valid
column name. The only restriction is that space is not allowed between a function name
and the ’(’ when using functions.
• All MySQL commands have --help or -? for help.
• The server currently supports error messages to clients in many languages. See section
What languages are supported by MySQL..
• The clients uses a TCP connection or unix socket when connecting to the MySQL server.
• User commands as show tables, show keys from table and show columns from Table

Problems starting MySQL
Check the log file to see if mysqld started up correctly.
cd <localstatedir default /usr/local/var>
tail <your host name>.log
To verify that MySQL is working, run the following tests:
> cd /usr/local/bin
> ./mysqlshow
+-----------+
| Databases |
+-----------+
| mysql |
+-----------+
> ./mysqlshow mysql
Database: mysql
+--------+
| Tables |
+--------+
| db |
| host |
| user |
+--------+
> ./mysql -e "select host,db,user from db" mysql
+------+--------+------+
| host | db | user |
+------+--------+------+
| % | test | |
| % | test_% | |
+------+--------+------+
There is also a benchmark suite so you can compare how MySQL perform on different platforms.
In the near future this will also be used to compare MySQL to other SQL databases.
> cd bench
> run-auto-increment-test
2

You can also run the tests in the test subdirectory. To run ‘auto_increment.tst’:
./mysql -vf test < ./tests/auto_increment.tst
24
Expected results are shown in the file ‘./tests/auto_increment.res’.

Automatic start/stop of MySQL
To start or stop MySQL use the following commands:
scripts/mysql.server stop
scripts/mysql.server start
You might want to add these start and stop commands in the appropriate places in your /etc/rc*
files
when you start using MySQL for production applications.

How to debug MySQL
If you are porting MySQL to an new system you should first try to get mysys/thr_lock and
mysys/thr_alarm to work. They shouldn’t core dump and not print any error (they also print a lot
of
other information). Also see the file ‘PORTING’ in the distribution.
By starting bin/safe_mysqld with --log you will get a log in
install-directory/var/’hostname’.log (the top level database directory). This log will
contain all successful connections and all commands issued to the MySQL server.
If you have compiled MySQL with --with-debug=yes you can can also get a very informative log
with:
libexec/mysqld --debug
which makes a large log in /tmp/mysql.trace. The default debug option is
d:i:t:o,/tmp/mysql.trace.
You can get a smaller log with:
libexec/mysqld --debug=d,info,query,general:o,/tmp/mysql.trace
or an even smaller (on stdout):
libexec/mysqld --debug=d,general,query
You can get more information about the debug switches by examining the file dbug/dbug.c.
If you have a problem with mysqld that it crashes and you want this quickly solved, you should
include a trace file with your mail if possible. Trace files can be posted directly to
@email{mysql-support@tcx.se} to avoid long messages to the standard mail list. If the trace file
is big
you should use ftp and send it to ftp://www.tcx.se/pub/mysql/secret/ together with a
mysqldump of any tables that you think can help to solve the problem. The context of the above
directory is hidden for outside users so no one except the TCX staff can look at what you send
into it.
Go to the first [p 5] , previous [p 19] , next [p 26] , last [p 87] section, table of contents [p 1] .
25
How does MySQL privileges work
MySQL has a advanded but non standard security/privileges system.
How does the privilege system work?
The MySQL privilege system makes sure that each user may do exactly the things that they are

3

supposed to be allowed to do. The system decides to grant different privileges depending on
which
user connects from which host to which database. The decision is based on the contents of
the three
tables in the MySQL database: user, host and db.
The grant tables privileges on rows are select, insert, update and delete.
The table and database privileges are create and drop. Create and drop are for both tables and
databases. Since a user with a drop grant can delete any table, this is the same thing as a drop
grant for
the database.
Other privileges give the right to use files (for LOAD DATA INFILE and SELECT INTO
OUTFILE)
and to use the administrative commands shutdown, reload, refresh and process, to get the
current
process list.
The privilege tables are read into mysqld with mysqladmin reload. If the privilege tables are
empty or non-existent or if the server is started with --skip-grant-tables, full access is
granted to everyone.
You can always test your privileges with the script mysqlaccess, which Yves Carlier has
provided
for the MySQL distribution. See section Why do I get ’Access denied’? [p 78]
The host and db fields may contain a SQL regexp with chars % and _. Leaving any of these
fields
empty is equivalent to setting it to ’%’.
A host may be localhost, a hostname, an IP number or an SQL expression. An empty host in
the db table means any host in the host table. An empty host in the host or user table means any
host that can create a TCP connection to your server.
A db is the name of a database or an SQL regexp.
An empty user field means any username will do.
An empty password means that the entry will only be used if no password is supplied.
The privileges from the user table are OR’ed to the db table. This means that a superuser only
needs to be in the user table with all privilege-flags set to Y.
Everything granted in the user table is valid for every database that cannot be found in the db
table.
For this reason, it might be wise to grant users (apart from superusers) privileges on a per-
database
basis only.
The host table is mainly there to maintain a list of "secure" servers. At TcX host contains a list of
all machines on the local network. These are granted all privileges.
26
The connecting user’s privileges are calculated by the following algorithm:
1. First sort all three tables by host by putting hosts without wildcards first, followed by hosts
with
wildcards and entries with host = "". Within each host, di.e. very much like GROUP BY
Host, sort by user using the same rules. Finally, in the db table, sort by db using the same rules.
In the steps below, we will look through the sorted tables and always use the first match found.
2. Get the privileges for the connecting user from the db table using the first match found. Call
this
set of privileges P.
4

3. If host = "" for the entry found in the db table, AND P with the privileges for the host in the
host table, i.e. remove all privileges that are not "Y" in both. (If host <> "", P is not
affected. In suchcases, host must have matched the connecting host’s name at least partially.
Therefor it can be assumed that the privileges found in this row match the connecting host’s
profile.)
4. OR (add) P with the privileges for the user from the user table, i.e. add all privileges that are
"Y" in user.
The connecting user gets the set of privileges P.
Let’s show an example of the sorting and matching! Suppose that the user table contains this:
+-----------+----------+-
| Host | User | ...
+-----------+----------+-
| % | root | ...
| % | jeffrey | ...
| localhost | root | ...
| localhost | | ...
+-----------+----------+-
Then the search order will be:
localhost/root
localhost/any
any/jeffrey
any/root
So jeffrey attempting to connect on localhost will be matched by the localhost/any line, not by
the any/jeffrey line. The first match found is used!
So if you have access problems, print out the user table, sort it by hand, and see where the match
is
being made.
Here follows an example to add a user ’custom’ that can connect from hosts ’localhost’,
’server.domain’ and ’whitehouse.gov’. He wants to have password ’stupid’. The database
’bankacount’ he only want to use from ’localhost’ and the ’customer’ database he wants to be
able to
reach from all three hosts.
shell> mysql mysql.
mysql> insert into users (host,user,password)
values(’localhost’,’custom’,password(’stupid’));
values(’server.domain’,’custom’,password(’stupid’));
values(’whitehouse.gov’,’custom’,password(’stupid’));
27
mysql> insert into db
(host,db,user,Select_priv,Insert_priv,Update_priv,Delete_priv,
Create_priv,Drop_priv)
values
(’localhost’,’bankacount’,’custom’,’Y’,’Y’,’Y’,’Y’,’Y’,’Y’);
mysql> insert into db
(host,db,user,Select_priv,Insert_priv,Update_priv,Delete_priv,
Create_priv,Drop_priv)
values
5

(’%’,’customers’,’custom’,’Y’,’Y’,’Y’,’Y’,’Y’,’Y’);
You can of course also use xmysqladmin, mysql_webadmin, mysqladmin and even xmysql to
insert/change and update values in the privilege tables. You can find these utilities in the Contrib
directory.
Adding new user privileges to MySQL
To add privileges to the MySQL database:
This assumes the current user has insert privileges for the mysql db table and reload privileges.
The
server (mysqld) has to be running. If it is not, start it with safe_mysqld --log &.
> mysql mysql
insert into user values
(’%’,’monty’,password(’something’),’Y’,’Y’,’Y’,’Y’,’Y’,’Y’,’Y’,’Y’,’Y’,’Y’) ;
insert into user (host,user,password) values(’localhost’,’dummy’,") ;
insert into user values (’%’,’admin’,",’N’,’N’,’N’,’N’,’N’,’N’,’Y’,’N’,’Y’,’Y’) ;
quit
> mysqladmin reload
This makes three new users:
Monty
Full superuser, but must use password when using MySQL.
admin
Doesn’t need a password but is only allowed to use mysqladmin reload, mysqladmin
refresh and mysqladmin processlist. May be granted individual database privileges
through table db.
dummy
Must be granted individual database privileges through table db.
Default privileges.
The default privileges (set in ‘scripts/mysql_install_db’) is that root can do anything. Any
user can do anything with any database whose name is ’test’ or starts with ’test_’. A normal user
can’t
use mysqladmin shutdown or mysqladmin processlist. See the script
(‘scripts/mysql_install_db’) for an example on how to add other users.
The privilege tables are read into mysqld with ’mysqladmin reload’. If the privilege tables are
empty
(or non-existent) full access are granted to everyone.
28
A example of permission setup.
A common mistake is to try something like:
INSERT INTO user VALUES
(’%’,’jeffrey’,’bLa81m0’,’Y’,’Y’,’Y’,’N’,’N’,’N’,’N’,’N’, ’N’,’N’);
Then (of course) a mysqladmin reload to make the authentication change take effect, then trying
to connect to the server:
$ ./mysql -h sqlserver -u jeffrey -p bLa81m0 test Access denied
Try this instead:
INSERT INTO user VALUES
(’%’,’jeffrey’,password(’bLa81m0’),’Y’,’Y’,’Y’,’N’,’N’,’N’,’N’,’N’,’N’,’N’);
And like before mysqladmin reload to make the authentication change take effect.
Now things should work.
29
6

MySQL language reference
Literals. How do you write strings and numbers?
STRINGS
A string may have ’ or " around it.
is a escape character. The following escape characters are recognised:
0 A ascii 0 character.
n A newline character.
t A tab character.
r A return character.
b A backspace character.
’ A ’ character.
" A " character.
A character.
% A % character. This is used in wild-card strings to search after %.
_ A _ character. This is used in wild-card strings to search after _.
A ’ inside a string started with ’ may be written as ".
A " inside a string started with " may be written as "".
MySQL> select ’hello’, "’hello’", ’""hello""’, "’h"e"l"l"o"’, "hel""lo";
1 rows in set (0.00 sec)
+-------+---------+-----------+-------------+--------+
| hello | ’hello’ | ""hello"" | ’h’e’l’l’o’ | hel"lo |
+-------+---------+-----------+-------------+--------+
| hello | ’hello’ | ""hello"" | ’h’e’l’l’o’ | hel"lo |
+-------+---------+-----------+-------------+--------+
mysql> select ’hello’, "hello", ’""hello""’, "’ello", ’e"l"lo’, ’’hello’;
+-------+-------+-----------+-------+--------+--------+
| hello | hello | ""hello"" | ’ello | e’l’lo | ’hello |
+-------+-------+-----------+-------+--------+--------+
| hello | hello | ""hello"" | ’ello | e’l’lo | ’hello |
+-------+-------+-----------+-------+--------+--------+
mysql> select "ThisnIsnFournlines";
+--------------------+
| This
Is
Four
lines |
+--------------------+
| This
30
Is
Four
lines |
+--------------------+
If you want to insert binary data into a blob the following characters must be represented by
escape
sequences:
7

0 Ascii 0. Should be replaced with "0" (A backslash and a 0 digit).
Ascii 92, backslash
’ Ascii 39, Single quote
" Ascii 33, Double quote
NUMBERS
Integers are just a sequence of digits. Floats use . as a decimal separator.
Examples of valid numbers are: 1221, 294.42, -32032.6809e+10.
NULL
When using the text file export formats, NULL may be represented by N. See section LOAD
DATA
INFILE syntax [p 45]
Column types.
The following column types are supported:
Name Description Size
TINYINT [(max display size)]
[UNSIGNED] [ZEROFILL]
A very small integer. Signed range -128 - 127. Unsigned
range 0 - 255.
1
SMALLINT [(max display
size)]. [UNSIGNED]
[ZEROFILL]
A small integer. Signed range -32768 - 32767. Unsigned
range 0 - 65535.
2
MEDIUMINT [(max display
size)] [UNSIGNED]
[ZEROFILL]
A medium integer. Signed range -8388608-8388607.
Unsigned range 0 - 16777215.
3
INT [(max display size)]
A normal integer. xSigned range -2147483648 -
2147483647. Unsigned range 0 - 4294967295.
4
BIGINT [(max display size)]
A large integer. Signed range -9223372036854775808 -
9223372036854775807. Unsigned Range 0 -
18446744073709551615.
8
31
FLOAT(Precision)
A small floating point number. Precision can be 4 or 8.
FLOAT(4) is a single precision number and FLOAT(8) is a
double precision number (se the DOUBLE entry). This
syntax is for ODBC compatibility. Range
-3.402823466E+38F - -1.175494351E-38, 0,
8

-1.175494351E-38 - 3.402823466E+38F.
4
FLOAT [(max display
size,number of decimals)]
A small floating point number. Cannot be unsigned. Range
-3.402823466E+38F - -1.175494351E-38, 0,
-1.175494351E-38 - 3.402823466E+38F.
4
DOUBLE PRECISION [(max
display size,number of
decimals)]
A normal floating point number. Cannot be unsigned.
Range -1.7976931348623157E+308 -
-2.2250738585072014E-308, 0,
2.2250738585072014E-308 - 1.7976931348623157E+308.
8
REAL [(length,decimals)] Same as DOUBLE 8
DECIMAL [(max display
size,number of decimals)]
An unpacked floating point number. Cannot be unsigned.
Currently the same range maximum range as a double.
Behaves as a CHAR column
M+D
NUMERIC [(length,decimals)] Same as DECIMAL M+D
TIMESTAMP [(display size)] An automatic timestamp. 4
DATE
A type to store date information. Uses the
"YYYY-MM-DD" syntax, but may be updated with a
number or a string. Understands at least the following
syntaxes: ’YY-MM-DD’, ’YYYY-MM-DD’, ’YYMMDD’,
’YYMM’, ’YY’. Range 0000-00-00 to 9999-12-31.
4
TIME
A type to store time information. Uses the "HH:MM:SS"
syntax, but may be updated with a number or a string.
Understands at least the following syntaxes: ’HH:MM:DD,
’HHMMDD’, ’HHMM’, ’HH’.
3
DATETIME
A type to store date and time information. Format
"YYYY-MM-DD HH:MM:SS". Takes 8 bytes. Range
’0000-01-01 00:00:00’ - ’9999-12-31 23:59:59’.
8
CHAR(max_length) [binary]
A fixed length string that is always filled up with spaces to
the specified length. Range 1 - 255 characters. All end
space are removed when retrieved. Is sorted and compared
case insensitively unless the binary keyword is given.
M
9

VARCHAR(max_length)
[binary]
A variable length string that is stored with its length.
Maximum range 1 - 255 characters. Is sorted and compared
case insensitively unless the binary keyword is given.
L+1
TINYTEXT and TINYBLOB A TEXT/BLOB with max length of 255 characters. L + 1
TEXT and BLOB A TEXT/BLOB with max length of 65535 characters. L + 2
32
MEDIUMTEXT and
MEDIUMBLOB
A TEXT/BLOB with max length of 1677216 characters. L + 3
LONGTEXT and LONGBLOB
A TEXT/BLOB with max length of 4294967295
characters.
L+4
ENUM(’value’,’value2’,...)
A string object that can have only one set of allowed
values. See section More about data types [p 33] .
1 or
2
SET(’value’,’value2’,....)
A string object that can have one or many values of a set of
allowed values. See section More about data types [p 33] .
1-8
More about data types
Database size info.
In the above table L means the actual length of a instance and M the maximum length. So L+1
for
"abcd" means 5 bytes in the database.
If you use any data type with a L in the length field you will get a variable length record format.
The numeric types
All integer types can have a optional argument unsigned. This can be used when you only want
to
allow positive numbers in the column or you need a little bigger numerical range for the column.
Also for all integer columns the optional argument ZEROFILL means that the column will be
padded
with zeroes upto the maximum length.
Max display size and decimals are for formating and calculation of max column width.
TIMESTAMP type
Has a range of 1 Dec 1970 time 0.00 to sometime in the year 2106 and a resolution of one
second.
Will be automatically updated if not used in a statement that updates a row or if set to NULL.
Can be a
(part of) a key. Note that if you have many timestamp fields in a row, then only the first
timestamp
field will be automatically updated. Any timestamp field will be set to the current time if set to
NULL.

10

Depending on the display size one gets one of the following formats: "YYYY-MM-DD
HH:MM:SS",
"YY-MM-DD HH:MM:SS", "YYYY-MM-DD" or "YY-MM-DD".
TEXT and BLOB types
These are objects that can have a variable length without upper limit. All TEXT and BLOB
objects are
stored with is length (saved in 1 to 4 bytes depending on the type of object). The maximum
TEXT and
BLOB length you can use is dependent on available memory and client buffers. The only
differences
between TEXT and BLOB is that TEXT is sorted and compared case insensitively while BLOB
is
compared case insensitive (by character values). TEXT and BLOB objects CANNOT be keys.
A BLOB is a binary large object which can hold any amount of data :) There are 4 kinds of blobs
See
section Column types. [p 31] . Normally one can regard a BLOB as a VARCHAR without a
specified
limit.
33
TEXT is a BLOB that is sorted and compared case insensitively.
There are some constraints because of the message buffer used. The default size of the buffer is
64K
for the server and 512K for the clients. To change the buffer length for the server, use mysqld -O
max_allowed_packet=max_blob_length. This allows the message buffer to grow up to this
limit when needed.
MyODBC defines BLOB:s as LONGVARBINARY and TEXT:s as LONGVARCHAR.
Restrictions for BLOB and TEXT columns:
1. A BLOB or TEXT cannot be a key or a part of a key
2. When one sorts or groups a BLOB or TEXT only the first max_sort_length (default 1024)
of the blob is used. This value can be changed by the -O option when starting the mysqld demon.
One can group on an expression involving a BLOB/TEXT: SELECT
id,SUBSTR(blob,1,100) GROUP BY 2
3. There is no end space truncation for BLOB and TEXT as there is for CHAR and VARCHAR.
ENUM type
A string object that can have only one set of allowed values. The value to be stored may be given
case
independently. If one tries to store a non existing value, "" is stored. If used in a number context
this
object returns/stores the value index. If there is less than 255 possible values this object occupies
1
byte, else two bytes (with a maximum of 65535 different values). Note that if a integer is put in
the
ENUM you get the corresponding string with the first counting as number zero. Sorting on
ENUM types
are done after the order of the strings in the enum.
For example the column test ENUM("one","two", "three") can have any off these values:
"one"
"two"
"three"
11

SET type
A string object that can have one or many values of a set of allowed values. Each value is
separated by
a ’,’. If used in a number context this object returns/stores the bit positions of the used values.
This
object occupies (number_of_different_values-1)/8+1 bytes, rounded up to 1,2,3,4 or 8. One can’t
have
more than 64 different values. Note that if a integer is put in the SET you get the corresponding
string
with the first bit corresponding to the first string.Sorting on SET types are done numericaly.
For example the column test SET("one","two") can have any off these values:
""
"one"
"two"
"one,two"
34
Column indexes
You can have indexes on all MysQL column types except BLOB and TEXT. Using indexes on
the
relevant columns is the most important thing is getting you selects to run fast.
For CHAR and VARCHAR columns you can have a index on a prefix. The example below show
how
to create a index for the first 10 characters of a column. This is much faster and requires less disk
space than having a index on the whole field.
CREATE TABLE test (
name CHAR(200),
KEY name (last_name(10));
Multiple field indexes
MySQL can have one index on parts of different columns.
A multiple column key can be considered as a sorted array where the columns are concatenated.
Suppose that you have a table:
CREATE TABLE test (
id INT,
last_name CHAR(30),
first_name CHAR(30),
PRIMARY KEY (id),
KEY name (last_name,first_name));
Then the key ’name’ is a key over last_name and first_name.
The ’name’ key will be used in the following queries:
SELECT * FROM test WHERE last_name="Widenius";
SELECT * FROM test WHERE last_name="Widenius" AND first_name="Michael";
SELECT * FROM test WHERE last_name="Widenius" AND (first_name="Michael" OR
first_name="Monty");
SELECT * FROM test WHERE last_name="Widenius" and first_name >="M" and first_name <
"N";
The ’name’ key will NOT be used in the following queries:
SELECT * FROM test WHERE first_name="Michael";
SELECT * FROM test WHERE last_name="Widenius" or first_name="Michael";
Type mapping (to ease moving tabel definitions between different
12

databases engines)
To support easier use of code from different SQL vendors, MySQL does supports the following
mappings:
35
binary(num) char(num) binary
char varying varchar
float4 float
float8 double
int1 tinyint
int2 smallint
int3 mediumint
int4 int
int8 bigint
long varbinary blob
long varchar text
middleint mediumint
varbinary(num) varchar(num) binary
CREATE TABLE syntax.
CREATE TABLE table_name ( create_definition,... )
create_definition:
column_name type [DEFAULT default_value] [NOT NULL | NULL] [ PRIMARY KEY ]
[reference_definition]
or PRIMARY KEY ( key_column_name,... )
or KEY [key_name] KEY( key_column_name,...)
or INDEX [key_name] ( key_column_name,...)
or UNIQUE [key_name] ( key_column_name,...)
or FOREIGN KEY key_name ( key_column_name,...) [reference_definition]
or CHECK (expr)
key_column_name:
column_name [ (length) ]
reference_definition:
REFERENCES table_name [( key_coulmn_name,...)]
[ MATCH FULL | MATCH PARTIAL]
[ ON DELETE reference_option]
[ ON UPDATE reference_option]
reference_option:
RESTRICT | CASCADE | SET NULL | NO ACTION | SET DEFAULT
The FOREIGN KEY, CHECK and REFERENCE syntax are only for compatibility. They don’t
actually do anything.
If a column doesn’t have a DEFAULT value and is not declared as NOT NULL, the default
value is
NULL.
36
ZEROFILL means that number is pre-zeroed to maximal length. With INT(5) ZEROFILL a
value of 5 is retrieved as 00005.
BINARY means that the column will be compared case sensitive. The default is that all strings
are compared case insensitive. BINARY is ’sticky’ which means that if a column marked
’binary’ is used in a expression, the whole expression is compared ’binary’.
INDEX is only a synonym for KEY.
13

If one doesn’t assign a name to a key, the key will get the same name as the first key_column
with an optional _# to make it unique.
Key columns and timestamp columns can’t be NULL. For these columns the NULL attribute is
silently removed.
With column_name(length) syntax one can specify a key which is only a part of a string
column. This can make the index file much smaller.
A number column may have the additional attribute AUTO_INCREMENT to automatically get
the largest value+1 for each insert where column value is NULL or 0. See section How can I get
the unique ID for the last row? [p 83]
One can insert NULL for timestamp and auto_increment columns. This results in the current
time
/ the next number.
Blob columns can’t be keys.
When one groups on a blob only the first ’max_sort_length’ bytes are used See section
Limitations of BLOB and TEXT types [p 21] .
Deleted records are in a linked list and subsequent inserts will reuse old positions. To get smaller
files one can use the isamchk utility to reorganise tables.
Each null column takes 1 bit extra, rounded up to the nearest byte.
The maximum record length can be calculated as follows: 1+ sum_of_column_lengths +
null_columns/8 + number of variable length columns.
In some cases an attribute may silently change after creation: VARCHAR columns with a length
of
1 or 2 are changed to CHAR. When using one VARCHAR column all CHAR columns longer
than 2
are changed to VARCHAR’s.
On INSERT/UPDATE all strings (CHAR and VARCHAR) are silently chopped/padded to the
maximal length given by CREATE. All end spaces are also automatically removed. For example
VARCHAR(10) means that the column can contain strings with a length up to 10 characters.
Something/0 gives a NULL value.
The regular expression function (REGEXP and RLIKE) uses ISO8859-1 (Latin1) when deciding
the type of a character.
ALTER TABLE syntax
ALTER [IGNORE] TABLE table_name alter_specification [, alter_specification ...]
alter_specification:
ADD [COLUMN] create_definition
or CHANGE [COLUMN] old_column_name create_definition
or ALTER [COLUMN] column_name { SET DEFAULT literal | DROP DEFAULT }
or DROP [COLUMN] column_name
or DROP PRIMARY KEY
DROP INDEX key_name
ALTER TABLE works by creating a temporary table and copying all information to it and then
the old table is deleted and the new one is renamed. This is done in such a way that all updates
are automatically redirect to the new table without any failed updates. While the ALTER
TABLE
is working, the old table is readable for other clients. Table updates/writes to the table are stalled
37
and only executed after the new table is ready.
If IGNORE isn’t specified then the copy will be aborted and rolled back if there exists any
duplicated unique keys in the new table. This is a MySQL extension.
14

The CHANGE column_name, DROP column_name and DROP INDEX are MySQL
extensions to ANSI SQL92.
The optional word COLUMN is a pure noise word and can be omitted.
The ADD and CHANGE takes the same create_definition as CREATE TABLE. See section
CREATE TABLE syntax. [p 36] .
ALTER COLUMN sets a new default value or removes the old default value for a column.
DROP INDEX removes an index. This is an MySQL extension.
The FOREIGN KEY syntax in MySQL exists only for compatibility with other SQL vendors
CREATE TABLE commands: It doesn’t do anything. See section How standards compatible are
MySQL [p 19]
If one drops a column_name which is part of some key, this key part is removed. If all key parts
are removed then the key is removed.
DROP PRIMARY KEY drops the first UNIQUE key in the table.
CHANGE tries to convert data to the new format as good as possible.
With mysql_info(MYSQL*) one can retrieve how many records were copied and how many
records were deleted because of multiple keys.
To use ALTER TABLE one has to have select, insert, delete, update, create and drop privileges
on the table.
DROP TABLE syntax.
DROP TABLE table_name [, table_name....]
Removes one or more tables. All the data and the definition is removed so take it easy with this
command!
DELETE syntax.
DELETE FROM table_name WHERE where_definition
Returns records affected.
If one does a delete without a where clause then the table is recreated which is much faster than
doing
a delete for each row. In these cases, the SQL command returns zero as affected records.
Using DELETE FROM table_name will work even if the data files do not exist as only
information from the table definition file, table_name.frm, is used.
All string comparisons are case independent with case according to ISO-8859-1 Latin1.
LIKE is allowed on numerical columns.
Compare with explicit NULL (column == NULL) is the same as if IS NULL was used (column
IS NULL). This is is done to be consistent with mSQL.
38
SELECT syntax
SELECT [STRAIGHT_JOIN] [DISTINCT | ALL] select_expression,... [INTO
OUTFILE ’file_name’ ...] [ FROM tables... [WHERE where_definition ]
[GROUP BY column,...] [ ORDER BY column [ASC | DESC] ,..] HAVING
where_definition [LIMIT [offset,] rows] [PROCEDURE procedure_name]]
Strings are automatically converted to numbers and numbers to strings when needed (ala perl). If
in a
compare operation ((=, <>, <= ,<, >=, >)) either if the arguments are numerical the
arguments are compared as numbers, else the arguments are compared as strings. All string
comparisons are by default done case-independent by ISO8859-1 (The Scandinavian letter set
which
also works excellent with English).
select 1 > "6x"; -> 0
select 7 > "6x"; -> 1
15

select 0 > "x6"; -> 0
select 0 = "x6"; -> 1
A column name does not have to have a table prefix if the given column name is unique.
A select expression may be given an alias which will be column name and can be used when
sorting and grouping or in the HAVING clause.
select concat(last_name,’ ’,first_name) as name from table order by name
In LIKE expressions % and _ may be preceded with ’’ to skip the wild-card meaning.
A DATE is a string with one of the following syntaxes:
YYMMDD (Year is assumed to be 2000 if YY < 70.)
YYYYMMDD
YY.MM.DD Where ’.’ may be any non numerical separator.
YYYY.MM.DD Where ’.’ may be any non numerical separator.
IFNULL() and IF() returns number or string value according to use.
ORDER and GROUP columns may be given as column names, column alias or column numbers
in
SELECT clauses.
The HAVING clause can take any columns or alias in the select_expressions. It is applied last,
just before items are sent to the client, without any optimisation. Don’t use it for items that
should be in the WHERE clause. You can’t write (yet):
SELECT user,MAX(salary) FROM users GROUP BY users HAVING max(salary)>10
Change it to:
SELECT user,MAX(salary) AS sum FROM users GROUP BY users HAVING sum > 10
STRAIGHT_JOIN forces the optimiser to join the tables in the same order that the tables are
given in the FROM clause. One can use this to get a query to be done more quickly if the
optimiser joins the tables in not optimal order. See section EXPLAIN syntax. Get information
about a SELECT. [p 47]
LIMIT takes one or two numerical arguments.
If one argument, the argument indicates the maximum number of rows in a result.
If two arguments, the first argument says the offset to the first row to return, the second is
39
the maximum number of rows.
INTO OUTFILE ’filename’ writes the given set to a file. The file can not exist from before. See
section LOAD DATA INFILE syntax [p 45] .
Functions
A select_expression or where_definition can consist of any expression using the
following functions:
Group functions.
(
) Parenthesis
Normal mathematical operations.
+
-
*
/ A division by zero results in a NULL.
Bit functions. These have a range of maximum 64 bits because MySQL uses longlong arithmetic.
|
&
BIT_COUNT()
Number of set bits in an argument.
16

Normal logical. Returns TRUE (1) or FALSE (0).
NOT
!
OR
AND
Comparison operators. Returns TRUE (1) or FALSE (0). These functions work for both numbers
and
strings.
= Equal
<> Not equal. Synonym: !=
<= Smaller or equal.
< Smaller.
>= Bigger or equal
> Bigger
ISNULL(A)
Returns 1 if A is NULL else 0. Same as ’( A == NULL ’).
A BETWEEN B AND C
A is bigger or equal as B and A is smaller or equal to C. Is the same as (A >= B AND A <= C).
40
String comparison functions.
expr IN (value,...)
Returns 1 if expr is any of the values in the IN list, else it returns 0. If all values are constants,
then all values are evaluated according to the type of expr and sorted. The search for them item is
then done by using a binary search. This means IN is very quick when used with constants in the
IN part.
expr LIKE expr
SQL simple regular expression comparison. Returns TRUE (1) or FALSE (0). With LIKE you
have two wild characters. ’%’ stands for any number of characters, even zero characters. ’_’
stands exactly one character. If you want to search after a ’%’ or a ’_’ you must precede it with
a ’’ For example the string "Monty was here" is matched by "Monty%", "%Monty ___ h%"
and "%was%".
expr NOT LIKE expr
Returns TRUE (1) or FALSE (0).
expr REGEXP expr
Checks string against extended regular expr.
expr NOT REGEXP expr
Checks string against extended regular expr.
STRCMP()
Returns 0 if the strings are the same. Otherwise return -1 if first argument is smaller according to
sort-order, otherwise it returns 1.
Control flow functions.
IFNULL(A,B)
If A is not null it returns A, else B.
IF(A,B,C)
If A is true (<> 0 and <> NULL) then return B, else return C.
Mathematical functions. These returns NULL in the case of a error.
- Sign
ABS()
SIGN()
17

Sign of argument. Returns -1, 0 or 1.
MOD()
% Module (like in C). Equivalent with MOD().
FLOOR()
CEILING()
ROUND()
EXP()
LOG()
LOG10()
POW()
SQRT()
PI()
COS()
SIN()
41
TAN()
TAN2()
ACOS()
ASIN()
ATAN()
RAND([integer_expr])
Returns a random float, 0 <= x <= 1.0, using integer_expr as the option seed value.
MIN()
Min value of arguments. Must have 2 or more arguments, else these are GROUP BY functions.
The arguments are compared as numbers.
MAX()
Max value of arguments. Must have 2 or more arguments, else these are GROUP BY functions.
The arguments are compared as numbers.
String functions.
ASCII(string_exp)
Returns the ASCII code value of the leftmost character of string_exp. If string_exp is NULL, it
then returns NULL.
CHAR(num,...)
Returns a string that consist of the characters given by the ASCII code values of the arguments.
NULLS are skipped.
CONCAT()
Concatenates strings. May have more than 2 arguments.
LENGTH()
Length of string. Synomyes: CHAR_LENGTH(), OCTET_LENGTH()
LOCATE(A,B)
Returns position of B substring in A. The first position is 1. Returns 9 if b is not in A.
LOCATE(A,B,C)
Returns position of B substring in A starting at C. Synonym: POSITION(B IN A)
LEFT(str,length)
Gets length characters from beginning of string.
RIGHT(str,right)
Gets length characters from end of string. Synonym: SUBSTRING(A FROM B)
LTRIM(str)
Removes space characters from the beginning of string.
18

RTRIM(str)
Removes space characters from the end of string.
TRIM([[ BOTH | LEADING | TRAILING] [ A ] FROM ] B)
Returns a character string with all A prefixes and/or suffixes removed from B. If BOTH,
LEADING and TRAILING isn’t used BOTH are assumed. If A is not given, then spaces are
removed.
SOUNDEX(str)
Gets a soundex string from str. Two strings that sound ’about the same’ should have identical
soundex strings. A ’standard’ soundex string is 4 long, but this function returns an arbitrary long
string. One can use SUBSTRING on the result to get a ’standard’ soundex string. All non alpha
characters are ignored in the given string. All characters outside the A-Z range are treated as
vocals.
SUBSTRING(A,B,C)
Returns substring from A starting at B with C chars.
42
SUBSTRING(A FROM B FOR C)
Same as SUBSTRING(A,B,C). This is ANSI SQL 92.
SUBSTRING(A FROM B)
Same as RIGHT(A,B). This is ANSI SQL 92.
SUBSTRING_INDEX(string,delimiter,count)
Returns the substring from ’string’ after ’count’ delimiters. If count is positive the strings are
searched from left else if count is negative the substrings are searched and returned from right.
substring("www.tcx.se",".",2) would return "www.tcx" and
substring("www.tcx.se",".",-2) would return "tcx.se".
REPLACE(A,B,C)
Replaces all occurrences of B in A with C.
REPEAT(string,count)
Repeats string count times. If count <= 0 returns a empty string. If string or count is NULL or
LENGTH(string)*count > max_allowed_size returns NULL.
INSERT(org,start,length,new)
Replaces substring org[start...length] with new. First position in string=1.
LCASE(A)
Changes A to lower case. Synonym: LOWER()
UCASE(A)
Changes a to upper case. Synonym: UPPER()
ENUM(A,a,b,c,d)
Returns 1 if A == a, 2 if A == b. If no match 0 is returned. A,a,b,c,d... are strings.
ELT(N,a,b,c,d,...)
Returns a if N == 1, b if N == 2. a,b,c,d are strings.
FIELD(A,a,b,c,d,...)
Returns index of A in the a,b,c... list. The complement of ELT().
Date and time functions.
PERIOD_ADD(P:N)
Adds N months to period P (of type YYMM or YYYYMM). Returns YYYYMM.
PERIOD_DIFF(A,B)
Returns months between periods A,B. A and B should be of format YYMM or YYYYMM.
Example: PERIOD_DIFF(9612,199712).
TO_DAYS(DATE)

19

Changes a DATE to a daynumber. DATE may be a date string, a datetime string, a timestamp([6
|
8 | 14]) or a number of format YYMMDD or YYYYMMDD.
FROM_DAYS()
Changes a daynumber to a DATE.
WEEKDAY(DATE)
Gets weekday for date (0 = Monday, 1 = Tuesday). DATE may be a date string, a datetime
string,
a timestamp([6 | 8 | 14]) or a number of format YYMMDD or YYYYMMDD.
CURDATE()
Returns date of today. In form YYYYMMDD or "YYYY-MM-DD" depending on whether
CURDATE() is used in a number or string context. Synonymes: CURRENT_DATE().
CURTIME()
Returns current time in form HHMMSS or "HH:MM:SS", depending on whether CURTIME() is
used in a number or string context. Synonymes: CURRENT_TIME().
43
NOW()
Returns current time. In format YYYYMMDDHHMMSS or "YYYY-MM-DD HH:MM:SS"
depending on whether NOW() is used in a number or string context. Synonymes:
CURENT_TIMESTAMP(), SYSDATE().
UNIX_TIMESTAMP([DATE])
If called without any arguments a unix timestamp (seconds in GMT since 1970.01.01 00:00:00)
is returned. Normally it is called with a timestamp column as an argument in which case it
returns the timestamp. DATE may also be a date string, a datetime string, or a number of format
YYMMDD or YYYMMDD in local time.
FROM_UNIXTIME(unix_timestamp)
Returns a string of the timestamp in YYYY-MM-DD HH:MM:SS or YYYYMMDDHHMMSS
format depending on context.
SEC_TO_TIME(seconds)
Returns the hours, minutes and seconds of the argument in H:MM:SS or HMMSS format
depending on context.
Miscellaneous functions.
DATABASE()
Returns current database name.
USER()
Returns current user name. Synonymes: SESSION_USER() and SYSTEM_USER().
PASSWORD()
Calculates a password string. This must be used to store a password in the ’user’ grant table.
ENCRYPT(string[,salt])
Crypt messsage with the unix crypt() command. The salt should be a string with 2 characters.
LAST_INSERT_ID()
Returns the last automaticly generated value that was set in a auto_increment column. See
section
How can I get the unique ID for the last row? [p 83]
FORMAT(nr,NUM)
Formats number to a format like ’#,###,###.##’ with NUM decimals.
Functions for GROUP BY clause.
count(expr)
Number of rows. count(*) is optimised to return very quickly if no other column is used in the
20

SELECT.
avg(expr)
Average value of expr.
min(expr)
max(expr)
Minimum/Maximum value of expr. min() and max() may take a string argument and will then
return the minimum/maximum string value.
sum(expr)
Sum of expr.
std(expr)
Standard derivation of expression. This is a extension to ANSI SQL.
bit_or(expr)
The logical OR of all bits in expr. Caclulation done with 64 bits precision.
44
bit_and(expr)
The logical AND of all bits in expr. Caclulation done with 64 bits precision.
MySQL has extended the use of GROUP BY. You can use columns or calculations in the
SELECT
expressions which doesn’t appear in the GROUP BY part. This stands for ’any possible value for
this
group’. By using this, one can get a higher performance by avoiding sorting and grouping on
unnecessary items. For example, in the following query one doesn’t need to sort on b.name:
SELECT a.id,b.name,COUNT(*) from a,b WHERE a.id=b.id GROUP BY a.id
INSERT syntax
INSERT INTO table [ (column_name,...) ] VALUES (expression,...)
or INSERT INTO table [ (column_name,...) ] SELECT ....
An expression may use any previous column in column_name list (or table if no column name
list is
given).
The following holds for a multi-row INSERT statement:
The query cannot contain an ORDER BY clause.
The target table of the INSERT statement cannot appear in the FROM clause of the query.
If one uses INSERT INTO ... SELECT ... then one can get the following info string with
the C API function mysql_info(). @result{Records: 100 Duplicates: 0 Warnings: 0}
Duplicates are number of rows which couldn’t be written because some key would be duplicated.
Warnings are columns which were set to NULL, but have been declared NOT NULL. These will
be set to their default value.
If one sets a time stamp value to anything other than NULL, the time stamp value will be copied
to the result table.
Auto increment fields works as usual.
LOAD DATA INFILE syntax
LOAD DATA INFILE ’table_interval.text’ [REPLACE | IGNORE] INTO TABLE
table_interval [FIELDS [TERMINATED BY ’,’ [OPTIONALLY] ENCLOSED BY
’"’ ESCAPED BY ’’ ]] [LINES TERMINATED BY ’n’] [(field list)]
This is used to read rows from a text file, located on the server, at a very high speed. The server-
client
protocol doesn’t yet support files over a connection. If you only have the file on the client, use
rcp or
ftp to copy it, possibly compressed, to the server before using LOAD DATA INFILE.
21

To write data to a text file, use the SELECT ... INTO OUTFILE ’interval.txt’ fields
terminated by ’,’ enclosed by ’"’ escaped by ’’ lines terminated
by ’n’ FROM .... syntax.
Normally you don’t have to specify any of the text file type options. The default is a compact
text file
with columns separated with tab characters and all rows end with a newline. Tabs, newlines and
inside fields are prefixed with a . NULL’s are read and written as N.
45
FIELDS TERMINATED BY has the default value of t.
FIELDS [OPTIONALLY] ENCLOSED BY has the default value of ".
FIELDS ESCAPED BY has the default value of ’’.
LINES TERMINATED BY has the default value of ’n’.
FIELDS TERMINATED BY and LINES TERMINATED BY may be more than 1 character.
If LINES TERMINATED BY is an empty string and FIELDS TERMINATED BY is not empty
then
lines are also terminated with FIELDS TERMINATED BY.
If FIELDS TERMINATED BY and FIELDS ENCLOSED BY both are empty strings (") then
this
gives a fixed row format ("not delimited" import format). With a fixed row size NULL values
are
outputed as a blank string. If you specify OPTIONALLY in ENCLOSED BY, then only strings
are
enclosed in ENCLOSED BY by the SELECT ... INTO statement.
Duplicated ENCLOSED BY chars are removed from strings that start with ENCLOSED BY. For
example: With ENCLOSED BY ’"’:
"The ""BIG"" boss" -> The "BIG" boss
The "BIG" boss -> The "BIG" boss
If ESCAPED BY is not empty then the following characters will be prefixed with the escape
character: ESCAPED BY, ASCII 0, and the first character in any of FIELDS TERMINATED
BY,
FIELDS ENCLOSED BY and LINES TERMINATED BY.
If FIELDS ENCLOSED BY is not empty then NULL is read as a NULL value. If FIELDS
ESCAPED
BY is not empty then N is also read as a NULL value.
If REPLACE is used, then the new row will replace all rows which have the same unique key. If
IGNORE is used, the row will then be skipped if there already exist a record with an identical
unique
key. If none of the above options are used an error will be issued. The rest of the text file will be
ignored if one gets a duplicate key error.
Some possible cases that’s not supported by LOAD DATA:
Fixed size rows (FIELDS TERMINATED BY and FIELDS ENCLOSED BY both are empty)
and BLOB fields.
If some of the separators are a prefix of another.
FIELDS ESCAPED BY is empty and the data contains LINES TERMINATED BY or FIELDS
ENCLOSED BY followed by FIELDS TERMINATED BY.
All rows are read into the table. If a row has too few fields the rest of the fields are set to default
values.
For security reasons the text file must either reside in the database directory or be readable by all.

22

For more information about the escaped syntax See section Literals. How do you write strings
and
numbers? [p 30] .
46
When the LOAD DATA query is done, one can get the following info string with the C API
function
mysql_info().
@result{Records: 1 Deleted: 0 Skipped: 0 Warnings: 0}
Warnings are incremented for each column which can’t be stored without loss of precision, for
each
column which didn’t get a value from the read text line (happens if the line is too short) and for
each
line which has more data than can fit into the given columns.
Example that loads all fields;
LOAD DATA INFILE ’persondata.text’ INTO TABLE persondata;
See section How should I arrange my table to be as fast/small as possible? [p 59]
UPDATE syntax
UPDATE table SET column=expression,... WHERE where_definition
All updates are done from left to right. If one accesses a column in the expression update will
then use
the current value (a given value or the default value) of the column.
UPDATE persondata SET count=count+1
SHOW syntax. Get information about names of columns.
SHOW DATABASES [LIKE wild]
or SHOW TABLES [FROM database] [LIKE wild]
or SHOW COLUMNS FROM table [FROM database] [LIKE wild]
Gives information about databases, tables or columns. The wild is a LIKE string. FIELDS may
be
used as an alias for COLUMNS.
EXPLAIN syntax. Get information about a SELECT.
EXPLAIN SELECT select_options
Gives information about how and in which order tables are joined. With the help of EXPLAIN
one can
see when one has to add more keys to tables to get a faster select that uses keys to find the
records.
You can also see if the optimiser joins the tables in an optimal order. One can force the optimiser
to
use a specific join order with the STRAIGHT_JOIN option to select.
The different join types are:
system
The table has only 1 record (= system table)
const
The table has at most 1 matching record which will be read at the start of the query. All columns
in this table will be regarded as constants by the rest of the optimiser.
eq_ref
One record will be read from this table for each combination of the previous tables.
47
ref
All records with matching keys will be read from this table for each combination of the previous
23

tables.
all
A full table scan will be done for each combination of the previous tables.
DESCRIBE syntax. Get information about columns.
(DESCRIBE | DESC) table [column]
Gives information about columns. This command is for Oracle compatibility. See section SHOW
syntax. Get information about names of columns. [p 47] Column may be a column name or a
string.
Strings may contain wild cards.
Lock tables syntax
LOCK TABLES table_name [AS alias] READ|WRITE [, table_name READ|WRITE ] ...
UNLOCK TABLES
Locks tables for this thread. Many threads may have a READ lock on the same table, but one can
use
a table with a WRITE lock. One can’t update a table on which one has a read LOCK. When one
uses
LOCK TABLES one must lock all tables one is going to use!
LOCK TABLES t READ, t as t2 READ
SELECT * from t,t2;
All tables are automaticly unlocked when one issues another LOCK TABLES or if the
connection to
the server is closed.
SET OPTION syntax.
SET OPTION SQL_VALUE_OPTION=value, ...
The used options remain in effect for the whole current session.
The different options are:
SQL_SELECT_LIMIT=value
The maximum number of records to return in any select. If a select has a limit clause it overrides
this statement. The default value for a new connection is ’unlimited’.
SQL_BIG_TABLES= 0 | 1
If set to 1 then all temporary tables are stored on disk based. This will be a little slower, but one
will not get the error The table ### is full anymore for big selects that require big
temporary tables. The default value for a new connection is 0 (use in memory temporary tables).
SQL_BIG_SELECTS= 0 | 1
If set to 1 then MySQL will aborts if one will try to make a select with probably will take a very
long time. This is nice to SELECTS with a wrong WHERE statement. A big query is defined as
a
SELECT that will probably have to examine more than max_join_size rows. The default
value for a new connection is 0 (allow all SELECT’s).
48
GRANT syntax. (Compatibility function).
GRANT (ALL PRIVILEGES | (SELECT, INSERT, UPDATE, DELETE,
REFERENCES (column list), USAGE))
ON table TO user,... [WITH GRANT OPTION]
This command doesn’t do anything. It is only in MySQL for compatibility reasons. Privileges in
MySQL is handled with the mysql grant tables. See section How does the privilege system work?
[p
26]

24

CREATE INDEX syntax (Compatibility function).
CREATE [UNIQUE] INDEX index_name ON table_name ( column_name,... )
This function doesn’t do anything. It is only in MySQL for compatibility reasons. You can create
new
index with ALTER TABLE. See section ALTER TABLE syntax [p 37]
DROP INDEX syntax (Compatibility function).
DROP INDEX index_name
This always succeeds. You can drop an index with ALTER TABLE. See section ALTER
TABLE
syntax [p 37]
Is MySQL picky about reserved words?
It seems that I can’t create a table with column names timestamp or group.
Functions don’t clash with table or column names. (For example ABS is an allowed column
name).
The only restriction is that space is not allowed between a function name and the ’(’ when using
functions.
The following are reserved words in MySQL. Most of them (for example) group, are forbidden
by
ANSI SQL92 as column and/or table names. A few are because MySQL needs them and is
(currently)
using a Yacc parser:
49
action add all alter and
as asc auto_increment between bigint
bit binary blob both by
cascade char character change check
column columns create data databases
date datetime dec decimal default
delete desc describe distinct double
drop escaped enclosed enum explain
fields float float4 float8 foreign
from for full grant group
having ignore in index infile
insert int integer interval int1
int2 int3 int4 int8 into
is key keys leading like
lines limit lock load long
longblob longtext match mediumblob mediumtext
mediumint middleint numeric no not
null on option optionally or
order outfile partial precision primary
procedure privileges read real references
regexp repeat replace restrict rlike
select set show smallint sql_big_tables
sql_big_selects sql_select_limit straight_join table tables
terminated text time timestamp tinyblob
tinytext tinyint trailing to unique
unlock unsigned update usage values
varchar varying with write where
25

zerofill
The following symbols (from the table above) are disallowed by ANSI SQL but allowed by
MySQL
as column/table names. This is because some of theese names are very natural names so a lot of
people has already used them.
ACTION
BIT
DATE
50
ENUM
NO
TEXT
TIME
TIMESTAMP
51
How safe/stable is MySQL
How stable is MySQL?
At TcX, MySQL has worked without any problems in our projects since nid 1996. When
released to a
wider public we noticed that there where some parts of ’untested code’ in MySQL that was
quickly
found by the new user group that don’t do queries exactly like we do at TcX. Each new release
have
had fewer non portability problems than the previous one, even though they have had a lot of
new
features, and we hope that one of the next releases will be possible to be labelled ’stable’.
Each release of MySQL has been usable and there has only been problems when users start to
use
code from ’the gray zones’. Naturally, outdoor users can’t know what the gray zones are and I
hope
this section will clarify those currently known.
We will here try to answer some of the more important questions that seems to concern a lot of
people
and clarify some issues. This section has been put together from the information that has come
forward in the mailing list which is very active in reporting bugs.
How stable is MySQL? Can I depend on MySQL in this project!
This is about the 3.21.x version of MySQL. All known and reported bugs are fixed in the latest
version with the exception of the bugs listed in the BUGS file which are things that are ’design’
related.
MySQL is written in multiple layers and different independent modules. Here is a list of the
different
modules and how tested each of them are.
The ISAM table handler. Stable
(ISAM) where all data is stored. In all MySQL releases there hasn’t been a single (reported) bug
in this code. The only known way to get a corrupted table is to kill the server in the middle of a
update and because all data is flushed to disk between each query even this is unlikely to destroy

26

any data beyond rescue. There hasn’t been a single bug report about lost data because of bugs in
MySQL either.
The parser and lexical analyser. Stable
There hasn’t been a single reported bug in this system for a couple of months.
The C client code. Stable
No known problems. In early 3.20 releases there were some limitations in the send/receive buffer
size. In 3.21.x the send/receive buffer is now dynamic up to a default of 512K.
mysql, mysqladmin and mysqlshow. Stable
The command line clients have had very few bugs.
mysqldump and mysqlimport. Beta
Rewritten for 3.21.
Basic SQL. Stable
The basic SQL function system and string classes and dynamic memory handling. Not a single
reported bug on this system.
52
Query optimiser. Gamma
Some changes in 3.21.
Range optimiser. Alpha
Totally rewritten for 3.21.x
Join optimiser. Gamma
Small changes for 3.21.
GROUP BY, ORDER BY and related function COUNT(). Beta
Rewritten for 3.21 and throughly tested.
Locking. Gamma
This is very system dependent. One some system there is big problems to use standard OS
locking (fcntl). In these case one should run the MySQL demon with the --skip-locking flag.
Known problems are some Linux systems and SUNOS when using NFS mounted file systems.
Linux threads. Gamma
The only found problem is with the fcntl() call, which is fixed by using --skip-locking. Some
people have reported lockup problems with the 0.5 release.
Solaris 2.5+ pthreads. Stable
We use this for all our production work.
MIT threads (Other systems). Beta
No reported bugs since 3.20.15 and no known bugs since 3.20.16. On some system there is
’misfeature’ where some operations are quote slow (a 1/20 second sleep is done between each
query).
Other thread implementions. Alpha
The ports to other systems is very new and may still have bugs, either in MySQL or most often
in the thread implementation itself.
LOAD DATA..., INSERT ... SELECT. Stable
Some people have thought they have found bugs in this but has turned up being
misunderstandings. So check the manual before reporting bugs!
ALTER TABLE. Gamma
Partly rewritten for 3.21.
mysqlperl. Stable
No bugs reported except a lot of compiling and linking problems.
DBD. Beta
Now maintained by Jochen Wiedmann . Thanks!
27

mysqlaccess. Beta
Written and maintained by Yves.Carlier@rug.ac.be. Thanks!
The Technical Documentation. Beta
It is improving.
MyODBC (uses ODBC SDK 2.5). Beta
It seams to work well with some programs. There is some known problems with Office97.
PowerBuilder doesn’t work yet. Error messages aren’t returned right when using a Microsoft
New ODBC 3.0 driver but this may well be a bug in the Microsoft ODBC 3.0 driver. It isn’t fully
up to ODBC level 2 yet in every regard. (My personal opinion is that the ODBC levels are
defined in such a way that it’s almost impossible to make a simple ODBC driver that works with
all products). Sometimes the MyODBC code must be ported to ODBC 3.0.
TcX provides email support for paying customers, but the MySQL mailing list usually provides
answers to all common questions. Bugs are usually fixed right away with a patch that usually
works
and for serious bugs there is almost always a new release.
53
Why are there is so many release of MySQL? Is it because
there are so many bugs?
Well, MySQL is evolving quite rapidly here at TcX and we want to share this with other MySQL
users. We try to make a release when we have a very useful feature that others seem to have a
need
for.
We also try to help out users who request features that are easy to implement. We also take note
on
what our licensed users want to have and we especially take notes on what our extended email
supported customers want and try to help them out.
No one has to download a new release. The NEWS file will tell you if the new release has
something
you really want.
If there is, by any chance, a fatal bug in a release we will make a new release as soon as possible,
that
fixes the problem. We would like other companies to do this too :)
The 3.21.x version has had many big changes for portability to a lot of different systems. When
the
3.21 release is stable we will remove the alpha/beta suffix and move active development to the
3.22.
Bugs will still be fixed in the stable version. We don’t believe in a complete freeze, as this also
leaves
out bug fixes and things that ’must be done’. ’Somewhat freezed’ means that we will maybe add
small
things that ’almost surely will not affect anything thats already working’.
Checking a table for errors.
If MySQL crashed (for example if the computer is turned off) when all data is not written to disk
the
tables may have become corrupted. To check a table use:
isamchk table_name
This finds 99.99 % of all errors. What it can’t find is when only the data file has been corrupted.
isamchk -e table_name

28

This goes through all data and does a complete check.
isamchk -ei table_name
As the above but it also prints some statistics.
We at TcX run a cron job on all our important tables once a week.
35 0 * * 0 /path/to/isamchk -s /path/to/dbs/*/*.ISM
This prints out any crashed tables so we can go and examine and repair them when needed.
As we haven’t had any unexpected crashed (without hardware trouble) tables for a couple of
years
now (this is really true), once a week is more than enough of us.
Of course, whenever the machine has done a reboot in the middle of a update one usually has to
check
all the tables that could have been affected. (This is a ’expected crashed table’)
We recommend that to start with one should do a isamchk -s on all updated tables each night
until
one gets to trust MySQL as much as we trust it.
54
Naturally, one could add a check to safe_mysql that, if there is an old pid file left after a reboot,
it
should check all tables that have been modified the last 24 hours.
How to repair tables.
The file format that MySQL uses to store data has been extensively tested, but there is always
instances (like a hard kill on the mysqld process in the middle of a write, a hardware error or a
unexpected shutdown of the computer) when some tables may be corrupted.
The sign of a corrupted table is usually when queries abort unexpectedly and one gets errors like:
table.frm is locked against change.
Can’t find file ’table.ISM’ (Errcode: ###)
Got error ### from table handler. (Error 135 is an exception in this case)
Unexpected end of file.
Record file is crashed.
In thess cases you have to repair your tables. The isamchk extern utility can usually detect and
fix
most things that can go wrong. See section The MySQL table check, optimize and repair
program [p
66] .
If you are going to use isamchk on very large files, you should first decide how much memory
you
want to give to isamchk. More memory gives more speed. For example, if you have more than
32M
ram, try:
isamchk -O sortbuffer=16M -O keybuffer=16M -O readbuffer=1M
-O writebuffer=1M .....
Part 1; Checking
Check the permissions of the table files. Make them readable for the user running mysqld.
cd to the database directory.
Run ’isamchk *.ISM’ or (’isamchk -e *.ISM’ if you have more time).
You only have to repair those tables that your isamchk gives an error. Use option -s to avoid
unnecessary information.
Part 2; Easy safe repair.
If you get weird errors when checking or repairing, like out of memory or if isamchk
29

crashes, go to part 3.
Try first ’isamchk -rq table’. This will try to repair the .ISM file without touching the
important data table (.ISD). If the data file (.ISD) contains everything and the delete links
point at the right places in the data file, this should work and the table is fixed. Start
repairing next table.
Make a backup of the data file (table.ISD) before continuing.
Use ’isamchk -r table’. This will remove wrong records and deleted records from the data
file and reconstruct the index (.ISM) file.
If the above fails, use ’isamchk -ro table’. This is a little slower but a crash-safer version of
the above.
Part 3; Hard repair.
This should only happen if the first 16K block in the .ISM file is destroyed, contains wrong
information or if the .ISM is missing.
In this case we have to create a new .ISM file. Do as follows:
55
Move the data file .ISD file to some safe place.
Recreate the .ISM file from the .frm file:
shell> mysql database
mysql> delete from table_name;
mysql> quite
Copy (don’t move) the data file (.ISD) back on the newly created .ISD file
Go back to Part 2. (This shouldn’t be a endless loop). isamchk -rq should now work.
Part 4; Very hard repair.
This can only happen if the descriptor file (.frm) also has crashed. This should never happen,
because the .frm file isn’t written to after the table is created.
Restore the .frm file from a backup and go back to Part 3. You can also restore the .ISM file
and go back to Part 2. In the latter case you should start with ’isamchk -r’.
If you don’t have a backup but know exactly how the table was created, create a copy of the
table in another database and copy the .frm and .ISM file from there to your crashed
database and go back to Part 2;
Is there anything special to do when upgrading/downgrading
MySQL?
One can always move the MySQL form and data files between different versions on the same
architecture as long as one has the same base version of MySQL. The current base version is of
course
3. If one changes the sort order by recompiling MySQL one has to do a isamchk -rq on the tables.
56
How to get maximum performance out of MySQL
How does MySQL use memory ?
You can get the currently used buffer sizes with:
> ./mysqld --help
This should result in a list of all mysqld options and configurable variables like the following.
See
section How can I change the buffer sizes of mysqld ? [p 61] .
Possibly variables to option --set-variable (-O) are:
back_log current value: 5
join_buffer current value: 131072
30

key_buffer current value: 1048568
max_allowed_packet current value: 65536
max_connections current value: 90
max_join_size current value: 4294967295
max_sort_length current value: 1024
net_buffer current value: 8192
record_buffer current value: 131072
table_cache current value: 64
tmp_table_size current value: 131072
sort_buffer current value: 2097144
All threads share the same base memory.
Nothing is memmaped yet (except compressed tables but that’s another story). This is because
32bit memory space of 4GB is not large enough for most large tables. When we get a system
with 64 bit address space we may add general support for memmaping.
When starting mysqld one can specify a key buffer. This will buffer all keys in all tables at
FIFO basis (variable keybuffer).
Each connection uses some thread space, a stack and a connection buffer (variable
net_buffer_length).
Each request doing a sequential scan over a table allocates a read buffer (variable record_buffer).
Each request doing a sort, allocates a sortbuffer and one or two temporary files. The maximum
extra disk-space needed is @math{(sort_key_length +sizeof(long))*2}.
All joins are done in one pass and most temporary tables are memory based (HEAP) tables.
Temporary tables with a big recordlength (= sum of all column length) or that contains BLOB’s
are stored on disk. One current problem is that if the HEAP table exceeds the size of
tmp_table_size, one will get the error: ’The table ### is full’. In the future we will fix this
by automatically changing the in memory (HEAP) table to a disk based (NISAM) table if
needed.
To go around this problem one can increase the -O tmp_table_size=# option to mysqld or
use the SQL option SQL_BIG_TABLES. See section SET OPTION syntax. [p 48] . In MySQL
3.20 the maximum size of the temporary table was recordbuffer*16, so if you are using this
version you have to increase recordbuffer. There also exists a patch to always store
temporary tables on disk, but this will affect the speed of all complicated queries.
Almost all memory used when parsing and calculating is done on a local memory store. No
memory overhead is needed for small items and the normal slow memory allocation/freeing is
avoided. Memory is only allocated for unexpectedly large strings (this is done with malloc/free).
57
Each index file is opened once and the data file is opened once for each concurrently running
thread. For each concurrent thread a table structure, column structures for each column and a
buffer that has the size of 3 * (maximum row length not counting BLOBS) is allocated. A BLOB
uses 5 to 8 bytes +length of blob data.
For each table having BLOBS, a buffer is enlarged dynamically to read in larger BLOB’s. If one
scans a table there will be a allocated buffer as large as the largest BLOB.
All used tables are saved in a cache and used by FIFO. Normally the cache is 64 tables. If a table
has been used by 2 running threads at the same time, there will be 2 entries of the table in the
cache.
A MySQLadmin refresh closes all tables that are not used and marks all used tables to be
closed when the running thread finishes. This will effectively free most used memory.
When running mysqld ps, other programs may report that is takes a lot of memory. This may be

31

caused by thread-stacks on different memory addresses. For example, the Solaris ps calculates
the
unused memory between stacks as used memory. You can verify this by checking available swap
with
’swap -s’. We have tested mysqld with commercial memory-leakage detectors so there should
not be
any memory leaks.
How does MySQL use keys?
All keys, PRIMARY, UNIQUE and INDEX(), are stored in B trees. Strings are automatically
prefixand
end-space compressed.
INDEX(col1, col2) creates a multiple index over the two columns. The key can be seen like a
concatenation of the given columns. If you use INDEX(col1), INDEX(col2) instead of
INDEX(col1,col2) you get two separate keys instead.
SELECT * FROM table WHERE col1=# AND col2=#
In a case of INDEX(col1,col2) the right row(s) can be fetched directly. In a case of
INDEX(col1),
INDEX(col2) the optimiser decides which index will find fewer rows and this index will be used
to
fetch the rows.
If the table has an index INDEX(col1,col2,col3....) the prefix of this can be used by the optimiser
to
find the rows. This means that the above gives you search capabilities on: INDEX(col1) and
INDEX(col1,col2) and INDEX(col1,col2,col3)...
MySQL can’t use a sub part of a key to locate rows through a key.
With the definition INDEX(col1,col2,col3):
SELECT * FROM table WHERE col1=#
SELECT * FROM table WHERE col2=#
SELECT * FROM table WHERE col2=# and col3=#
only the first query will use keys.
58
How does MySQL open & close tables?
The cache of open tables can grow to a maximum of table-cache (default 64, changeable with -O
table_cache=#). A table is never closed, except when the cache is full and another thread tries to
open
a table or if one uses ’mysqladmin refresh’.
When the limit is reached, MySQL closes as many tables as possible, until the cache size has
been
reached or there are no more unused tables. This means that if all tables are in use by some
threads,
there will be more open tables than the cache limit, but the extra tables will be closed eventually.
Tables are closed according to last-used order.
A table is opened (again) for each concurrent access. This means that if one has two threads
running
on the same table or access the table twice in the same query (with AS) the table needs to be
opened
twice. The first open of any table takes 2 file descriptors, each following use of the table takes
only
one file descriptor.
32

How should I arrange my table to be as fast/small as possible?
Use NOT NULL if possible. It makes everything faster and you save 1 bit per column.
All columns have default values. Only insert when the default value isn’t acceptable. You don’t
have to insert the columns of a timestamp or an autoincremented key in the insert statement. See
section How can I get the unique ID for the last row? [p 83]
Use the smaller INT types if possible to get smaller tables. For example, MEDIUMINT is often
better than INT.
If you don’t have any VARCHAR columns, a fixed size record format will be used. This is much
faster but may unfortunately waste some space. See section What are the different row formats?
Or when to use VARCHAR/CHAR? [p 63] .
To make MySQL optimise queries better, run isamchk -a on the table once it is loaded with
relevant data. This updates a value for each index that tells how many rows that have the same
value for this key in average. Of course, this is always 1 for unique indexes.
To sort an index and data according to an index use isamchk -Sir1 (if you want so sort on
index 1). If you have a unique key that you want to read all records from in numeric order this is
a good way to make that faster.
When loading a table with data use LOAD DATA FROM INFILE. This is usually 20 times
faster than using a lot of INSERTs. If the text file isn’t on the server, rcp it to the server first. See
section LOAD DATA INFILE syntax [p 45] . You can even get more speed when loading data to
tables with many keys by doing:
Create the table in mysql or perl with CREATE TABLE....
Do a mysqladmin refresh.
Use isamchk -k0 database/table_name. This will remove all use of keys from
the table.
Insert data into the table with LOAD DATA INFILE....
If you have pack_isam and want to compress the table, run pack_isam on it.
Recreate the keys with isamchk -rq database/table_name.
Do mysqladmin refresh.
The other possibility to get some more speed to both LOAD DATA FROM INFILE and
INSERT is to enlarge the key buffer. This can be done with the -O key_buffer=# option to
(safe)mysqld. For example 16M should be a good value if you have much RAM :)
When dumping data as textfiles to other programs, use SELECT ... INTO OUTFILE. See
59
section LOAD DATA INFILE syntax [p 45] .
When doing many inserts/updates in a row, you can get more speed by using LOCK TABLES on
the tables. ...FROM INFILE... and ...INTO OUTFILE... are atomic so you don’t have
to use LOCK TABLES when using these. See section Lock tables syntax [p 48] .
To check how you are doing, run isamchk -evi on the .ISM file.
What affects the speed of the INSERT statement?
The time to insert a record consists of:
Connect: (3)
Sending query to server: (2)
Parsing query: (2)
Inserting record: (1 x size of record)
Inserting keys: (1 x keys)
Close: (1)
Where (number) is proportional time. This does not take into calculation the initial overhead to
open
tables (which is done once for each simultaneous running query).
33

The size of the table slows down the insert of keys with N log N (B-trees).
What affects the speed of DELETE statement?
The delete time of a record is exactly proportional to the number of keys. To increase the seed of
deletes you can increase the size of the key cache. The default key cache is 1M and to get faster
deletes it should be increased a couple of times (try 16M if you have enough memory).
What kind of optimisation is done on the WHERE clause?
(Incomplete, MySQL does a lot of optimisations.)
Brace removal (all unnecessary braces are removed). ((a AND b) AND c OR (((a AND
b) AND (c AND d)))) -> (a AND b) OR (a AND b AND c AND d)
Constant folding. (a<b AND b=c) AND a=5 -> b>5 AND b=c A=5
Constant condition removal (needed because of constant folding). (b>=5 AND b=5) OR
(b=6 and 5=5) or (B=7 and 5=6) -> B=5 or B=6
All constant expressions used by keys are evaluated only once.
Return zero rows if impossible select.
HAVING is merged with WHERE if one doesn’t use GROUP BY or group functions.
For each sub join a simpler WHERE is constructed to get a fast WHERE evaluation for each sub
join and also to skip records as soon as possible.
Find all keys that may be used. Use the key which finds least records. A key is used for the
following cases: =, >, >=, <, <=, BETWEEN and a LIKE with a character prefix like
’something%’.
Remove keys that don’t span all AND levels.
have all preceding key_parts specified.
key = 1 or A = 10 -> NULL (can’t use key.)
60
key = 1 or A = 10 and key=2 -> key = 1 OR key = 2
key_part_1 = const and key_part_3 = const -> key_part_1 = const
Read all constant tables. A constant table is:
1. A table with 0 or 1 record.
2. A table which uses only other const tables and constants on a full unique key.
const_table.key = constant const_table.key_part_1 =
const_table2.column and const_table.key_part_2 = constant
Find the best join combination to join the tables, by trying all possibilities :(. If all columns in
ORDER BY and in GROUP comes from the same table, then this table is preferred first when
joining.
If there is a order clause and a different group clause or if the order or group contains columns
from other tables than the first table in the join cue, a temporary table is created.
For each table use a ranged key, if possible, to read records. Each table index is queried and if
there exists a key range that spans < 30% of the records then a key is used. If no such key can be
found a quick table scan is used.
Before each record is output-ed, skip those that match the HAVING clause.
How can I change the buffer sizes of mysqld ?
With the mysqld -O variable=size command. Example run:
> mysqld --help (or > mysqld -?)
>mysqld --help
/my/monty/sql/mysqld Ver 3.21.0-alpa for SOLARIS 2.5 (SPARCstation)
TcX Datakonsult AB, by Monty. This is FPL, not free software
This software comes with NO WARRANTY: see the file PUBLIC for details.
Starts the mysql server
Usage: /my/monty/sql/mysqld [OPTIONS]
34

-b, --basedir=path path to installation directory
-h, --datadir=path path to the database root
-#, --debug=... output debug log. Often this is ’d:t:o,filename‘
-T, --debug-info print some debug info at exit
-?, --help display this help and exit
-L, --language=... client error messages in given language
-l, --log[=filename] log connections and queries to file
--log-update[=filename]
log updates to filename.# where # is a unique number
if not given.
--log-isam[=filename]
log all isam changes to file
-P, --port=... Port number to use for connection
-O, --set-variable var=option
give a variable an value. --help lists variables
--skip-new-routines don’t use new possible wrong routines.
--skip-grant-tables start without grant tables. This gives anyone FULL
ACCESS to all tables!
--skip-locking don’t use system locking. To use isamchk one has
to shut down the server.
--skip-name-resolve Don’t resolve hostnames.
All hostnames are IP’s or ’localhost’
--skip-new-routines don’t use new possible wrong routines.
--skip-unsafe-select skip unsafe select optimisations.
61
--socket=... Socket file to use for connection
-V, --version output version information and exit
Current base_dir: /my/monty
Current data_dir: data/
Current language: english/
Possibly variables to option --set-variable (-O) are:
back_log current value: 5
join_buffer current value: 131072
key_buffer current value: 1048568
max_allowed_packet current value: 65536
max_connections current value: 90
max_join_size current value: 4294967295
max_sort_length current value: 1024
net_buffer current value: 8192
record_buffer current value: 131072
table_cache current value: 64
tmp_table_size current value: 131072
sort_buffer current value: 2097144
> safe_mysqld -O key_buffer=512k -O sort_buffer=100k -O record_buffer=100k &
The key_buffer is shared by all threads, the rest are allocated when needed.
mysqld demon starts with a cd to ’mysql-data-dir’. After this, mysqld-data-dir is changed to ’./’
(current dir). All paths (databases, pid file, and log file) are prefixed with ’./’.
What options to use to get MySQL to run at full speed?
More memory gives more speed. You can try something like:
35

mysqld -O key_buffer=16m -O sort_buffer=1m -O record_buffer=512k
--skip-locking
--skip-locking disables file locking between SQL requests. If this is used then the following can
happen:
If mysqld or the machine crashes a table has a higher risk of being corrupted. Tables should at
least be checked with isamchk *.ISM after a crash.
One MUST flush all tables with mysqladmin refresh before one tries to check/repair tables
with isamchk. (isamchk -d table_name is always allowed).
The --skip-locking is default when compiling with MIT threads. This is because flock() isn’t
fully support by MIT threads on all platforms.
How to get MySQL to run as fast as possible with little
memory?
If there are very many connections, ’swapping problems’ may occur, unless mysqld hasn’t been
configured to use very little memory for each connection.
62
For example, for 200 open connections one should have a table cache of at least 200 *
(max_number
of tables in join).
Of course MIT threads may slow down everything a bit, but for key based selects a select is
usually
down in one time frame so there shouldn’t be a mutex locking/thread juggling.
If updates are a problem, updates can be delayed and then do many updates in a row later. Many
updates done in a row are much quicker than one at a time.
If the problem is with MIT threads and one is using FreeBSD x.x. upgrading to FreeBSD 3.0.(or
higher:) should help. This gives a possibility to use sockets (quicker than the current TCP/IP with
MIT
threads) and the thread package is much more integrated.
What are the different row formats? Or when to use
VARCHAR/CHAR?
Actually using no VARCHAR or BLOB types results in a fixed row size. Otherwise CHAR and
VARCHAR are the same.
You can check the format used in a table with isamchk -d.
MySQL has three different table formats:
1. Fixed length tables;
This is the default format.
All non packed columns are space filled.
Very quick.
Easy to cash.
Easy to reconstruct if crashed, of course, this only theoretical :-) as records are in fixed
positions.
Don’t have to be reorganised unless a huge number of records are deleted.
2. Dynamic tables
Is used if there exists any VARCHAR, TEXT or BLOB columns in a table.
All strings are dynamic (except if length < 3).
Each record is preceded with a bitmap for which columns are not empty (this isn’t the same
as null columns).
Each string is saved with a length byte + string. If string is zero length or a number is zero it
takes no extra space, just the zero length bit for each column.
Each record is uses the exact record space required. If a record becomes larger it is split into
36

as many pieces as required.
Takes little disk space.
If records are changed a lot, isamchk -r should be run now and then to reorganise the
table. This is to get a better layout. Use isamchk -ei table_name for some statistics.
Not as easy to reconstruct because a record may be in many pieces and a link may be
missing.
The expected row length for dynamic sized records is: 3 + (number_of_columns + 7) / 8 +
(number of char columns) + packed_size_of_number_columns + length_of_strings +
(null_columns + 7) / 8. There will be a penalty of 6 bytes for each link. A dynamic record
will be linked whenever a update causes a enlargement of the record. Each new link will be
63
at least 20 bytes, so the next enlargement will probably go in the same link. If not, there will
be another link. You may check how many links there are with isamchk -ed. All links
may be removed with isamchk -r.
3. Compressed tables (this is only with UNIREG/pack_isam)
Read only tables.
Takes very little disk space. Minimises disk usage.
Each record is compressed separately (very little access overhead).
Can handle fixed or dynamic length records, but no BLOBs.
Can be uncompressed with isamchk.
Why so many open tables?
When you run mysqladmin status you get something like:
Uptime: 426 Running threads: 1 Questions: 11082 Reloads: 1 Open tables: 12
This can be somewhat perplexing if you only have 6 tables.
As MySQL is multithreaded it may have many queries on the same table at once. To minimise
the
problem with two threads having different states on the same file, I open the table again for each
concurrent thread. This takes some memory and one extra file descriptor for the data file. The
index
file descriptor is shared all threads.
64
MySQL Utilites
Overview of the different MySQL programs
All MySQL client that communicates with the server uses the following environment variables:
Name Description
MYSQL_UNIX_PORT The default socket. Used with ’localhost’.
MYSQL_TCP_PORT The default TCP port.
MYSQL_PWD The default password.
mysql
A SQL shell (with gnu readline). Supports interactive use or as a non interactive query tool.
When used interactively result is given in a ascii-table format, but when used as a filter the result
is a tab-separated output.
mysqlaccess
Script to check the privileges for a host, user and database combination.
mysqladmin
Administration utility. Create/Drop of databases, reload (read new users) and refresh (flush
tables
37

to disk, reopen log files). Also gives version and process info.
mysqld
The SQL daemon. This should always be running.
mysqldump
Dump a MySQL database into a file with SQL statements or tab separated text files. Enchanted
freeware originally by Igor Romanenko.
mysqlimport
Imports text-file(s) into respective table(s). Can use all formats supported by LOAD DATA
INFILE. See section LOAD DATA INFILE syntax [p 45]
mysqlshow
Shows information about database, tables, fields and keys.
mysqlbug
This script should always be used when filing a bug report to the MySQL list.
mysql_install_db
Creates the MySQL grant tables with default privileges. This is usually only executed when
installing the first MySQL release on a new system.
isamchk
Check, optimise and repair MySQL tables.
make_binary_release
Makes a binary release of a compiled MySQL. This could be sent by ftp to
www.tcx.se/pub/mysql/Incoming for the convenience of other MySQL users.
msql2mysql
A shell script to convert a mSQL program to MySQL. Doesn’t handle all cases but gives a good
start when converting.
65
replace
Binary used for msql2mysql. Utility program to change strings in place in files or on stdin. Uses
a finite state machine to match longer strings first. Can be used to swap strings, for example
’replace a b b a -- files’ swaps ’a’ and ’b’ in the given files.
safe_mysqld
Starts the mysqld demon with some safety features. Restarts on error and has logging of runtime
information to a log file.
The MySQL table check, optimize and repair program
For infromation about how to use isamchk to repair a crached table: See section How to repair
tables.
[p 55] .
Getting low level table information
To get a description/statistics from a table use the forms below. We will explain some of the
information in more detail later.
isamchk -d table_name
isamchk in ’describe mode’. If one uses ’--skip-locking’ isamchk may report an error for a table
that is updated while isamchk runs, but there isn’t any risk of destroying data. A short form.
ISAM file: company.ISM
Data records: 1403698 Deleted blocks: 0
Recordlength: 226
Record format: Fixed length
table description:
Key Start Len Index Type

38

1 2 8 unique double
2 15 10 multip. text packed stripped
3 219 8 multip. double
4 63 10 multip. text packed stripped
5 167 2 multip. unsigned short
6 177 4 multip. unsigned long
7 155 4 multip. text
193 1 text
For explanations see below.
isamchk -d -v table_name
A little more verbose.
ISAM file: company.ISM
Isam-version: 2
Creation time: 1996-08-28 11:44:22
Recover time: 1997-01-12 18:35:29
Datafile: Parts: 1403698 Deleted data: 0
Datafilepointer (bytes): 3 Keyfile pointer (bytes): 3
Max datafile length: 3791650815 Max keyfile length: 4294967294
Recordlength: 226
r
table description:
66
Key Start Len Index Type Root Blocksize Rec/key
1 2 8 unique double 15845376 1024 1
2 15 10 multip. text packed stripped 25062400 1024 2
3 219 8 multip. double 40907776 1024 73
4 63 10 multip. text packed stripped 48097280 1024 5
5 167 2 multip. unsigned short 55200768 1024 4840
6 177 4 multip. unsigned long 65145856 1024 1346
7 155 4 multip. text 75090944 1024 4995
193 1 text
ISAM file
Name of isam file.
Isam-version
Version of isam format. Currently always 2.
Creation time
When was the data file created.
Recover time
When was the index/data file last reconstructed.
Data records
How many records/rows.
Deleted blocks
How many deleted blocks still have reserved space. See section How to repair tables. [p 55]
39

.
Datafile: Parts
For dynamic record format this shows how many data blocks there are. For a optimised table
without splits this is the same as Data records.
Deleted data
How many bytes of non reclaimed deleted data.
Datafile pointer
How many bytes the datafile pointer has. It is usually 2, 3 or 4 bytes. Most tables manage
with 2 bytes but this cannot be controlled from MySQL yet. For fixed tables this is a record
address. For dynamic tables this is a byte address.
Keyfile pointer
How many bytes has the datafile pointer. It is usually 1, 2 or 3 bytes. Most tables manage
with 2 bytes but this is calculated automatically by MySQL. It is always a block address.
Max datafile length
How long (in bytes) can the table’s data file (.ISD) get.
Max keyfile length
How long (in bytes) can the table’s key file (.ISM) get.
Recordlength
How much space does each record/row take.
Record format
Which format does each record/row have. This example uses Fixed length.
table description
A list of all keys in the table. For each key some low level information is presented.
Key
This key’s number.
Start
Where in the record/row does this index-part start.
67
Len
How long is this index-part. For packed numbers this should always be the full length
of the field. For strings it may be shorter than the full length.
Index
unique or multip.. If one value can exist multiple times in this index.
Type
What data-type does this index part have. This is a C data-type with the options packed,
stripped or empty.
Root
Address of the root index block.
Blocksize
The size of each index block. This is by default 1024 but may be changed a compile
time.
Rec/key
This is a statistical value used by the optimiser. It tells how many records there are per
value for this key. A unique key always has a value of 1. This may be updated after a
table is loaded (or greatly changed) with isamchk -a. If this is not updated at all, a
default value of 30 is given.
The 9th key is a multiple part key with two parts.
isamchk -eis table_name
Shows only the most important information from a table. Slow since it must read the whole table.
40

Checking ISAM file: company.ISM
Key: 1: Keyblocks used: 97% Packed: 0% Max levels: 4
Total: Keyblocks used: 98% Packed: 17%
Records: 1403698 M.recordlength: 226 Packed: 0%
Recordspace used: 100% Empty space: 0% Blocks/Record: 1.00
Recordblocks: 1403698 Deleteblocks: 0
Recorddata: 317235748 Deleted data: 0
Lost space: 0 Linkdata: 0
User time 1626.51, System time 232.36
Maximum resident set size 0, Integral resident set size 0
Non physical pagefaults 0, Physical pagefaults 627, Swaps 0
Blocks in 0 out 0, Messages in 0 out 0, Signals 0
Voluntary contexts switches 639, Involuntary contexts switches 28966
Keyblocks used
How many percent of the keyblocks are used. Since this table has just been reorganised with
isamchk the values are very high (very near theoretical maximum).
Packed
MySQL tries to pack keys with a common suffix. This can only be used for
CHAR/VARCHAR/DECIMAL keys. For long strings like names, this can significantly reduce
the space used. In the above example the 4 key is 10 characters long and gets a 60%
reduction in space.
68
Max levels
How deep is the btree for this key. Large tables with long keys get high values.
Records
How many rows does the table have.
M.recordlength
Average recordlength. For fixed tables this is the recordlength.
Packed
MySQL strips spaces from the end of strings. How many percent did we save by doing this.
Recordspace used
How many percent of the datafile is used.
Empty space
How many percent of the datafile is unused.
Blocks/Record
How many blocks are there per record. This is always 1 for fixed format tables. This value
should stay as close to 1.0 as possible. If it gets to great you can reorganise the table with
isamchk. See section How to repair tables. [p 55] .
Recordblocks
How many blocks are used. For fixed format, this is the same as the number of records.
Deleteblocks
41

How many blocks are deleted.
Recorddata
How many bytes of actual user data there are in the datafile.
Deleted data
How many bytes of deleted data there are in the datafile.
Lost space
If a record is updated to a shorter length some space is lost. This is the sum of all such
losses.
Linkdata
When the dynamic format is used, blocks are linked with pointers (length 4-7 bytes). This is
the sum of all such pointers.
isamchk -eiv table_name
Same as above but tells you what it is being done.
Checking ISAM file: company.ISM
- check file-size
- check delete-chain
index 1:
index 2:
index 3:
index 4:
index 5:
index 6:
index 7:
index 8:
index 9:
No recordlinks
- check index reference
- check data record references index: 1
69
Total: Keyblocks used: 9% Packed: 17%
- check records and index references
42

[LOTS OF ROW NUMBERS DELETED]
Records: 1403698 M.recordlength: 226 Packed: 0%
Recordspace used: 100% Empty space: 0% Blocks/Record: 1.00
Recordblocks: 1403698 Deleteblocks: 0
Recorddata: 317235748 Deleted data: 0
Lost space: 0 Linkdata: 0
User time 1639.63, System time 251.61
Maximum resident set size 0, Integral resident set size 0
Non physical pagefaults 0, Physical pagefaults 10580, Swaps 0
Blocks in 4 out 0, Messages in 0 out 0, Signals 0
Voluntary contexts switches 10604, Involuntary contexts switches 122798
Here is the data file sizes of the table used above.
-rw-rw-r-- 1 monty tcx 317235748 Jan 12 17:30 company.ISD
-rw-rw-r-- 1 davida tcx 96482304 Jan 12 18:35 company.ISM
The MySQL compressed read only table generator
pack_isam is an extra that you get when ordering more that 10 licenses or extended support.
Since
pack_isam is distributed binary only, pack_isam is only available on some platforms.
Of course, all future updates to pack_isam is included in the price. pack_isam may at some time
be included as standard when we get some kind of turnover for MySQL.
pack_isam works by compressing each column in the table separately. The information needed to
decompress is read into memory when the table is opened. This gives a much better performance
when accessing individual records as one only has to uncompress exactly one record, not a much
larger disk block like when using stacker on MSDOS.
MySQL uses memory mapping (mmap) on compressed tables and falls back to normal
read/write file
usage if mmap did not work.
Usually, pack_isam packs the datafile 40-70%.
There is currently two limitations with pack_isam:
After packing, the table is read only.
It can’t pack blobs, yet.
70
Fixing these limitations is on our TODO but with low priority.
pack_isam options:
> pack_isam --help
pack_isam Ver 5.0 for SOLARIS 2.5 on SPARCstation
Copyright (C) 1994-1997 TcX AB & Monty Program KB & Detron HB.
This is not free software. You must have a license to use this program
This software comes with ABSOLUTELY NO WARRANTY
Pack a ISAM-database to take much smaller space
Keys are not updated, one must run isamchk -rq on datafile afterwards
Usage: pack_isam [OPTIONS]
-b, --backup Make a backup of the table as table_name.OLD
-f, --force Force packing of table even if it’s gets bigger or
tempfile exists.
-j, --join=# Join all given tables into table.
All tables MUST be identical.
-p, --packlength=# Force storage size of recordlength (1,2 or 3)
-s, --silent Be more silent.
43

-t, --test Don’t pack table only test packing it
-v, --verbose Write info about progress and packing result
-w, --wait Wait and retry if table is in use
-T, --tmp_dir=# Use temporary directory to store temporary table
-#, --debug=... output debug log. Often this is ’d:t:o,filename‘
-?, --help display this help and exit
-V, --version output version information and exit
Typical run:
(/my/monty/tmp) ls -l station.*
-rw-rw-r-- 1 monty my 994128 Apr 17 19:00 station.ISD
-rw-rw-r-- 1 monty my 53248 Apr 17 19:00 station.ISM
-rw-rw-r-- 1 monty my 5767 Apr 17 19:00 station.frm
(/my/monty/tmp) isamchk -dvv station
ISAM file: station
Isam-version: 2
Creation time: 1996-03-13 10:08:58
Recover time: 1997-02-02 3:06:43
Recordlength: 834
table description:
1 2 4 unique unsigned long 1024 1024 1
2 32 30 multip. text 10240 1024 1
Field Start Length Type
111
224
364
4 10 1
5 11 20
6 31 1
71
7 32 30
8 62 35
9 97 35
10 132 35
11 167 4
12 171 16
13 187 35
14 222 4
15 226 16
16 242 20
17 262 20
18 282 20
19 302 30
20 332 4
44

21 336 4
22 340 1
23 341 8
24 349 8
25 357 8
26 365 2
27 367 2
28 369 4
29 373 4
30 377 1
31 378 2
32 380 8
33 388 4
34 392 4
35 396 4
36 400 4
37 404 1
38 405 4
39 409 4
40 413 4
41 417 4
42 421 4
43 425 4
44 429 20
45 449 30
46 479 1
47 480 1
48 481 79
49 560 79
50 639 79
51 718 79
52 797 8
53 805 1
54 806 1
55 807 20
56 827 4
57 831 4
Compressing station.ISD: (1192 records)
- Calculating statistics
normal: 20 empty-space: 16 empty-zero: 12 empty-fill: 11
pre-space: 0 end-space: 12 intervall-fields: 5 zero: 7
Original trees: 57 After join: 17
- Compressing file
87.14%
(/my/monty/tmp) ls -l station.*
72
-rw-rw-r-- 1 monty my 127874 Apr 17 19:00 station.ISD
-rw-rw-r-- 1 monty my 55296 Apr 17 19:04 station.ISM
-rw-rw-r-- 1 monty my 5767 Apr 17 19:00 station.frm
45

(my/monty/tmp) isamchk -dvv station
ISAM file: station
Isam-version: 2
Creation time: 1996-03-13 10:08:58
Recover time: 1997-04-17 19:04:26
Recordlength: 834
Record format: Compressed
table description:
1 2 4 unique unsigned long 10240 1024 1
2 32 30 multip. text 54272 1024 1
Field Start Length Type Huff tree Bits
1 1 1 constant 1 0
2 2 4 zerofill(1) 2 9
3 6 4 no zeros, zerofill(1) 2 9
4 10 1 3 9
5 11 20 table-lockup 4 0
6 31 1 3 9
7 32 30 no endspace, not_always 5 9
8 62 35 no endspace, not_always, no empty 6 9
9 97 35 no empty 7 9
11 167 4 zerofill(1) 2 9
14 222 4 zerofill(1) 2 9
16 242 20 no endspace, not_always 8 9
17 262 20 no endspace, no empty 8 9
20 332 4 allways zero 2 9
22 340 1 3 9
26 365 2 2 9
30 377 1 3 9
32 380 8 no zeros 2 9
46

37 404 1 2 9
38 405 4 no zeros 2 9
73
40 413 4 no zeros 2 9
42 421 4 no zeros 2 9
44 429 20 no empty 3 9
45 449 30 no empty 3 9
46 479 1 14 4
47 480 1 14 4
49 560 79 no empty 2 9
50 639 79 no empty 2 9
51 718 79 no endspace 16 9
52 797 8 no empty 2 9
53 805 1 17 1
54 806 1 3 9
55 807 20 no empty 3 9
74
Adding functions to MySQL
Adding new functions to MySQL
If you need it as sql function (like SOUNDEX()), its real easy:
1. Add one line in sql_lex.cc defining the function name in the sql_functions array.
2. Add two lines in sql_yacc.y. On defines the preprocessor symbol yacc can define (this should
be
added at the beginning of the file). Then define the function parameters and create an ’item’ with
these parameters. Check, for example, all occurrences of SOUNDEX in sql_yacc.y
3. In item_func.h declare a class inheriting from Item_num_func or Item_str_func depending on
whether your function returns a number or a string.
4. In ‘item_func.cc’ add: double *Item_func_newname::val() If you are defining
a number function or String *Item_func_newname::Str(String *str) If you are
defining a string function.
5. You should probably also define the following function: void
Item_func_newname::fix_length_and_dec() This should at least calcutate
max_length based on the given arguments. max_length is the maximal number of chars the
function may return. If the function can’t return a NULL, one should set maybe_null = 0.
About string functions:
1. For string functions the ’String *str’ argument provides a string buffer that may be used to
hold
the result.
47

2. A string function should return the string that holds the result.
3. All current string functions tries to avoid to alloc any memory unless absolutely necessary!
75
ODBC
Which operating systems does MySQL ODBC support?
MySQL ODBC is a 32 bit ODBC (2.50) level 0 for Windows95 and NT. We hope somebody
will port
it to Windows 3.x.
How should I report problems with MySQL ODBC?
We have only tested ODBC with Admndemo, some C programs, Msquery and Excel.
To give some light about any problem we would like to have the log file from the ODBC
manager (the
log you get when requesting logs from ODBCADMIN) and a MYODBC log.
To get a MYODBC log, please put this MYSQL_DEBUG=d:t:O,filename in your
AUTOEXEC.BAT and restart.
The log will be written to file ‘filename’.
Programs known to work with MyODBC.
Most programs should work with MyODBC, but these we tested ourselves or got a conformation
from
some user that is works:
Program Comment
Access
Wants a primary key and a timestamp in each record to allow updates. Fails if
comparing with single floats.
Excel Works.
odbcadmin Test program for ODBC.
Delphi One must use DBE 3.2 or newer.
C++Builder
Tested with BDE 3.0. The only known problem is that when the table schema changes,
query fields are not updated. BDE however does not seem to recognise primary keys,
only the index PRIMARY, though this has not been a problem.
How do I fill in the various fields in the ODBC administrator
program?
There are three possibilities for specifying the server name on windows95:
76
1. Use the IP.
2. Add a file lmhosts with the following info: ip hostname For example: 194.216.84.21 my
3. Configure the PC to use DNS.
Example of how to fill in the ’ODBC setup’.
Data Source Name: test
Description: This is my test database
Server: 194.216.84.21
User: monty
Password: my
Port:
These are default values to be given when prompting for a Driver connect. You don’t have to
give
48

’server’, ’user’ or ’password’ in this screen. If port is not given the default port (3333) is used.
When connection to a ODBC source you have the option to change the server, user, password
and
port.
77
Problems
Why do I get ’Access denied’?
Have you installed the MySQL grant tables with the script ’mysql_install_db’? Test by executing
’mysql -u root test’. This should not give an error. You can also check if you have a file
’user.ISD’ in the mysql database directory (ordinary install_dir/var/mysql/user.ISD).
Remember that you have to do ’mysqladmin reload’ each time you change the grant tables.
Otherwise the old tables are still used!
For testing, you should start the mysqld demon with the --without-grant-tables option.
Now you can change the MySQL grant tables and use the script mysqlaccess to check if your
grants works. mysqladmin reload tells the mysqld demon to start using the new grant
tables.
Even if you have access problems with perl, python, or ODBC, always test your privilege
problems with mysql -u user database or mysql -uuser -ppassword
database.
If you get the error ’Access denied’ when trying to connect to the database with mysql -u
user database then you have a problem with the ’user’ table. Check this by doing mysql
-u root mysql and select * from user. You should get an entry with ’hostname’ and
’user’ matching your computers hostname and your username. If the client and the server is
running on the same host and you haven’t used the --host option to mysql and you are not
using MIT threads, ’localhost’ is a synonym for your hostname.
The ’Access denied’ error message will tell you who you are trying to log in as, from which host
you are trying to log in from and if you was using a password or not. You should normally have
one entry in the user table that exactly matches your host and user, exactly as given in the error
message.
If mysql -u root test works but mysql -h your_hostname -u root test
gives ’Access denied’ then you don’t have the right name for your host in the user table. For
example if you have an entry with host ’tcx’ in the ’user’ table, but your DNS tells MySQL that
your hostname is ’tcx.subnet.se’ then the entry will not work. Test by adding a record with the IP
of your host in the ’user’ table. You can, natuarally, also add a host with a wildcard (for example
’tcx%’) in the ’user table’ (but using hostnames ending with % is pretty insecure).
When using MIT threads, localhost is never used. All connections to the mysqld demon is done
by TCP IP and you must have your real hostname in ’user’ even if you are running the client on
the same host as the server.
If you get the error ’Access to database denied’ then you have a problem with the db table. If the
used entry in the db table has an empty hostname, check also the corresponding entry in the
’host’ table.
If mysql -u user database works on the server machine, but mysql -u host -u
user database doesn’t work on another client machine, then you don’t have the client
machine in the ’user’ or the ’db’ table.
If you can’t get password to work, remember that passwords must be inserted with the
PASSWORD function. See section A example of permission setup. [p 29]
If mysql -u user test works but mysql -u user other_database doesn’t work,
49

you don’t have the other_database in the ’db’ table.
78
If you get ’Access to database denied’ when using the SELECT ... INTO OUTFILE or
LOAD DATA SQL commands then you probably don’t have the file_priv privilege set for you
in
the user table.
If everything else fails, start the mysqld demon with: --debug=d,general,query. This
will print info about the host and user that tries to connect and also information about each
command issued. See section How to debug MySQL [p 25] .
If you have any other problems with the MySQL grant tables and feel you must post the problem
to the list, always add a dump of the MySQL grant tables. You can dump the tables with the
’mysqldump mysql’ command. Post your problem with the mysqlbug script.
See section How does the privilege system work? [p 26]
How to run MySQL as a normal user.
mysqld (the MySQL server) can run as any user. In order to change mysqld to run as user USER,
you’d have to the following:
Stop the server if its running
Change the database directories so that USER has read and write privliges priviliges to them:
shell> chown -R USER /your/path/to/mysql/var
start the server as user USER
If you are using mysql.server to start mysqld when the system is rebooted, you should change
mysql.server to call ’su’ to user USER.
You don’t have to do anything to safe_mysqld to run as not root.
At this point, your mysqld process would be running fine and dandy as user ’USER’. One thing
hasn’t
changed though - the access permissions. By default (right after running the permissions table
install
script), only user ’root’ has access permission to the database. Unless you have changed that, it’s
still
true. This shouldn’t stop you from accessing MySQL when you’re logged in under a user other
than
root, just specify -u root to the client program. Note that accessing MySQL as root, by supplying
-u
root in the command line, doesn’t have ANYTHING to do with MySQL running as root, as a
user or
as anyone else. The access permissions and userbase of MySQL are completely separate from the
UNIX users. The only connection to the UNIX users us if you don’t use the -u option to clients.
In this
case the client will try to login into MySQL with your UNIX login name. If your UNIX box itself
isn’t secured, you should probably at least put a password on the root users in the MySQL access
tables, since any johndoe user can run ’mysql -u root dbname’ and do whatever he likes.
Problems with file permissions
If you have problems with file permissions, for example when creating a table mysql gives:
"ERROR:
Can’t find file: ’path/with/filename.frm’ (Errcode: 13)", then you might have the wrong value for
environment variable UMASK. Default umask is 0664. Fix:
UMASK=432
export UMASK

50

./bin/safe_mysqld
79
Problems using DATE fields.
The format of DATE is ’YYYY-MM-DD’. Actually nothing else is allowed (ANSI SQL). One
should
use this format to update or in the WHERE clause, ie select * from table_1 where idate >=
’1997-05-05’;
As a convenience, MySQL automatically converts the date to a number if used in a number
context. It
is also smart enough to allow a ’relaxed’ string form when updating and in a WHERE with a
compare
to a TIMSTAMP, DATE or a DATETIME field.
This means that the following works:
insert into table_1 (idate) values (19970505) ;
insert into table_1 (idate) values (’19970505’) ;
insert into table_1 (idate) values (’1997-05-05’);
insert into table_1 (idate) values (’1997.05.05’);
insert into table_1 (idate) values (’1997 05 05’);
select idate from table_1 where idate >= ’1997-05-05’;
select idate from table_1 where idate >= 19970505;
select mod(idate,100)1 from table_1 where idate >= 19970505;
select idate from table_1 where idate >= ’19970505’;
The following will not work:
select idate from table_1 where strcmp(idate,’19970505’)=0;
Because ’19970505’ is compared as a string to ’1997-05-05’.
80
MySQL client tools and API’s
MySQL C API
The C API is distributed with MySQL. It is included in the libmysqlclinet library.It allows C
programs to access a database.
int mysql_affected_rows(MYSQL *mysql)
Retrieves the number of affected rows by
the last UPDATE, DELETE or INSERT.
void mysql_close(MYSQL *mysql) Closes a server connection.
MYSQL *mysql_connect(MYSQL *mysql, const char
*host, const char *user, const char *passwd)
Establish a connection to a MySQL
server.
int mysql_create_db(MYSQL *mysql, const char *db) Create a database.
void mysql_data_seek(MYSQL_RES *res, uint offset)
Seeks to an arbitrary row in a query
result set.
int mysql_drop_db(MYSQL *mysql, const char *db) Drop a database.
int mysql_eof(MYSQL_RES *) Determine if last row has been read.
char *mysql_error(MYSQL *mysql)
The error message from last MySQL
51

function.
MYSQL_FIELD *mysql_fetch_field(MYSQL_RES
*handle)
Find out what type a table field is.
unsigned int *mysql_fetch_lengths(MYSQL_RES
*mysql)
Returns the length of all columns in a
query result set.
MYSQL_ROW mysql_fetch_row(MYSQL_RES
*mysql)
Fetch the ’next’ row in the query result.
void mysql_field_seek(MYSQL_RES *result, int field)
Put the column cursor on column number
field.
void mysql_free_result(MYSQL_RES *result) Free memory used to store a query result.
char *mysql_get_client_info(void)
Return version information for the
current client library.
char *mysql_get_host_info(MYSQL *mysql) Returns name of server host.
int mysql_get_proto_info(MYSQL *mysql) Get protocol version used by connection.
char *mysql_get_server_info(MYSQL *mysql) Returns the version number of the server.
81
int mysql_insert_id(MYSQL *mysql)
Returns ID generated for
AUTO_INCREMENT field.
MYSQL_RES *mysql_list_dbs(MYSQL *mysql, const
char *wild)
Return maching database names.
MYSQL_RES *mysql_list_fields(MYSQL *mysql,
const char *table, const char *wild)
Return maching field names.
MYSQL_RES *mysql_list_processes(MYSQL *mysql) Get a list of the current server threads.
MYSQL_RES *mysql_list_tables(MYSQL *mysql,
const char *wild)
Return maching table names.
int mysql_num_fields(MYSQL_RES *result)
Return the number of columns in a result
set.
int mysql_num_rows(MYSQL_RES *result) Returns the number of rows in result set.
int mysql_query(MYSQL *mysql, const char *query) Executes a SQL query.
int mysql_real_query(MYSQL *mysql, const char
*query, uint length)
Executes a SQL query with length
information.
int mysql_reload(MYSQL *mysql)
Reload the user permissions table in the
server.
int mysql_select_db(MYSQL *mysql, const char *db) Connect to a database.
52

int mysql_shutdown(MYSQL *mysql) Shut down the database server.
char *mysql_stat(MYSQL *mysql) Return server status in a string.
MYSQL_RES *mysql_store_result(MYSQL *mysql) Reads a result set to the client.
MYSQL_RES *mysql_use_result(MYSQL *mysql) Initiate a dynamic result set for each row.
Why is it that after mysql_query() returns success,
mysql_store_result() sometimes returns NULL?
It means one of the following:
1. Malloc failure.
2. The data couldn’t be read (Error on connection).
3. The statement was a statement which never returns data (INSERT or UPDATE or DELETE).
You can always check if the statement should have given a result by checking that
mysql_num_columns(MYSQL *) isn’t 0. If this is 0 the last query was a statement that does not
return values INSERT, DELETE....
You have got an error if mysql_error(MYSQL *) isn’t empty!
82
What results can I get from a query?
mysql_affected_rows(MYSQL *) returns the number of affected rows in the last query when
doing an INSERT, UPDATE or DELETE. Except, if DELETE is used without a where clause
then the
table is truncated which is much faster!. In this case it returns records affected as zero.
mysql_insert_id(MYSQL *) returns the given ID of the last query when inserting a row into a
table with a AUTO_INCREMENT key. See section How can I get the unique ID for the last
row? [p 83]
Some queries, LOAD DATA INFILE... and INSERT INTO ... SELECT ..., return
additional info. The result is returned in mysql_info(MYSQL *). mysql_info() returns a null
pointer if there is no additional information.
How can I get the unique ID for the last row?
If you insert a record that has a AUTO_INCREMENT key then you can get the given id with
mysql_insert_id(MYSQL *).
The last value is also stored in the server and can be retrieved with the LAST_INSERT_ID()
function.
You can check if an auto_increment key is used by the following code. This also checks if the
query
was a insert with a auto_increment key.
if (mysql_error(MYSQL)[0] == 0 &&
mysql_num_fields(MYSQL_RESULT) == 0 &&
mysql_insert_id(MYSQL) != 0)
used_id = mysql_insert_id(MYSQL);
When using mysql_perl you can do something like:
$std=$dbh->query("insert into foo (auto,text) values(NULL,’text’)");
print $std->insert_id;
With DBD-mysql you can do something like:
$dbh->do("INSERT INTO foo (auto,text) VALUES(NULL,’text’)");
print $dbh->func("_InsertID");
With ODBC you can do something like this:
INSERT INTO foo (auto,text) VALUES(NULL,’text’)
select LAST_INSERT_ID()
or even

53

INSERT INTO foo (auto,text) VALUES(NULL,’text’)
INSERT INTO foo2 (id,text) VALUES(LAST_INSERT_ID(),’text’)
What is the difference between mysql_use_result() and
mysql_store_result() modes?
83
mysql_use_results reads the result directly from the server without storing it in a local differ.
This is somewhat faster and uses much less memory. One shouldn’t use mysql_use_results if
there is a lot of processing being done for each row at the client side, or if the output is sent to a
screen
on which the user may do a ^S (stop scroll). Doing this would tie up the server and then other
threads
can’t update the used tables. One can’t use mysql_data_seek when using
mysql_use_result.
Problems linking with C API.
When linking with the C API you can get the following errors on some systems:
gcc -g -o client test.o -L/usr/local/lib/mysql -lmysqlclient -lsocket -lnsl
Undefined first referenced
symbol in file
floor /usr/local/lib/mysql/libmysqlclient.a(password.o)
ld: fatal: Symbol referencing errors. No output written to client
This means that on your system you have to include the math library (-lm) last in the
compile/link
line.
How to make a threadsafe client
The client is ’almost’ thread-safe. The biggest problem is that net.c (the subroutines that read
from
sockets) are not interrupt safe. This was done with the thought that one may want to have one’s
own
alarm that can break a long read to a server.
The standard client libraries are not compiled with the thread options.
To get a thread safe client use the -lmysys, -lstring and -ldbug libraries and net_serv.o that the
server
uses.
When using a threaded client I think one can have great use of the thr_alarm.c routine. If you are
using
the mysys routines, the only thing one has to remember is to call my_init() first!
Making a threadsafe client
All functions except mysql_connect() are currently thread safe.
To get connect thread_safe you have to do the following:
Recompile the client with:
CPPFLAGS=-DTHREAD_SAFE_CLIENT ./configure ....
You may get some errors because of undefined symbols when linking the standard client as the
pthread libraries are not included by default.
The resulting libmysqld.a library is now thread safe.
Two treads can’t use the same handle (returned by mysql_connect()) at the same time, even if
two
threads can use different MYSQL_RES handles that was created with mysql_store_result().
84

54

When using a threaded client one can have great use of the thr_alarm.c routine. If you are using
the
mysys routines, the only thing one has to remember is to call my_init() first!
What is the difference between different thread packages?
There are at least three types of thread packages.
User threads in a single process; Thread switching is managed with alarms and the threads
library
manages all not threadsafe functions with locks. Read, write and select are usually managed
with a thread-specific select that switches to another thread if the running threads have to wait
for
data. If the user thread packages are integrated in the standard libs (FreeBSD and BSDI threads)
the thread package requires less overhead than thread packages that have to map all unsafe calls
(MIT-threads, FSU-threads and RTS threads). In some environments (for example SCO), all
system calls are threadsafe so the mapping can be done very easily (FSU-threads on SCO).
Downside: All mapped calls take a little time and it’s quite tricky to be able to handle all
situations. There are usually also some system calls that are not handled by the thread package
(like MIT-threads and sockets). Thread scheduling isn’t always optimal.
User threads in separate processes; Thread switching is done by the kernel and all data is shared
between threads. The thread package manages the standard thread calls to allow sharing data
between threads. Linuxthreads is using this method. Downside: Lots of processes. Thread
creating is slow. If one thread dies the rest are usually left hanging and one has to kill them all
before restarting. Thread switching is somewhat expensive.
Kernel threads; Thread switching is handled by the thread library or the kernel and is very fast.
Everything is done in one process but ’ps’ may on some systems show the different threads. If
one thread aborts the whole process aborts. Most system calls are threadsafe and should require
very little overhead. Solaris, HPUX, AIX and OSF1 has kernel threads.
In some systems kernel threads are managed by integrating user level threads in the system
libraries.
In such cases, the thread switching can only be done by the thread library and the kernel isn’t
really
’thread aware’.
MySQL Perl API’s
mysqlperl
DBD::mysql
MySQL JAVA connectivity (JDBC)
MySQL C++ API’s
MySQL TCL API’s
85
MySQL Python API’s
MySQL Python API’s
86
Go to the first [p 5] , previous [p 81] , next, last section, table of contents [p 1] .
How does MySQL compare with other databases
How does MySQL compare with mSQL
This section has been written by the MySQL developers so it should be read with that in mind.
But
there are NO factual error that we know of.

55

Performance.
For a true comparison of speed so the growing MySQL benchmarks suite. It is meant to be a
benchmark that will tell any user which SQL is good at what. Because of no thread creation
overhead, small parser, few features and simple security mSQL should be quicker at:
Test that do a lot of connect/disconnect with a very simple query.
Inserts into very simple tables with few fields and keys.
CREATE TABLE and DROP TABLE.
SELECT one something that isn’t a key. (A table scan is very easy)
As these operations are so simple, it is hard to be better at them when you have a higher start
overhead. After the connection is established MySQL should perform much better. MySQL on
the other hand is much faster than mSQL and most other SQL implementions on the following:
Retrieving large results (MySQL has a better, faster and safer protocol)
Tables with variable length strings since MySQL has more efficent handling and can have
keys on varchars.
Handling tables with many fields (columns).
Handling tables with large record lengths.
SELECT with many expressions.
SELECT on large tables.
Handle many connections at the same time. MySQL handles with multiple threads. mSQL
will take care of one connection at a time. This problem only exists on MySQL if you make
long update statements (like updating the whole table) and even in those cases other threads
can continue to work with other tables.
Joins (I have seen speed differences of 50 times on this one).
ORDER BY and GROUP BY.
DISTINCT.
Using TEXT or BLOB columns.
Disk space efficiency
That is, how small can you make your tables. MySQL has very precise types so that use can
create tables that take very little space. An example of a useful MySQL datatype is the
mediumint that is 3 bytes long. If you have 10.000.0000 records even saving one byte per record
is very important. Since mSQL2 only has 4 types (char,text,int,real) it is hard to get small tables.
Stability
This is harder to judge objectively. For MySQL stability see See section How stable is MySQL?
[p 52] . We have no experience with mSQL stability so we can not say anything about that.
Price
Another important issue is of course the license. MySQL has a more flexible license than mSQL
and is also cheaper than mSQL. Remember to at least consider paying for a license or email
support for whatever product you choose to use. If you sell a product with MySQL you are of
87
course required to get a license for this.
Perl interfaces
MySQL has basically the same interfaces to perl as mSQL with some added features.
JDBC (Java)
MySQL has a java interface by GWE technologies that has been improved by Xiaokun Kelvin
ZHU. We know that mSQL has one but we have too little experience with it to compare.
Rate of development
MySQL has a very small team of developers, but we are quite use to coding C and C++ very
rapidly. Since threads, functions, group by and so on are still not implemented in mSQL, it has a
56

lot of catching up to do. To get some perspective on this you can view the mSQL HISTORY file
for the last year and compare it with MySQL NEWS file. It should be pretty obvious which one
has developed most rapidly.
Utility programs
Both mSQL and MySQL has a lot of interesting third party tools. Since it is very easy to port
upwars (mSQL -> MySQL) MySQL has almost all interesting mSQL applications. MySQL
comes with a simple msql2mysql program that fixes the different spelling of the most used
functions. A conversion of a client program from mSQL to mySQL usually takes a couple of
minutes.
How about mSQL tools like msql-tcl, msqljava?
According to our experience it would just take a few hours to convert a tool using the mSQL C
API to
the MySQL C API.
The procedure:
1. Run the shell script msql2mysql on the source. This needs the binary program replace, which
is
distributed with MySQL.
2. Compile
3. Fix all compiler errors:
Differences between the MySQL and mSQL C API’s.
MySQL uses MySQL as a connection type (mSQL uses an int).
MySQL connect takes a pointer to a MYSQL structure as a parameter. It is easy to define one
globally or use malloc to get one.
mysql_connect takes 2 extra parameters. You may set these to NULL,NULL for default use (user
and password).
mysql_error takes the MYSQL structure as a parameter. Just add the parameter to your old
msql_error code if you are porting old code.
Some incompatibilities exist as a result of MySQL supporting multiple connections to the server
from the same process.
MySQL returns a error number and a text for all errors. mSQL only returns a text error
How different from mSQL are the MySQL client/server
communications protocols?
88
There are enough differences that it is impossible (at least not easy) to support both.
The greatest differences between MySQL and mSQL are:
A message buffer may contain many result rows.
The message buffers are dynamically enlarged if the question or the result is bigger than the
current buffer up to a configurable server and client limit.
All packages are numbered to catch out dupplicated or missing packets.
All columns are sent in ASCII, the length of columns and rows are sent in packed binary coding
(1,2 or 3 bytes).
MySQL can read in the result unbuffered (without having to store the full set in the client).
If a single write/read takes more than 30 seconds then the server closes the connection.
If nothing has happened on a connection for 8 hours the server closes the connection.
We don’t think it makes sense to give a full specification of the connection protocol as we have
plans
to optimize this in the near future.
What are the differences in the SQL syntax between MySQL
57

& mSQL 2.0?
CREATE TABLE
MySQL
Has the following column option types: See section CREATE TABLE syntax. [p 36]
UNSIGNED for integer columns.
ZEROFILL for integer columns.
AUTO_INCREMENT for integer columns. See section How can I get the unique ID
for the last row? [p 83]
DEFAULT values for all columns.
MySQL also have many more types than ANSI SQL.
mSQL
To create indexes.
MySQL
All indexes have to be given to CREATE TABLE. Indexes can not be removed without
recreating the table. See section ALTER TABLE syntax [p 37] .
mSQL
Indexes must be created by a separate CREATE INDEX clause. Indexes may be removed
with DROP INDEX.
To get a unique identifier in a table at insert.
MySQL
Use autoincrement as a column type specifier. See section How can I get the unique ID
for the last row? [p 83]
mSQL
Create a SEQUENCE on a table and use the __seq function to get a unique index.
Group functions (functions that can be used in a SQL group by clause).
MySQL
count(), avg(), min(), max(), sum() and std(). min() and max() may take
string arguments. count(*) is optimised to return very quickly if this is the only thing in
the query:
89
SELECT COUNT(*) FROM table_name
mSQL
Does not support GROUP BY yet.
How do I search independent of character case?
MySQL
LIKE is always case independent. If possible MySQL uses indexes if the like argument
doesn’t start with a wild-card.
mSQL
Use CLIKE.
How do I search with regular expressions?
MySQL
Use REGEXP or RLIKE. The syntax should be basically POSIX.
mSQL
Use RLIKE.
What are the differences in the WHERE statement?
MySQL
MySQL does everything according to mathematical priorities (AND is evaluated before
OR). To get mSQL behaviour in MySQL, use brackets: select * from table

58

where a=1 and b=2 or a=3 and b=4 -> select * from table where
(a=1 and (b=2 or (a=3 and (b=4)))).
mSQL
Evaluates everything from left to right. That means that some logical calculation with more
that 3 arguments can not be expressed.
Qualifying column names
MySQL
If a column name is unique you do not have to use the full qualifier.
mSQL
When using more than one table in a select you must use full table qualifiers.
Aliasing
MySQL
Has table and column aliasing.
mSQL
Has table aliasing.
Insert/update with calculations.
MySQL
Full calculation support. In a insert you may use values of preceding columns.
mSQL
Only insert update with constants.
Which formulas may be used in the select statement
MySQL
Too many to print here. See section SELECT syntax [p 39] .
mSQL
No functions are supported yet.
SQL HAVING clause.
MySQL
Supported, but can only use calculation on the selected columns. To select on a calculated
value one must use a column alias. Ex: SELECT COUNT(*) AS id_count,id FROM
groups GROUP BY id HAVING id_count > 10.
90
mSQL
Not supported.
Access Control
MySQL
Has tables to store grant options per user, host and db. See section How does the privilege
system work? [p 26]
mSQL
Has a file ‘mSQL.acl’ where you can grant read/write privileges for users.
Threads
MySQL
Is multi-threaded. Each connection has it is own thread, which means that none of them has
to wait for the other (unless a data-modifying query is executed).
mSQL
All other connections have to wait until the first, doesn’t matter whether the query is long or
short, is executed and finished. After that the next connection can be served, while the others
wait again, etc.
Speed
MySQL is significantly quicker on complex selects than mSQL.
59

Problems with AND and OR priority
Remember that AND has higher priority in boolean algebra than OR. It is like 1+2*3 is = 1+
(2*3) = 7.
Always when you are in doubt:
1. Remove all distinct keywords to make the query easier to debug.
2. Write all fields from all tables to really see what you are doing.
This will make it easier to see what goes wrong. Distinct may also be a lot slower than a normal
select.
Go to the first [p 5] , previous [p 81] , next, last section, table of contents [p 1] .
91

60

My sql technical reference manual

More Related Content

What's hot (14)

Viewers also liked (7)

Similar to My sql technical reference manual (20)

More from Mir Majid (16)

My sql technical reference manual