Merge jcole@work.mysql.com:/home/bk/mysql

* Estimating performance::      Estimating query performance

* SELECT speed::                Speed of @code{SELECT} queries

* Where optimizations::         How MySQL optimizes @code{WHERE} clauses

* LEFT JOIN optimization::      How MySQL optimizes @code{LEFT JOIN}

* LEFT JOIN optimization::      How MySQL optimizes @code{LEFT JOIN} and @code{RIGHT JOIN}

* LIMIT optimization::          How MySQL optimizes @code{LIMIT}

* Insert speed::                Speed of @code{INSERT} queries

* Update speed::                Speed of @code{UPDATE} queries

@c Added 980312

@c @image{Flags/usa}

USA [Hurricane Electric/San Jose] @

@uref{http://mysql.he.net, WWW}

@uref{http://mysql.he.net/, WWW}

@item

@c EMAIL: ask@valueclick.com (Ask Bjoern Hansen)

@c Added 20000925

@c @image{Flags/usa}

USA [ValueClick, Los Angeles CA] @

@uref{http://mysql.valueclick.com/, WWW}

@uref{ftp://mysql.valueclick.com/mysql/, FTP}

@c @item

@c Not ok 20000919; Non-existent (Matt)

@code{SUM()} now returns @code{NULL}, instead of 0, if there is no matching

rows. This is according to ANSI SQL.

@item

New restricted words: @code{CASE, THEN, WHEN, ELSE and END}

New restricted words: @code{CASE, THEN, WHEN, ELSE, END and RIGHT}

@item An @code{AND} or @code{OR} with @code{NULL} values will now return

@code{NULL} instead of 0. This mostly affects queries that uses @code{NOT}

on an @code{AND/OR} expression as @code{NOT NULL} = @code{NULL}.

@code{DROP DATABASE} drops all tables in the database and deletes the

database.  @strong{Be VERY careful with this command!}

@code{DROP DATABASE} returns the number of files that were removed from the

database directory.  Normally, this is three times the number of tables,

because each table corresponds to a @file{.MYD} file, a @file{.MYI} file and a

@file{.frm} file.

@code{DROP DATABASE} returns the number of files that were removed from

the database directory.  Normally, this is three times the number of

tables, because normally each table corresponds to a @file{.MYD} file, a

@file{.MYI} file and a @file{.frm} file.

The @code{DROP DATABASE} command removes from the given database

directory all files with the following extensions:

@multitable @columnfractions .25 .25 .25 .25

@item .BAK @tab .DAT @tab .HSH @tab .ISD

@item .ISM @tab .ISM @tab .MRG @tab .MYD

@item .MYI @tab .db  @tab .frm

@end multitable

All sub directories that consists of 2 digits (@code{RAID} directories)

are also removed.

In @strong{MySQL} 3.22 or later, you can use the keywords @code{IF EXISTS} to

prevent an error from occurring if the database doesn't exist.

@findex LEFT OUTER JOIN

@findex NATURAL LEFT JOIN

@findex NATURAL LEFT OUTER JOIN

@findex RIGHT JOIN

@findex RIGHT OUTER JOIN

@findex NATURAL RIGHT JOIN

@findex NATURAL RIGHT OUTER JOIN

@findex STRAIGHT_JOIN

@node JOIN, INSERT, SELECT, Reference

@section @code{JOIN} syntax

table_reference LEFT [OUTER] JOIN table_reference

table_reference NATURAL [LEFT [OUTER]] JOIN table_reference

@{ oj table_reference LEFT OUTER JOIN table_reference ON conditional_expr @}

table_reference RIGHT [OUTER] JOIN table_reference join_condition

table_reference RIGHT [OUTER] JOIN table_reference

table_reference NATURAL [RIGHT [OUTER]] JOIN table_reference

@end example

Where @code{table_reference} is defined as

This example finds all rows in @code{table1} with an @code{id} value that is

not present in @code{table2} (i.e., all rows in @code{table1} with no

corresponding row in @code{table2}).  This assumes that @code{table2.id} is

declared @code{NOT NULL}, of course.

declared @code{NOT NULL}, of course. @xref{LEFT JOIN optimization}.

@item

The @code{USING} @code{(column_list)} clause names a list of columns that must

with a @code{USING} clause that names all columns that exist in both

tables.

@item

@code{RIGHT JOIN} works analogously as @code{LEFT JOIN}.  To keep code

portable across databases, it's recommended to use @code{LEFT JOIN}

instead of @code{RIGHT JOIN}.

@item

@code{STRAIGHT_JOIN} is identical to @code{JOIN}, except that the left table

is always read before the right table. This can be used for those (few)

host to the server host.  On the other hand, you do not need the

@strong{file} privilege to load local files.

Note that you can't read from a FIFO with @code{LOAD DATA INFILE}; If

you need to read from a FIFO (for example the output from gunzip), use

@code{LOAD DATA LOCAL INFILE} instead.

@c old version

If you are using @strong{MySQL} before 3.23.24 you can't read from a

FIFO with @code{LOAD DATA INFILE}; If you need to read from a FIFO (for

example the output from gunzip), use @code{LOAD DATA LOCAL INFILE}

instead.

@cindex @code{mysqlimport}

You can also load data files by using the @code{mysqlimport} utility; it

to do this is to use @code{SELECT ... INTO OUTFILE} into another file and

compare this to your original input file.

If you need @code{LOAD DATA INFILE} to read from a pipe, you can use the

If you need @code{LOAD DATA} to read from a pipe, you can use the

following trick:

@example

mysql -e "LOAD DATA INFILE 'x' INTO TABLE x" x

@end example

If you are using and older @strong{MySQL} version than 3.23.25

you can only do the above with @code{LOAD DATA LOCAL INFILE}.

For more information about the efficiency of @code{INSERT} versus

@code{LOAD DATA INFILE} and speeding up @code{LOAD DATA INFILE},

@xref{Insert speed}.

@end example

@node LEFT JOIN optimization, LIMIT optimization, Where optimizations, Query Speed

@subsection How MySQL optimizes @code{LEFT JOIN}

@subsection How MySQL optimizes @code{LEFT JOIN} and @code{RIGHT JOIN}

@code{A LEFT JOIN B} is in @strong{MySQL} implemented as follows:

matches the @code{LEFT JOIN} condition.

@end itemize

The table read order forced by @code{LEFT JOIN} and @code{STRAIGHT JOIN} will help

the join optimizer (which calculates in which order tables should be joined) to do

its work much more quickly as there are fewer table permutations to check.

@code{RIGHT JOIN} is implemented analogously as @code{LEFT JOIN}.

The table read order forced by @code{LEFT JOIN} and @code{STRAIGHT JOIN}

will help the join optimizer (which calculates in which order tables

should be joined) to do its work much more quickly as there are fewer

table permutations to check.

Note that the above means that if you do a query of type:

SELECT * FROM a,b LEFT JOIN c ON (c.key=a.key) LEFT JOIN d (d.key=a.key) WHERE b.key=d.key

@end example

Then @strong{MySQL} will do a full scan on @code{b} as the @code{LEFT JOIN} will

force it to be read before @code{d}.

Then @strong{MySQL} will do a full scan on @code{b} as the @code{LEFT

JOIN} will force it to be read before @code{d}.

The fix in this case is to change the query to:

If you need more connections than the default (100), then you should restart

@code{mysqld} with a bigger value for the @code{max_connections} variable.

Note that @code{mysqld} actually allows (@code{max_connections}+1) clients to connect.

The last connection is reserved for a user with the @strong{process} privilege.

By not giving this privilege to normal users (they shouldn't need this), an

administrator with this privilege can login and use @code{SHOW PROCESSLIST}

to find out what could be wrong. @xref{SHOW}.

Note that @code{mysqld} actually allows (@code{max_connections}+1)

clients to connect.  The last connection is reserved for a user with the

@strong{process} privilege.  By not giving this privilege to normal

users (they shouldn't need this), an administrator with this privilege

can login and use @code{SHOW PROCESSLIST} to find out what could be

wrong. @xref{SHOW}.

The maximum number of connects @strong{MySQL} is depending on how good

the thread library is on a given platform. Linux or Solaris should be

able to support 500-1000 simultaneous connections, depending on how much

RAM you have and what your clients are doing.

@node Out of memory, Packet too large, Too many connections, Common errors

@subsection @code{Out of memory} error

shell> /path/to/safe_mysqld &

@end example

In @strong{MySQL} 3.23.25 and above, @strong{MySQL} assumes that the

value for @code{UMASK} and @code{UMASK_DIR} is in octal if it starts

with a zero.

@xref{Environment variables}.

@node Not enough file handles, Using DATE, File permissions , Problems

@appendixsubsec Changes in release 3.23.25

@itemize @bullet

@item

Fixed that databasename works as second argument to @code{mysqlhotcopy}.

@item

@code{UMASK} and @code{UMASK_DIR} can now be specified in octal.

@item

Added @code{RIGHT JOIN}. This makes @code{RIGHT} a reserved word.

@item

Added @code{@@@@IDENTITY} as a synonym for @code{LAST_INSERT_ID()}.

(for Visual Basic)

@item

Fixed bug in myisamchk and @code{REPAIR} when using @code{FULLTEXT} index.

@item

@code{LOAD DATA INFILE} now works for FIFO's.

(Patch by Toni L. Harbaugh-Blackford).

@item

@code{FLUSH LOGS} broke replication if one had @code{log-bin} with 

a log with explicit extension 

@item

@code{strace} or @code{/usr/proc/bin/pstack} to examine where

@code{mysqld} has hung.

@example

strace /tmp/log libexec/mysqld

@end example

If @code{mysqld} starts to eat up CPU or memory or if it ``hangs'', you

can use @code{mysqladmin processlist status} to find out if someone is

executing a query that takes a long time.  It may be a good idea to

  MYSQL_FIELD *field;

  mysql_field_seek(result,0);

  printf("<TABLE BORDER=1>\n");

  printf("<TR>\n");

  fputs("<TABLE BORDER=1><TR>",stdout);

  if (!skip_column_names)

  {

    while((field = mysql_fetch_field(result)))

    {

      printf("<TH>%s</TH>",field->name ? (field->name[0] ? field->name:" &nbsp; "):"NULL");

      printf("<TH>%s</TH>", (field->name ? (field->name[0] ? field->name : 

					    "   ") :

			     "NULL"));

    }

    puts("\n</TR>");

    puts("</TR>");

  }

  while ((cur = mysql_fetch_row(result)))

  {

    puts("<TR>");

    fputs("<TR>",stdout);

    for (uint i=0; i < mysql_num_fields(result); i++)

    {

      ulong *lengths=mysql_fetch_lengths(result);

      safe_put_field(cur[i],lengths[i]);

      fputs("</TD>",stdout);

    }

    puts("\n</TR>");

    puts("</TR>");

  }

  puts("</TABLE>");

}

/* Return error-text for system error messages and nisam messages */

#define PERROR_VERSION "2.3"

#define PERROR_VERSION "2.4"

#include <global.h>

#include <my_sys.h>

  { 141,"Duplicate unique on write or update"},

  { 142,"Unknown character set used"},

  { 143,"Conflicting table definition between MERGE and mapped table"},

  { 144,"Table is crashed and last repair failed"},

  { 0,NullS },

};

#define HA_OPEN_ABORT_IF_LOCKED		0	/* default */

#define HA_OPEN_WAIT_IF_LOCKED		1

#define HA_OPEN_IGNORE_IF_LOCKED	2

#define HA_OPEN_TMP_TABLE		4

#define HA_OPEN_DELAY_KEY_WRITE		8

#define HA_OPEN_TMP_TABLE		4	/* Table is a temp table */

#define HA_OPEN_DELAY_KEY_WRITE		8	/* Don't update index  */

#define HA_OPEN_ABORT_IF_CRASHED	16

	/* The following is parameter to ha_rkey() how to use key */

#define HA_OPTION_TEMP_COMPRESS_RECORD	((uint) 16384)	/* set by isamchk */

#define HA_OPTION_READ_ONLY_DATA	((uint) 32768)	/* Set by isamchk */

	/* Bits in flag to ni_create() */

	/* Bits in flag to create() */

#define HA_DONT_TOUCH_DATA	1	/* Don't empty datafile (isamchk) */

#define HA_PACK_RECORD		2	/* Request packed record format */

#define HA_ERR_FOUND_DUPP_UNIQUE 141	/* Dupplicate unique on write */

#define HA_ERR_UNKNOWN_CHARSET	 142	/* Can't open charset */

#define HA_ERR_WRONG_TABLE_DEF	 143

#define HA_ERR_CRASHED_ON_REPAIR 144	/* Last (automatic?) repair failed */

	/* Other constants */

#define MY_LINK_WARNING 32	/* my_redel() gives warning if links */

#define MY_COPYTIME	64	/* my_redel() copys time */

#define MY_HOLD_ORIGINAL_MODES 128  /* my_copy() holds to file modes */

#define MY_REDEL_MAKE_BACKUP 256

#define MY_SEEK_NOT_DONE 32	/* my_lock may have to do a seek */

#define MY_DONT_WAIT	64	/* my_lock() don't wait if can't lock */

#define MY_ZEROFILL	32	/* my_malloc(), fill array with zero */

  const char **type_names;

} TYPELIB;

enum cache_type {READ_CACHE,WRITE_CACHE,READ_NET,WRITE_NET};

enum cache_type {READ_CACHE,WRITE_CACHE,READ_FIFO,READ_NET,WRITE_NET};

enum flush_type { FLUSH_KEEP, FLUSH_RELEASE, FLUSH_IGNORE_CHANGED,

		  FLUSH_FORCE_WRITE};