sql_update.h File Reference

#include <sys/types.h>
#include "mem_root_deque.h"
#include "my_alloc.h"
#include "my_sqlcommand.h"
#include "my_table_map.h"
#include "sql/query_result.h"
#include "sql/sql_cmd_dml.h"
#include "sql/sql_list.h"
#include "sql/thr_malloc.h"

Go to the source code of this file.

Classes
class	Query_result_update
class	Sql_cmd_update

Functions
bool	records_are_comparable (const TABLE *table)
	True if the table's input and output record buffers are comparable using compare_records(TABLE*). More...
bool	compare_records (const TABLE *table)
	Compares the input and output record buffers of the table to see if a row has changed. More...
bool	should_switch_to_multi_table_if_subqueries (const THD *thd, const Query_block *select, const Table_ref *table_list)
	Decides if a single-table UPDATE/DELETE statement should switch to the multi-table code path, if there are subqueries which might benefit from semijoin or subquery materialization, and if no feature specific to the single-table path are used. More...
table_map	GetImmediateUpdateTable (const JOIN *join, bool single_target)
	Find out which of the target tables can be updated immediately while scanning. More...
bool	FinalizeOptimizationForUpdate (JOIN *join)
	Makes the TABLE and handler objects ready for being used in an UPDATE statement. More...
unique_ptr_destroy_only< RowIterator >	CreateUpdateRowsIterator (THD *thd, MEM_ROOT *mem_root, JOIN *join, unique_ptr_destroy_only< RowIterator > source)
	Creates an UpdateRowsIterator which updates the rows returned by the given "source" iterator. More...

Function Documentation

compare_records()

bool compare_records

(

const TABLE *

table

)

Compares the input and output record buffers of the table to see if a row has changed.

The algorithm iterates over updated columns and if they are nullable compares NULL bits in the buffer before comparing actual data. Special care must be taken to compare only the relevant NULL bits and mask out all others as they may be undefined. The storage engine will not and should not touch them.

Parameters

table

The table to evaluate.

Returns

true if row has changed.

false otherwise.

CreateUpdateRowsIterator()

unique_ptr_destroy_only< RowIterator > CreateUpdateRowsIterator	(	THD *	thd,
		MEM_ROOT *	mem_root,
		JOIN *	join,
		unique_ptr_destroy_only< RowIterator >	source
	)

Creates an UpdateRowsIterator which updates the rows returned by the given "source" iterator.

FinalizeOptimizationForUpdate()

bool FinalizeOptimizationForUpdate

(

JOIN *

join

)

Makes the TABLE and handler objects ready for being used in an UPDATE statement.

Called at the beginning of each execution.

Parameters

join	The top-level JOIN object of the UPDATE operation.

Returns

true on error.

GetImmediateUpdateTable()

table_map GetImmediateUpdateTable	(	const JOIN *	join,
		bool	single_target
	)

Find out which of the target tables can be updated immediately while scanning.

This is used by the old optimizer after the plan has been created. The hypergraph optimizer does not use this function, as it makes the decision about immediate update during planning, not after planning.

Parameters

join	The top-level JOIN object of the UPDATE statement.
single_target	True if the UPDATE statement has exactly one target table.

Returns

Map of tables to update while scanning.

records_are_comparable()

bool records_are_comparable

(

const TABLE *

table

)

True if the table's input and output record buffers are comparable using compare_records(TABLE*).

should_switch_to_multi_table_if_subqueries()

bool should_switch_to_multi_table_if_subqueries	(	const THD *	thd,
		const Query_block *	qb,
		const Table_ref *	table_list
	)

Decides if a single-table UPDATE/DELETE statement should switch to the multi-table code path, if there are subqueries which might benefit from semijoin or subquery materialization, and if no feature specific to the single-table path are used.

Parameters

thd	Thread handler
qb	Query block
table_list	Table to modify

Returns

true if we should switch