目录
in和or在表数据量比较大的情况会走索引,在表记录不多的情况下会选择全表扫描
第二阶段:SQL优化阶段(分析具体用哪个索引,以及分析使用的原因)
示例表结构
mysql-version : [8.0.27]
CREATE TABLE `test_index` (
`id` int NOT NULL,
`name` varchar(20) DEFAULT NULL,
`score` int DEFAULT NULL,
`addr` varchar(30) DEFAULT NULL,
`ctime` timestamp NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
PRIMARY KEY (`id`),
KEY `union_name_score_addr` (`name`,`score`,`addr`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb3;
示例索引
-
联合索引首字段范围查找失效
正常情况下,范围之后的索引失效,但是如果范围查询的是首个索引字段,则所有索引均失效
-
强制索引
我们有时可以强制某个查询用索引。用法如下:force index;虽然查询由上面的全表扫描变成了范围查询,而且扫描的行数也减少了,但是最终cost的时间未必最少。其内部有成本分析。
explain select * from test_index t force index (union_name_score_addr) where t.name > '北京' and Score=1 and addr='010北京'
-
in和or在表数据量比较大的情况会走索引,在表记录不多的情况下会选择全表扫描
explain select * from test_index t where t.name in( '北京' )and Score=1 and addr='010北京' -- ken_len=161 explain select * from test_index t where t.name in( '北京' ,'上海')and Score=1 and addr='010北京' -- ken_len=161 explain select * from test_index t where t.name ='北京' or name='上海' and Score=1 and addr='010北京' -- ken_len=161
我们复制一份表,表记录仅保留10行记录,看效果
-
LIKE 'XX%'走索引(数据量大时)
当联合索引的首字段为like时,走索引
索引下推
首先该理论仅限于二级索引。(可能由于InnoDb引擎库中聚簇索引包含了所有数据,所以效率不高)
索引下推:即在联合索引中根据第一个索引值筛选出结果后,同时在接下来的索引中进行条件判断,同时满足条件时进行回表,从而减少回表的次数。
下图为5.6版本前未索引下推以及索引下推的差异。
认识trace工具
我们正常一个sql语句如何走的索引,为什么同样的语句不同的数据量,有的走索引有的不走?
带着这个疑惑我们来认识一下trace工具,最后附上完成分析代码。
首先看一下如何使用:
set session optimizer_trace="enabled=on",end_markers_in_json=on; -- 开启trace。切记该开关仅临时使用,测试完记得关闭,影响性能。
select * from test_index where name > '北' order by addr; -- 执行语句
SELECT * FROM information_schema.OPTIMIZER_TRACE; -- 分析数据
第一阶段:SQL准备阶段,格式化sql
{
"join_preparation": { -- 第一阶段:SQL准备阶段,格式化sql
"select#": 1,
"steps": [
{
"expanded_query": "/* select#1 */ select `test_index`.`id` AS `id`,`test_index`.`name` AS `name`,`test_index`.`score` AS `score`,`test_index`.`addr` AS `addr`,`test_index`.`ctime` AS `ctime` from `test_index` where (`test_index`.`name` > '北') order by `test_index`.`addr`"
}
] /* steps */
} /* join_preparation */
}
第二阶段:SQL优化阶段(分析具体用哪个索引,以及分析使用的原因)
条件处理
{
/* begin-condition_processing */
"condition_processing": { -- --条件处理
"condition": "WHERE",
"original_condition": "(`test_index`.`name` > '北')",
"steps": [
{
"transformation": "equality_propagation",
"resulting_condition": "(`test_index`.`name` > '北')"
},
{
"transformation": "constant_propagation",
"resulting_condition": "(`test_index`.`name` > '北')"
},
{
"transformation": "trivial_condition_removal",
"resulting_condition": "(`test_index`.`name` > '北')"
}
] /* steps */
} /* end-condition_processing */
}
表依赖详情
/* begin-table_dependencies */
{
"table_dependencies": [ -- 表依赖详情
{
"table": "`test_index`",
"row_may_be_null": false,
"map_bit": 0,
"depends_on_map_bits": [
] /* depends_on_map_bits */
}
] /* end-table_dependencies */
}
预估表的访问成本
/* begin-rows_estimation */
{
"rows_estimation": [ -- 预估表的访问成本
{
"table": "`test_index`",
"range_analysis": {
"table_scan": { -- 全表扫描情况
"rows": 106, -- 扫描行数
"cost": 12.95 -- 扫描成本
} /* table_scan */,
/* begin-potential_range_indexes */
"potential_range_indexes": [ -- 查询可能使用的索引
{
"index": "PRIMARY",
"usable": false,
"cause": "not_applicable"
},
{
"index": "union_name_score_addr", -- 辅助联合索引
"usable": true,
"key_parts": [
"name",
"score",
"addr",
"id"
] /* key_parts */
}
] /* end-potential_range_indexes */,
"setup_range_conditions": [
] /* setup_range_conditions */,
"group_index_range": {
"chosen": false,
"cause": "not_group_by_or_distinct"
} /* group_index_range */,
"skip_scan_range": {
"potential_skip_scan_indexes": [
{
"index": "union_name_score_addr",
"usable": false,
"cause": "query_references_nonkey_column"
}
] /* potential_skip_scan_indexes */
} /* skip_scan_range */,
/* begin-analyzing_range_alternatives */
"analyzing_range_alternatives": { -- 分析各个索引使用成本
"range_scan_alternatives": [
{
"index": "union_name_score_addr",
"ranges": [
"北 < name" -- 索引使用范围
] /* ranges */,
"index_dives_for_eq_ranges": true,
"rowid_ordered": false, -- 使用该索引获取的记录是否按照主键排序
"using_mrr": false,
"index_only": false, -- 是否使用覆盖索引
"in_memory": 1,
"rows": 93, -- 索引扫描行数
"cost": 32.81, -- 索引使用成本
"chosen": false, -- 是否选择该索引
"cause": "cost"
}
] /* range_scan_alternatives */,
"analyzing_roworder_intersect": {
"usable": false,
"cause": "too_few_roworder_scans"
} /* analyzing_roworder_intersect */
} /* end-analyzing_range_alternatives */
} /* range_analysis */
}
] /* end-rows_estimation */
}
确定执行计划(索引使用情况)
{
"considered_execution_plans": [
{
"plan_prefix": [
] /* plan_prefix */,
"table": "`test_index`",
"best_access_path": { -- 最优访问路径
"considered_access_paths": [ -- 最终选择的访问路径
{
"rows_to_scan": 106,
"access_type": "scan", -- 访问类型:为scan,全表扫描
"resulting_rows": 106,
"cost": 10.85,
"chosen": true, -- 确定选择
"use_tmp_table": true
}
] /* considered_access_paths */
} /* best_access_path */,
"condition_filtering_pct": 100,
"rows_for_plan": 106,
"cost_for_plan": 10.85,
"sort_cost": 106,
"new_cost_for_plan": 116.85,
"chosen": true
}
] /* considered_execution_plans */
}
第三阶段:SQL执行阶段 (此处判断filesort)
{
"join_execution": { -- 第三阶段:SQL执行阶段
"select#": 1,
"steps": [
{
"sorting_table": "test_index",
"filesort_information": [
{
"direction": "asc",
"expression": "`test_index`.`addr`"
}
] /* filesort_information */,
"filesort_priority_queue_optimization": {
"usable": false,
"cause": "not applicable (no LIMIT)"
} /* filesort_priority_queue_optimization */,
"filesort_execution": [
] /* filesort_execution */,
"filesort_summary": {
"memory_available": 262144,
"key_size": 61,
"row_size": 230,
"max_rows_per_buffer": 106,
"num_rows_estimate": 106,
"num_rows_found": 93,
"num_initial_chunks_spilled_to_disk": 0,
"peak_memory_used": 33792,
"sort_algorithm": "std::sort",
"sort_mode": "<fixed_sort_key, packed_additional_fields>"
} /* filesort_summary */
}
] /* steps */
} /* join_execution */
}
完整分析代码
{
"steps": [
{
"join_preparation": { -- 第一阶段:SQL准备阶段,格式化sql
"select#": 1,
"steps": [
{
"expanded_query": "/* select#1 */ select `test_index`.`id` AS `id`,`test_index`.`name` AS `name`,`test_index`.`score` AS `score`,`test_index`.`addr` AS `addr`,`test_index`.`ctime` AS `ctime` from `test_index` where (`test_index`.`name` > '北') order by `test_index`.`addr`"
}
] /* steps */
} /* join_preparation */
},
{
"join_optimization": { -- 第二阶段:SQL优化阶段(分析具体用哪个索引,以及分析使用的原因)
"select#": 1,
"steps": [
{
/* begin-condition_processing */
"condition_processing": { -- --条件处理
"condition": "WHERE",
"original_condition": "(`test_index`.`name` > '北')",
"steps": [
{
"transformation": "equality_propagation",
"resulting_condition": "(`test_index`.`name` > '北')"
},
{
"transformation": "constant_propagation",
"resulting_condition": "(`test_index`.`name` > '北')"
},
{
"transformation": "trivial_condition_removal",
"resulting_condition": "(`test_index`.`name` > '北')"
}
] /* steps */
} /* end-condition_processing */
},
{
"substitute_generated_columns": {
} /* substitute_generated_columns */
},
/* begin-table_dependencies */
{
"table_dependencies": [ -- 表依赖详情
{
"table": "`test_index`",
"row_may_be_null": false,
"map_bit": 0,
"depends_on_map_bits": [
] /* depends_on_map_bits */
}
] /* end-table_dependencies */
},
{
"ref_optimizer_key_uses": [
] /* ref_optimizer_key_uses */
},
/* begin-rows_estimation */
{
"rows_estimation": [ -- 预估表的访问成本
{
"table": "`test_index`",
"range_analysis": {
"table_scan": { -- 全表扫描情况
"rows": 106, -- 扫描行数
"cost": 12.95 -- 扫描成本
} /* table_scan */,
/* begin-potential_range_indexes */
"potential_range_indexes": [ -- 查询可能使用的索引
{
"index": "PRIMARY",
"usable": false,
"cause": "not_applicable"
},
{
"index": "union_name_score_addr", -- 辅助联合索引
"usable": true,
"key_parts": [
"name",
"score",
"addr",
"id"
] /* key_parts */
}
] /* end-potential_range_indexes */,
"setup_range_conditions": [
] /* setup_range_conditions */,
"group_index_range": {
"chosen": false,
"cause": "not_group_by_or_distinct"
} /* group_index_range */,
"skip_scan_range": {
"potential_skip_scan_indexes": [
{
"index": "union_name_score_addr",
"usable": false,
"cause": "query_references_nonkey_column"
}
] /* potential_skip_scan_indexes */
} /* skip_scan_range */,
/* begin-analyzing_range_alternatives */
"analyzing_range_alternatives": { -- 分析各个索引使用成本
"range_scan_alternatives": [
{
"index": "union_name_score_addr",
"ranges": [
"北 < name" -- 索引使用范围
] /* ranges */,
"index_dives_for_eq_ranges": true,
"rowid_ordered": false, -- 使用该索引获取的记录是否按照主键排序
"using_mrr": false,
"index_only": false, -- 是否使用覆盖索引
"in_memory": 1,
"rows": 93, -- 索引扫描行数
"cost": 32.81, -- 索引使用成本
"chosen": false, -- 是否选择该索引
"cause": "cost"
}
] /* range_scan_alternatives */,
"analyzing_roworder_intersect": {
"usable": false,
"cause": "too_few_roworder_scans"
} /* analyzing_roworder_intersect */
} /* end-analyzing_range_alternatives */
} /* range_analysis */
}
] /* end-rows_estimation */
},
{
"considered_execution_plans": [
{
"plan_prefix": [
] /* plan_prefix */,
"table": "`test_index`",
"best_access_path": { -- 最优访问路径
"considered_access_paths": [ -- 最终选择的访问路径
{
"rows_to_scan": 106,
"access_type": "scan", -- 访问类型:为scan,全表扫描
"resulting_rows": 106,
"cost": 10.85,
"chosen": true, -- 确定选择
"use_tmp_table": true
}
] /* considered_access_paths */
} /* best_access_path */,
"condition_filtering_pct": 100,
"rows_for_plan": 106,
"cost_for_plan": 10.85,
"sort_cost": 106,
"new_cost_for_plan": 116.85,
"chosen": true
}
] /* considered_execution_plans */
},
{
"attaching_conditions_to_tables": {
"original_condition": "(`test_index`.`name` > '北')",
"attached_conditions_computation": [
] /* attached_conditions_computation */,
"attached_conditions_summary": [
{
"table": "`test_index`",
"attached": "(`test_index`.`name` > '北')"
}
] /* attached_conditions_summary */
} /* attaching_conditions_to_tables */
},
{
"optimizing_distinct_group_by_order_by": {
"simplifying_order_by": {
"original_clause": "`test_index`.`addr`",
"items": [
{
"item": "`test_index`.`addr`"
}
] /* items */,
"resulting_clause_is_simple": true,
"resulting_clause": "`test_index`.`addr`"
} /* simplifying_order_by */
} /* optimizing_distinct_group_by_order_by */
},
{
"reconsidering_access_paths_for_index_ordering": {
"clause": "ORDER BY",
"steps": [
] /* steps */,
"index_order_summary": {
"table": "`test_index`",
"index_provides_order": false,
"order_direction": "undefined",
"index": "unknown",
"plan_changed": false
} /* index_order_summary */
} /* reconsidering_access_paths_for_index_ordering */
},
{
"finalizing_table_conditions": [
{
"table": "`test_index`",
"original_table_condition": "(`test_index`.`name` > '北')",
"final_table_condition ": "(`test_index`.`name` > '北')"
}
] /* finalizing_table_conditions */
},
{
"refine_plan": [
{
"table": "`test_index`"
}
] /* refine_plan */
},
{
"considering_tmp_tables": [
{
"adding_sort_to_table": "test_index"
} /* filesort */
] /* considering_tmp_tables */
}
] /* steps */
} /* join_optimization */
},
{
"join_execution": { -- 第三阶段:SQL执行阶段
"select#": 1,
"steps": [
{
"sorting_table": "test_index",
"filesort_information": [
{
"direction": "asc",
"expression": "`test_index`.`addr`"
}
] /* filesort_information */,
"filesort_priority_queue_optimization": {
"usable": false,
"cause": "not applicable (no LIMIT)"
} /* filesort_priority_queue_optimization */,
"filesort_execution": [
] /* filesort_execution */,
"filesort_summary": {
"memory_available": 262144,
"key_size": 61,
"row_size": 230,
"max_rows_per_buffer": 106,
"num_rows_estimate": 106,
"num_rows_found": 93,
"num_initial_chunks_spilled_to_disk": 0,
"peak_memory_used": 33792,
"sort_algorithm": "std::sort",
"sort_mode": "<fixed_sort_key, packed_additional_fields>"
} /* filesort_summary */
}
] /* steps */
} /* join_execution */
}
] /* steps */
}
FileSort两种排序方式
select * from test_index where name > '北' order by addr;
单路排序
将我们通过where筛序出的结果集,全部放入到sort buffer中,按照order by 的条件进行排序,最终直接返回数据集合。
缺点:所有字段全部放入sort buffer中占用空间
-- mysql 8.0.27 版本
"filesort_summary": { -- 文件排序信息
"memory_available": 262144,
"key_size": 61,
"row_size": 230,
"max_rows_per_buffer": 106,
"num_rows_estimate": 106,
"num_rows_found": 93,
"num_initial_chunks_spilled_to_disk": 0,
"peak_memory_used": 33792,
"sort_algorithm": "std::sort",
"sort_mode": "<fixed_sort_key, packed_additional_fields>"
} /* filesort_summary */
-- mysql 5.6 版本
"filesort_summary": {
"rows": 93, -- 预计扫描行数
"examined_rows": 106, -- 参与排序的行
"number_of_tmp_files": 0,-- 使用临时文件的个数,这个值如果为0代表全部使用的sort_buffer内存排序,否则使用的磁盘文件排序
"sort_buffer_size": 234112,-- 排序缓存的大小,单位Byte
"sort_mode": "<sort_key, additional_fields>" -- 排序方式,这里用的单路排序
} /* filesort_summary */
双路排序
将我们通过where筛序出的结果集当中的主键id(或许是伪列rowid)以及order by的字段放入sort buffer当中,按照order by 的条件进行排序,然后按照排序后的主键id(或许是伪列rowid)进行回表查询出满足所有条件的记录,最终直接返回数据集合。
优点:当sort buffer空间有限时适合使用,因为放入的字段少,占用空间小。
-- mysql 8.0.27 版本 效果等同同mysql版本的单路排序(已失效)
"filesort_summary": {
"memory_available": 262144,
"key_size": 61,
"row_size": 230,
"max_rows_per_buffer": 106,
"num_rows_estimate": 106,
"num_rows_found": 93,
"num_initial_chunks_spilled_to_disk": 0,
"peak_memory_used": 33792,
"sort_algorithm": "std::sort",
"sort_mode": "<fixed_sort_key, packed_additional_fields>"
}
-- mysql 5.6版本
"filesort_summary": {
"rows": 93,
"examined_rows": 106,
"number_of_tmp_files": 0,
"sort_buffer_size": 72416,
"sort_mode": "<sort_key, rowid>" -- 双路排序
} /* filesort_summary */
选择依据
MySQL 通过比较系统变量 max_length_for_sort_data(mysql8.0默认4096字节,mysql5.6默认1024字节) 的大小和需要查询的字段总大小来判断使用哪种排序模式。
- 如果字段的总长度小于max_length_for_sort_data ,那么使用 单路排序模式。
- 如果字段的总长度大于max_length_for_sort_data ,那么使用 双路排序模式。
结束语
记得最后关闭trace工具哟!
-- 关闭trace
set session optimizer_trace="enabled=off",end_markers_in_json=OFF;