Deep into MySQL QEP with optimizer_trace
2013-01-05 19:32 心中无码 阅读(564) 评论(0) 编辑 收藏 举报
Deep into MySQL QEP with optimizer_trace
Louis Hust
0 Preface
MySQL5.6 add a new option to show QEP more deeply-optimizer_trace. This option is very different from EXPLAIN which just showes the table access method, index using, table join order and so on. But optimizer_trace output every steps of how the Optimizer processes query, such as join prepare, join optimizer and join exec. Each of the three steps contains many substeps such as expanded_query, equality_propagation, trivial_condition_removal, etc.
0 Show QEP in deep
# Turn tracing on (it's off by default): SET optimizer_trace="enabled=on"; SELECT * FROM t1 WHERE c1=1 or c1=100;# your query here SELECT * FROM INFORMATION_SCHEMA.OPTIMIZER_TRACE; # possibly more queries... # When done with tracing, disable it: SET optimizer_trace="enabled=off";
An example below:
{ "steps": [ { "join_preparation": { "select#": 1, "steps": [ { "expanded_query": "/* select#1 */ select `t1`.`c1` AS `c1`,`t1`.`c2` AS `c2` from `t1` where ((`t1`.`c1` = 1) or (`t1`.`c1` = 100))" } ] } }, { "join_optimization": { "select#": 1, "steps": [ { "condition_processing": { "condition": "WHERE", "original_condition": "((`t1`.`c1` = 1) or (`t1`.`c1` = 100))", "steps": [ { "transformation": "equality_propagation", "resulting_condition": "(multiple equal(1, `t1`.`c1`) or multiple equal(100, `t1`.`c1`))" }, { "transformation": "constant_propagation", "resulting_condition": "(multiple equal(1, `t1`.`c1`) or multiple equal(100, `t1`.`c1`))" }, { "transformation": "trivial_condition_removal", "resulting_condition": "(multiple equal(1, `t1`.`c1`) or multiple equal(100, `t1`.`c1`))" } ] } }, { "table_dependencies": [ { "table": "`t1`", "row_may_be_null": false, "map_bit": 0, "depends_on_map_bits": [ ] } ] }, { "ref_optimizer_key_uses": [ ] }, { "rows_estimation": [ { "table": "`t1`", "range_analysis": { "table_scan": { "rows": 10157, "cost": 2057.5 }, "potential_range_indices": [ { "index": "c1", "usable": true, "key_parts": [ "c1" ] } ], "setup_range_conditions": [ ], "group_index_range": { "chosen": false, "cause": "not_group_by_or_distinct" }, "analyzing_range_alternatives": { "range_scan_alternatives": [ { "index": "c1", "ranges": [ "1 <= c1 <= 1", "100 <= c1 <= 100" ], "index_dives_for_eq_ranges": true, "rowid_ordered": false, "using_mrr": false, "index_only": false, "rows": 2, "cost": 4.41, "chosen": true } ], "analyzing_roworder_intersect": { "usable": false, "cause": "too_few_roworder_scans" } }, "chosen_range_access_summary": { "range_access_plan": { "type": "range_scan", "index": "c1", "rows": 2, "ranges": [ "1 <= c1 <= 1", "100 <= c1 <= 100" ] }, "rows_for_plan": 2, "cost_for_plan": 4.41, "chosen": true } } } ] }, { "considered_execution_plans": [ { "plan_prefix": [ ], "table": "`t1`", "best_access_path": { "considered_access_paths": [ { "access_type": "range", "rows": 2, "cost": 4.81, "chosen": true } ] }, "cost_for_plan": 4.81, "rows_for_plan": 2, "chosen": true } ] }, { "attaching_conditions_to_tables": { "original_condition": "((`t1`.`c1` = 1) or (`t1`.`c1` = 100))", "attached_conditions_computation": [ ], "attached_conditions_summary": [ { "table": "`t1`", "attached": "((`t1`.`c1` = 1) or (`t1`.`c1` = 100))" } ] } }, { "refine_plan": [ { "table": "`t1`", "pushed_index_condition": "((`t1`.`c1` = 1) or (`t1`.`c1` = 100))", "table_condition_attached": null, "access_type": "range" } ] } ] } }, { "join_explain": { "select#": 1, "steps": [ ] } } ] }
As seen above, we can get almost every step in Optimizer processing, from setup conditions to plan choosing. The result is output according to order of the code executing, but EXPLAIN is just a print of JOIN_TAB. I can not explain every step, cause I do not know every steps.
Actually the Optimizer code of MySQL is hard to read, with the trace, we can read the code more easily.
References
File translated from
TEX
by
TTH,
version 4.03.
On 5 Jan 2013, 19:30.
【推荐】国内首个AI IDE,深度理解中文开发场景,立即下载体验Trae
【推荐】编程新体验,更懂你的AI,立即体验豆包MarsCode编程助手
【推荐】抖音旗下AI助手豆包,你的智能百科全书,全免费不限次数
【推荐】轻量又高性能的 SSH 工具 IShell:AI 加持,快人一步
· Linux系列:如何用 C#调用 C方法造成内存泄露
· AI与.NET技术实操系列(二):开始使用ML.NET
· 记一次.NET内存居高不下排查解决与启示
· 探究高空视频全景AR技术的实现原理
· 理解Rust引用及其生命周期标识(上)
· 阿里最新开源QwQ-32B,效果媲美deepseek-r1满血版,部署成本又又又降低了!
· 单线程的Redis速度为什么快?
· 展开说说关于C#中ORM框架的用法!
· SQL Server 2025 AI相关能力初探
· Pantheons:用 TypeScript 打造主流大模型对话的一站式集成库