UNION · k8s原理与实践

[TOC] # UNION UNION 算子用于将两个查询的结果集进行并集运算。 OceanBase 数据库支持的 UNION 算子包括 UNION ALL、 HASH UNION DISTINCT 和 MERGE UNION DISTINCT。 ## UNION ALL UNION ALL 用于直接对两个查询结果集进行合并输出。如下示例中，Q1 对两个查询使用 UNION ALL 进行联接，使用 UNION ALL 算子进行并集运算。算子执行时依次输出左右子节点所有输出结果。 ~~~ obclient>CREATE TABLE t1(c1 INT PRIMARY KEY, c2 INT); Query OK, 0 rows affected (0.12 sec) obclient>INSERT INTO t1 VALUES(1,1); Query OK, 1 rows affected (0.12 sec) obclient>INSERT INTO t1 VALUES(2,2); Query OK, 1 rows affected (0.12 sec) Q1: obclient>EXPLAIN SELECT c1 FROM t1 UNION ALL SELECT c2 FROM t1\G; *************************** 1. row *************************** Query Plan: ==================================== |ID|OPERATOR |NAME|EST. ROWS|COST| ------------------------------------ |0 |UNION ALL | |4 |74 | |1 | TABLE SCAN|T1 |2 |37 | |2 | TABLE SCAN|T1 |2 |37 | ==================================== Outputs & filters: ------------------------------------- 0 - output([UNION(T1.C1, T1.C2)]), filter(nil) 1 - output([T1.C1]), filter(nil), access([T1.C1]), partitions(p0) 2 - output([T1.C2]), filter(nil), access([T1.C2]), partitions(p0) ~~~ 上述示例中，执行计划展示中的 outputs & filters 详细列出了 UNION ALL 算子的输出信息如下： <div data-card-value="data:%7B%22rows%22%3A3%2C%22cols%22%3A2%2C%22html%22%3A%22%3Ctable%20class%3D%5C%22lake-table%5C%22%20style%3D%5C%22width%3A%20748px%3B%5C%22%3E%3Ccolgroup%3E%3Ccol%20span%3D%5C%221%5C%22%20width%3D%5C%22139%5C%22%20%2F%3E%3Ccol%20span%3D%5C%221%5C%22%20width%3D%5C%22609%5C%22%20%2F%3E%3C%2Fcolgroup%3E%3Ctbody%3E%3Ctr%20style%3D%5C%22height%3A%2033px%3B%5C%22%3E%3Ctd%20style%3D%5C%22vertical-align%3A%20top%3B%5C%22%3E%3Cp%20data-lake-id%3D%5C%22b92b778e0b6d42b6559f0dbdfd18725c%5C%22%3E%3Cspan%20class%3D%5C%22lake-fontsize-11%5C%22%20style%3D%5C%22color%3A%20%23262626%3B%5C%22%3E%3Cstrong%3E%E4%BF%A1%E6%81%AF%E5%90%8D%E7%A7%B0%3C%2Fstrong%3E%3C%2Fspan%3E%3C%2Fp%3E%3C%2Ftd%3E%3Ctd%20rowspan%3D%5C%221%5C%22%20colspan%3D%5C%221%5C%22%20style%3D%5C%22vertical-align%3A%20top%3B%5C%22%3E%3Cspan%20class%3D%5C%22lake-fontsize-11%5C%22%20style%3D%5C%22color%3A%20%23262626%3B%5C%22%3E%3Cstrong%3E%3Cspan%3E%E5%90%AB%E4%B9%89%3C%2Fspan%3E%3C%2Fstrong%3E%3C%2Fspan%3E%3C%2Ftd%3E%3C%2Ftr%3E%3Ctr%20style%3D%5C%22height%3A%2033px%3B%5C%22%3E%3Ctd%20rowspan%3D%5C%221%5C%22%20colspan%3D%5C%221%5C%22%20style%3D%5C%22vertical-align%3A%20top%3B%5C%22%3E%3Cspan%20class%3D%5C%22lake-fontsize-11%5C%22%3Eoutput%3C%2Fspan%3E%3C%2Ftd%3E%3Ctd%20rowspan%3D%5C%221%5C%22%20colspan%3D%5C%221%5C%22%20style%3D%5C%22vertical-align%3A%20top%3B%5C%22%3E%3Cspan%20class%3D%5C%22lake-fontsize-11%5C%22%3E%E8%AF%A5%3Cspan%3E%E7%AE%97%E5%AD%90%3C%2Fspan%3E%3Cspan%3E%E7%9A%84%E8%BE%93%E5%87%BA%E8%A1%A8%E8%BE%BE%E5%BC%8F%3C%2Fspan%3E%E3%80%82%3C%2Fspan%3E%3C%2Ftd%3E%3C%2Ftr%3E%3Ctr%20style%3D%5C%22height%3A%2033px%3B%5C%22%3E%3Ctd%20rowspan%3D%5C%221%5C%22%20colspan%3D%5C%221%5C%22%20style%3D%5C%22vertical-align%3A%20top%3B%5C%22%3E%3Cspan%20class%3D%5C%22lake-fontsize-11%5C%22%3Efilter%3C%2Fspan%3E%3C%2Ftd%3E%3Ctd%20rowspan%3D%5C%221%5C%22%20colspan%3D%5C%221%5C%22%20style%3D%5C%22vertical-align%3A%20top%3B%5C%22%3E%3Cspan%20class%3D%5C%22lake-fontsize-11%5C%22%3E%3Cspan%3E%3Cspan%3E%E8%AF%A5%E7%AE%97%E5%AD%90%E4%B8%8A%E7%9A%84%E8%BF%87%E6%BB%A4%E6%9D%A1%E4%BB%B6%E3%80%82%3C%2Fspan%3E%3C%2Fspan%3E%3Cspan%3E%E7%94%B1%E4%BA%8E%3C%2Fspan%3E%3Cspan%20class%3D%5C%22lake-fontsize-11%5C%22%3E%3Cspan%3Eunion%C2%A0%3C%2Fspan%3E%3C%2Fspan%3E%3Cspan%3E%3Cspan%3E%E7%AE%97%E5%AD%90%E6%B2%A1%E6%9C%89%20filter%3C%2Fspan%3E%3C%2Fspan%3E%3Cspan%3E%EF%BC%8C%E6%89%80%E4%BB%A5%E4%B8%BA%20nil%3C%2Fspan%3E%E3%80%82%3C%2Fspan%3E%3C%2Ftd%3E%3C%2Ftr%3E%3C%2Ftbody%3E%3C%2Ftable%3E%22%2C%22margin%22%3Afalse%2C%22hideBorder%22%3Afalse%2C%22id%22%3A%22ZczYL%22%7D" id="div-5rc-giy-1k5" data-tag="div" class="div"><table data-tag="table" id="table-33a-t1s-284" class="table"><colgroup span="1" width="139" data-tag="col" id="col-454-p5l-esl" colwidth="1*" colnum="1" colname="col1" style="width:50%" class="col"></colgroup><colgroup span="1" width="609" data-tag="col" id="col-e77-y81-2qz" colwidth="1*" colnum="2" colname="col2" style="width:50%" class="col"></colgroup><thead id="thead-jy6-28n-fbj" class="thead"><tr id="tr-xji-2kt-m4g"><th id="td-4hb-7e6-1ey"><p id="p-h3q-7vg-osj"><b>信息名称</b></p></th><th id="td-jsf-ljj-liz"><p id="p-zee-dts-gk5"><b>含义</b></p></th></tr></thead><tbody data-tag="tbody" id="tbody-tcu-dsz-2di" class="tbody"><tr data-tag="tr" id="tr-x6p-1bh-68a" class="tr"><td data-tag="td" id="td-o8f-4i7-jvn" class="td"><p id="p-rpa-0xl-f04">output</p></td><td data-tag="td" id="td-6m2-vhb-zy7" class="td"><p id="p-7fc-42p-4w8">该算子的输出表达式。</p></td></tr><tr data-tag="tr" id="tr-l4x-d51-vih" class="tr"><td data-tag="td" id="td-v2v-7im-s36" class="td"><p id="p-prr-a56-e1l">filter</p></td><td data-tag="td" id="td-2kc-val-yzj" class="td"><p id="p-283-lci-ad2">该算子上的过滤条件。</p><p id="p-0yy-04c-scm">由于示例中 UNION ALL 算子没有设置 filter，所以为 nil。</p></td></tr></tbody></table></div> ## MERGE UNION DISTINCT MERGE UNION DISTINCT 用于对结果集进行并集、去重后进行输出。如下示例中，Q2 对两个查询使用 UNION DISTINCT 进行联接， c1 有可用排序，0 号算子生成 MERGE UNION DISTINCT 进行取并集、去重。由于 c2 无可用排序，所以在 3 号算子上分配了 SORT 算子进行排序。算子执行时从左右子节点读取有序输入，进行合并得到有序输出并去重。 ~~~ Q2: obclient>EXPLAIN SELECT c1 FROM t1 UNION SELECT c2 FROM t1\G; *************************** 1. row *************************** Query Plan: ============================================= |ID|OPERATOR |NAME|EST. ROWS|COST| --------------------------------------------- |0 |MERGE UNION DISTINCT| |4 |77 | |1 | TABLE SCAN |T1 |2 |37 | |2 | SORT | |2 |39 | |3 | TABLE SCAN |T1 |2 |37 | ============================================= Outputs & filters: ------------------------------------- 0 - output([UNION(T1.C1, T1.C2)]), filter(nil) 1 - output([T1.C1]), filter(nil), access([T1.C1]), partitions(p0) 2 - output([T1.C2]), filter(nil), sort_keys([T1.C2, ASC]) 3 - output([T1.C2]), filter(nil), access([T1.C2]), partitions(p0) ~~~ 上述示例的执行计划展示中的 outputs & filters 详细列出了 MERGE UNION DISTINCT 算子的输出信息，字段的含义与 UNION ALL 算子相同。 ## HASH UNION DISTINCT HASH UNION DISTINCT 用于对结果集进行并集、去重后进行输出。如下示例中，Q3 对两个查询使用 UNION DISTINCT 进行联接，无可利用排序，0 号算子使用 HASH UNION DISTINCT 进行并集、去重。算子执行时读取左右子节点输出，建立哈希表进行去重，最终输出去重后结果。 ~~~ Q3: obclient>EXPLAIN SELECT c2 FROM t1 UNION SELECT c2 FROM t1\G; *************************** 1. row *************************** Query Plan: ============================================ |ID|OPERATOR |NAME|EST. ROWS|COST| -------------------------------------------- |0 |HASH UNION DISTINCT| |4 |77 | |1 | TABLE SCAN |T1 |2 |37 | |2 | TABLE SCAN |T1 |2 |37 | ============================================ Outputs & filters: ------------------------------------- 0 - output([UNION(T1.C2, T1.C2)]), filter(nil) 1 - output([T1.C2]), filter(nil), access([T1.C2]), partitions(p0) 2 - output([T1.C2]), filter(nil), access([T1.C2]), partitions(p0) ~~~ 上述示例的执行计划展示中的 outputs & filters 详细列出了 HASH UNION DISTINCT 算子的输出信息，字段的含义与 UNION ALL 算子相同。