{"id":378,"date":"2020-10-01T10:00:07","date_gmt":"2020-10-01T02:00:07","guid":{"rendered":"https:\/\/www.mengqingzhong.com\/?p=378"},"modified":"2024-01-14T11:37:31","modified_gmt":"2024-01-14T03:37:31","slug":"postgresql-index-bloom-10","status":"publish","type":"post","link":"https:\/\/www.mengqingzhong.com\/2020\/10\/01\/postgresql-index-bloom-10\/","title":{"rendered":"PostgreSQL\u7d22\u5f15(10) \u2013 Bloom"},"content":{"rendered":"<p class=info>\n\u672c\u6587\u7ffb\u8bd1\u81ea<a target='_blank' href=\"https:\/\/habr.com\/en\/company\/postgrespro\/blog\/452968\/\" rel=\"noopener noreferrer\">Egor Rogov\u7684\u82f1\u6587\u535a\u5ba2<\/a>\uff0c\u4e14\u5df2\u5f81\u5f97Egor Rogov\u7684\u540c\u610f\u3002\u8bd1\u8005\u505a\u4e86\u5c11\u91cf\u4fee\u6539\u3002\n<\/p>\n<p class=info>\n\u5ba1\u6821\uff1a\u7eaa\u6600\u7ea2\uff0c\u4e2d\u56fd\u4eba\u6c11\u5927\u5b66\u4fe1\u606f\u5b66\u9662\u5728\u8bfb\u535a\u58eb\u751f\u3002\n<\/p>\n<p>\u5728\u524d\u9762\u51e0\u7bc7\u6587\u7ae0\u4e2d\uff0c\u6211\u4eec\u8ba8\u8bba\u4e86\u901a\u7528\u7d22\u5f15\u5f15\u64ce\u3001AM\u7684\u63a5\u53e3\u548c\u516d\u79cdAM\uff08Hash\u7d22\u5f15\u3001B-tree\u7d22\u5f15\u3001GiST\u7d22\u5f15\u3001SP-GiST\u7d22\u5f15\u3001GIN\u7d22\u5f15\u3001RUM\u7d22\u5f15\u3001BRIN\u7d22\u5f15\uff09\u3002\u8fd9\u7bc7\u6587\u7ae0\u5c06\u4ecb\u7ecdBloom\u7d22\u5f15\u3002<\/p>\n<h1>1. \u57fa\u672c\u6982\u5ff5<\/h1>\n<p>\u4e00\u4e2a\u7ecf\u5178\u7684Bloom filter\u662f\u4e00\u4e2a\u6570\u636e\u7ed3\u6784\uff0c\u4f7f\u7528\u5b83\u53ef\u4ee5\u5feb\u901f\u7684\u68c0\u67e5\u4e00\u4e2a\u5143\u7d20\u662f\u5426\u5c5e\u4e8e\u4e00\u4e2a\u96c6\u5408\u3002\u8fc7\u6ee4\u5668\u88ab\u9ad8\u5ea6\u538b\u7f29\uff0c\u4f46\u662f\u5b83\u53ef\u80fd\u4f1a\u4ea7\u751f\u5047\u9633\u6027\u7684\u7ed3\u679c\uff1a\u5b83\u53ef\u80fd\u4f1a\u8bef\u8ba4\u4e3a\u4e00\u4e2a\u5143\u7d20\u5728\u96c6\u5408\u4e2d\u5b58\u5728\uff0c\u5b9e\u9645\u8fd9\u4e2a\u5143\u7d20\u5728\u96c6\u5408\u4e2d\u5e76\u4e0d\u5b58\u5728\u3002\u5982\u679c\u5b83\u8ba4\u4e3a\u4e00\u4e2a\u5143\u7d20\u5728\u96c6\u5408\u4e2d\u4e0d\u5b58\u5728\uff0c\u90a3\u4e48\u8fd9\u4e2a\u5143\u7d20\u5728\u96c6\u5408\u4e2d\u80af\u5b9a\u4e0d\u5b58\u5728\u3002<\/p>\n<p>\u8fc7\u6ee4\u5668\u662f\u4e00\u4e2a\u7531\u6bd4\u7279\u4f4d\u7ec4\u6210\u7684\u6570\u7ec4\uff08\u4e5f\u88ab\u79f0\u4e3a\u7b7e\u540d\uff09\uff0c\u6240\u6709\u6bd4\u7279\u6700\u521d\u90fd\u88ab\u521d\u59cb\u5316\u6210\u96f6\u3002\u5f53\u5411\u96c6\u5408\u4e2d\u6dfb\u52a0\u4e00\u4e2a\u5143\u7d20\u65f6\uff0c\u6211\u4eec\u4f1a\u4f7f\u7528\u591a\u4e2ahash\u51fd\u6570\u5bf9\u8fd9\u4e2a\u5143\u7d20\u8ba1\u7b97hash\u503c\uff0c\u628a\u6bcf\u4e2ahash\u503c\u5bf9\u5e94\u7684\u6bd4\u7279\u4f4d\u7f6e\u4e3a1\u3002\u6240\u4ee5\uff0c\u5f53\u4e00\u4e2a\u5143\u7d20\u5bf9\u5e94\u7684\u6240\u6709\u6bd4\u7279\u4f4d\u90fd\u4e3a1\u65f6\uff0c\u8fd9\u4e2a\u5143\u7d20\u53ef\u80fd\u5c5e\u4e8e\u8fd9\u4e2a\u96c6\u5408\u3002\u4f46\u662f\u5f53\u4efb\u610f\u4e00\u4f4d\u4e3a0\u65f6\uff0c\u5219\u8fd9\u4e2a\u5143\u7d20\u80af\u5b9a\u4e0d\u5c5e\u4e8e\u8fd9\u4e2a\u96c6\u5408\u3002<\/p>\n<p>\u901a\u8fc7\u6539\u53d8\u7b7e\u540d\u7684\u957f\u5ea6\uff0c\u6211\u4eec\u53ef\u4ee5\u5728\u7d22\u5f15\u4f53\u79ef\u548c\u5047\u9633\u6027\u7684\u6982\u7387\u4e4b\u95f4\u5f97\u5230\u4e00\u4e2a\u5e73\u8861\u3002Bloom\u7d22\u5f15\u7684\u5e94\u7528\u975e\u5e38\u5e7f\u6cdb\uff0c\u76f8\u5f53\u591a\u7684\u5bbd\u8868\u90fd\u5728\u6bcf\u4e2a\u5b57\u6bb5\u4e0a\u4f7f\u7528\u8fc7\u6ee4\u5668\u3002\u8fd9\u79cdAM\uff0c\u548cBRIN\u7c7b\u4f3c\uff0c\u4e5f\u53ef\u4ee5\u88ab\u770b\u4f5c\u662f\u987a\u5e8f\u626b\u63cf\u7684\u52a0\u901f\u5668\uff1a\u7d22\u5f15\u8fd4\u56de\u7684\u6240\u6709\u5143\u7ec4\uff0c\u90fd\u9700\u8981\u88abrecheck\uff0c\u4f46\u662f\u6709\u53ef\u80fd\u907f\u514d\u626b\u63cf\u5927\u90e8\u5206\u4e0d\u6ee1\u8db3\u6761\u4ef6\u7684\u5143\u7ec4\u3002<\/p>\n<h1>2. \u7ed3\u6784<\/h1>\n<p>\u6211\u4eec\u5728GiST\u7d22\u5f15\u4e2d\u8ba8\u8bba\u8fc7\u7b7e\u540d\u3002\u548c\u8fd9\u4e9b\u6811\u4e0d\u540c\uff0cBloom filter\u662f\u4e00\u4e2a\u5e73\u5766\u7684\u7ed3\u6784\u3002\u5b83\u7684\u7b2c0\u9875\u662f\u5143\u9875\u9762\uff0c\u540e\u9762\u6240\u6709\u7684\u9875\u9762\u90fd\u662f\u5305\u542b\u7d22\u5f15\u884c\u7684\u5e38\u89c4\u9875\u9762\u3002\u6bcf\u4e2a\u7d22\u5f15\u884c\u5305\u542b\u4e00\u4e2a\u7b7e\u540d\u548c\u4e00\u4e2a\u6307\u5411\u8868\u4e2d\u884c\u7684TID\uff0c\u5982\u4e0b\u56fe\u6240\u793a\uff1a<\/p>\n<p><a class=\"wp-editor-md-post-content-link\" href=\"https:\/\/www.mengqingzhong.com\/wp-content\/uploads\/2020\/12\/wp_editor_md_22810dc55c16e9d7539d51ba5965998b.jpg\"><img decoding=\"async\" src=\"https:\/\/www.mengqingzhong.com\/wp-content\/uploads\/2020\/12\/wp_editor_md_22810dc55c16e9d7539d51ba5965998b.jpg\" alt=\"\" \/><\/a><\/p>\n<h2>2.1 \u53c2\u6570<\/h2>\n<p>\u5728\u521b\u5efaBloom\u7d22\u5f15\u65f6\uff0c\u53ef\u4ee5\u6307\u5b9a\u6bcf\u4e2a\u7b7e\u540d\u7684\u603b\u957f\u5ea6\uff0c\u4e5f\u53ef\u4ee5\u6307\u5b9a\u6bcf\u4e2a\u5b57\u6bb5\u9700\u8981\u5c06\u7b7e\u540d\u4e2d\u591a\u5c11\u4f4d\u7f6e1\u3002<\/p>\n<pre><code class=\"language-sql \">create index on ... using bloom(...) with (length=..., col1=..., col2=..., ...);\n<\/code><\/pre>\n<p>\u6211\u4eec\u5e94\u8be5\u600e\u6837\u9009\u53d6\u5408\u9002\u7684\u503c\u5462\uff1f<a class=\"wp-editor-md-post-content-link\" href=\"https:\/\/en.wikipedia.org\/wiki\/Bloom_filter\">\u7406\u8bba\u8868\u660e<\/a>\uff0c\u7ed9\u5b9a\u4e00\u4e2a\u5047\u9633\u6027\u7684\u6982\u7387\uff0c\u7b7e\u540d\u7684\u6700\u4f18\u957f\u5ea6\u53ef\u4ee5\u7528$m=-n\\log_2 p\/\\ln 2$\u4f30\u8ba1\uff0c\u5176\u4e2d$n$\u662f\u7d22\u5f15\u4e2d\u5b57\u6bb5\u7684\u4e2a\u6570\uff0c\u7b7e\u540d\u4e2d\u9700\u8981\u8bbe\u7f6e\u4e3a1\u7684\u4f4d\u6570$k$\u4e3a$k=-\\log_2 p$\u3002<\/p>\n<p>\u7b7e\u540d\u7684\u957f\u5ea6\u4f1a\u88ab\u5411\u4e0a\u53d6\u6574\u52302\u5b57\u8282\u7684\u500d\u6570\u3002<\/p>\n<p>\u9700\u8981\u6ce8\u610f\u51e0\u70b9\uff1a<\/p>\n<ul>\n<li>\u5047\u9633\u6027\u7684\u6982\u7387p\u4e0e\u4e00\u4e2afilter\u76f8\u5173\uff0c\u6211\u4eec\u671f\u671b\u5728\u4e00\u6b21\u626b\u63cf\u4e2d\u5f97\u5230Np\u4e2a\u5047\u9633\u6027\u7684\u7ed3\u679c\u3002\u4f8b\u5982\u4e00\u4e2a\u8868\u6709100\u4e07\u884c\uff0c\u6982\u7387\u4e3a0.01\uff0c\u5e73\u5747\u6765\u8bb2\uff0c\u53ef\u4ee5\u89c2\u5bdf\u5230Rows Removed by Index Recheck\u7684\u6570\u91cf\u4e3a10000\u3002<\/li>\n<li>Bloom filter\u662f\u4e00\u4e2a\u6982\u7387\u7ed3\u6784\u3002\u53ea\u6709\u5728\u5927\u91cf\u6570\u636e\u7684\u573a\u666f\u4e0b\uff0c\u8fd9\u79cd\u6982\u7387\u624d\u6709\u610f\u4e49\u3002<\/li>\n<li>\u4e0a\u9762\u7684\u4f30\u8ba1\u57fa\u4e8e\u4e00\u4e2a\u7406\u60f3\u7684\u6570\u5b66\u6a21\u578b\u548c\u4e00\u4e9b\u5047\u8bbe\u3002\u5b9e\u9645\u4e2d\uff0c\u7ed3\u679c\u8981\u7cdf\u7cd5\u4e00\u4e9b\u3002\u6240\u4ee5\uff0c\u4e0d\u8981\u8fc7\u5206\u4f9d\u8d56\u8fd9\u4e2a\u516c\u5f0f\uff1a\u5b83\u4eec\u4ec5\u4ec5\u63d0\u4f9b\u4e86\u4e00\u79cd\u9009\u53d6\u521d\u59cb\u503c\u7684\u65b9\u6cd5\u3002<\/li>\n<li>\u5bf9\u6bcf\u4e2a\u5b57\u6bb5\u6765\u8bf4\uff0c\u8fd9\u4e2aAM\u5141\u8bb8\u6211\u4eec\u9009\u62e9\u5728\u7b7e\u540d\u4e2d\u7f6e1\u7684\u4f4d\u6570\u3002\u5b9e\u9645\u4e2d\u6700\u4f18\u7684\u4f4d\u6570\u4e0e\u6570\u636e\u7684\u5206\u5e03\u6709\u5173\u7cfb\u3002\u5982\u679c\u60f3\u6df1\u6316\uff0c\u8bf7\u53c2\u8003<a class=\"wp-editor-md-post-content-link\" href=\"http:\/\/blog.coelho.net\/database\/2016\/12\/11\/postgresql-bloom-index.html\">\u8fd9\u7bc7\u6587\u7ae0<\/a>\u3002\u4f46\u662f\u4e5f\u522b\u5fd8\u4e86\u4e0a\u4e00\u6761\u3002<\/li>\n<\/ul>\n<h2>2.2 \u66f4\u65b0<\/h2>\n<p>\u5f53\u5411\u8868\u4e2d\u63d2\u5165\u4e00\u884c\u6570\u636e\u65f6\uff0c\u7b7e\u540d\u88ab\u8ba1\u7b97\u51fa\u6765\uff1a\u5bf9\u6240\u6709\u88ab\u7d22\u5f15\u7684\u5b57\u6bb5\uff0c\u4e0e\u5b83\u4eec\u5bf9\u5e94\u7684\u4f4d\u5168\u90e8\u88ab\u7f6e\u4e3a1\u3002\u7406\u8bba\u4e0a\u6765\u8bf4\uff0c\u6211\u4eec\u9700\u8981k\u4e2a\u4e0d\u540c\u7684hash\u51fd\u6570\uff0c\u5b9e\u9645\u4e2d\uff0c\u53ea\u9700\u8981\u8fde\u7eed\u751f\u6210k\u4e2a\u968f\u673a\u6570\u5373\u53ef\u3002\u968f\u673a\u6570\u7684\u79cd\u5b50\u6709hash\u503c\u51b3\u5b9a\u3002<\/p>\n<p>\u5f53\u4e00\u884c\u6570\u636e\u88ab\u5220\u9664\u65f6\uff0c\u6574\u4e2a\u7b7e\u540d\u5168\u90e8\u88ab\u5220\u9664\u3002<\/p>\n<p>\u66f4\u65b0\u64cd\u4f5c\u7531\u4e24\u4e2a\u64cd\u4f5c\u7ec4\u5408\u800c\u6210\uff1a\u5220\u9664\u65e7\u7248\u672c\uff0c\u63d2\u5165\u65b0\u7248\u672c\u3002<\/p>\n<h2>2.3 \u626b\u63cf<\/h2>\n<p>\u56e0\u4e3aBloom\u53ea\u80fd\u7528\u6765\u68c0\u67e5\u67d0\u4e2a\u5143\u7d20\u662f\u5426\u5c5e\u4e8e\u67d0\u4e2a\u96c6\u5408\uff0c\u6240\u4ee5Bloom\u7d22\u5f15\u652f\u6301\u7684\u64cd\u4f5c\u53ea\u6709\u7b49\u503c\u68c0\u67e5\uff0c\u5c31\u50cfHash\u7d22\u5f15\u4e00\u6837\u3002<\/p>\n<p>\u4e0a\u6587\u5df2\u7ecf\u63d0\u5230\uff0cBloom\u7d22\u5f15\u662f\u4e00\u4e2a\u5e73\u5766\u7684\u7ed3\u6784\uff0c\u6240\u4ee5\u5728\u8bbf\u95ee\u5b83\u7684\u65f6\u5019\uff0c\u603b\u9700\u8981\u628a\u5b83\u6309\u987a\u5e8f\u5b8c\u6574\u7684\u8bfb\u51fa\u6765\u3002\u5728\u8bfb\u53d6\u7684\u8fc7\u7a0b\u4e2d\uff0c\u5c06\u4f4d\u56fe\u6784\u9020\u51fa\u6765\uff0c\u8fd9\u4e2a\u4f4d\u56fe\u7528\u6765\u626b\u63cf\u57fa\u8868\u3002<\/p>\n<p>\u8bfb\u53d6Bloom\u7d22\u5f15\u65f6\uff0c\u5b9e\u9645\u5c31\u662f\u5bf9\u7d22\u5f15\u8fdb\u884c\u5168\u90e8\u626b\u63cf\u3002\u4e3a\u4e86\u9632\u6b62\u626b\u63cf\u8fc7\u7a0b\u628a\u4e00\u4e9b\u6709\u7528\u7684\u7f13\u5b58\u6362\u51fa\uff0c\u5728\u8bfb\u53d6\u65f6\u4f1a\u4f7f\u7528\u4e00\u4e2a\u5c0f\u7684\u7f13\u51b2\u533a\u73af\uff0c\u5c31\u50cf\u5bf9\u57fa\u8868\u8fdb\u884c\u5168\u8868\u626b\u63cf\u4e00\u6837\u3002<\/p>\n<p>\u6211\u4eec\u9700\u8981\u6ce8\u610f\u5230\uff0cBloom\u7d22\u5f15\u7684\u4f53\u79ef\u8d8a\u5927\uff0c\u4f18\u5316\u5668\u5bf9\u5b83\u8d8a\u4e0d\u611f\u5174\u8da3\u3002\u8fd9\u4e2a\u4f9d\u8d56\u662f\u7ebf\u6027\u7684\uff0c\u8fd9\u548c\u6811\u72b6\u7684\u7d22\u5f15\u4e0d\u540c\u3002<\/p>\n<h1>3. \u793a\u4f8b<\/h1>\n<h2>3.1 \u8868<\/h2>\n<p>\u6211\u4eec\u7ee7\u7eed\u4f7f\u7528\u4e0a\u4e00\u7bc7\u6587\u7ae0\u4e2d\u7684\u8868flights_bi\u3002\u8868\u7684\u5927\u5c0f\u7ea64GB\uff0c\u603b\u5171\u5305\u542b\u5927\u7ea63000\u4e07\u884c\u3002\u8868\u7684\u5b9a\u4e49\u5982\u4e0b\uff1a<\/p>\n<pre><code class=\"\">\\d flights_bi\n                          Table \"bookings.flights_bi\"\n       Column       |           Type           | Collation | Nullable | Default \n--------------------+--------------------------+-----------+----------+---------\n airport_code       | character(3)             |           |          | \n airport_coord      | point                    |           |          | \n airport_utc_offset | interval                 |           |          | \n flight_no          | character(6)             |           |          | \n flight_type        | text                     |           |          | \n scheduled_time     | timestamp with time zone |           |          | \n actual_time        | timestamp with time zone |           |          | \n aircraft_code      | character(3)             |           |          | \n seat_no            | character varying(4)     |           |          | \n fare_conditions    | character varying(10)    |           |          | \n passenger_id       | character varying(20)    |           |          | \n passenger_name     | text                     |           |          | \n<\/code><\/pre>\n<pre><code class=\"language-sql \">create extension bloom;\n<\/code><\/pre>\n<p>\u5728BRIN\u7d22\u5f15\u4e2d\uff0c\u6211\u4eec\u4f7f\u7528\u4e863\u4e2a\u5b57\u6bb5(scheduled_time, actual_time, airport_utc_offset)\u3002\u56e0\u4e3aBloom\u7d22\u5f15\u4e0d\u9700\u8981\u4f9d\u8d56\u6570\u636e\u5728\u7269\u7406\u4e0a\u6709\u5e8f\uff0c\u6211\u4eec\u5c1d\u8bd5\u5728\u7d22\u5f15\u4e2d\u4f7f\u7528\u8868\u4e2d\u51e0\u4e4e\u6240\u6709\u7684\u5b57\u6bb5\u3002\u5f53\u7136\uff0c\u6211\u4eec\u4e5f\u8981\u6392\u9664\u65f6\u95f4\u5b57\u6bb5\uff08scheduled_time\uff0cactual_tim\uff09\uff1a\u56e0\u4e3a\u5b83\u53ea\u652f\u6301\u7b49\u503c\u67e5\u8be2\uff0c\u5728\u4e00\u4e2a\u7cbe\u786e\u7684\u65f6\u95f4\u4e0a\u505a\u7b49\u503c\u67e5\u8be2\u57fa\u672c\u6ca1\u6709\u610f\u4e49\u3002\u867d\u7136\u53ef\u4ee5\u521b\u5efa\u8868\u8fbe\u5f0f\u7d22\u5f15\uff0c\u4f46\u6211\u4eec\u8fd8\u662f\u4e0d\u8d39\u529b\u4e86\u3002\u6211\u4eec\u6392\u9664\u673a\u573a\u7684\u5730\u7406\u5750\u6807airport_coord\uff1apoint\u7c7b\u578b\u8fd8\u4e0d\u88ab\u652f\u6301\u3002<\/p>\n<p>\u4e3a\u4e86\u8ba1\u7b97\u53c2\u6570\u7684\u65f6\uff0c\u6211\u4eec\u628a\u5047\u9633\u6027\u7684\u6982\u7387\u8bbe\u7f6e\u4e3a0.01\uff0c\u5e26\u5165\u4e0a\u9762\u7684\u516c\u5f0f\uff0c\u53ef\u4ee5\u8ba1\u7b97\u51fa\uff0c\u7b7e\u540d\u9700\u898196\u4e2a\u6bd4\u7279\uff0c\u5efa\u8bae\u5bf9\u6bcf\u4e2a\u5b57\u6bb5\u628a\u7b7e\u540d\u4e2d7\u4f4d\u7f6e1\u3002\u7d22\u5f15\u4f30\u8ba1\u4e3a515MB\uff08\u5927\u7ea6\u4e3a\u57fa\u8868\u5927\u5c0f\u7684\u516b\u5206\u4e4b\u4e00\uff09\u3002<\/p>\n<p>\u867d\u7136\u4f7f\u752816\u4f4d\u7b7e\u540d\uff0c\u7d22\u5f15\u7684\u4f53\u79ef\u66f4\u5c0f\uff0c\u4f46\u662f\u8fd9\u4f1a\u5bfc\u81f450%\u7684\u5047\u9633\u6027\uff0c\u8fd9\u592a\u5dee\u4e86\u3002<\/p>\n<h2>3.2 \u64cd\u4f5c\u7b26\u7c7b<\/h2>\n<p>\u9996\u5148\u521b\u5efa\u7d22\u5f15\uff1a<\/p>\n<pre><code class=\"language-sql \">create index flights_bi_bloom on flights_bi\nusing bloom(airport_code, airport_utc_offset, flight_no, flight_type, aircraft_code, seat_no, fare_conditions, passenger_id, passenger_name)\nwith (length=96, col1=7, col2=7, col3=7, col4=7, col5=7, col6=7, col7=7, col8=7, col9=7);\n<\/code><\/pre>\n<pre><code class=\"\">ERROR:  data type character has no default operator class for access method \"bloom\"\nHINT:  You must specify an operator class for the index or define a default operator class for the data type.\n<\/code><\/pre>\n<p>\u4e0d\u5e78\u7684\u662f\uff0c\u5b83\u53ea\u652f\u6301\u4e24\u4e2a\u64cd\u4f5c\u7b26\u7c7b\uff1a<\/p>\n<pre><code class=\"language-sql \">select opcname, opcintype::regtype\nfrom pg_opclass\nwhere opcmethod = (select oid from pg_am where amname = 'bloom')\norder by opcintype::regtype::text;\n<\/code><\/pre>\n<pre><code class=\"\"> opcname  | opcintype\n----------+-----------\n int4_ops | integer\n text_ops | text\n(2 rows)\n<\/code><\/pre>\n<p>\u5e78\u8fd0\u7684\u662f\uff0c\u4e3a\u5176\u5b83\u6570\u636e\u7c7b\u578b\u521b\u5efa\u76f8\u4f3c\u7684\u64cd\u4f5c\u7b26\u7c7b\u975e\u5e38\u7b80\u5355\u3002Bloom\u7d22\u5f15\u7684\u6bcf\u4e2a\u64cd\u4f5c\u7b26\u7c7b\u90fd\u5fc5\u987b\u5305\u542b\u4e00\u4e2a\u64cd\u4f5c\u7b26\uff08\u76f8\u7b49\uff09\u548c\u4e00\u4e2a\u8f85\u52a9\u51fd\u6570\uff08\u54c8\u5e0c\u51fd\u6570\uff09\u3002\u5bfb\u627e\u64cd\u4f5c\u7b26\u548c\u54c8\u5e0c\u51fd\u6570\u6700\u7b80\u5355\u7684\u65b9\u6cd5\u662f\u67e5\u627eHash\u7d22\u5f15\u7684\u64cd\u4f5c\u7b26\u7c7b\uff1a<\/p>\n<pre><code class=\"language-sql \">select distinct\n       opc.opcintype::regtype::text,\n       amop.amopopr::regoperator,\n       ampr.amproc\nfrom pg_am am, pg_opclass opc, pg_amop amop, pg_amproc ampr\nwhere am.amname = 'hash'\n   and opc.opcmethod = am.oid\n   and amop.amopfamily = opc.opcfamily\n   and amop.amoplefttype = opc.opcintype\n   and amop.amoprighttype = opc.opcintype\n   and ampr.amprocfamily = opc.opcfamily\n   and ampr.amproclefttype = opc.opcintype\norder by opc.opcintype::regtype::text;\n<\/code><\/pre>\n<pre><code class=\"\"> opcintype |       amopopr        |    amproc    \n-----------+----------------------+--------------\n abstime   | =(abstime,abstime)   | hashint4\n aclitem   | =(aclitem,aclitem)   | hash_aclitem\n anyarray  | =(anyarray,anyarray) | hash_array\n anyenum   | =(anyenum,anyenum)   | hashenum\n anyrange  | =(anyrange,anyrange) | hash_range\n ...\n<\/code><\/pre>\n<p>\u6211\u4eec\u521b\u5efa\u4e24\u4e2a\u64cd\u4f5c\u7b26\u7c7b\uff1a<\/p>\n<pre><code class=\"language-sql \">CREATE OPERATOR CLASS character_ops\nDEFAULT FOR TYPE character USING bloom AS\n  OPERATOR  1  =(character,character),\n  FUNCTION  1  hashbpchar;\n\nCREATE OPERATOR CLASS interval_ops\nDEFAULT FOR TYPE interval USING bloom AS\n  OPERATOR  1  =(interval,interval),\n  FUNCTION  1  interval_hash;\n<\/code><\/pre>\n<p>point\u7c7b\u578b\u6ca1\u6709\u5bf9\u5e94\u7684hash\u51fd\u6570\uff0c\u7d22\u5f15\u4e0d\u80fd\u4e3a\u5b83\u521b\u5efaBloom\u7d22\u5f15\uff0c\u4e5f\u4e0d\u80fd\u5bf9\u8fd9\u79cd\u7c7b\u578b\u4f7f\u7528hash\u8fde\u63a5\u3002<\/p>\n<p>\u518d\u6b21\u5c1d\u8bd5\u4e00\u4e0b\uff1a<\/p>\n<pre><code class=\"language-sql \">create index flights_bi_bloom on flights_bi\nusing bloom(airport_code, airport_utc_offset, flight_no, flight_type, aircraft_code, seat_no, fare_conditions, passenger_id, passenger_name)\nwith (length=96, col1=7, col2=7, col3=7, col4=7, col5=7, col6=7, col7=7, col8=7, col9=7);\n<\/code><\/pre>\n<pre><code class=\"\">CREATE INDEX\n<\/code><\/pre>\n<p>\u7d22\u5f15\u7684\u5927\u5c0f\u4e3a526MB\uff0c\u6bd4\u4f30\u8ba1\u7684\u7a0d\u5927\u4e00\u4e9b\u3002\u8fd9\u662f\u56e0\u4e3a\u6211\u4eec\u5728\u4f30\u7b97\u65f6\u6ca1\u6709\u8003\u8651\u9875\u9762\u4e0a\u7684\u5176\u5b83\u5f00\u9500\u3002<\/p>\n<pre><code class=\"language-sql \">select pg_size_pretty(pg_total_relation_size('flights_bi_bloom'));\n<\/code><\/pre>\n<pre><code class=\"\"> pg_size_pretty\n----------------\n 526 MB\n(1 row)\n<\/code><\/pre>\n<h2>3.3 \u67e5\u8be2<\/h2>\n<p>\u6211\u4eec\u4f7f\u7528\u5404\u79cd\u6807\u51c6\uff0c\u5bf9\u7d22\u5f15\u8fdb\u884c\u67e5\u8be2\u3002<\/p>\n<p>\u5c31\u50cf\u5df2\u7ecf\u63d0\u5230\u7684\u90a3\u6837\uff0cBloom filter\u662f\u4e00\u4e2a\u6982\u7387\u6027\u7684\u6570\u636e\u7ed3\u6784\uff0c\u56e0\u6b64\u5b83\u9ad8\u5ea6\u4f9d\u8d56\u5177\u4f53\u7684\u573a\u666f\u3002\u4f8b\u5982\uff0c\u6211\u4eec\u67e5\u770b\u4e0e\u4e24\u4e2a\u4e58\u5ba2\u76f8\u5173\u7684\u884c\u3002Miroslav Sidorov\uff1a<\/p>\n<pre><code class=\"language-sql \">explain(costs off,analyze)\nselect * from flights_bi where passenger_name='MIROSLAV SIDOROV';\n<\/code><\/pre>\n<pre><code class=\"\">                                            QUERY PLAN\n--------------------------------------------------------------------------------------------------\n Bitmap Heap Scan on flights_bi (actual time=2639.010..3010.692 rows=2 loops=1)\n   Recheck Cond: (passenger_name = 'MIROSLAV SIDOROV'::text)\n   Rows Removed by Index Recheck: 38562\n   Heap Blocks: exact=21726\n   -&gt;  Bitmap Index Scan on flights_bi_bloom (actual time=1065.191..1065.191 rows=38564 loops=1)\n         Index Cond: (passenger_name = 'MIROSLAV SIDOROV'::text)\n Planning time: 0.109 ms\n Execution time: 3010.736 ms\n<\/code><\/pre>\n<p>Marfa Soloveva:<\/p>\n<pre><code class=\"language-sql \">explain(costs off,analyze)\nselect * from flights_bi where passenger_name='MARFA SOLOVEVA';\n<\/code><\/pre>\n<pre><code class=\"\">                                            QUERY PLAN\n---------------------------------------------------------------------------------------------------\n Bitmap Heap Scan on flights_bi (actual time=9980.884..10142.684 rows=2 loops=1)\n   Recheck Cond: (passenger_name = 'MARFA SOLOVEVA'::text)\n   Rows Removed by Index Recheck: 3950168\n   Heap Blocks: exact=45757 lossy=67332\n   -&gt;  Bitmap Index Scan on flights_bi_bloom (actual time=1037.588..1037.588 rows=212972 loops=1)\n         Index Cond: (passenger_name = 'MARFA SOLOVEVA'::text)\n Planning time: 0.157 ms\n Execution time: 10142.730 ms\n<\/code><\/pre>\n<p>\u5728\u7b2c\u4e00\u4e2a\u67e5\u8be2\u4e2d\uff0c\u53ea\u670940000\u4e2a\u5047\u9633\u6027\u7ed3\u679c\uff0c\u4f46\u662f\u5728\u7b2c\u4e8c\u4e2a\u67e5\u8be2\u4e2d\u67094000000\u4e2a\u3002\u6267\u884c\u65f6\u95f4\u4e5f\u4e0d\u540c\u3002<\/p>\n<p>\u4e0b\u9762\u4f7f\u7528\u4e58\u5ba2\u7684id\u8fdb\u884c\u67e5\u8be2\uff0c\u800c\u4e0d\u662f\u59d3\u540d\uff1a<\/p>\n<pre><code class=\"language-sql \">explain(costs off,analyze)\nselect * from flights_bi where passenger_id='5864 006033';\n<\/code><\/pre>\n<pre><code class=\"\">                                           QUERY PLAN\n-------------------------------------------------------------------------------------------------\n Bitmap Heap Scan on flights_bi (actual time=13747.305..16907.387 rows=2 loops=1)\n   Recheck Cond: ((passenger_id)::text = '5864 006033'::text)\n   Rows Removed by Index Recheck: 9620258\n   Heap Blocks: exact=50510 lossy=165722\n   -&gt;  Bitmap Index Scan on flights_bi_bloom (actual time=937.202..937.202 rows=426474 loops=1)\n         Index Cond: ((passenger_id)::text = '5864 006033'::text)\n Planning time: 0.110 ms\n Execution time: 16907.423 ms\n<\/code><\/pre>\n<p>Marfa:<\/p>\n<pre><code class=\"language-sql \">explain(costs off,analyze)\nselect * from flights_bi where passenger_id='2461 559238';\n<\/code><\/pre>\n<pre><code class=\"\">                                            QUERY PLAN\n--------------------------------------------------------------------------------------------------\n Bitmap Heap Scan on flights_bi (actual time=3881.615..3934.481 rows=2 loops=1)\n   Recheck Cond: ((passenger_id)::text = '2461 559238'::text)\n   Rows Removed by Index Recheck: 30669\n   Heap Blocks: exact=27513\n   -&gt;  Bitmap Index Scan on flights_bi_bloom (actual time=1084.391..1084.391 rows=30671 loops=1)\n         Index Cond: ((passenger_id)::text = '2461 559238'::text)\n Planning time: 0.120 ms\n Execution time: 3934.517 ms\n<\/code><\/pre>\n<p>\u8fd9\u4e24\u4e2a\u67e5\u8be2\u7684\u6548\u7387\u4e5f\u6709\u5f88\u5927\u5dee\u8ddd\u3002<\/p>\n<p>\u6ce8\u610f\u5230\uff0c\u5982\u679c\u53ef\u4ee5\u540c\u65f6\u4f7f\u7528\u4e24\u4e2a\u7d22\u5f15\uff0c\u5219\u5047\u9633\u6027\u7684\u6982\u7387\u53ef\u4ee5\u964d\u4f4e\u5230\u4e07\u5206\u4e4b\u4e00\u3002<\/p>\n<pre><code class=\"language-sql \">explain(costs off,analyze)\nselect * from flights_bi\nwhere passenger_name='MIROSLAV SIDOROV'\n  and passenger_id='5864 006033';\n<\/code><\/pre>\n<pre><code class=\"\">                                                     QUERY PLAN\n--------------------------------------------------------------------------------------------------------------------\n Bitmap Heap Scan on flights_bi (actual time=872.593..877.915 rows=2 loops=1)\n   Recheck Cond: (((passenger_id)::text = '5864 006033'::text)\n               AND (passenger_name = 'MIROSLAV SIDOROV'::text))\n   Rows Removed by Index Recheck: 357\n   Heap Blocks: exact=356\n   -&gt;  Bitmap Index Scan on flights_bi_bloom (actual time=832.041..832.041 rows=359 loops=1)\n         Index Cond: (((passenger_id)::text = '5864 006033'::text)\n                   AND (passenger_name = 'MIROSLAV SIDOROV'::text))\n Planning time: 0.524 ms\n Execution time: 877.967 ms\n<\/code><\/pre>\n<p>\u4f46\u662f\uff0cBloom\u7d22\u5f15\u6839\u672c\u4e0d\u652f\u6301or\u3002\u8fd9\u662f\u4f18\u5316\u5668\u7684\u9650\u5236\uff0c\u800c\u4e0d\u662fBloom filter\u7684\u9650\u5236\u3002\u5f53\u7136\uff0c\u5176\u4e2d\u4e00\u4e2a\u9009\u62e9\u5c31\u662f\u8bfb\u53d6\u4e24\u4e2a\u7d22\u5f15\uff0c\u8ba1\u7b97\u4e24\u4e2a\u4f4d\u56fe\uff0c\u7136\u540e\u6c42\u4ea4\u96c6\uff0c\u4f46\u662f\u4ee3\u4ef7\u5f88\u5927\u3002<\/p>\n<h1>4. \u4e0eBRIN\u548cHash\u7d22\u5f15\u7684\u6bd4\u8f83<\/h1>\n<p>\u5f88\u660e\u663e\uff0cBloom\u7d22\u5f15\u4e0eBRIN\u7684\u5e94\u7528\u573a\u666f\u6709\u4ea4\u96c6\u3002\u5b83\u4eec\u90fd\u9762\u5bf9\u9700\u8981\u4f7f\u7528\u4e0d\u540c\u5b57\u6bb5\u5bf9\u5927\u8868\u8fdb\u884c\u626b\u63cf\u7684\u573a\u666f\uff0c\u4ee5\u635f\u5931\u7cbe\u5ea6\u6362\u53d6\u7a7a\u95f4\u4e0a\u7684\u538b\u7f29\u3002<\/p>\n<p>BRIN\u7d22\u5f15\u66f4\u52a0\u7d27\u51d1\uff0c\u53ef\u4ee5\u5feb\u901f\u6392\u9664\u4e00\u4e2a\u533a\u95f4\u4e0a\u6240\u6709\u7684\u5143\u7ec4\uff0c\u4f46\u662f\u57fa\u4e8e\u4e00\u4e2a\u5f88\u5f3a\u7684\u5047\u8bbe\uff1a\u6570\u636e\u7684\u987a\u5e8f\u4e0e\u7269\u7406\u4f4d\u7f6e\u76f8\u5173\u3002Bloom\u7d22\u5f15\u8981\u5927\u4e00\u4e9b\uff0c\u4f46\u662f\u5b83\u9664\u4e86\u9700\u8981\u4e00\u4e2a\u5408\u9002\u7684Hash\u51fd\u6570\u4e4b\u5916\uff0c\u6ca1\u6709\u4efb\u4f55\u5047\u8bbe\u3002<\/p>\n<p>\u548cBloom\u7d22\u5f15\u76f8\u4f3c\uff0cHash\u7d22\u5f15\u4e5f\u53ea\u652f\u6301\u7b49\u503c\u67e5\u8be2\u3002Hash\u7d22\u5f15\u7684\u7cbe\u5ea6\u6bd4Bloom\u7d22\u5f15\u9ad8\uff0c\u4f46\u662f\u9700\u8981\u7684\u7a7a\u95f4\u4e5f\u5927\u3002<\/p>\n<h1>5. \u5c5e\u6027<\/h1>\n<p>\u548c\u5f80\u5e38\u4e00\u6837\uff0c\u770b\u4e00\u4e0bBloom filter\u7684\u5c5e\u6027\uff0c\u67e5\u8be2\u89c1\u7b2c\u4e8c\u7bc7\u6587\u7ae0\u3002<\/p>\n<p>AM\u5c42\u9762\u7684\u5c5e\u6027\uff1a<\/p>\n<pre><code class=\"\"> amname |     name      | pg_indexam_has_property\n--------+---------------+-------------------------\n bloom  | can_order     | f\n bloom  | can_unique    | f\n bloom  | can_multi_col | t\n bloom  | can_exclude   | f\n<\/code><\/pre>\n<p>\u5b83\u652f\u6301\u591a\u5217\u7d22\u5f15\u3002<\/p>\n<p>\u67d0\u4e2a\u7d22\u5f15\u5c42\u9762\u7684\u5c5e\u6027\uff1a<\/p>\n<pre><code class=\"\">     name      | pg_index_has_property\n---------------+-----------------------\n clusterable   | f\n index_scan    | f\n bitmap_scan   | t\n backward_scan | f\n<\/code><\/pre>\n<p>\u5b83\u652f\u6301\u4f4d\u56fe\u626b\u63cf\u3002<\/p>\n<p>\u5217\u5c42\u9762\u7684\u5c5e\u6027\uff1a<\/p>\n<pre><code class=\"\">        name        | pg_index_column_has_property \n--------------------+------------------------------\n asc                | f\n desc               | f\n nulls_first        | f\n nulls_last         | f\n orderable          | f\n distance_orderable | f\n returnable         | f\n search_array       | f\n search_nulls       | f\n<\/code><\/pre>\n","protected":false},"excerpt":{"rendered":"<p>\u672c\u6587\u4ecb\u7ecdBloom filter\uff0c\u5b83\u53ef\u4ee5\u7528\u6765\u68c0\u67e5\u4e00\u4e2a\u5143\u7d20\u5728\u96c6\u5408\u4e2d\u662f\u5426\u5b58\u5728\u3002\u5982\u679c\u5b83\u5224\u65ad\u5143\u7d20\u5728\u96c6\u5408\u4e2d\u4e0d\u5b58\u5728\uff0c\u5219\u8fd9\u4e2a\u5143\u7d20\u5728\u96c6\u5408\u4e2d\u4e00\u5b9a\u4e0d\u5b58\u5728\u3002\u5982\u679c\u5b83\u5224\u65ad\u5143\u7d20\u5728\u96c6\u5408\u4e2d\u5b58\u5728\uff0c\u5219\u53ef\u80fd\u5b58\u5728\uff0c\u4e5f\u53ef\u80fd\u4e0d\u5b58\u5728\u3002\u5373\u5b83\u53ef\u80fd\u8fd4\u56de\u5047\u9633\u6027\u7684\u7ed3\u679c\uff0c\u4f46\u4e0d\u4f1a\u8fd4\u56de\u5047\u9634\u6027\u7684\u7ed3\u679c\u3002<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[2],"tags":[],"_links":{"self":[{"href":"https:\/\/www.mengqingzhong.com\/wp-json\/wp\/v2\/posts\/378"}],"collection":[{"href":"https:\/\/www.mengqingzhong.com\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.mengqingzhong.com\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.mengqingzhong.com\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.mengqingzhong.com\/wp-json\/wp\/v2\/comments?post=378"}],"version-history":[{"count":10,"href":"https:\/\/www.mengqingzhong.com\/wp-json\/wp\/v2\/posts\/378\/revisions"}],"predecessor-version":[{"id":501,"href":"https:\/\/www.mengqingzhong.com\/wp-json\/wp\/v2\/posts\/378\/revisions\/501"}],"wp:attachment":[{"href":"https:\/\/www.mengqingzhong.com\/wp-json\/wp\/v2\/media?parent=378"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.mengqingzhong.com\/wp-json\/wp\/v2\/categories?post=378"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.mengqingzhong.com\/wp-json\/wp\/v2\/tags?post=378"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}