Question

出于某种原因，Postgres没有使用我定义的索引 - 我可能会遗漏一些显而易见的东西，但是我已经在墙上撞了一会儿并且非常喜欢你的建议。

表格定义：

CREATE TABLE "public"."useractivity" (
  "id" int4 NOT NULL DEFAULT nextval('useractivity_id_seq'::regclass),
  "userid" int8 NOT NULL,
  "created_at" timestamp(6) NOT NULL DEFAULT now(),
  "channel" varchar(32) NOT NULL COLLATE "default",
  "action" varchar(64) NOT NULL COLLATE "default",
  "details" varchar(256) COLLATE "default",
  CONSTRAINT "useractivity_pkey" PRIMARY KEY ("id") NOT DEFERRABLE INITIALLY IMMEDIATE,
  CONSTRAINT "action_user_fk" FOREIGN KEY ("userid") REFERENCES "public"."user" ("id") ON UPDATE CASCADE ON DELETE CASCADE NOT DEFERRABLE INITIALLY IMMEDIATE)
  WITH (OIDS=FALSE);

CREATE INDEX  "activity_user_idx" ON "public"."useractivity" USING btree(userid ASC NULLS LAST);
CREATE INDEX  "idx_useractivity_action" ON "public"."useractivity" USING btree(userid ASC NULLS LAST, "action" COLLATE "default" ASC NULLS LAST);
CREATE INDEX  "idx_useractivity_action2" ON "public"."useractivity" USING btree("action" COLLATE "default" ASC NULLS LAST, userid ASC NULLS LAST);
CREATE INDEX  "idx_useractivity_action_date" ON "public"."useractivity" USING btree(userid ASC NULLS LAST, created_at ASC NULLS LAST);

查询：

SELECT useractivity.userid,
        count(*) AS num
       FROM useractivity
      WHERE (useractivity.action= 'sent')
      GROUP BY useractivity.userid

查询的EXPLAIN输出：

HashAggregate  (cost=15846.78..15880.29 rows=3351 width=8)
->  Seq Scan on useractivity  (cost=0.00..14987.83 rows=171791 width=8)
    Filter: ((action)::text = 'sent'::text)

为什么Postgres没有采用任何索引，而是对171,000行进行序列扫描？

Postgres不使用索引

0 个答案: