在postgresql中导入CSV而不复制行

将CSV文件导入到表中时，我的查询重复行：

+----------+----------------+---------------------+
| fullname | formattedvalue | recordTime          |
+----------+----------------+---------------------+
| text1    | 170.01346      | 09/02/2020 21:45:00 |
+----------+----------------+---------------------+
| text2    | 24.153432536   | 09/02/2020 21:45:00 |
+----------+----------------+---------------------+
| text3    | 3.583432424    | 09/02/2020 21:45:00 |
+----------+----------------+---------------------+
| text1    | 170.01346      | 08/02/2020 21:45:00 |
+----------+----------------+---------------------+
| text2    | 24.153432536   | 08/02/2020 21:45:00 |
+----------+----------------+---------------------+
| text3    | 3.583432424    | 08/02/2020 21:45:00 |
+----------+----------------+---------------------+

以及查询：

CREATE TEMP TABLE tmp_x
(
"fullname" varchar,
"formattedvalue" double precision,
"recordtime" timestamp
);

COPY tmp_x FROM PROGRAM 'more +1 "D:\MEAS_20200308x.csv"' (FORMAT csv, DELIMITER ',');

--UPDATE tmp_x
--SET    formattedvalue = ROUND( CAST(formattedvalue as numeric), 3 );

insert into meas_kanal select * from (
select x.*
from tmp_x x
left outer join meas_kanal t on t.fullname = x. fullname AND t. recordtime = x. recordtime
where t. fullname is null AND  t. recordtime is null
) as missing;


DROP TABLE tmp_x;

我的逻辑是检查cloumn组合中的重复项：

fullname

+

recordtime

当我再次启动查询时，它再次插入相同的行

知道我错在哪里吗

编辑2：

我试着用同样的问题来解决这个问题：

INSERT INTO meas_kanal
SELECT x.*
FROM tmp_x x
    LEFT OUTER JOIN meas_kanal t ON (t. fullname = x. fullname AND t. recordtime = x. recordtime)
WHERE t.fullname IS NULL AND t. recordtime IS NULL;

编辑3： 还有一次失败

INSERT INTO meas_kanal
SELECT *
FROM tmp_x
WHERE NOT EXISTS(SELECT * 
                 FROM meas_kanal 
                 WHERE (tmp_x.fullname=meas_kanal.fullname and
                       tmp_x.recordtime=meas_kanal.recordtime)
                 );

我认为问题出在别的地方

编辑4：可能的解决方案

顺便说一句，我忘了提。我没有主键。 现在我提出两点：

CREATE TEMP TABLE tmp_x
(
"fullname" varchar,
"formattedvalue" double precision,
"recordtime" timestamp,
UNIQUE (fullname, recordtime)
);

CREATE TEMP TABLE tmp_x
(
"fullname" varchar,
"formattedvalue" double precision,
"recordtime" timestamp,
UNIQUE (fullname, recordtime)
);

并在此插入：

insert into meas_kanal(fullname, formattedvalue,recordtime) 
    SELECT fullname, formattedvalue,recordtime FROM tmp_x x
ON CONFLICT DO NOTHING;

insert into meas_kanal(fullname, formattedvalue,recordtime) 
    SELECT fullname, formattedvalue,recordtime FROM tmp_x x
ON CONFLICT DO NOTHING;

---

现在它的工作方式和我预期的一样。如果没有人给出更好的解决方案，我将写下此解决方案作为答案。

您可以使用

分组方式

或

独特方式

假设当前查询结果与显示的表一致

INSERT INTO meas_kanal
SELECT DISTINCT x.fullname, x.formattedvalue, x.recordtime
FROM tmp_x x

---

更新1

根据您最近的评论，如果希望插入最新记录时间，请使用

GROUP BY

INSERT INTO meas_kanal
SELECT x.fullname, x.formattedvalue, MAX(x.recordtime) recordtime
FROM tmp_x x
GROUP BY x.fullname, x.formattedvalue

---

我有两个主键：

并在此插入：

insert into meas_kanal(fullname, formattedvalue,recordtime) 
    SELECT fullname, formattedvalue,recordtime FROM tmp_x x
ON CONFLICT DO NOTHING;

insert into meas_kanal(fullname, formattedvalue,recordtime) 
    SELECT fullname, formattedvalue,recordtime FROM tmp_x x
ON CONFLICT DO NOTHING;

---

现在它的工作方式和我预期的一样。

你应该使用

分组方式
@SILENT你能举个例子吗？我的答案有用吗？@SILENT不，没有。仍在进行双重记录。您的问题是，在任何查询之前，原始表的外观如何？你能告诉我tmp_x中有哪些列吗？我假设有一个id列，对吗？