查询返回不同的行数，导致交叉表错误

Question

我的查询出现此错误，但我无法弄清楚它有什么问题：

查询指定的返回元组有 135 列，但交叉表返回 295。

这是我的查询：（以粗体突出显示的部分在 pgAdmin 中单独运行时返回相同数量的行。）

SELECT X.*, pi.productcode, pi.productitemdesc, pi.retailsalesprice, cat.productcategorydesc FROM ( 
SELECT * FROM crosstab ( 
'SELECT a.productitem AS productitemid, l.locationcode, (CASE WHEN SUM(a.netamount) IS NOT NULL THEN SUM(a.netamount) ELSE 0 END) || ''#'' || (CASE WHEN SUM(a.quantity) IS NOT NULL THEN SUM(a.quantity) ELSE 0 END) AS sales_qty FROM invoiceitem a INNER JOIN invoiceinfo b ON a.invoice = b.invoiceid INNER JOIN locationinfo l ON b.location = l.locationid WHERE b.status !=2 AND l.locationtype = 1 AND l.status = 1 AND TO_CHAR (b.invoicedate, ''YYYY-MM-DD'')>=''2018-03-01'' AND TO_CHAR (b.invoicedate, ''YYYY-MM-DD'')<=''2018-03-03'' GROUP BY a.productitem, l.locationcode ORDER BY a.productitem', 
'SELECT l.locationcode FROM locationinfo l INNER JOIN invoiceinfo b ON b.location = l.locationid
 WHERE b.status !=2 AND l.locationtype = 1 AND l.status = 1 AND TO_CHAR (b.invoicedate, ''YYYY-MM-DD'')>=''2018-03-01'' AND TO_CHAR (b.invoicedate, ''YYYY-MM-DD'')<=''2018-03-03'' GROUP BY l.locationcode order by l.locationcode') 
 AS (productitemid int, "0007" text,"BE101" text,"BE1013" text,"BE1014" text,"BE102" text,"BE103" text,"BE1034" text,"BE104" text,"BE1040" text,"BE1043" text,"BE1044" text,"BE1045" text,"BE1046" text,"BE105" text,"BE106" text,"BE107" text,"BE108" text,"BE109" text,"BE110" text,"BE111" text,"BE112" text,"BE123" text,"BE1265" text,"BE1266" text,"BE1271" text,"BE1272" text,"BE1273" text,"BE1274" text,"BE1279" text,"BE1280" text,"BE1281" text,"BE1282" text,"BE1351" text,"BE1400" text,"BE1401" text,"BE1404" text,"BE141" text,"BE142" text,"BE193" text,"BE194" text,"BE2125" text,"BE2126" text,"BE2127" text,"BE2128" text,"BE3001" text,"BE3002" text,"BE3005" text,"BE3006" text,"BE3009" text,"BE3010" text,"BE3031" text,"BE3032" text,"BE3121" text,"BE3122" text,"BE3123" text,"BE3124" text,"BE3127" text,"BE3128" text,"BE3131" text,"BE3132" text,"BE3203" text,"BE3204" text,"BE325" text,"BE3253" text,"BE3254" text,"BE326" text,"BE332" text,"BE3503" text,"BE3504" text,"BE355" text,"BE356" text,"BE365" text,"BE366" text,"BE381" text,"BE382" text,"BE383" text,"BE384" text,"BE400" text,"BE401" text,"BE402" text,"BE403" text,"BE405" text,"BE406" text,"BE408" text,"BE409" text,"BE411" text,"BE412" text,"BE4311" text,"BE4316" text,"BE4401" text,"BE4402" text,"BE4521" text,"BE4522" text,"BE4551" text,"BE4552" text,"BE470" text,"BE473" text,"BE475" text,"BE481" text,"BE482" text,"BE601" text,"BE604" text,"BE609" text,"BE610" text,"BE7040" text,"BE7043" text,"BE7045" text,"BE7046" text,"BE7048" text,"BE7049" text,"BE708" text,"BE7111" text,"BE7112" text,"BE7127" text,"BE7128" text,"BE7217" text,"BE7218" text,"BE7307" text,"BE7308" text,"BE7351" text,"BE7352" text,"BE801" text,"BE802" text,"BE803" text,"BE804" text,"BE831" text,"BE832" text,"BE860" text,"BE861" text,"BE862" text,"BE863" text,"BE865" text,"BE981" text,"BE982" text 
 )) X
 LEFT JOIN productitem pi ON X.productitemid = pi.productitemid
 LEFT JOIN productcategory cat ON pi.productcategory = cat.productcategoryid

的粗体部分

productitemid int, "0007" text,"BE101" text,"BE1013" text,"BE1014" text,"BE102" text,"BE103" text,"BE1034" text,"BE104" text,"BE1040" text,"BE1043" text,"BE1044" text,"BE1045" text,"BE1046" text,"BE105" text,"BE106" text,"BE107" text,"BE108" text,"BE109" text,"BE110" text,"BE111" text,"BE112" text,"BE123" text,"BE1265" text,"BE1266" text,"BE1271" text,"BE1272" text,"BE1273" text,"BE1274" text,"BE1279" text,"BE1280" text,"BE1281" text,"BE1282" text,"BE1351" text,"BE1400" text,"BE1401" text,"BE1404" text,"BE141" text,"BE142" text,"BE193" text,"BE194" text,"BE2125" text,"BE2126" text,"BE2127" text,"BE2128" text,"BE3001" text,"BE3002" text,"BE3005" text,"BE3006" text,"BE3009" text,"BE3010" text,"BE3031" text,"BE3032" text,"BE3121" text,"BE3122" text,"BE3123" text,"BE3124" text,"BE3127" text,"BE3128" text,"BE3131" text,"BE3132" text,"BE3203" text,"BE3204" text,"BE325" text,"BE3253" text,"BE3254" text,"BE326" text,"BE332" text,"BE3503" text,"BE3504" text,"BE355" text,"BE356" text,"BE365" text,"BE366" text,"BE381" text,"BE382" text,"BE383" text,"BE384" text,"BE400" text,"BE401" text,"BE402" text,"BE403" text,"BE405" text,"BE406" text,"BE408" text,"BE409" text,"BE411" text,"BE412" text,"BE4311" text,"BE4316" text,"BE4401" text,"BE4402" text,"BE4521" text,"BE4522" text,"BE4551" text,"BE4552" text,"BE470" text,"BE473" text,"BE475" text,"BE481" text,"BE482" text,"BE601" text,"BE604" text,"BE609" text,"BE610" text,"BE7040" text,"BE7043" text,"BE7045" text,"BE7046" text,"BE7048" text,"BE7049" text,"BE708" text,"BE7111" text,"BE7112" text,"BE7127" text,"BE7128" text,"BE7217" text,"BE7218" text,"BE7307" text,"BE7308" text,"BE7351" text,"BE7352" text,"BE801" text,"BE802" text,"BE803" text,"BE804" text,"BE831" text,"BE832" text,"BE860" text,"BE861" text,"BE862" text,"BE863" text,"BE865" text,"BE981" text,"BE982" text

与

SELECT l.locationcode FROM locationinfo l INNER JOIN invoiceinfo b ON b.location = l.locationid
 WHERE b.status !=2 AND l.locationtype = 1 AND l.status = 1 AND TO_CHAR (b.invoicedate, ''YYYY-MM-DD'')>=''2018-03-01'' AND TO_CHAR (b.invoicedate, ''YYYY-MM-DD'')<=''2018-03-03'' GROUP BY l.locationcode order by l.locationcode

单独运行时，我得到 295 个结果，这是正确的。但是，将它放在整个查询中会出现错误。

Answer 1

The manual:

剩余的输出列必须具有 source_sql 查询结果的最后一列的类型，并且它们的数量必须与 @987654327 中的行数完全相同@ 查询结果。

crosstab(text, text) 使用动态 (!) SELECT 查询作为第二个参数很容易出错，因为输出列列表是静态的。您还应该动态生成输出列列表。示例：

• Dynamic alternative to pivot with CASE and GROUP BY

除此之外，从同一查询返回的行数不同，很可能是由于两个不同会话中的 timezone 设置不同。

您的谓词中有表达式TO_CHAR(b.invoicedate, 'YYYY-MM-DD')。如果b.invoicedate 是键入timestamptz，则结果取决于timezone 设置。考虑：

SET timezone = '+10';
SELECT TO_CHAR(timestamptz '2018-04-07 23:30+0', 'YYYY-MM-DD');

to_char
----------
2018-04-08

SET timezone = '-10';
SELECT TO_CHAR(timestamptz '2018-04-07 23:30+0', 'YYYY-MM-DD');

to_char
----------
2018-04-07

解决方案

要消除对timezone 设置的依赖，请使用绝对值。

• to_char() 在这个地方开始是昂贵的并发症。删除它并调整谓词。
• 提供timestamptz 值以匹配您的列。这也允许使用索引（如果适用）。
  • Return Table Type from A function in PostgreSQL
• 使用时区名称（不是偏移量或缩写）以避免更多误解。
  • Ignoring timezones altogether in Rails and PostgreSQL
• 使用美元报价来清理报价混乱。
  • Insert text with single quotes in PostgreSQL

我建议将此作为第二个参数：

$$
SELECT l.locationcode
FROM   locationinfo l
JOIN   invoiceinfo  b ON b.location = l.locationid
WHERE  b.status <> 2 AND l.locationtype = 1 AND l.status = 1
AND    b.invoicedate >=  timestamptz '2018-03-01 Europe/Vienna'
AND    b.invoicedate <   timestamptz '2018-03-04 Europe/Vienna'
GROUP  BY 1
ORDER  BY 1
$$

将我的示例中的 Europe/Vienna 替换为定义日期的时区名称。
请注意 2018-03-04 以按照您的意图包含所有 2018-03-03。