在sqlite中多个LIKE

Question

我正在尝试创建一个搜索功能。

如果搜索输入字段是“foo bar”，我将其拆分为两个关键字然后执行此查询：

SELECT p.* FROM p_extra_fields as x INNER JOIN products as p ON x.product = p.id
  WHERE x.type = "1"
  AND
  (
     (x.key = "model" AND x.value LIKE "%foo%")
  OR (x.key = "model" AND x.value LIKE "%bar%")
  OR (x.key = "color" AND x.value LIKE "%foo%")
  OR (x.key = "color" AND x.value LIKE "%bar%")
  OR (x.key = "make" AND x.value LIKE "%foo%")
  OR (x.key = "make" AND x.value LIKE "%bar%")
  )      

GROUP BY x.product LIMIT 0, 50

关键字的数量可能更高，因此我可能需要更多“喜欢”。 “钥匙”的数量也可以增加:)

有什么办法可以简化这个查询吗？我可以做像LIKE("%foo%", "%bar%")这样的事吗？

Answer 1

如果启用了SQLite FTS3 and FTS4 Extensions，则可以利用全文搜索（FTS）功能。您需要将p_extra_fields表重新创建为VIRTUAL表。然后你可以在搜索词之间插入OR并使用MATCH操作符...

SELECT p.* 
FROM p_extra_fields x
JOIN products p ON p.id = x.product
WHERE x.key IN ('model', 'color', 'make')
AND x.type = '1'
AND x.value MATCH 'foo OR bar'
GROUP BY x.product LIMIT 0, 50;

好信息here也。点击here，在SQL Fiddle中查看它的运行情况。

Answer 2

我认为这个where条款更简单：

  WHERE x.type = "1" and
        x.key in ('model', 'color', 'make') and
        (x.value like '%foo%' or x.value like '%bar%')

Answer 3

我有同样的要求，我正在寻找一个像REGEXP "A|B|C"匹配的机制，这意味着匹配A，B，C。

所以最后这是我提出的解决方案：

WITH words(str, strSubString, hasComma) AS (
  VALUES ('', "foo,bar", 1)
  UNION ALL SELECT
              SUBSTR(strSubString, 0,
                     CASE WHEN INSTR(strSubString, ',')
                       THEN INSTR(strSubString, ',')
                     ELSE LENGTH(strSubString) + 1 END),
              LTRIM(SUBSTR(strSubString, INSTR(strSubString, ',')), ','),
    INSTR(strSubString, ',')
      FROM ssgPaths
      WHERE hasComma
  )

SELECT p.* FROM p_extra_fields as x INNER JOIN products as p ON x.product = p.id
JOIN words AS w ON x.value LIKE '%' || w.str || '%' AND w.str != ''
WHERE x.type = "1" and x.key in ('model', 'color', 'make');

匹配标准相当于@ Gordon的答案：

WHERE x.type = "1" and
        x.key in ('model', 'color', 'make') and
        (x.value like '%foo%' or x.value like '%bar%')

但是这使您可以灵活地根据查询参数动态匹配值（您可以提取出"foo,bar"作为参数）。

例如，在不更改您的查询的情况下，您可以通过"foo,bar,boo"直到匹配类似于正则表达式匹配："foo|bar|boo"