带有子表达式的 MySQL 布尔全文搜索答案

【问题标题】：MySQL boolean full text search with subexpressions带有子表达式的 MySQL 布尔全文搜索
【发布时间】：2016-04-04 16:54:24
【问题描述】：

我想在 MySQL 中编写一个布尔型全文搜索查询。但我想按这个顺序得到结果：

Exact Matches > Including all the words > Including some of the words

（> 表示更高。）

通过以下查询，我得到更高的完全匹配：

SELECT *, MATCH(name,description,address) 
AGAINST('>("word1 word2 word3") <(word1 word2 word3)' IN BOOLEAN MODE) 
AS score FROM samples WHERE MATCH(name,description,address) 
AGAINST('>("word1 word2 word3") <(word1 word2 word3)' IN BOOLEAN MODE) ORDER BY score DESC

通过以下查询，我得到的文档包含的所有单词都高于包含查询的某些部分的单词：

SELECT *, MATCH(name,description,address) 
AGAINST('>(+word1 +word2 +word3) <(word1 word2 word3)' IN BOOLEAN MODE) 
AS score FROM samples WHERE MATCH(name,description,address) 
AGAINST('>(+word1 +word2 +word3) <(word1 word2 word3)' IN BOOLEAN MODE) ORDER BY score DESC

但我想结合这两个查询。这是迄今为止我取得的最好成绩，但我没有得到我想要的结果。

SELECT *, MATCH(name,description,address) 
AGAINST('>("word1 word2 word3") <(>(+word1 +word2 +word3) <(word1 word2 word3))' IN BOOLEAN MODE) 
AS score FROM samples WHERE MATCH(name,description,address) 
AGAINST('>("word1 word2 word3") <(>(+word1 +word2 +word3) <(word1 word2 word3))' IN BOOLEAN MODE) 
ORDER BY score DESC

没有 UNION 有没有办法做到这一点？

【问题讨论】：

如果要匹配的单词超过 2 个，例如 match(..) against('>frist second third')，则无法执行此操作。您可能只需要使用union。

标签： mysql database full-text-search information-retrieval

【解决方案1】：

这行得通吗？

  SELECT *,
         MATCH (name,description,address) 
           AGAINST ('"word1 word2 word3"' IN BOOLEAN MODE) score_exact,
         MATCH (name,description,address) 
           AGAINST ('+word1 +word2 +word3' IN BOOLEAN MODE) score_all,
         MATCH (name,description,address) 
           AGAINST ('word1 word2 word3' IN BOOLEAN MODE) score_some
    FROM samples
   WHERE MATCH (name,description,address) 
           AGAINST ('word1 word2 word3' IN BOOLEAN MODE)
ORDER BY score_exact DESC, score_all DESC, score_some DESC

【讨论】：

哇，很有创意！我认为它会起作用，我必须对其进行测试。

【解决方案2】：

( SELECT *,
         1 AS sort_order,       -- top priority to exact match
         1.0 AS extra_ordering
      ... WHERE txt = 'aaaa bbbb cccc' )
UNION ALL
( SELECT *,
         2 AS sort_order,      -- next to all words exist
         MATCH(txt) AGAINST("+aaaa +bbbb +cccc" IN BOOLEAN MODE)
      ... WHERE MATCH(txt) AGAINST("+aaaa +bbbb +cccc" IN BOOLEAN MODE)
UNION ALL
( SELECT *,
         3 AS sort_order,      -- finally "any" words
         MATCH(txt) AGAINST("aaaa bbbb cccc" IN BOOLEAN MODE)
      ... WHERE MATCH(txt) AGAINST("aaaa bbbb cccc" IN BOOLEAN MODE)
ORDER BY sort_order ASC, extra_ordering DESC
LIMIT 1;  -- to get the 'best'

【讨论】：

我的联合问题是它返回重复记录，因为尽管主键相等，但分数列不同！
由于我添加的列不会有任何重复：sort_order。我添加了LIMIT 1。
我必须得到更多的结果，并且每个子查询的排序顺序是不同的，如果一行出现在多个子查询中，它将被重复。
你想要txt，还是想要*？（这对如何获得多个结果的复杂性产生了很大影响。）
我也想要其他列，我想我可以使用 group by id 来消除重复，效率高吗？