按年份保留一种水果组合

Question

我有一桌水果：

 name1  name2 year1 year2
 apple   pear  2010  2001
 apple   pear  2011  2002
  pear  apple  2010  2003
  pear  apple  2011  2004
 apple   null  2009  2005
  pear orange  2008  2006
 apple   pear  2010  2007
 apple  grape  2010  2008

问题：每年1，我只希望名字出现一次......例如苹果梨 2010 与梨苹果 2010 相同。也就是说，当存在重复项时...我只想保留每个重复项的第一次出现（例如第一次出现）

我认为正确的输出应该是这样的：

 name1  name2 year1 year2
 apple   pear  2010  2001
 apple   pear  2011  2002
 apple   null  2009  2005
  pear orange  2008  2006
 apple  grape  2010  2008

我尝试了以下代码：

  SELECT 
    name1,
    name2,
    year1,
    year2,
    ROW_NUMBER() OVER (PARTITION BY name1, name2, year1 ORDER BY year2) AS rn
  FROM 
    fruits
)
SELECT 
  name1,
  name2,
  year1,
  year2
FROM 
  ranked_names
WHERE 
  rn = 1;

但这并没有产生正确的结果：

 name1  name2 year1 year2
 apple  grape  2010  2008
 apple   null  2009  2005
 apple   pear  2010  2001
 apple   pear  2011  2002
  pear  apple  2010  2003
  pear  apple  2011  2004
  pear orange  2008  2006

例如(apple pear 2010 2001) 和 ( pear apple 2010 2003) 出现两次，即使只应出现其中一个。

有人可以告诉我如何纠正这个问题吗？

谢谢！

替代想法？

WITH ranked_names AS (
    SELECT 
        name1,
        name2,
        year1,
        year2,
        ROW_NUMBER() OVER (PARTITION BY year1, 
                                       CASE WHEN name1 < name2 THEN name1 ELSE name2 END, 
                                       CASE WHEN name1 < name2 THEN name2 ELSE name1 END 
                          ORDER BY year2) AS rn
    FROM 
        fruits
)
SELECT 
    name1,
    name2,
    year1,
    year2
FROM 
    ranked_names
WHERE 
    rn = 1;

Answer 1

WITH ranked_names AS (
  SELECT 
    name1,
    name2,
    year1,
    year2,
    ROW_NUMBER() OVER (PARTITION BY LEAST(name1, name2), GREATEST(name1, name2), year1 ORDER BY year2) AS rn
  FROM 
    fruits
)
SELECT 
  name1,
  name2,
  year1,
  year2
FROM 
  ranked_names
WHERE 
  rn = 1;

演示：https://dbfiddle.uk/dQk17bSS

浏览 SQL 中的最小和最大概念。