标签：
有的时候需要在不同的数据库实例之间做集合操作，这就无法直接使用SQL语句的join，left join了。相同类型的数据库之间虽然也有类似于DBLINK和FEDERATED之类的东西，但一来这些东西不同的数据库之间是不能通用的，还有就是这些都要在数据库服务端上操作的，即使是dba要走这么一个申请流程也是非常麻烦的。所以就想自己写几个常用的：
inner_join：内连接操作，t1 join t2 on t1.col=t2.col，返回连接字段相等的行，方式包括下面的外连接用的都是最简单的nested loop join。
left_join：左连接操作，t1 left join t2 on t1.col=t2.col，返回左边所有的行以及右表中连接字段相等的行，注意的是右边没有的要以NULL填充。
right_join：右连接操作，t1 right join t2 on t1.col=t2.col，和左连接一样，就是左右的位置换了一下。
union_all：两个集合合并。
union：和union_all一样，就是要除去重复的。
代码如下：
#!/usr/local/bin/python # -*- coding: utf8 -*- ‘‘‘ Created on 2016年5月23日 @author: PaoloLiu ‘‘‘ class sql_merge(object): ‘‘‘ classdocs ‘‘‘ def __init__(self): ‘‘‘ Constructor ‘‘‘ def inner_join(self, left_table, right_table, left_join_col, right_join_col): pass new_result = [] for left_row in left_table: for right_row in right_table: if left_row[left_join_col] == right_row[right_join_col]: new_row = left_row + right_row new_result.append(new_row) return new_result def left_join(self, left_table, right_table, left_join_col, right_join_col): pass new_result = [] addnull = None if len(right_table) > 0: for i in range (1, len(right_table[0])): addnull = (addnull, None) for left_row in left_table: n = 0 for right_row in right_table: if left_row[left_join_col] == right_row[right_join_col]: new_row = left_row + right_row new_result.append(new_row) else: n = n + 1 if n == len(right_table): new_result.append(left_row + addnull) else: new_result = left_table return new_result def right_join(self, left_table, right_table, left_join_col, right_join_col): return self.left_join(right_table, left_table, right_join_col, left_join_col) def union_all(self, left_table, right_table): return left_table + right_table def union(self, left_table, right_table): result = self.union_all(left_table, right_table) new_result = [] for row in result: if row not in new_result: new_result.append(row) return new_result def order_asc(self, result, col): result.sort(key=lambda x:x[col]) return result def order_desc(self, result, col): result.sort(key=lambda x:x[col], reverse=True) return result def test(): pass left_result = [(1, "aaa"), (2, "bbb"), (3, "ccc"), (6, None), (1, "111")] right_result = [(1, "111"), (2, "2222"), (4, "444"), (5, "abc")] print "left_result=" + str(left_result) print "right_result=" + str(right_result) print "==========================================================" sqlmerge = sql_merge() print "inner_join=" + str(sqlmerge.inner_join(left_result, right_result, 0, 0)) print "left_join=" + str(sqlmerge.left_join(left_result, right_result, 0, 0)) print "right_join=" + str(sqlmerge.right_join(left_result, right_result, 0, 0)) print "union_all=" + str(sqlmerge.union_all(left_result, right_result)) print "union=" + str(sqlmerge.union(left_result, right_result)) print "order_asc=" + str(sqlmerge.order_asc(left_result, 0)) print "order_desc=" + str(sqlmerge.order_desc(left_result, 0)) if __name__ == "__main__": test()
测试记录运行结果：
left_result=[(1, ‘aaa‘), (2, ‘bbb‘), (3, ‘ccc‘), (6, None), (1, ‘111‘)] right_result=[(1, ‘111‘), (2, ‘2222‘), (4, ‘444‘), (5, ‘abc‘)] ========================================================== inner_join=[(1, ‘aaa‘, 1, ‘111‘), (2, ‘bbb‘, 2, ‘2222‘), (1, ‘111‘, 1, ‘111‘)] left_join=[(1, ‘aaa‘, 1, ‘111‘), (2, ‘bbb‘, 2, ‘2222‘), (3, ‘ccc‘, None, None), (6, None, None, None), (1, ‘111‘, 1, ‘111‘)] right_join=[(1, ‘111‘, 1, ‘aaa‘), (1, ‘111‘, 1, ‘111‘), (2, ‘2222‘, 2, ‘bbb‘), (4, ‘444‘, None, None), (5, ‘abc‘, None, None)] union_all=[(1, ‘aaa‘), (2, ‘bbb‘), (3, ‘ccc‘), (6, None), (1, ‘111‘), (1, ‘111‘), (2, ‘2222‘), (4, ‘444‘), (5, ‘abc‘)] union=[(1, ‘aaa‘), (2, ‘bbb‘), (3, ‘ccc‘), (6, None), (1, ‘111‘), (2, ‘2222‘), (4, ‘444‘), (5, ‘abc‘)] order_asc=[(1, ‘aaa‘), (1, ‘111‘), (2, ‘bbb‘), (3, ‘ccc‘), (6, None)] order_desc=[(6, None), (3, ‘ccc‘), (2, ‘bbb‘), (1, ‘aaa‘), (1, ‘111‘)]
最后貼一张我认为最能说明SQL JOIN关系的图。还有要注意的是not in子查询是可以用左边从上往下第二个left join所代替的。
标签：
原文地址：http://www.cnblogs.com/paololiu/p/5649277.html
