Spark SQL执行报错EOL while scanning string literal 语法问题求助

该EOL while scanning string literal语法错误属于Python语法规则触发的报错，和你写的Spark SQL本身逻辑无关：
Python中普通双引号"包裹的字符串不支持直接跨行书写，你将多行SQL语句用单个双引号包裹，Python解析到第一行字符串结尾的换行符时，还未扫描到匹配的闭合双引号，就会抛出该类行尾扫描错误。

推荐优先使用第一种方案，改动最小且易维护：

• 方案1：将包裹SQL的单个双引号替换为三重双引号"""，无需调整原有SQL的换行格式即可正常识别多行字符串：

results = spark.sql(
                    """
                        select b.* 
                        from ( 
                                select a.*, dense_rank() over (partition by nm order by total 
                                desc) as rnk
                                from (
                                        select dt, code, nm, sum(num) as total
                                        from cases
                                        group by dt, code, nm   
                                    )a
                            )b
                                where b.rnk = 1
                                order by total desc, dt
                    """
)

• 方案2：保留单个双引号的前提下，在每一行SQL字符串的末尾添加反斜杠\转义换行符，该方案容易漏加转义符，不推荐使用：

results = spark.sql(
                    " \
                        select b.* \
                        from ( \
                                select a.*, dense_rank() over (partition by nm order by total \
                                desc) as rnk \
                                from ( \
                                        select dt, code, nm, sum(num) as total \
                                        from cases \
                                        group by dt, code, nm   \
                                    )a \
                            )b \
                                where b.rnk = 1 \
                                order by total desc, dt \
                    "
)

修改完成后可先打印字符串验证SQL内容是否完整，确认无误后再执行查询即可。

内容的提问来源于stack exchange，提问作者NM24