select a.dt, count(distinct a.account_id) role, count(distinct b.account_id) ret2 from ( select 'involve' as type,account_id,dt from src_loginrole_day where dt between 20180604 and 20180606 )a left join ( select 'involve' as type,account_id,dt from src_loginrole_day where dt between 20180605 and 20180607 )b on a.account_id = b.account_id and datediff(from_unixtime(unix_timestamp(b.dt,'yyyyMMdd')),from_unixtime(unix_timestamp(a.dt,'yyyyMMdd'))) = 1 group by a.dt order by a.dt;哪里有错

这个 SQL 语句中存在一个错误，因为 `from src_loginrole_day` 后面缺少表的别名。正确的语句如下：

select a.dt, count(distinct a.account_id) role, count(distinct b.account_id) ret2  
from (
    select 'involve' as type, account_id, dt  
    from src_loginrole_day  
    where dt between 20180604 and 20180606 
) a  
left join (
    select 'involve' as type, account_id, dt  
    from src_loginrole_day  
    where dt between 20180605 and 20180607  
) b  
on a.account_id = b.account_id and datediff(from_unixtime(unix_timestamp(b.dt,'yyyyMMdd')), from_unixtime(unix_timestamp(a.dt,'yyyyMMdd'))) = 1 
group by a.dt  
order by a.dt;

在子查询中，给 `src_loginrole_day` 表起一个别名，例如 `s`，然后在 `select` 子句中使用这个别名来指代该表。

相关问题

select a.dt,count(distinct b.account_id)/count(distinct a.account_id)as rate from ( select account_id,dt from src_logoutrole_day where dt between 20180604 and 20180606 )a left join ( select account_id,dt from src_logoutrole_day where dt between 20180605 and 20180607 )b on a.account_id = b.account_id and datediff(from_unixtime(unix_timestamp(b.dt,'yyyyMMdd')),from_unixtime(unix_timestamp(a.dt,'yyyyMMdd'))) = 1 group by a.dt order by a.dt;有错吗

这个 SQL 语句看起来没有语法错误，但是可能存在逻辑或数据问题。这个 SQL 语句是为了计算在一段时间内，前一天和后一天都有登录的玩家占总登录玩家数的比例。需要确认以下几点：

1. 数据表 src_logoutrole_day 中是否包含 account_id 和 dt 两个字段。
2. 时间范围是否正确，确保时间范围内有数据。
3. 是否存在同一天内多次登录的玩家，如果存在，需要去重。
4. 是否存在某个日期的登录数据没有被匹配到对应的前一天或后一天数据，这种情况下会导致计算不准确。

如果以上问题都排除了，那么可以执行该 SQL 语句并检查输出结果来验证逻辑是否正确。

select a.dt,count(b.account_id)/count(a.account_id) from ( select distinct account_id from src_loginrole_day where dt <=20180606 and dt>= 20180604 )a left JOIN ( select distinct account_id from src_loginrole_day where dt <=20180607 and dt>= 20180605 )b on a.account_id = b.account_id and datediff(from_unixtime(unix_timestamp(b.dt,'yyyyMMdd')),from_unixtime(unix_timestamp(a.dt,'yyyyMMdd'))) = 1 group by a.dt;哪里有错

这个 SQL 查询语句中存在语法错误。经过查看，发现在子查询中的 `src_loginrole_day` 表缺少别名。

以下是修正后的 SQL 查询语句：

SELECT a.dt, COUNT(b.account_id) / COUNT(a.account_id)
FROM (
    SELECT DISTINCT account_id, dt
    FROM src_loginrole_day
    WHERE dt <= 20180606 AND dt >= 20180604
) a
LEFT JOIN (
    SELECT DISTINCT account_id, dt
    FROM src_loginrole_day
    WHERE dt <= 20180607 AND dt >= 20180605
) b ON a.account_id = b.account_id AND DATEDIFF(FROM_UNIXTIME(UNIX_TIMESTAMP(b.dt,'yyyyMMdd')),FROM_UNIXTIME(UNIX_TIMESTAMP(a.dt,'yyyyMMdd'))) = 1
GROUP BY a.dt;

在这个修正后的查询语句中，我为子查询中的 `src_loginrole_day` 表添加了别名 `a` 和 `b`，并且在 SELECT 子句中添加了 `dt` 字段，以便在后面的 JOIN 操作中使用。