Superset实现动态SQL查询功能

原作者: [db:作者] 来自: [db:来源] 收藏邀请

使用自定义参数方式实现 superset 实现SQL动态查询

1、启用参数：config.py 设置"ENABLE_TEMPLATE_PROCESSING": True

2、当前superset v1.2版本支持的参数包括：

{{ current_username() }} 当前登录用户名
{{ current_username(add_to_cache_keys=False) }} 不从缓存中获取登录用户名，默认从缓存获取
{{ current_user_id()}} 当前登录用户ID
{{ current_user_id(add_to_cache_keys=False) }} 不从缓存中获取登录用户ID，默认从缓存获取
{{ url_param('custom_variable') }} url 参数，比如127.0.0.1：8001\dashboard?abc=123，参数就是{{ url_param('abc') }} 结果就是123
{{ cache_key_wrapper() }} 还没有弄明白啥用
{{ filter_values("字段名") }} 获取dashboard filter_box组件对某个字段的筛选结果
{{ from_dttm }} 获取dashboard filter_box组件日期筛选的开始时间
{{ to_dttm }} 获取dashboard filter_box组件日期筛选的结束时间
{{ get_filters() }} 暂时没有弄明白

除此之外，还可以自定义参数，自定义参数方法：

①修改superset/jinja_context.py文件，修改三个地方：

regex = re.compile(
        r"\{\{.*("
        r"current_user_id\(.*\)|"
        r"current_username\(.*\)|"
        r"current_userroles\(.*\)|"
        r"isadmin\(.*\)|"
        r"cache_key_wrapper\(.*\)|"
        r"url_param\(.*\)"
        r").*\}\}"
    )

↑↑↑↑注意此处的 current_userroles 和 isadmin 是我自定义的，源文件没有

def current_user_id(self, add_to_cache_keys: bool = True) -> Optional[int]:
        """
        Return the user ID of the user who is currently logged in.

        :param add_to_cache_keys: Whether the value should be included in the cache key
        :returns: The user ID
        """

        if hasattr(g, "user") and g.user:
            if add_to_cache_keys:
                self.cache_key_wrapper(g.user.get_id())
            return g.user.get_id()
        return None

    def current_username(self, add_to_cache_keys: bool = True) -> Optional[str]:
        """
        Return the username of the user who is currently logged in.

        :param add_to_cache_keys: Whether the value should be included in the cache key
        :returns: The username
        """

        if g.user and hasattr(g.user, "username"):
            if add_to_cache_keys:
                self.cache_key_wrapper(g.user.username)
            return g.user.username
        return None
    def current_userroles(self, add_to_cache_keys: bool = True) -> Optional[str]:
        """
        Return the roles of the user who is currently logged in.

        :param add_to_cache_keys: Whether the value should be included in the cache key
        :returns: The userroles
        """

        if g.user and hasattr(g.user, "roles"):
            if add_to_cache_keys:
                user_roles = "/".join([role.name.lower() for role in list(g.user.roles)])
                self.cache_key_wrapper(user_roles)
                print(user_roles)
                return user_roles
                """admin in user_roles"""
        return None

    def isadmin(self, add_to_cache_keys: bool = True) -> Optional[str]:
        """
        Return the roles of the user who is currently logged in.

        :param add_to_cache_keys: Whether the value should be included in the cache key
        :returns: The userroles
        """

        if g.user and hasattr(g.user, "roles"):
            if add_to_cache_keys:
                user_roles = [role.name.lower() for role in list(g.user.roles)]
                return "admin" in user_roles
        return None

↑↑↑↑仿照系统自带的 current_username 编造自己的函数，我写了current_userroles 和 isadmin

class JinjaTemplateProcessor(BaseTemplateProcessor):
    def set_context(self, **kwargs: Any) -> None:
        super().set_context(**kwargs)
        extra_cache = ExtraCache(self._extra_cache_keys)
        self._context.update(
            {
                "url_param": partial(safe_proxy, extra_cache.url_param),
                "current_user_id": partial(safe_proxy, extra_cache.current_user_id),
                "current_username": partial(safe_proxy, extra_cache.current_username),
                "current_userroles": partial(safe_proxy, extra_cache.current_userroles),
                "isadmin": partial(safe_proxy, extra_cache.isadmin),
                "cache_key_wrapper": partial(safe_proxy, extra_cache.cache_key_wrapper),
                "filter_values": partial(safe_proxy, filter_values),
            }
        )

↑↑↑↑仿照系统自带的 current_username 编造自己的函数，我写了current_userroles 和 isadmin

就是这3个地方，但是注意，自己在第二步早的函数，返回值必须是：

ALLOWED_TYPES = (
    NONE_TYPE,
    "bool",
    "str",
    "unicode",
    "int",
    "long",
    "float",
    "list",
    "dict",
    "tuple",
    "set",
)

否则会提示错误，或者自己修改这个types，我是转换，比如上面那个g.user.roles 返回的结果就不是上面类型，导致我一直不成功，最后修改了下，才可以

3、判断是否自定义成功：

在superset sql lab中执行如下代码，如果能被解析，就说明成功

4、应用案例：

在dataset里面，动态访问数据源，数据源添加where语句：select * from sales where salesname =' {{current_username()}}'

dashboard里面，通过获取筛选器的结果，然后获取其他表应当显示的数据范围：

select  DATE,risktype,sum(num) as num from
(SELECT date , customerid,product,risktype ,count(*) as num
from v_superset_forecast_risk group by date , customerid,product,risktype ) a
join
(select distinct customer_code,product from v_superset_access
where name='{{ current_username() }}' )access
on a.customerid=access.customer_code
and a.product=access.product
and DATE_FORMAT(date,'%Y-%m')> DATE_FORMAT(date_sub(STR_TO_DATE(concat( {{ "'" + "', '".join(filter_values('yearmonthend')) + "'" }},'-01'), '%Y-%m-%d'), interval 12 month),'%Y-%m')
and DATE_FORMAT(date,'%Y-%m')<={{ "'" + "', '".join(filter_values('yearmonthend')) + "'" }}
group by DATE,risktype

因为sql里面可以使用jinja 表达式，比如判断筛选当前没有筛选的时候，获取什么数据

注意{% %} 内部使用参数的时候，不需要加{{}}，否则报错

通过筛选器实现模糊查询