SQL SERVER SSAS：如何处理事实表中的 NULL 日期值，以便可以无错误地处理时间维度？
我有一个事实表，其中有一列包含从 MS Access 源加载的日期。问题是，某些值是 NULL，SSAS 不允许我将 DATE 维度引用到它。
是在SSIS阶段解决比较好还是在SSAS阶段有解决办法？
非常感谢您的帮助。
I have a fact table that has a column with dates loaded from an MS Access source. The thing is, some of the values are NULL and the SSAS won't let me refer my DATE dimension to it.
Is it better to solve it at the SSIS stage or is there some solution at the SSAS?
Thank you very much for you help.
如果你对这篇内容有疑问，欢迎到本站社区发帖提问 参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱，如果其他用户或者作者回复了您的评论，将不能在第一时间通知您！
发布评论
评论（3）
最佳实践是事实表中不要有任何 NULL 键（即维度键）值。
相反，请在“日期”维度表中创建一个“最大日期”（或“未知值”，例如 -1）并为其设置键。
Best practice is not to have any NULL key (i.e. Dimension key) values in a Fact table.
Instead, create a MAX date in the Date dimension table (or an 'UnknownValue', -1 for instance) and key to that.
有时，由于非技术原因，不希望修复 DSV 或源系统中的空值。在这种情况下，您可以使用未知成员和 null 处理属性来解决此问题：
http://technet.microsoft.com/en-us/library/ms170707.aspx
我在尝试突出数据质量问题或出于快速原型设计目的时这样做了。
Sometimes it is undesirable for non-technical reasons to fix the nulls in the DSV or source system. In that case you can use the unknown member and null processing properties to work around this issue:
http://technet.microsoft.com/en-us/library/ms170707.aspx
I have done this when trying to highlight the data qualities problems or for fast prototyping purposes.
层次结构的每个成员都有一个属性“HideMemberIf”。将其设置为“NoName”，应在维度浏览器中隐藏空值并允许处理多维数据集。
您还可以在数据集市视图中创建命名计算。命名计算将使用 ISNULL 函数，该函数可以填充值来代替空值。然后当然可以根据这些命名计算而不是原始数据字段构建时间维度。
同样，最好不要在数据中包含任何空值，但通常可以在多维数据集中修复此问题。
Each member of a hierarchy has a property "HideMemberIf". Setting this to "NoName", should hide the null values from the Dimension Browser and allow processing of the cube.
Also you could created Named Calculations in the Datamart View. The Named Calculation would use the ISNULL function, which can fill in values in place of nulls. Then of course build your Time Dimension off of these Named Calculations instead of the raw data fields.
Again, it's better not to have any nulls in your data altogether, but you can usually fix this inside the Cube.