Hacker News new | past | comments | ask | show | jobs | submit login

Null values are so important in representing data. But they cause so much confusion in a) unexpected behaviors in queries and b) inconsistent handling across various engines... I sometimes wish <whisper> that they hadn't been included in the spec at all </whisper>. But then I come to my senses again, and go fix yet another bugged query for an analyst who didn't account for nulls in the data.



Does it make sense to coalesce them away in a view? I thought most analysts are given star schemas implemented by views or ETL'd data anyway.


Depends on the level of sophistication of the analyst, and if nulls have a meaning or value to the result.

Also, at a certain point, knowing that nulls are present gives you yet another measure of dq: not knowing if they are present and hidden vs. visible and countable can be the difference between a wrong answer vs. just an uncertain one.




Consider applying for YC's Spring batch! Applications are open till Feb 11.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: