Snowflake SQL: Making Schema-on-Read a Reality (Part 1)
This is my 1st official post on the Snowflake blog in my new role as their Technical Evangelist. It discusses getting results from semi-structured JSON data using our extensions to ANSI SQL.
Schema? I don’t need no stinking schema!
Over the last several years, I have heard this phrase schema-on-read used to explain the benefit of loading semi-structured data into a Big Data platform like Hadoop. The idea being you could delay data modeling and schema design until long after the data was loaded (so as to not slow down getting your data while waiting for those darn data modelers).
Every time I heard it, I thought (and sometimes said) – “but that implies there is a knowable schema.” So really you are just delaying the inevitable need to understand the structure in order to derive some business value from that data. Pay me now or pay me later.
Why delay the pain?
Check out the rest of the post here:
The Data Warrior