So what about JSON?

JSON has been on the headline for the last few years. Some may attribute MongoDB for much of this (though technically that’s BSON but let’s not quibble). But if you look beyond the NoSQL realm, JSON has quietly entered the mainstream: Postgres, Teradata, IBM and Oracle have all either released or are on the verge of releasing support for JSON as a first-class data type.

We don’t need to mention that JSON is a first class citizen in the Hadoop eco-system.

So let’s roll up our sleeves and work out what the above means.

For a start, JSON is at now least addressable in its aggregate form. The next step that everyone will want is to manipulate it. For example, Postgres shows one approach that uses functions to “unroll” the JSON’s content. But there are different choices one can make. For instance, Simba’s MongoDB ODBC driver uses sampling to surface the JSON structure out.

I have not surveyed all the various approaches out there but this is likely an area that will be highly debated in the coming days and years to come.