Is CQL limiting Cassandra from being 'schema free'?

Question

As I've understood it, one feature of Cassandra is that it's supposedly 'schema free', that is (coming from a RDBS perspective) a table could contain several rows with varying columns.

In fact I never found the word 'table' used in the official documentation, instead Cassandra was supposed to use SuperColumns and ColumnFamilies: http://en.wikipedia.org/wiki/Column_family

But when reading up on CQL, the replacement language for SQL, creating these kinds of structures seems imposible, instead it closely mirrors a relation database structure that requires the creation and updating of tables and prohibiting on the fly attaching a new column to an row if it's not already defined in the table.

This is not at all what I need for my current project, as I need to store a lot of different information for each entity in the database (it's an experiment combining logic programming with probabilistic information and machine learning)

I only just begun reading up CQL, Am I missing something or is Cassandra perhaps not the right choice for my project?

Answer 1

For documentation you will most likely find more up to date information at the Datastax Website as there is a paid documentation team constantly working on Cassandra Docs. Datastax C* Docs

The simplest solution for having arbitrary information in a row would be to use a MAP, LIST, or SET type in your logical row.

CQL Datatypes Reference

But if you want an arbitrary layout you can create a table in CQL like

CREATE TABLE entityvalue ( key blob, column_name blob, value blob , PRIMARY KEY (key, column_name)) 

and as a final option you could just use any of the old thrift drivers and have the old style if you like but there has been a vote to stop thrift development in Cassandra.