Oops, sorry about double post. Alain
2013/7/31 Alain RODRIGUEZ <arodr...@gmail.com> > I like to point to this article from Sylvain, which is really well written. > > http://www.datastax.com/dev/blog/thrift-to-cql3 > > It explains a lot of things and is really interesting for Cassandra users > pre-CQL3. > > Actually, old dynamic columns were defined this way : > > CREATE TABLE test ( > key text, > column1 text, > value text, > PRIMARY KEY (key, column1) > ) WITH COMPACT STORAGE; > > This is still doable with CQL3, column1 would be your column name, value, > the value of your column. As the primary key is composed by key + column1, > you can add as much columns as you want. > > An other way to do it is to dynamically add columns (Alter table ..., > afaik, this is lock free and does not slow performance too much). > > > 2013/7/31 Jon Ribbens <jon-cassan...@unequivocal.co.uk> > >> I thought that part of the point of Cassandra was that, unlike a >> standard relational database, each row does not have to have the same >> set of columns. I don't understand how this squares with CQL. If I want >> to have a table ("column family"?) with a few fixed columns that are >> relevant to every row, I can create that with CQL's CREATE TABLE, but >> if I then want to set extra columns with arbitrary names on various >> rows, how do I tell CQL what type those columns are? Or is this >> feature of Cassandra now deprecated? >> > >