Added support for Collections and Secondary Indexes

[bsbodden]

One of my projects needed support for the new CQL collections so I've done a very simple implementation of it, alongside with a simple way to crate secondary indexes (need to add a way to removed an index also).

Missing items:

• ability to pass option hash to further configure the types for collections, e.g. list, set, map<key_type,value_type>

I fixed a set of typos when raising exceptions, e.g. raise Foo('bar') versus raise Foo, 'bar'.

[bsbodden]

I've done a bit more work on this area. I'll send another pull request in a few days with more refinements. So you can safely ignore this one if you want.

[hsgubert]

It seems great bsbodden! I'll not do the merge right know but I will take a look further on the code later.

About your comment of removing secondary indexes, do you know if it is possible with Cassandra 1.2? I was baffled the other day when I discovered I could't remove a column. I mean CQL supports it, but Cassandra 1.2 does not accept the operation.

I also never used collections either (I use Cassandra in a fairly straightforward way mainly because of its scaling and safety). I see you included migration support, but what about loading support for these collections? Is it already working?

Thx

[bsbodden]

You're right, I think we need some sort of matrix of features in CQL for C* 1.2 versus 2.0. I do have loading and saving support implemented for collections but since some one decided to use { } and [] for sets and lists respectively you need to have the column type available when assembling the CQL statement. The only way I could figure out how to do it was to do a query with limit=1 for that specific column and get the type from the result set metadata. Also the is the choice of adding to a set/list/map versus replacing the whole map. So I'm thinking of implementing/overloading the << operator to build that CQL correctly.

Examples:

Update sets:

Adding...

UPDATE users SET emails = emails + {'fb@friendsofmordor.org'} WHERE user_id = 'frodo';

Removing...

UPDATE users SET emails = emails - {'fb@friendsofmordor.org'} WHERE user_id = 'frodo';

Update lists:

Replacing...

UPDATE users SET top_places = [ 'rivendell', 'rohan' ] WHERE user_id = 'frodo';

Adding ...

UPDATE users SET top_places = [ 'the shire' ] + top_places WHERE user_id = 'frodo';

Removing...

UPDATE users SET top_places = top_places + [ 'mordor' ] WHERE user_id = 'frodo';

[bsbodden]

I also wonder if cassandra_migrations should check the target version that's running and throw exceptions when a used feature is not supported? I'll take a look at other migration/ORMs in Ruby and see how they deal with different target versions of the driver/db.

[bsbodden]

I've updated the implementation to support inserting/updating the different collection types and removing secondary indexes along with a new battery of tests.

[bsbodden]

I'll test this for a week or so in the app we're building and then I'll ping you with my comments.

[hsgubert]

Great! I'll review the code through the weekend meanwhile.

[hsgubert]

Merged.

bsbodden, I would ask you to briefly review de readme file and see if any addition is necessary (in respect to using collections).

Thanks for your great work!

[hsgubert]

Also released version 0.0.5 with this modifications already.

[bsbodden]

I will. Thanks.