How to create complex structure in Cassandra with CQL3

Question

I have problem with presenting complex data structure in cassandra. JSON example of data :

{
  "A": {
    "A_ID" : "1111"
    "field1": "value1",
    "field2": "value2",
    "field3": [
      {
        "id": "id1",
        "name": "name1",
        "segment": [
          {
            "segment_id": "segment_id_1",
            "segment_name": "segment_name_1",
            "segment_value": "segment_value_1"
          },
          {
            "segment_id": "segment_id_2",
            "segment_name": "segment_name_2",
            "segment_value": "segment_value_2"
          },
          ...
        ]
      },
      {
        "id": "id2",
        "name": "name2",
        "segment": [
          {
            "segment_id": "segment_id_3",
            "segment_name": "segment_name_3",
            "segment_value": "segment_value_3"
          },
          {
            "segment_id": "segment_id_4",
            "segment_name": "segment_name_4",
            "segment_value": "segment_value_4"
          },
          ...
        ]
      },
      ...
    ]
  }
}

Will be used only one query: Find by A_ID.

I think this data should store in one TABLE (Column Family) and without serialization/deserialization operations for more efficiency. How can I do this if CQL does not support nested maps and lists?

Answer 1

As you will be using it just as a key/value, you could actually store it either as JSON, or for saving data more efficiently, something like BSON or event Protobuf.

I personally would store it in the Protobuf record, as it doesn't save the field names which may be repeating in your case.

Answer 2

Cassandra 2.1 adds support for nested structures: https://issues.apache.org/jira/browse/CASSANDRA-5590

The downside to "just store it as a json/protobuf/avro/etc blob" is that you have to read-and-rewrite the entire blob to update any field. So at the very least you should pull your top level fields into Cassandra columns, leveraging collections as appropriate.