For what reasons it is not preferable to use Thrift API for Cassandra access?

Please enumerate reasons for why its not advisable to use Thrift 开发者_StackOverflow社区interface for Cassandra? What are the possible advantages and disadvantages?

If you use the raw Thrift apis the cons will be:

• no connection pooling
• no monitoring
• no object oriented interface (not entirely true)
• no failover support

To continue Schildmeijer's good start:

• No batch interface.
• No chunking of get_range_slices() or get_indexed_slices() so you can easily swamp Cassandra
• Non-string types must be packed into binary strings yourself
• You'll probably mess up timestamp precision
• Exception messages are generally useless
• Thrift is broken by default in some languages. See the PHP C extension, for example.
• Because the code is generated, it's not intuitive, especially regarding super columns, SlicePredicates and batch_mutate().
• Schema modification commands don't wait for schema agreement among all the nodes in the cluster