For legacy reasons, non-string types support "empty" values, e.g. you can have
an empty int, distinct from NULL. This case used to trigger deserialization errors,
so now it's simply a fallback to NULL. Another option is to introduce a new "empty"
state that a value can be in, but it's semantically no different from NULL,
even though it's serialized in a different way.
psarna
commented
2 years ago
For legacy reasons, non-string types support "empty" values, e.g. you can have an empty int, distinct from NULL. This case used to trigger deserialization errors, so now it's simply a fallback to NULL. Another option is to introduce a new "empty" state that a value can be in, but it's semantically no different from NULL, even though it's serialized in a different way.
Ref: https://github.com/apache/cassandra/blob/7b58b79fe24c612f8bbf7984c325e88496f482d6/doc/native_protocol_v4.spec#L861-L863
Fixes #59