CassandraFilterExpressionConverter bug with `ExpressionType.IN` with collections

[soheilrahsaz]

When using a filter expression on a collection in CassandraVectorStore.similaritySearch like this
new Filter.Expression(Filter.ExpressionType.IN, new Filter.Key("currencies"), new Filter.Value(List.of("BTC")));

In CassandraFilterExpressionConverter, it will go through doListValue and then tries to encode each value using doValue, But in doValue method, it is trying to get the codec for the column itself, which is a list, and then fails to encode the string value in the list and throws exception:
java.lang.ClassCastException: class java.lang.String cannot be cast to class java.util.List (java.lang.String and java.util.List are in module java.base of loader 'bootstrap')

It seems that for collections, it should use the elementType of the dataType and not the dataType of the column itself. (CassandraFilterExpressionConverter line 124)

[sobychacko]

@soheilrahsaz PR submitted to address this specific issue here: #3197.

Just wanted to clarify a limitation we're dealing with in Cassandra's query language. When we tried using the IN operator on collection columns (like currencies IN ('BTC')), Cassandra throws an error because it fundamentally doesn't support this syntax. In Cassandra, you can't use IN to check if a collection contains specific values - instead, you'd need to use the CONTAINS operator with the syntax flipped around ('BTC' IN currencies). Our fix here prevents Java exceptions when building the query, but to fully support filtering on collection columns, we'd need to implement proper support for the CONTAINS operator in the Spring AI Cassandra vector store. For now, Cassandra will correctly reject these queries with a clear error message rather than causing application errors.

[soheilrahsaz]

Well, it needs to be currencies CONTAINS ('BTC') as the documents say: https://cassandra.apache.org/doc/latest/cassandra/developing/cql/indexing/sai/sai-working-with.html
And the Contains_key would be for map types. =

[sobychacko]

Ya, thanks for that clarification. Regardless, that needs to wait until we properly implement the CONTAINS operator in CassandraVectorStore.