Lucene 9.5 BaseCompositeReader provides support to read stored fields values from fdt files for a specific document through StoredFields class. However I couldn't find the the same for vector fields in CompositeReader for vec files.
I have read the vector field value of a specific document using the CodecReader and FloatVectorValues. What I noticed is each retrieval involves disk seek compared to disk seek in stored fields where a block of data is seeked and subsequent neighbour retrievals are faster.
Is there any better approach to read vectors from Index? What is the reason behind not including the vector reader in CompositeReader?