There is not always a 1 to 1 mapping between unit objects in the data source, and documents in the index. For example, suppose you are writing a connector for a data source dealing with emails. Should it be possible for a user to find emails based on the content of their attachments? Most probably yes, therefore this connector is probably going to map an email and all its attachments with a single document. Should it also be possible to find a whole thread of discussion, query with quotes from an email? If so, then the connector will probably push along with the previous documents, 1 document per thread, in which the content of all emails will have been mapped. For example:
|