This could apply to all ADABAS versions and platforms.
During the customer story on 6/21 with Michael S and St. of Delaware, they noted that connx queries for null values or non-keyed searches on million record files take minutes to respond and are not suitable for customer response time.
These types of queries are common for Oracle (and other) SQL dbs.
While it would be possible to create many DEs on every field, the subsequent update / delete processing may become slow to maintain all the ASSO space / values.
How do SQL dbs handle such queries internally - it does not appear as though they read the entire DB / file of data? And then joins add another level of complexity / overhead.
Would it be possible for ADABAS to use a user exit to maintain a separate "SEARCH PATH" (using the ORacle term for "show path") - that could keep updates to data indexes in an async mode - not exactly up to the second - and not delaying response times for ADD / UPDATE / DELETE commands which must maintain NI structures. But reduce the response time on million record files of non-keyed searches to seconds instead of minutes.
|extend ADABAS connx queries to be more dynamic and not practically limited to keyed selection criteria