Open mikepizzo opened 8 years ago
See Yes, we find them very useful, see Issue 2.
The help both in dynamically generating user frontends as well as query optimization when you have your own SQL engine using metadata from the backing databases to choose the right join strategies and query plan.
Side note:
Problem at the moment is that many new cloud platforms are multi-tenant and have different metadata per partition (per rented office). So if you need to provide an ODBC driver that has the ability to run queries across multiple partitions (multiple tenants) then you need to be able to get the metadata across all partitions in an efficient way and determine what are additions and what metadata is available in all. SQL currently offers little to no support for dynamic metadata and given the foundation in set theory, I don't think it should, so the possible scenarios are either using the intersect or union of the metadata when querying across multiple partitions. And with a union of the metadata let's hope that identically considered columns have the same semantics :-)
ANSI also supports DEGREE as a descriptor field for number of columns in ROW (and CARDINALITY for max rows in an array)
Are those useful? How does the degree of a row change with dynamic columns? How do you specify “unknown” cardinality (or degree)?