On Tue, Nov 21, 2000 at 08:31:29PM +0200, Michael Widenius wrote:
> barries> Is this the start of computed function indexes (I hope)?
> No; Computed index are in a sence easier to handle than this :)
> The problems with computed index are:
> - You can't use a standalone tool like myisamchk to check the table.
> - It will be a problem for the optimizer to pattern-match the query
> for when to use the index.
This sounds like an excuse for me to bring up my suggestion for
"truncated resolution" indexes again...
Purpose: To make indexes on high-resolution values more space efficient.
Summary: Similar in concept to partial indexing of blobs but applied
to numeric and timestamp types. The index would store values truncated
to some lower level of resolution.
Example: Web logs with an index on a timestamp field. The resolution of
the index could be set to, say, minutes (/60) or hours (/3600). The
number of 'duplicate' values in the index would increase thus saving
space (especially if the index is packed, as I recall).