Re: Not-So-New Bit-Mapped Indices (WAS Re: Bit-Mapped Indexing Routines---Beta Testers Wanted)

Barry Johnson <BJohnson_at_WorldBank.Org> wrote in article <32AEC55B.1844_at_WorldBank.Org>...
> David Walrath (walrath_at_sybase.com) wrote:
> > Incidentally, I wonder what the original poster meant by bitmap indexes
> > being "relatively new" in databases. Low cardinality bitmap indexes
> > were in Model 204 well over 20 years ago, so maybe he/she meant
> > relative to punch-card databases ?
> > ...
> > Opinions are my own; my employer doesn't want them.
>
> FWIW, Burroughs' DMSII also had them by the time I got to it in '77...
>
> --
> Barry Johnson - BJohnson_at_WorldBank.Org

Actually bitmap indexes existed in the punch card era. They were called (I believe) "Keysort" a trade name of (I believe) Royal-McBee. Holes were punched around the edges of a card upon which was written the information. The "bits" were turned from "zero" to "one" by notching the edge of the card to the hole. A query was made by running a sorting needle through the stack of cards and shaking the deck. All cards which matched the query attribute would fall out. The process would then be repeated on the fallen cards to refine the query. The final results of the query could then be read from the remaining "dropped" cards. This is the origin of the term "false drop" for a query match that should not have occurred or occurred by accident.

Needless to say these were not Hollerith (IBM) punched cards.

--Dave Received on Wed Dec 18 1996 - 00:00:00 CST