Re: indexed views, some clarifying terminology

In article <un1l42jxl.fsf_-__at_fmr.com>, steve.tolkin_at_fmr.com writes
>
>In the hope of clarifying the subject let me introduce some
>terminology.
>
>A "view" (aka a viewed table) is a table whose contents correspond to
>a select query. The dbms may impose limitations on what language
>constructs can be be used in the view's defining query.

OK.

>
>A "materialized view" is a view table that stores its contents,
>i.e. the results of running the view's query.

That's a plausible definition.

>
>A "pure view" or "virtual view" is a view that is not a materialized
>view. This is the usual connotation of the word view. The increased
>use of materialized views has forced us to invent this term. (I do
>not recall the name for this linguistic phenomenon, e.g. the need to
>add the adjective in analog watch, whale oil, etc.)

That's also a plausible definition.

These definitions describe ideal situations and real-world situations may not conform to either. For instance I can imagine a caching solution where a server actualises a subset of a view, predicting what the client will request next. Rows that have already been passed to the client are discarded to save space.

That's neither a pure nor a materialised view.

This is a theory newsgroup so it's fine to talk about theoretical models of database design. In reality RDBMS suppliers aren't constrained to these theoretical ideals. There is a spectrum between the ideals and they can position their products anywhere in the spectrum.

-- 
Bernard Peek
bap_at_shrdlu.com
bap_at_shrdlu.co.uk
bap_at_shrdlu.org.uk