Clarify TIMEWINDOW semantics

Description

Create a better more coherent idea of TIMEWINDOW, probably based on clinical (real world) timing window, and with a new concept to represent the idea of a system time (i.e. versioning time) moment that you want to query up until - @Seref called this ATTIME.

Thomas:

Here is the last Slack discussion on TIMEWINDOW. There was also a discussion in the old SEC list. @bna, @Seref and @pablo had various useful things to say in both . The outcome of the discussions was approximately that:

- we agreed that ‘window’ is not really the correct concept, and it should be understood as a single point in ‘system time’ (i.e. version commit time, the time the system knew of any given data);

- we possibly should replace it with something like AT_TIME Tc, meaning execute the query on the DB commits up to Time Tc (i.e. ignoring anything later);

- we might want a ‘time window’ concept for world time (i.e. clinical time) as well.

We don’t need to work all this out right now for AQL 1.0.1. I would propose to either obsolete or just remove TIMEWINDOW from the spec (it is not in the grammar, and AFAIK, not in anyone’s implementation) then we put something new into AQL 1.1.0. How does this sound?

 

 

see https://discourse.openehr.org/t/aql-1-0-1-progress/345

Activity

Show:

Details

Reporter

Raised By

Pablo Pazos

Components

Affects versions

Created February 14, 2020 at 2:29 PM
Updated December 7, 2020 at 3:24 PM