Supervision#

Purpose#

This component provides a supervision mechanism over tshistory.

The central use case is as follows:

time series from third party providers are stored as is
from time to time, a data analyst will want to override (bogus) values, and need to know what values were overriden
when upstream fix its bogus values the fixed values must override the data analyst fixes
the complete series history must retain information on the supervision activity

The basic tshistory usage is described on its own documentation.

An override is made as follows:

>>> tsa.update('my-series', series, 'analyst@corp.com', manual=True)

This also works with the replace api call.

Inserted values will show up in the next .get call.

Specific API calls exist to provide a standard workflow:

.edited returns a couple series and markers where series is like the output of a standard .get call and markers a boolean series indicating whether any data point is from upstream or a manual edition.

The series metadata grows a new supervision-status attribute, whose value can be one of unsupervised, supervised and hand-crafted.

Hand-crafted is for series that are entirely made of manual updates.