time-series databases fit naturaly with Pandas dataframes. Plan is to make this a PR against Pandas when underlying backend(s) is released.
The dataframe column names that start with _ will be treated as metadata that do not need to be saved to database unless as tags/labels, e.g. _pk can be used as primary key to group different measurements together.
Column names should follow a namespace notation with dots (.) e.g. activity.calories is measurement about calories but has namesapce of activity.
Namespaces can be automatically added to columns prior to recording to database, and taken out when reading results.
At this point the only supported backend is InfluxDB, but there is at least one more we will add later.
BSD