add mechanism to scan a time series (TS) key, N data points at a time

[kmuthukk]

YugaByte's Redis API introduces a timeseries (TS) type with operations such as: TSAdd, TSRangeByTime, and TSGet and TSRem.

In particular, TSRangeByTime can be used to retrieve a slice of the timeseries corresponding to the specified range of timestamps. In addition to TSRangeByTime, it'll be useful to add a mechanism to fetch the data points of a timeseries N elements a time.

Given that the natural order in which the data points are stored in DocDB is from highest timestamp to lowest, the proposal is to add a:

TSRevRangeByTime key low_ts high_ts [LIMIT N]

This will return the results in high to low TS order, but up to a max of N elements.

When the response returns fewer than N data points, it signals the end of the scan.

Using the above, it will be possible to scan through the data points of a timeseries key in the following manner:

Say you want tofetch data from highest timestamp to lowest timestamp 100 elements at time, all that an app would have to do would be to start with:

TSRevRangeByTime key -inf +inf LIMIT 100

and then follow it up with subsequent calls:

TSRevRangeByTime key -inf (<lowest_ts_received_from_previous_call> LIMIT 100

where the left-paren "(" signifies an exclusive bound (which is a standard Redis idoim).

(Note: For symmetry, we should eventually also enhance TSRangeByTime with a LIMIT clause; but we could defer that till we actually implement reverse scans in DocDB for our Redis data model.)

[kmuthukk]

Reassigning to you @hectorgcr (as per our offline discussion).