TSDataset.describe and TSDataset.info show end_timestamp but in most places we use time-series until the ts.index[-1]

[d-a-bunin]

Description

We have TSDataset.describe and TSDataset.info that for each segment show end_timestamp without taking into account missing values at the end of the series. However, during forecasting time-series we use ts.index[-1] as end timestamp for each segment that contradicts with these methods.

We should probably fill end_timestamp with ts.index[-1] and may be rename current value into something else like end_timestamp_without_missing or smth like that.

Thoughts

There are probably places when we ignore missing values at the end of time-series. For example, there are multiple places where we use last_valid_index method: link.

We probably need more consistency in this regard.

Related

No response