This release contains more experimental features, along with additional functionality for prior features.
New Experimental features:
- ASAP Smoothing – A graph smoothing algorithm that highlights changes.
- Counter Aggregates – Tools to ease working with reset-able counters.
- Largest Triangle Three Buckets – A downsampling algorithm that tries to preserve visual similarity.
- Time Bucket Range – A version of
time_bucket()that outputs the[start, end)times of the bucket.
In addition, we've updated UddSketch with an aggregate that merges multiple UddSketchs and various internal improvements.
First release. Currently all features are experimental, and may change by future versions; do not use them if you will unable to handle that. In particular, the schema containing the experimental features will be dropped and recreated in the upgrade script, so do not create any objects (views, continuous aggregates, tables, functions, etc.) that rely on features in the experimental schema unless you are ok with said objects being dropped on upgrade.
There are no stable features added in this release.
Notable Experimental features in this release include:
- Hyperloglog – An approximate
COUNT DISTINCTbased on hashing that provides reaonable accuracy in constant space. - T-Digest A quantile estimate sketch optimized to provide more accurate estimates near the tails (i.e. 0.001 or 0.995) than conventional approaches.
- UddSketch – A quantile estimate sketch which provides a guaranteed maximum relative error.