[PHOENIX-418] Support approximate COUNT DISTINCT

• Type: New Feature
• Status: Resolved
• Priority: Blocker
• Resolution: Fixed
• Fix Version/s: 4.12.0
• Labels: gsoc2016
• old issue number: 287
• Assignee: Ethan Wang
• Reporter: James R. Taylor
• Created: 10/Mar/14 23:47
• Updated: 19/Sep/17 23:40
• Resolved: 31/Aug/17 00:37

Description

Support an "approximation" of count distinct to prevent having to hold on to all distinct values (since this will not scale well when the number of distinct values is huge). The Apache Drill folks have had some interesting discussions on this [here](http://mail-archives.apache.org/mod_mbox/incubator-drill-dev/201306.mbox/%3CJIRA.12650169.1369931282407.88049.1370645900553%40arcas%3E). They recommend using [Welford's method](http://en.wikipedia.org/wiki/Algorithms_for_calculating_variance_Online_algorithm). I'm open to having a config option that uses exact versus approximate. I don't have experience implementing an approximate implementation, so I'm not sure how much state is required to keep on the server and return to the client (other than realizing it'd be much less that returning all distinct values and their counts).

Update:
Syntax of using approximate count distinct as:
select APPROX_COUNT_DISTINCT(name) from person
select APPROX_COUNT_DISTINCT(address||name) from person

It is equivalent of Select COUNT(DISTINCT ID) from person. Implemented using hyperloglog, see discuss below.

Source code patch link below, co-authorred with [e-mail]
https://git-wip-us.apache.org/repos/asf?p=phoenix.git;a=commitdiff;h=d6381afc3af976ccdbb874d4458ea17b1e8a1d32

Issue Links

• duplicates: PHOENIX-3225 Distinct Queries are slower than expected at scale.
• duplicates: PHOENIX-3390 Custom UDAF for HyperLogLogPlus
• relates to: CALCITE-1588 Add SQL syntax to allow approximate LIMIT and distinct-COUNT
• relates to: PHOENIX-4160 research for a proper hash size set for APPROX_COUNT_DISTINCT