Michael Kleen

Jdk path points to the wrong location when run latest-nightly

3

The jdk path points to the wrong location when starting up `latest-nightly` in `cr8 0.18.1.dev5+g74d5c9f.d20200427` on mac os x. ``` > cr8 run-crate latest-nightly Skipping download, tarball alrady extracted at...

Extend Genbank parser to provide references to publications

2

Hello, I would like to extract references to publications from genbank entries, which is unfortunately not supported right now from the current genbank parser. I am happy to implement this...

Fix handling of correlated subqueries in join conditions

5

## Summary of the changes / Why this improves CrateDB This makes correlated sub-queries work when used in a join condition. In case that any other relation than the left...

## Summary of the changes / Why this improves CrateDB Follow-up https://github.com/crate/crate/pull/15517#discussion_r1497147130 ## Checklist - [x] Added an entry in the latest `docs/appendices/release-notes/.rst` for user facing changes - [x] Updated...

Performance Regression in outer join benchmark

11

We have a performance regression on the query: ``` select e.name as employee, o.name as office from employees e left join offices o on e.office_id = e.id limit 50000 ```...

The idea is a generalization of the `MoveConstantJoinConditionsBeneathNestedLoop` rule and taken from Apache Calcite: Assume the following logical plan: ``` Join[INNER | (a = b) AND (a > 1)] ├...

Join optimizations when sources are empty

3

If a logical plan returns 0 rows, the plan can be replaced by an empty relation. An empty relation is a placeholder which represents a logical plan without any rows...

Support Hash-Join execution for Outer-Joins

1

## Problem Statement We have various issues reporting poor performance with Outer-Joins such as https://github.com/crate/crate/issues/11851 or https://github.com/crate/crate/issues/14101. Outer-Joins are executed in CrateDB as Nested-Loop-Joins which makes them slow because of...

Push aggregations past joins with optimizer rule

6

Aggregates can be pushed through a join which reduces the intermediate data to improve performance significantly. Data setup ``` create table devices (deviceid int primary key,att1 text,att2 text); insert into...

Introduce Cost-based Query Optimization

6

### Problem Description There are sometimes different ways a query could be executed. For example a join can be executed either via nested-loop, or as hash-join. CrateDB currently uses a...