drill-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Hanifi Gunes (JIRA)" <j...@apache.org>
Subject [jira] [Created] (DRILL-2147) Refactor ValueVector design
Date Tue, 03 Feb 2015 02:21:34 GMT
Hanifi Gunes created DRILL-2147:

             Summary: Refactor ValueVector design
                 Key: DRILL-2147
                 URL: https://issues.apache.org/jira/browse/DRILL-2147
             Project: Apache Drill
          Issue Type: Bug
            Reporter: Hanifi Gunes
            Assignee: Hanifi Gunes

The overall design of value vectors has become unclear and inconsistent with additions from
multiple contributors over the time. Also we need proper documentation for the abstractions
made for consistently communicating with developers. 

There are many instances that indicate possible design issues.

For instance, ValueVector implements Iterator<ValueVector>. This seems to assume all
vectors are somewhat hierarchical. This does not truly capture scalar vectors as they have
no child.

Similarly, RepeatedVector has the following interface definition:
interface RepeatedVector {
  RepeatedFixedWidthVector.RepeatedAccessor getAccessor()

Yet, RepeatedFixedWidthVector implements RepeatedVector as follows
interface RepeatedFixedWidthVector extends ValueVector, RepeatedVector {
  interface RepeatedAccessor extends Accessor {...}
  interface RepeatedMutator extends Mutator {...}

A super-type that is aware of its sub-type hints a need for re-design.

Examples could be multiplied here: some method names are not self-explaining or wrongly named
or seems to be misplaced. There are couple of more places where design is not capturing the
nature of vectors such like missing abstractions for Repeated vs Composite vectors. We should
consider a design refactoring.

This is an umbrella issue for tracking ValueVector design refactoring.

This message was sent by Atlassian JIRA

View raw message