[
https://issues.apache.org/jira/browse/DRILL-5356?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Paul Rogers updated DRILL-5356:
-------------------------------
Description:
The Parquet record reader class is a key part of Drill that has evolved over
time to become somewhat hard to follow.
A number of us are working on Parquet-related tasks and find we have to spend
an uncomfortable amount of time trying to understand the code. In particular,
this writer needs to figure out how to convince the reader to provide
higher-density record batches.
Rather than continue to decypher the complex code multiple times, this ticket
requests to refactor the code to make it functionally identical, but
structurally cleaner. The result will be faster time to value when working with
this code.
This is a lower-priority change and will be coordinated with others working on
this code base. This ticket is only for the record reader class itself; it does
not include the various readers and writers that Parquet uses since another
project is actively modifying those classes.
was:
The Parquet record reader class is a key part of Drill that has evolved over
time to become somewhat hard to follow.
A number of us are working on Parquet-related tasks and find we have to spend
an uncomfortable amount of time trying to understand the code. In particular,
this writer needs to figure out how to convince the reader to provide
higher-density record batches.
Rather than continue to decypher the complex code multiple times, this ticket
requests to refactor the code to make it functionally identical, but
structurally cleaner. The result will be faster time to value when working with
this code.
This is a lower-priority change and will be coordinated with others working on
this code base.
> Refactor Parquet Record Reader
> ------------------------------
>
> Key: DRILL-5356
> URL: https://issues.apache.org/jira/browse/DRILL-5356
> Project: Apache Drill
> Issue Type: Improvement
> Affects Versions: 1.10.0, 1.11.0
> Reporter: Paul Rogers
> Assignee: Paul Rogers
> Priority: Minor
> Fix For: 1.11.0
>
>
> The Parquet record reader class is a key part of Drill that has evolved over
> time to become somewhat hard to follow.
> A number of us are working on Parquet-related tasks and find we have to spend
> an uncomfortable amount of time trying to understand the code. In particular,
> this writer needs to figure out how to convince the reader to provide
> higher-density record batches.
> Rather than continue to decypher the complex code multiple times, this ticket
> requests to refactor the code to make it functionally identical, but
> structurally cleaner. The result will be faster time to value when working
> with this code.
> This is a lower-priority change and will be coordinated with others working
> on this code base. This ticket is only for the record reader class itself; it
> does not include the various readers and writers that Parquet uses since
> another project is actively modifying those classes.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
