Attendees: Chris Westin, Edmon Bergoli, Sudheesh Katkam, Daniel Barclay,
Jinfeng Ni, Venki Korukanti, Julien Le Dem

(*) Edmon has outstanding requests on mailing lists
- 1. plug-in for reading excel files
- 2. plug-in to read EDI files

Venki offered some suggestions re using the format plug-in to write these
plug-ins
- Venki will answer some of these on the dev list

Compute resources coming
- Sudheesh: Drill doesn't currently have any special handling for
co-processors
- available machines have up to 256G RAM
- have done some Spark testing in the past, so the support folks know
something
  about how to set up JVMs for this

(*) Daniel wants to talk about how empty files should be handled.
Message posted to dev list, but not answered.
- a filter pushdown made it appear like a file was empty (because nothing
passed the filter), so the schema information (OK_NEW_SCHEMA) was never
returned.

>From previous discussions: we're not good at handling empty files in
general.

The record handling protocol needs better definition; Daniel has come up
with
docs with help with Steven, but he's waiting for review and approval
DRILL-2888 (current bug) and DRILL-3641 (primarily the doc changes).

(*) Re upcoming release: Jacques shading change; Zookeeper used a different
version of Jackson than the one in the shaded jar.
- If using the jdbc-all jar, can't connect to Zookeeper, get ClassNotFound
- ticket currently assigned to Jacques

(*) Shaded test that isn't working
- might be easier to do that if we used maven integration tests
  - would that make it easier to run form an IDE?
- is someone at Dremio working on integration tests?

(*) A few people are going to Strata NYC next week

(*) Edmon wants to know the status of Java 1.8 Support.
- Venki: issue with order of values stored in the HashMap
- Chris: here's the umbrella ticket for the known issues:
https://issues.apache.org/jira/browse/DRILL-1491
That currently has one open sub-task left:
https://issues.apache.org/jira/browse/DRILL-1489

Reply via email to