[
https://issues.apache.org/jira/browse/DRILL-5928?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Paul Rogers updated DRILL-5928:
-------------------------------
Description:
Consider the following test framework query:
{noformat}
/root/drillAutomation/mapr/framework/resources/Functional/data-shapes/wide-columns/general/q2.q
{noformat}
Defined as follows:
{noformat}
select count(*) from `data-shapes/wide-columns/general/100000.tbl`
{noformat}
The data file ({{100000.tbl}}) contains a single row, single field of 100K
width. The query suceeds and returns 1.
However, if we did the following:
{noformat}
select * from `data-shapes/wide-columns/general/100000.tbl`
{noformat}
then the query would fail with an oversize field exception. (Before Drill 1.13
the error is "Tried to write something large in a field", while 1.13 and after
the error is "Text column is too large."
The question is, should the query succeed in the {{COUNT\(*)}} case if it would
fail in the {{SELECT *}} case?
Similarly, one can invent a table with more than 64K columns. A {{SELECT *}}
query would fail with an index out of bounds exception (Drill 1.12 or earlier)
or a "Too many fields" error (Drill 1.13 and later.) But, a {{COUNT(*)}} query
would succeed. Is this desired?
was:
Consider the following test framework query:
{noformat}
/root/drillAutomation/mapr/framework/resources/Functional/data-shapes/wide-columns/general/q2.q
{noformat}
Defined as follows:
{noformat}
select count(*) from `data-shapes/wide-columns/general/100000.tbl`
{noformat}
The data file ({{100000.tbl}}) contains a single row, single field of 100K
width. The query suceeds and returns 1.
However, if we did the following:
{noformat}
select * from `data-shapes/wide-columns/general/100000.tbl`
{noformat}
then the query would fail with an oversize field exception. (Before Drill 1.13
the error is "Tried to write something large in a field", while 1.13 and after
the error is "Text column is too large."
The question is, should the query succeed in the {{COUNT\(*)}} case if it would
fail in the {{SELECT *}} case?
> Drill allows a COUNT(*) over records that are too wide to load
> --------------------------------------------------------------
>
> Key: DRILL-5928
> URL: https://issues.apache.org/jira/browse/DRILL-5928
> Project: Apache Drill
> Issue Type: Bug
> Affects Versions: 1.11.0
> Reporter: Paul Rogers
> Priority: Minor
>
> Consider the following test framework query:
> {noformat}
> /root/drillAutomation/mapr/framework/resources/Functional/data-shapes/wide-columns/general/q2.q
> {noformat}
> Defined as follows:
> {noformat}
> select count(*) from `data-shapes/wide-columns/general/100000.tbl`
> {noformat}
> The data file ({{100000.tbl}}) contains a single row, single field of 100K
> width. The query suceeds and returns 1.
> However, if we did the following:
> {noformat}
> select * from `data-shapes/wide-columns/general/100000.tbl`
> {noformat}
> then the query would fail with an oversize field exception. (Before Drill
> 1.13 the error is "Tried to write something large in a field", while 1.13 and
> after the error is "Text column is too large."
> The question is, should the query succeed in the {{COUNT\(*)}} case if it
> would fail in the {{SELECT *}} case?
> Similarly, one can invent a table with more than 64K columns. A {{SELECT *}}
> query would fail with an index out of bounds exception (Drill 1.12 or
> earlier) or a "Too many fields" error (Drill 1.13 and later.) But, a
> {{COUNT(*)}} query would succeed. Is this desired?
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
