[
https://issues.apache.org/jira/browse/DRILL-5492?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16002062#comment-16002062
]
Paul Rogers commented on DRILL-5492:
------------------------------------
A variation on this bug. Consider this input file:
{code}
____,__,__
a,b,c
d,e,f
{code}
Where, again, the underscores represent spaces. The header parser returns:
{code}
[" "," "," "]
{code}
That is, three fields, each some number of spaces. And, query output is:
{code}
,
a,c
d,f
{code}
Strangely, we pass three field headings into the {{FieldVarCharOutput}}
constructor (all strings with blanks), but we only get the first and third
fields in the output.
Expected the query to just fail due to blank headers.
> CSV with spaces for header uses spaces as field name
> ----------------------------------------------------
>
> Key: DRILL-5492
> URL: https://issues.apache.org/jira/browse/DRILL-5492
> Project: Apache Drill
> Issue Type: Bug
> Reporter: Paul Rogers
> Priority: Minor
>
> Consider the same test case as in DRILL-5491, but with a slightly different
> input file:
> {code}
> ___
> a,b,c
> d,e,f
> {code}
> The underscores represent three spaces: use spaces in the real test.
> In this case, the code discussed in DRILL-5491 finds some characters and
> happily returns the following array:
> {code}
> [" "]
> {code}
> The field name of three blanks is returned to the client to produce the
> following bizarre output:
> {code}
> 2 row(s):
>
> a
> d
> {code}
> The blank line is normally the header, but the header here was considered to
> be three blanks. (In fact, the blanks are actually printed.)
> Since the blanks were considered to be a field, the file is assumed to have
> only one field, so only the first column was returned.
> The expected behavior is that spaces are trimmed from field names, so the
> field name list would be empty and a User Error thrown. (That is, it is
> confusing to the user why a blank line produces NPE, some produce the
> {{ExecutionSetupException}} shown in DRILL-5491, and some produce blank
> headings. Behavior should be consistent.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
