Mailing-List: contact issues-help@drill.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@drill.apache.org
Date: Thu, 8 Jan 2015 04:27:35 +0000 (UTC)
From: "Sean Hsuan-Yi Chu (JIRA)" <jira@apache.org>
To: issues@drill.apache.org
Message-ID: <JIRA.12762043.1418748850000.35059.1420691255111@Atlassian.JIRA>
In-Reply-To: <JIRA.12762043.1418748850000@Atlassian.JIRA>
References: <JIRA.12762043.1418748850000@Atlassian.JIRA>
 <JIRA.12762043.1418748850555@arcas>
Subject: [jira] [Updated] (DRILL-1874) Text reader should allow treating
 empty fields as NULLs instead of empty string
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit


     [ https://issues.apache.org/jira/browse/DRILL-1874?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Sean Hsuan-Yi Chu updated DRILL-1874:
-------------------------------------
    Attachment: DRILL-1874.3.patch

This patch passed unit tests, functional and tpch

> Text reader should allow treating empty fields as NULLs instead of empty string
> -------------------------------------------------------------------------------
>
>                 Key: DRILL-1874
>                 URL: https://issues.apache.org/jira/browse/DRILL-1874
>             Project: Apache Drill
>          Issue Type: Improvement
>          Components: Storage - Text & CSV
>    Affects Versions: 0.7.0
>            Reporter: Aman Sinha
>            Assignee: Sean Hsuan-Yi Chu
>             Fix For: 0.8.0
>
>         Attachments: DRILL-1874.1.patch, DRILL-1874.2.patch, DRILL-1874.3.patch
>
>
> The text reader treats a record as a repeated varchar and if there is an empty field it produces an empty string.   This prevents CASTing the field to a numeric type because one can get a NumberFormatException.   This is a pain point for many use cases because empty fields occur frequently in text files. 
> Ideally, based on some configuration setting, we should allow treating empty fields as NULL values instead of empty string.  This will allow cast to work correctly.  
> In the meantime, here's a rewrite that can be used as workaround:   
> {code}
> 0: jdbc:drill:zk=local> select case when columns[17] = '' then cast(null as int) else cast(columns[17] as int) end  from dfs.tmp.`sample2.tbl`;
> +------------+
> |   EXPR$0   |
> +------------+
> | 5          |
> | null       |
> +------------+
> {code}


--
This message was sent by Atlassian JIRA
(v6.3.4#6332)