[
https://issues.apache.org/jira/browse/FLINK-4506?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15510680#comment-15510680
]
Fabian Hueske commented on FLINK-4506:
--------------------------------------
I'm not sure about that. The CsvInputFormat supports tuple and Pojo data types,
tuples being the common choice. However, tuples do not support null values and
hence a null value will cause a NPE during serialization.
Apart from that, I'm not in favor of changing the behavior of stable declared
public interface without very good reason.
> CsvOutputFormat defaults allowNullValues to false, even though doc and
> declaration says true
> --------------------------------------------------------------------------------------------
>
> Key: FLINK-4506
> URL: https://issues.apache.org/jira/browse/FLINK-4506
> Project: Flink
> Issue Type: Bug
> Components: Batch Connectors and Input/Output Formats, Documentation
> Reporter: Michael Wong
> Assignee: Kirill Morozov
> Priority: Minor
>
> In the constructor, it has this
> {code}
> this.allowNullValues = false;
> {code}
> But in the setAllowNullValues() method, the doc says the allowNullValues is
> true by default. Also, in the declaration of allowNullValues, the value is
> set to true. It probably makes the most sense to change the constructor.
> {code}
> /**
> * Configures the format to either allow null values (writing an empty
> field),
> * or to throw an exception when encountering a null field.
> * <p>
> * by default, null values are allowed.
> *
> * @param allowNulls Flag to indicate whether the output format should
> accept null values.
> */
> public void setAllowNullValues(boolean allowNulls) {
> this.allowNullValues = allowNulls;
> }
> {code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
