[
https://issues.apache.org/jira/browse/FLINK-2021?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15117069#comment-15117069
]
ASF GitHub Bot commented on FLINK-2021:
---------------------------------------
Github user fhueske commented on the pull request:
https://github.com/apache/flink/pull/1536#issuecomment-174959098
The examples handle inputs differently than most other DataSet programs
because they use default input data if no input paths are specified. This adds
complexity which is not present in common programs.
I agree that reading input is an important aspect of a DataSet program, but
adding 15 lines for each input to each example seems unnecessary to me.
Extracting code into methods can be good practice, even if a method is only
called once. Reading code can become a lot easier if it is clear what a method
does (good method name + comments) because it can hide all little nasty details
and helps to focus on the relevant parts.
The only question is how do we define relevance ;-)
@stefanobaghino, I think this is a good idea. I would keep the parameter
parsing in the main function of the examples, but only move the code to either
read a CSV file or get a collection DataSource to a method.
> Rework examples to use ParameterTool
> ------------------------------------
>
> Key: FLINK-2021
> URL: https://issues.apache.org/jira/browse/FLINK-2021
> Project: Flink
> Issue Type: Improvement
> Components: Examples
> Affects Versions: 0.9
> Reporter: Robert Metzger
> Priority: Minor
> Labels: starter
>
> In FLINK-1525, we introduced the {{ParameterTool}}.
> We should port the examples to use the tool.
> The examples could look like this (we should maybe discuss it first on the
> mailing lists):
> {code}
> public static void main(String[] args) throws Exception {
> ParameterTool pt = ParameterTool.fromArgs(args);
> boolean fileOutput = pt.getNumberOfParameters() == 2;
> String textPath = null;
> String outputPath = null;
> if(fileOutput) {
> textPath = pt.getRequired("input");
> outputPath = pt.getRequired("output");
> }
> // set up the execution environment
> final ExecutionEnvironment env =
> ExecutionEnvironment.getExecutionEnvironment();
> env.getConfig().setUserConfig(pt);
> {code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
