[
https://issues.apache.org/jira/browse/FLINK-7947?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Till Rohrmann updated FLINK-7947:
---------------------------------
Description:
The {{ParameterTool}} directly implements the {{GlobalJobParameters}}
interface. Additionally it has grown over time to not only store the
configuration parameters but also to record which parameters have been
requested and what default value was set. This information is irrelevant on the
server side when setting a {{GlobalJobParameters}} object via
{{ExecutionConfig#setGlobalJobParameters}}.
Since we don't separate the {{ParameterTool}} logic and the actual data view,
users ran into problems when reusing the same {{ParameterTool}} to start
multiple jobs concurrently (see FLINK-7943). I think it would be a much clearer
separation of concerns if we would actually split the {{GlobalJobParameters}}
from the {{ParameterTool}}.
Furthermore, we should think about whether {{ParameterTool#get}} should have
side effects or not as it does right now.
was:
The {{ParameterTool}} directly implements the {{GlobalJobParameters}}
interface. Additionally it has grown over time to not only store the
configuration parameters but also to record which parameters have been
requested and what default value was set. This information is irrelevant on the
server side when setting a {{GlobalJobParameters}} object via
{{ExecutionConfig#setGlobalJobParameters}}.
Since we don't separate the {{ParameterTool}} logic and the actual data view,
users ran into problems when reusing the same {{ParameterTool}} to start
multiple jobs concurrently (see FLINK-7943). I think it would be a much clearer
separation of concerns if we would actually split the {{GlobalJobParameters}}
from the {{ParameterTool}}.
> Let ParameterTool return a dedicated GlobalJobParameters object
> ---------------------------------------------------------------
>
> Key: FLINK-7947
> URL: https://issues.apache.org/jira/browse/FLINK-7947
> Project: Flink
> Issue Type: Improvement
> Components: Client
> Affects Versions: 1.4.0
> Reporter: Till Rohrmann
>
> The {{ParameterTool}} directly implements the {{GlobalJobParameters}}
> interface. Additionally it has grown over time to not only store the
> configuration parameters but also to record which parameters have been
> requested and what default value was set. This information is irrelevant on
> the server side when setting a {{GlobalJobParameters}} object via
> {{ExecutionConfig#setGlobalJobParameters}}.
> Since we don't separate the {{ParameterTool}} logic and the actual data view,
> users ran into problems when reusing the same {{ParameterTool}} to start
> multiple jobs concurrently (see FLINK-7943). I think it would be a much
> clearer separation of concerns if we would actually split the
> {{GlobalJobParameters}} from the {{ParameterTool}}.
> Furthermore, we should think about whether {{ParameterTool#get}} should have
> side effects or not as it does right now.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
