[
https://issues.apache.org/jira/browse/DEFT-142?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13070117#comment-13070117
]
Roger Schildmeijer commented on DEFT-142:
-----------------------------------------
Oh sweet lord. Even worse than I could imagine. Reflection is slow. QED.
Regarding clone: I guess that means that we will have the clone method abstract
in the RequestHandler class (to make it "easier" for third parties)? Like:
protected abstract Object clone() throws CloneNotSupportedException;
One different approach is to let the concrete RequestHandlers be, and use the
fact that HttpRequest is unique for each request and providoe something similar
to HttpSessions. Thoughts? (pros: less impact to existing code)
Do you mind attach the current reflection patch?
> Dynamic creation of RequestHandlers
> -----------------------------------
>
> Key: DEFT-142
> URL: https://issues.apache.org/jira/browse/DEFT-142
> Project: Deft
> Issue Type: New Feature
> Assignee: Johnathan Meehan
>
> The outcome after a discussion with github.com/inferno-:
> (In the current design) it's volatile to have fields in user defined
> RequestHandlers in conjunction with asynchronous calls. (intermediate request
> (to same RH) could change the state of the RH)
> We should investigate the performance impact for dynamic creation of RH.
> Proposed solution:
> 1, Map<String, RequestHandler> => Map<String, Class<RequestHandler>>
> 2, Application.getHandler should create the appropriate RequestHandler using
> reflection.
> 3, UT / benchmark
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
