flink-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Aljoscha Krettek (JIRA)" <j...@apache.org>
Subject [jira] [Created] (FLINK-5929) Allow Access to Per-Window State in ProcessWindowFunction
Date Mon, 27 Feb 2017 16:33:45 GMT
Aljoscha Krettek created FLINK-5929:

             Summary: Allow Access to Per-Window State in ProcessWindowFunction
                 Key: FLINK-5929
                 URL: https://issues.apache.org/jira/browse/FLINK-5929
             Project: Flink
          Issue Type: Improvement
          Components: DataStream API
            Reporter: Aljoscha Krettek

Right now, the state that a {{WindowFunction}} or {{ProcessWindowFunction}} can access is
scoped to the key of the window but not the window itself. That is, state is global across
all windows for a given key.

For some use cases it is beneficial to keep state scoped to a window. For example, if you
expect to have several {{Trigger}} firings (due to early and late firings) a user can keep
state per window to keep some information between those firings.

The per-window state has to be cleaned up in some way. For this I see two options:
 - Keep track of all state that a user uses and clean up when we reach the window GC horizon.
 - Add a method {{cleanup()}} to {{ProcessWindowFunction}} which is called when we reach the
window GC horizon that users can/should use to clean up their state.

On the API side, we can add a method {{windowState()}} on {{ProcessWindowFunction.Context}}
that retrieves the per-window state and {{globalState()}} that would allow access to the (already
available) global state. The {{Context}} would then look like this:
 * The context holding window metadata
public abstract class Context {
     * @return The window that is being evaluated.
    public abstract W window();

     * State accessor for per-key and per-window state.
    KeyedStateStore windowState();

     * State accessor for per-key global state.
    KeyedStateStore globalState();

This message was sent by Atlassian JIRA

View raw message