[
https://issues.apache.org/jira/browse/STORM-1369?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15670386#comment-15670386
]
Magnus Koch edited comment on STORM-1369 at 12/14/16 3:33 PM:
--------------------------------------------------------------
Picked this up, because we need it internally.
Implementation is available in pull requests
https://github.com/apache/storm/pull/1781 (for master) and
https://github.com/apache/storm/pull/1782 (for 1.0.x-branch).
The README shows the basic usage of it, and there is a unit test that
implements the word count example for opaque, transactional and
nontransactional map states.
The approach was to reuse much of the existing features (e.g. CQL tuple
mappers) and only adding a few things to wire them together with an
IBackingMap.
Notably:
* There is now a StateMapper which converts the map state (which is generic
and can be opaque/transactional) to/from a tuple format.
* There is a new async execution option with an AsyncResultSetHandler to
collect the results - this is used by the MapState implementation which by
default queries Cassandra for multiple keys in parallel.
* There is a builder class to simplify construction, which builds the CQL
statements for you. This is not strictly necessary but it really does help
usage and readability for most cases.
was (Author: mkoch1):
Picked this up, because we need it internally.
Implementation is available at
https://github.com/ef-labs/storm/tree/feature/STORM-1369-on-master/external/storm-cassandra
The README shows the basic usage of it, and there is a unit test that
implements the word count example for opaque, transactional and
nontransactional map states.
The approach was to reuse much of the existing features (e.g. CQL tuple
mappers) and only adding a few things to wire them together with an
IBackingMap.
Notably:
* There is now a StateMapper which converts the map state (which is generic
and can be opaque/transactional) to/from a tuple format.
* There is a new async execution option with an AsyncResultSetHandler to
collect the results - this is used by the MapState implementation which by
default queries Cassandra for multiple keys in parallel.
* There is a builder class to simplify construction, which builds the CQL
statements for you. This is not strictly necessary but it really does help
usage and readability for most cases.
> Add trident MapState to Cassandra
> ---------------------------------
>
> Key: STORM-1369
> URL: https://issues.apache.org/jira/browse/STORM-1369
> Project: Apache Storm
> Issue Type: Improvement
> Components: storm-cassandra
> Reporter: Satish Duggana
> Assignee: Satish Duggana
> Time Spent: 1h 40m
> Remaining Estimate: 0h
>
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
