[infinispan-issues] [JBoss JIRA] (ISPN-6845) A JDBC-based CacheLoader/CacheWriter that breaks down keys and values into multiple table columns

[Krzysztof Sobolewski (JIRA)]

    [ https://issues.jboss.org/browse/ISPN-6845?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13262577#comment-13262577 ] 

Krzysztof Sobolewski edited comment on ISPN-6845 at 7/7/16 7:30 AM:
--------------------------------------------------------------------

I have a prototype implementation of this, but it's a direct extension of the existing classes/interfaces and the result is not really merge-worthy. It does the followng things:

* The configuration of key/data name/type is changed from a single String to a List<String> so that we can declare multiple columns for the key and the value
* The configuration builder uses varargs so source compatibility (not binary compatibility) is preserved
* The Key2StringMapper interface is extended so that it can return multiple values (Objects, not String) for the key
* The Key2StringMapper is also applied to the values
* The TableManagers are modified to generate multi-column statements, which is a little more involved than a simple String.format() :)

Overall, it works, but is a little awkward (Key2StringMapper applied to values? Key2StringMapper that generates Object[]? The mapper for values takes an extra "parameter" that is crammed in front of the array ob objects extracted from the database. The boolean fetchValues and boolean fetchMetadata parameters are ignored. Etc).

The code is in https://github.com/ksobolew/infinispan/tree/ISPN-6845 if someone wishes to take a look.


was (Author: ksobolewski):
I have a prototype implementation of this, but it's a direct extension of the existing classes/interfaces and the result is not really merge-worthy. It does the followng things:

* The configuration of key/data name/type is changed from a single String to a List<String> so that we can declare multiple columns for the key and the value
* The configuration builder uses varargs so source compatibility (not binary compatibility) is preserved
* The Key2StringMapper interface is extended so that it can return multiple values (Objects, not String) for the key
* The Key2StringMapper is also applied to the values

Overall, it works, but is a little awkward (Key2StringMapper applied to values? Key2StringMapper that generates Object[]? The mapper for values takes an extra "parameter" that is crammed in front of the array ob objects extracted from the database. The boolean fetchValues and boolean fetchMetadata parameters are ignored. Etc).

The code is in https://github.com/ksobolew/infinispan/tree/ISPN-6845 if someone wishes to take a look.

> A JDBC-based CacheLoader/CacheWriter that breaks down keys and values into multiple table columns
> -------------------------------------------------------------------------------------------------
>
>                 Key: ISPN-6845
>                 URL: https://issues.jboss.org/browse/ISPN-6845
>             Project: Infinispan
>          Issue Type: Feature Request
>          Components: Loaders and Stores
>    Affects Versions: 9.0.0.Alpha2
>            Reporter: Krzysztof Sobolewski
>
> Even in the JBoss Cache days we (at my $DAYJOB) were not really happy with how the JDBC cache loader stores the data. Yes, marshalling the value into a BLOB is universal and portable, but nevertheless we created our own variant that increases visibility into the data by storing each attribute of the value in its own database table, in a human-readable format. It also very slightly improves performance, but that's not the primary goal. I think something like this would be useful in Infinispan.



--
This message was sent by Atlassian JIRA
(v6.4.11#64026)