[jira] [Updated] (OAK-12340) Simplified index management improvements

["Thomas Mueller (Jira)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/OAK-12340?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Thomas Mueller updated OAK-12340:
---------------------------------
    Description: 
There are various changes and improvements needed for simplified index management:

(a)
Prevent that migrating to simplified index management disables an index. 
If indexes were configured previously with version numbers (abc-1-custom-1), but now are migrated to simplified index management, then and then migrated to simplified, there were cases where the old and new version exactly match, and so no new index was created. Because the old one was then no longer active, in reality migration caused a gap (the index definition was still there, but the index is not active). For this, we need to ignore disabled indexes.

(b)
Support for empty nodes. Currently, empty nodes (without any property) will get "type": "disabled". There is a workaround: set a primary type. But this is ugly.

(c)
There was some code that in theory could result in null-pointer exceptions in theory. It makes sense to protect against this.

  was:
There are various changes and improvements needed for simplified index management:

(a)
Prevent that migrating to simplified index management disables an index. 
If indexes were configured previously with version numbers (abc-1-custom-1), but now are migrated to simplified index management, then and then migrated to simplified, there were cases where the old and new version exactly match, and so no new index was created. Because the old one was then no longer active, in reality migration caused a gap (the index definition was still there, but the index is not active). For this, we need to ignore disabled indexes.

(b)
A small change for unversioned indexes: if the old index name was "abc" (without version number), then we should probably create an index named "abc-1-custom-1" instead of just "abc-custom-1", for clarity.

(c)
Support for empty nodes. Currently, empty nodes (without any property) will get "type": "disabled". There is a workaround: set a primary type. But this is ugly.

(d)
There was some code that in theory could result in null-pointer exceptions in theory. It makes sense to protect against this.


> Simplified index management improvements
> ----------------------------------------
>
>                 Key: OAK-12340
>                 URL: https://issues.apache.org/jira/browse/OAK-12340
>             Project: Jackrabbit Oak
>          Issue Type: Improvement
>            Reporter: Thomas Mueller
>            Priority: Major
>
> There are various changes and improvements needed for simplified index management:
> (a)
> Prevent that migrating to simplified index management disables an index. 
> If indexes were configured previously with version numbers (abc-1-custom-1), but now are migrated to simplified index management, then and then migrated to simplified, there were cases where the old and new version exactly match, and so no new index was created. Because the old one was then no longer active, in reality migration caused a gap (the index definition was still there, but the index is not active). For this, we need to ignore disabled indexes.
> (b)
> Support for empty nodes. Currently, empty nodes (without any property) will get "type": "disabled". There is a workaround: set a primary type. But this is ugly.
> (c)
> There was some code that in theory could result in null-pointer exceptions in theory. It makes sense to protect against this.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)