[
https://issues.apache.org/jira/browse/LUCENE-10143?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17423890#comment-17423890
]
Robert Muir commented on LUCENE-10143:
--------------------------------------
I really think we need to devise a different strategy on this issue that
doesn't involve rewriting all of o.a.l.store (which is dangerous!).
Things I feel should be not be tackled here at all, instead separate JIRA
issues:
* positional behavior on exceptions: this is unrelated to this change. if you
want to look at what it takes to make things consistent, start with
{{readBytes}} which will be different for e.g. NIOFSDirectory vs MMAPdirectory
(because of how they read byte array in chunks). So it really does not need to
be tied into this issue. personal opinion: we will just make things slow trying
to add such guarantees and nothing needs these guarantees.
* major changes to implementing subclasses. If we have to do anything more than
drop in an obvious 1/2/3-line change, it should really be left as a TODO and
dealt with as a separate issue. Otherwise we run the risk of introducing bugs.
Perhaps, we should be even more pessimistic and make a "rote" change here that
just makes methods abstract and then adds the 3 methods (falling back to
existing slow code) to every subclass, and defer ALL changes to separate issues.
There's a long history of changes to o.a.l.store causing horrible bugs. it
would be really bad to do something like this the wrong way right before a
release.
> RateLimitedIndexOutput should delegate writeShort/writeInt/writeLong
> --------------------------------------------------------------------
>
> Key: LUCENE-10143
> URL: https://issues.apache.org/jira/browse/LUCENE-10143
> Project: Lucene - Core
> Issue Type: Improvement
> Reporter: Adrien Grand
> Priority: Minor
> Time Spent: 3h
> Remaining Estimate: 0h
>
> Otherwise merges are not taking advantage of LUCENE-10125.
> cc [~uschindler]
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
