[
https://issues.apache.org/jira/browse/LUCENE-4536?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13491479#comment-13491479
]
Michael McCandless commented on LUCENE-4536:
--------------------------------------------
bq. These were my first ideas, but the truth is that I was very scared to break
something (for example doc values rely on the assumption that after reading the
last value of a direct array, the whole stream is consumed).
It's hard to know what's best :)
I like the explicitness / transparency / no sneaky code solution of
.skipTrailingBytes().
But then I don't like that skipTrailingBytes would only be for back
compat (ie, we will remove it eventually, unless somehow we go back to
wasted trailing bytes) ... annoying to add essentially a deprecated
API.
But then really it's "presumptuous" of the consumers of PackedInts to
expect all bytes are consumed after iterating all values ... like
that's making a sometimes invalid assumption about the file format of
PackedInts.
And this is an internal API so we are free to change things ..
But net/net I think we should stick w/ your current patch?
> Make PackedInts byte-aligned?
> -----------------------------
>
> Key: LUCENE-4536
> URL: https://issues.apache.org/jira/browse/LUCENE-4536
> Project: Lucene - Core
> Issue Type: Task
> Reporter: Adrien Grand
> Assignee: Adrien Grand
> Priority: Minor
> Fix For: 4.1
>
> Attachments: LUCENE-4536.patch
>
>
> PackedInts are more and more used to save/restore small arrays, but given
> that they are long-aligned, up to 63 bits are wasted per array. We should try
> to make PackedInts storage byte-aligned so that only 7 bits are wasted in the
> worst case.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
