[
https://issues.apache.org/jira/browse/LUCENE-5824?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Robert Muir updated LUCENE-5824:
--------------------------------
Attachment: LUCENE-5824.patch
Simple patch and test to encode as A << 8 + B (and also check the values are
really within range: they should be two ascii characters).
This bug currently impacts the more complicated dictionaries using this
encoding type (russian, arabic, hebrew, etc)
> hunspell FLAG LONG implemented incorrectly
> ------------------------------------------
>
> Key: LUCENE-5824
> URL: https://issues.apache.org/jira/browse/LUCENE-5824
> Project: Lucene - Core
> Issue Type: Bug
> Reporter: Robert Muir
> Attachments: LUCENE-5824.patch
>
>
> If you have more than 256 flags, you run out of 8-bit characters, so you have
> to use another flag type to get 64k:
> * UTF-8: 16-bit BMP flags
> * long: two-character flags like 'AB'
> * num: decimal numbers like '10234'
> But our implementation for 'long' is wrong, it encodes as 'A+B', which means
> it cant distinguish between 'AB' and 'BA' and causes overgeneration.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
