[
https://issues.apache.org/jira/browse/SOLR-4452?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jan Høydahl updated SOLR-4452:
------------------------------
Attachment: SOLR-4452-caseinsensitivemerge.patch
This extra patch takes care of this corner case where two lines are same word
but different case.
For case sensitive dictionary we will treat these as two distinct words. But
for case insensitive dictionaries we'll merge them as before.
Sounds good? Or would it be better to treat these also as duplicates and use
only the last one? If so, we should modify the unit test instead.
> Hunspell stemmer should not merge duplicate dictionary entries
> --------------------------------------------------------------
>
> Key: SOLR-4452
> URL: https://issues.apache.org/jira/browse/SOLR-4452
> Project: Solr
> Issue Type: Bug
> Components: Schema and Analysis
> Reporter: Jan Høydahl
> Assignee: Jan Høydahl
> Labels: hunspell
> Fix For: 5.0, 4.4
>
> Attachments: SOLR-4452-caseinsensitivemerge.patch, SOLR-4452.patch,
> SOLR-4452.patch
>
>
> Hunspell dictionaries are on the form
> {noformat}
> lucene/ABC
> mahout/X
> {noformat}
> Each word is listed once with is base form, and the flags after the / define
> allowed prefixes and suffixes.
> In HunspellDictionary's parsing logic, if the same base word should appear
> multiple times in the file, the flags from the duplicate entry are added to
> the flags from the existing entry.
> However, HunSpellStemFilterFactory allows for a comma-separated list of
> dictionary files to be passed in, the idea being that you can have your own
> custom extensions and not need to modify the "standard" ones which may change
> upstream once in a while. This feature now works only for NEW words, not for
> overriding existing entries from the first dictionary.
> Would like to change this behavior, so that the last line read overwrites any
> previous one. This will both fix the custom dictionary issue and also fix
> unintentional wrong original dictionaries, where someone added a word
> definition at the end without realizing there was another already.
> For the en_UK.dic there are no duplicates. For en_US.dic there is one
> duplicate, so I argue this behavior is a bug and not a feature dictionary
> authors depend upon.
--
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]
