[ 
https://issues.apache.org/jira/browse/LUCENE-2653?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12912227#action_12912227
 ] 

Robert Muir commented on LUCENE-2653:
-------------------------------------

bq. Could that have been a bw break since it did not do what it claimed to do?

I dont understand the question. ThaiWordFilter has always been broken this way, 
it is broken by design.

> ThaiAnalyzer assumes things about your jre
> ------------------------------------------
>
>                 Key: LUCENE-2653
>                 URL: https://issues.apache.org/jira/browse/LUCENE-2653
>             Project: Lucene - Java
>          Issue Type: Bug
>          Components: contrib/analyzers
>    Affects Versions: 3.1, 4.0
>            Reporter: Robert Muir
>         Attachments: LUCENE-2653.patch
>
>
> The ThaiAnalyzer/ThaiWordFilter depends on the fact that 
> BreakIterator.getWordInstance(new Locale("th")) returns a dictionary-based 
> break iterator that can segment thai phrases into words (it does not use 
> whitespace).
> But this is non-standard that the JRE will specialize this locale in this 
> way, its nice, but you can't depend on it.
> For example, if you are running on IBM JRE, this analyzer/wordfilter is 
> completely "broken" in the sense it won't do what it claims to do.
> At the minimum, we need to document this and suggest users look at 
> ICUTokenizer for thai, which always has this breakiterator and is not 
> jre-dependent.
> Better, would be to check statically that the thing actually works.
> when creating a new ThaiWordFilter we could clone() the BreakIterator, which 
> is often cheaper than making a new one anyway.
> we could throw an exception, if its not supported, and add a boolean so the 
> user knows it works.
> and we could refer to this boolean with Assert.assume in its tests.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to