Re: AnalyzingInfixSuggester questions

2014-08-15 Thread Michael Sokolov
On 8/14/2014 5:48 PM, Michael McCandless wrote: I think we should expose commit? Can you open an issue? I will And fix the NPEs to be a more informative exc? E.g. IllegalStateException w/ a reasonable message ... You should be able to init AIS on a previously built suggester, not .build()

Re: WhiteSpaceTokenizer

2014-08-15 Thread Jack Krupansky
Yeah, it should be documented better, and configurable. Some discussion of related issues here: https://issues.apache.org/jira/browse/LUCENE-1118 https://issues.apache.org/jira/browse/SOLR-4148 I actually filed a Jira for this already. No action so far, but PLEASE feel free to comment on it:

Re: WhiteSpaceTokenizer

2014-08-15 Thread Sheng
Thanks, Jack. I haven't added myself to the contributor list yet, will do that and then login and comment on that ticket. One quick comment: wouldn't it be more reasonable to throw exception it a token length is more than 255, if relaxing that limit is still debatable? This way user would know

Re: WhiteSpaceTokenizer

2014-08-15 Thread Jack Krupansky
Sure, that should be a configurable option. Oh, and I neglected to mention a workaround: use the pattern tokenizer, which doesn't have a limit (yet.) But it might be slower. -- Jack Krupansky -Original Message- From: Sheng Sent: Friday, August 15, 2014 8:13 AM To:

Re: AnalyzingInfixSuggester questions

2014-08-15 Thread Michael Sokolov
I opened LUCENE-5889. While I was writing that I remembered another gap: it would be useful to have a delete() method. Callers might like to be able to include or exclude terms based on their frequency, which can change over time, and at the very least when a term no longer occurs in the