Right, I'm loading models and caching them. Then each active use makes its own ME and detector/tokenizer/etc. The model object is just a passive bucket. It is essentially "final".
As to JIRAs: I'm used to much larger projects and JIRA pollution becomes a problem. It's not like this is Open Relevance :) On Sat, May 19, 2012 at 3:21 AM, Jim - FooBar(); <[email protected]> wrote: > On 19/05/12 09:34, Jörn Kottmann wrote: >> >> On 05/19/2012 05:52 AM, Lance Norskog wrote: >>> >>> Are the Model classes (subclasses of >>> opennlp.tools.util.model.BaseModel) guaranteed to be thread-safe? That >>> is, can I re-use the same SentenceModel in two different >>> SentenceDetectors and be sure there is no concurrency problems? >>> >>> If this is true, please add this to the BaseModel javadoc. >>> >> >> They are thread safe. You are welcome to open a jira issue >> and provide a patch to fix this. >> >> Jörn > > > Hi Jorn, > > If they are thread-safe, what is there to fix? Are you just referring to the > documentation? > > So, the models are thread-safe, but the detector,finder, tokenizer objects > are not? > > Jim > > -- Lance Norskog [email protected]
