Marshall Schor wrote: > > Thilo Goetz wrote: >> It's probably not important, but I don't understand what the >> difference is between "creating a top level pearSpecifier kind >> of AE" (your words) and "create AEs ... from top-level pear >> descriptors" (my words). >> > > Hmmm, if we're doing the same thing, then it should have been failing > for your situations. Do you have a sample / test case or whatever I > could try? I'll back out the change and test why it *isn't* failing. > That should shed some light on my understanding in this area :-)
How about the whitespace tokenizer and the regex annotator from the sandbox, they come pre-packaged as pears and you can run them directly, they have no dependencies. I did run some regression tests after your changes (with proprietary pear files) and everything has worked so far. --Thilo > > -Marshall >> Marshall Schor wrote: >> >>> The thing that was failing was creating a top level pearSpecifier kind >>> of AE... regular AE's worked just fine ;-) . >>> >>> -Marshall >>> >>> Thilo Goetz (JIRA) wrote: >>> >>>> [ >>>> https://issues.apache.org/jira/browse/UIMA-1465?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12736669#action_12736669 >>>> ] >>>> >>>> Thilo Goetz commented on UIMA-1465: >>>> ----------------------------------- >>>> >>>> I create AEs like that all the time, from top-level pear descriptors. >>>> It's what the CVD does, for example (unless the datapath is set). This >>>> most definitely used to work. >>>> >>>> >>>> >>>> >>>>> running a pearSpecifier as a top level component fails with >>>>> NullPointerException >>>>> -------------------------------------------------------------------------------- >>>>> >>>>> Key: UIMA-1465 >>>>> URL: https://issues.apache.org/jira/browse/UIMA-1465 >>>>> Project: UIMA >>>>> Issue Type: Bug >>>>> Components: Core Java Framework >>>>> Affects Versions: 2.2.2 >>>>> Reporter: Marshall Schor >>>>> Assignee: Marshall Schor >>>>> Fix For: 2.3 >>>>> >>>>> >>>>> This happens when produceAnalysisEngine call is done with no additional >>>>> parameters, for top level pears. Our previous testing has always had the >>>>> pear contained in an aggregrate wrapper, which I think works. >>>>> >>>>> >>>> >>>> >> >>