[jira] [Created] (CTAKES-411) Fix Unit tests with LVG (URI is not hierarchal in mvn package) test

["Pei Chen (JIRA)" <ji...@apache.org>]

Pei Chen created CTAKES-411:
-------------------------------

             Summary: Fix Unit tests with LVG (URI is not hierarchal in mvn package) test
                 Key: CTAKES-411
                 URL: https://issues.apache.org/jira/browse/CTAKES-411
             Project: cTAKES
          Issue Type: Improvement
          Components: ctakes-dependency-parser, ctakes-regression-test
    Affects Versions: 3.2.2, 3.1.1, 3.2.0
            Reporter: Pei Chen
            Priority: Minor
             Fix For: 3.2.3




---------- Forwarded message ----------
From: Pei Chen <pe...@wiredinformatics.com>
Date: Thu, Dec 15, 2016 at 6:30 PM
Subject: Re: Infrastructures questions.
To: dev@ctakes.apache.org


I can recreate what Tim suggested.
1) Comment out LVG from the dependency parser test/pipeline
2) The same thing would need to be done in the Regression Test (comment out
LVG in the test)
3) Update the regression test suite newly generated output -> expected

I'll commit the changes. Basically with the above, junit tests won't run
into the URI hierarchal issue when the "mvn package" command is issued.
A bit of background- I remember digging deeper into this some time ago:
Even if we fixed our impl/LVG wrapper, there are hard coded references
inside LVG itself that uses file:// instead of ResourceAsStream forcing
resources to be unpacked. This doesn't jive well during the package phase
of the maven build cycle where it references your .m2 repo.  mvn test works
fine because it will reference the explicit local unpacked lvg resources.

       Pei Chen
Wired Informatics <http://bit.ly/1pHmTcL>
265 Franklin St Ste 1702
Boston, MA 02110
tel: (617) 433-7544
Pei.Chen@wiredinformatics.com

On Wed, Dec 14, 2016 at 9:57 AM, Miller, Timothy <
Timothy.Miller@childrens.harvard.edu> wrote:

> Dependency tests pass with my change; new test error in regression test
> module that I'm not familiar with and error type I've never seen before
> -- reaching out for help debugging:
>
>
> > Exception in thread "BaseCPMImpl-Thread" junit.framework.AssertionFailedError:
> Verifying Test Output: testpatient_plaintext_2.txt.
> xmlorg.custommonkey.xmlunit.Diff
> > [different] Expected number of element attributes '7' but was '6' -
> comparing <org.apache.ctakes.typesystem.type.syntax.NewlineToken...> at
> /CAS[1]/org.apache.ctakes.typesystem.type.syntax.NewlineToken[1] to
> <org.apache.ctakes.typesystem.type.syntax.NewlineToken...> at
> /CAS[1]/org.apache.ctakes.typesystem.type.syntax.NewlineToken[1]
> >
> >       at junit.framework.Assert.fail(Assert.java:50)
> >       at junit.framework.Assert.assertTrue(Assert.java:20)
> >       at org.apache.ctakes.regression.test.RegressionPipelineTest.
> compareXMLOutput(RegressionPipelineTest.java:147)
> >       at org.apache.ctakes.regression.test.RegressionPipelineTest$
> StatusCallbackListenerImpl.collectionProcessComplete(
> RegressionPipelineTest.java:200)
> >       at org.apache.uima.collection.impl.cpm.BaseCPMImpl.run(
> BaseCPMImpl.java:538)
> >       at java.lang.Thread.run(Thread.java:745)
>
>
> Thanks
> Tim
>
> On Tue, 2016-12-13 at 16:15 +0000, Miller, Timothy wrote:
> > Quick followup - the test passes in eclipse, both with and without LVG
> > enabled. Can someone try to replicate at the command line and see if mvn
> > package works with LVG commented out? This is line 130 in
> > WriteClearNLPDescriptors.java. Otherwise I can try this afternoon.
> > Tim
> >
> > On Tue, 2016-12-13 at 15:57 +0000, Miller, Timothy wrote:
> > > Pretty sure this particular issue is caused by LVG being part of the
> > > test pipeline and the "URI is not hierarchical" bug from not having its
> > > files unpacked from the jar. A simple fix is to disable that test in
> > > code; a slightly more complex fix is to run the test with a modified
> > > pipeline that doesn't include LVG.
> > > Tim
> > >
> > >
> > > On Tue, 2016-12-13 at 10:51 -0500, Pei Chen wrote:
> > > > That's right.  mvn compile and test should work fine. The benign test
> > > > failed error from junit tests is coming from install/package; it's
> > > > been there since the beginning of time [1].  It would be a nice to
> > > > have and remove the benign warning messages.  If a proposed critical
> > > > patch release passes the regression tests, doesn't break any existing
> > > > behavior, enhances the project, and we have volunteers for RM, I do
> > > > not see these superious reasons as valid to block releases and keep
> > > > things moving along.
> > > > Sean:  it would great if you can open a Jira and apply the patch; we
> > > > can always cut another release next time- I'll be happy to be RM for
> > > > that one whenever you feel it's' ready.
> > > >



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)