Uploaded image for project: 'Terrier Core'
  1. Terrier Core
  2. TR-476

Java.lang.RuntimeException: trec_eval ended with non-zero exit code (35584)

    Details

      Description

      I have a runtime error whenever i use "bin\trec_terrier -e" to run my qrels.

      Java.lang.RuntimeException: trec_eval ended with non-zero exit code (35584)

      RunttimeException: trec_eval ended with non-zero exit code (35584)
      uk.ac.gla.terrier.jtreceval.trec_eval.runAndGetOutput(trec_eval.java:151)
      org.terrier.evaluation.TrecEvalEvaluation.evaluate(TrecEvalEvaluation.java:40)
      org.terrier.applications.TrecTerrier.run(TrecTerrier.java:527)
      org.terrier.applications.TrecTerrier.applyOptions(TrecTerrier.java:584)
      org.terrier.applications.TrecTerrier.main(TrecTerrier.java:229)

        Attachments

          Activity

          Hide
          krew dary added a comment -

          It doesn't matter now. I got it to work fine.

          Show
          krew dary added a comment - It doesn't matter now. I got it to work fine.
          Hide
          craigm Craig Macdonald added a comment -

          I think my point was to understand what the bug was. I think simply there was an indexing fault.

          But the trec_eval error message should have been visible. I will generate a test case for this, and adjust the error message if so.

          Craig

          Show
          craigm Craig Macdonald added a comment - I think my point was to understand what the bug was. I think simply there was an indexing fault. But the trec_eval error message should have been visible. I will generate a test case for this, and adjust the error message if so. Craig
          Hide
          gpapadop79 Georgios Papadopoulos added a comment -

          I confirm that this problem is due to duplicate documents at indexing phase, but the error message is cryptic.
          So due to a bug in a dataset there was a file named EP-0400071.txt that had a duplicate of the data, so it got indexed twice.
          It was like this:
          <DOC>
          <DOCNO>EP-0400021</DOCNO>
          <TEXT>
          ......
          </TEXT>
          </DOC>
          <DOC>
          <DOCNO>EP-0400021</DOCNO>
          <TEXT>
          ......
          </TEXT>
          </DOC>

          So in the result file appeared:
          EP-1225285-A2 Q0 EP-0400021 607 9.447041727144423 TF_IDF
          EP-1225285-A2 Q0 EP-0400021 608 9.447041727144423 TF_IDF

          Maybe a warning about duplicate docno's during indexing would be useful

          Show
          gpapadop79 Georgios Papadopoulos added a comment - I confirm that this problem is due to duplicate documents at indexing phase, but the error message is cryptic. So due to a bug in a dataset there was a file named EP-0400071.txt that had a duplicate of the data, so it got indexed twice. It was like this: <DOC> <DOCNO>EP-0400021</DOCNO> <TEXT> ...... </TEXT> </DOC> <DOC> <DOCNO>EP-0400021</DOCNO> <TEXT> ...... </TEXT> </DOC> So in the result file appeared: EP-1225285-A2 Q0 EP-0400021 607 9.447041727144423 TF_IDF EP-1225285-A2 Q0 EP-0400021 608 9.447041727144423 TF_IDF Maybe a warning about duplicate docno's during indexing would be useful
          Hide
          craigm Craig Macdonald added a comment -

          I will try to confirm this with a unit test. We could adjust TRECDocnoOutputFormat to report duplicates, but that would be a cost for every run going forward. I think the key thing here is why does jtreceval "swallow" the message from trec_eval instead of printing it to screen.

          Craig

          Show
          craigm Craig Macdonald added a comment - I will try to confirm this with a unit test. We could adjust TRECDocnoOutputFormat to report duplicates, but that would be a cost for every run going forward. I think the key thing here is why does jtreceval "swallow" the message from trec_eval instead of printing it to screen. Craig
          Hide
          craigm Craig Macdonald added a comment -

          New jtreceval version shipped with 5.0 and 4.4 no longer swallow errors

          Show
          craigm Craig Macdonald added a comment - New jtreceval version shipped with 5.0 and 4.4 no longer swallow errors

            People

            • Assignee:
              craigm Craig Macdonald
              Reporter:
              krew dary
            • Watchers:
              3 Start watching this issue

              Dates

              • Created:
                Updated:
                Resolved: