OPENNLP-1369 NPE when serializing a TokenNameFinder model trained with POSTaggerNameFeatureGeneratorFactory

- adds first reproducer to check if 1369 is broken => yes it is
- confirms basic workaround is curing the issue, however this is not pretty.

Author: mawiesne

opennlp-tools/src/main/java/opennlp/tools/util/model/BaseModel.java

@@ -596,6 +596,11 @@ public final void serialize(OutputStream out) throws IOException {
       zip.putNextEntry(new ZipEntry(name));
 
       Object artifact = entry.getValue();
+      // TODO Discuss if this is the correct location to have this workaround in place
+      if ("generator.featuregen".equals(name) && artifact == null) {
+        // An old model format was detected, skipping the process for this entry, see: OPENNLP-1369
+        continue;
+      }
 
       ArtifactSerializer serializer = getArtifactSerializer(name);
 

opennlp-tools/src/test/java/opennlp/tools/namefind/TokenNameFinderModelTest.java

@@ -51,13 +51,17 @@ void testNERWithPOSModel() throws IOException {
     // create a resources folder
     Path resourcesFolder = Files.createTempDirectory("resources").toAbsolutePath();
 
+    // TODO Restore old test and provide a separate one which does it with the pt-pos-perceptron.bin
     // save a POS model there
     POSModel posModel = POSTaggerMETest.trainPOSModel(ModelType.MAXENT);
-    File posModelFile = new File(resourcesFolder.toFile(), "pos-model.bin");
+    File posModelFile = new File("pt-pos-perceptron.bin");
+    Files.copy(posModelFile.toPath(), resourcesFolder.resolve("pt-pos-perceptron.bin"));
 
-    posModel.serialize(posModelFile);
+    // posModel.serialize(posModelFile);
 
     Assertions.assertTrue(posModelFile.exists());
+    Assertions.assertTrue(resourcesFolder.resolve("pt-pos-perceptron.bin").toFile().exists());
+    // end TODO
 
     // load feature generator xml bytes
     InputStream fgInputStream = this.getClass().getResourceAsStream("ner-pos-features.xml");

opennlp-tools/src/test/resources/opennlp/tools/namefind/ner-pos-features.xml

@@ -30,7 +30,8 @@
     <int name="prevLength">2</int>
     <int name="nextLength">2</int>
     <generator class="opennlp.tools.util.featuregen.POSTaggerNameFeatureGeneratorFactory">
-      <str name="model">pos-model.bin</str>
+      <!-- FIXME Restore old state for TokenNameFinderModelTest -->
+      <str name="model">pt-pos-perceptron.bin</str>
     </generator>
   </generator>
   <generator class="opennlp.tools.util.featuregen.PreviousMapFeatureGeneratorFactory"/>