[uima-uimaj] 01/01: [UIMA-6295] CAS transportable Java object not serialised or deserialised with compressed binary

[re...@apache.org]

This is an automated email from the ASF dual-hosted git repository.

rec pushed a commit to branch UIMA-6295-CAS-transportable-Java-object-not-serialised-or-deserialised-with-compressed-binary
in repository https://gitbox.apache.org/repos/asf/uima-uimaj.git

commit 15733de9437184deb7c00067bfd3d9b7b1b8c3cf
Author: Richard Eckart de Castilho <re...@apache.org>
AuthorDate: Fri Dec 4 08:22:24 2020 +0100

    [UIMA-6295] CAS transportable Java object not serialised or deserialised with compressed binary
    
    - Added a set tracking whether _save_to_cas_data() has already been called on an FS during the serialization process to avoid _save_to_cas_data() being called twice and potentially creating new FSes in that way that are then not discovered and properly linked up by the serialization process anymore.
---
 .../org/apache/uima/cas/impl/BinaryCasSerDes6.java | 27 ++++++++++++++--------
 1 file changed, 18 insertions(+), 9 deletions(-)

diff --git a/uimaj-core/src/main/java/org/apache/uima/cas/impl/BinaryCasSerDes6.java b/uimaj-core/src/main/java/org/apache/uima/cas/impl/BinaryCasSerDes6.java
index b5d7b39..00ebcf9 100644
--- a/uimaj-core/src/main/java/org/apache/uima/cas/impl/BinaryCasSerDes6.java
+++ b/uimaj-core/src/main/java/org/apache/uima/cas/impl/BinaryCasSerDes6.java
@@ -63,6 +63,7 @@ import org.apache.uima.internal.util.IntListIterator;
 import org.apache.uima.internal.util.IntVector;
 import org.apache.uima.internal.util.Misc;
 import org.apache.uima.internal.util.PositiveIntSet;
+import org.apache.uima.internal.util.PositiveIntSet_impl;
 import org.apache.uima.jcas.JCas;
 import org.apache.uima.jcas.cas.BooleanArray;
 import org.apache.uima.jcas.cas.ByteArray;
@@ -386,7 +387,12 @@ public class BinaryCasSerDes6 implements SlotKindsConstants {
    * FSs being serialized. For delta, just the deltas above the delta line.
    * Constructed from indexed plus reachable, above the delta line.
    */
-  private List<TOP> fssToSerialize; 
+  private List<TOP> fssToSerialize;
+  
+  /**
+   * Set of FSes on which UimaSerializable _save_to_cas_data has already been called.
+   */
+  private PositiveIntSet uimaSerializableSavedToCas;
   
   /**
    * FSs being processed, including below-the-line deltas.
@@ -735,6 +741,8 @@ public class BinaryCasSerDes6 implements SlotKindsConstants {
       
       os = new OptimizeStrings(doMeasurements);
    
+      uimaSerializableSavedToCas = new PositiveIntSet_impl(1024, 1, 1024);
+      
       /******************************************************************
        * Find all FSs to be serialized via the indexes
        *   including those FSs referenced  
@@ -1491,8 +1499,9 @@ public class BinaryCasSerDes6 implements SlotKindsConstants {
             }
           }
         } else {
-          if (fs instanceof UimaSerializable) {
+          if (fs instanceof UimaSerializable && !uimaSerializableSavedToCas.contains(fs._id)) {
             ((UimaSerializable)fs)._save_to_cas_data();
+            uimaSerializableSavedToCas.add(fs._id);
           }
           final BitSet featuresModified = changedFs.featuresModified;
           int next = featuresModified.nextSetBit(0);
@@ -2838,7 +2847,9 @@ public class BinaryCasSerDes6 implements SlotKindsConstants {
                             isTypeMapping ? fs -> isTypeInTgt(fs) : null, 
                             isTypeMapping ? typeMapper            : null)
                   .getAllFSsAllViews_sofas_reachable();
-            ;
+        // AllFSs internally already causes _save_to_cas_data() to be called, so we have to add all
+        // the FSes that are returned here to the uimaSerializableSavedToCas tracking set
+        allFSs.getAllFSs().forEach(fs -> uimaSerializableSavedToCas.add(fs._id));
       }
       fssToSerialize = CASImpl.filterAboveMark(allFSs.getAllFSsSorted(), mark);
       foundFSs = allFSs.getAllNew();
@@ -3037,12 +3048,10 @@ public class BinaryCasSerDes6 implements SlotKindsConstants {
       fsStartIndexes.addItemId(fs, nextTgtId, isIncludedType);  // maps src heap to tgt seq
       
       if (isIncludedType) {
-        // Disabled the following additional call to _save_to_cas_data, since it would cause a bug. See
-        // [UIMA-6295] CAS transportable Java object not serialised or deserialised with compressed binary.
-        //
-        // if (fs instanceof UimaSerializable) {
-        //     ((UimaSerializable)fs)._save_to_cas_data();
-        // }
+        if (fs instanceof UimaSerializable && !uimaSerializableSavedToCas.contains(fs._id)) {
+          ((UimaSerializable)fs)._save_to_cas_data();
+          uimaSerializableSavedToCas.add(fs._id);
+        }
 
         // for features in type - 
         //    strings: accumulate those strings that are in the target, if optimizeStrings != null