[ 
https://issues.apache.org/jira/browse/PARQUET-2292?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17715556#comment-17715556
 ] 

ASF GitHub Bot commented on PARQUET-2292:
-----------------------------------------

wgtmac commented on code in PR #1078:
URL: https://github.com/apache/parquet-mr/pull/1078#discussion_r1174720814


##########
parquet-avro/src/main/java/org/apache/parquet/avro/AvroWriteSupport.java:
##########
@@ -400,9 +400,18 @@ private Binary fromAvroString(Object value) {
     return Binary.fromCharSequence(value.toString());
   }
 
-  private static GenericData getDataModel(Configuration conf) {
+  private static GenericData getDataModel(Configuration conf, Schema schema) {
+    if (conf.get(AVRO_DATA_SUPPLIER) == null) {
+      final GenericData modelForSchema = 
AvroRecordConverter.getModelForSchema(schema);

Review Comment:
   What if `schema == null`?



##########
parquet-avro/src/main/java/org/apache/parquet/avro/AvroRecordConverter.java:
##########
@@ -169,6 +172,46 @@ public void add(Object value) {
     }
   }
 
+  /**
+   * Returns the specific data model for a given SpecificRecord schema by 
reflecting the underlying
+   * Avro class's `MODEL$` field, or Null if the class is not on the classpath 
or reflection fails.
+   */
+  static SpecificData getModelForSchema(Schema schema) {
+    final Class<?> clazz;
+
+    if (schema != null && (schema.getType() == Schema.Type.RECORD || 
schema.getType() == Schema.Type.UNION)) {
+      clazz = SpecificData.get().getClass(schema);
+    } else {
+      return null;
+    }
+
+    final SpecificData model;
+    try {
+      final Field modelField = clazz.getDeclaredField("MODEL$");
+      modelField.setAccessible(true);
+
+      model = (SpecificData) modelField.get(null);

Review Comment:
   Should it return if model is set successfully?



##########
parquet-avro/src/main/java/org/apache/parquet/avro/AvroReadSupport.java:
##########
@@ -149,12 +149,22 @@ private static <T> RecordMaterializer<T> 
newCompatMaterializer(
         parquetSchema, avroSchema, model);
   }
 
-  private GenericData getDataModel(Configuration conf) {
+  private GenericData getDataModel(Configuration conf, Schema schema) {
     if (model != null) {
       return model;
     }
+
+    if (conf.get(AVRO_DATA_SUPPLIER) == null) {
+      final GenericData modelForSchema = 
AvroRecordConverter.getModelForSchema(schema);
+
+      if (modelForSchema != null) {
+        return modelForSchema;
+      }
+    }
+
     Class<? extends AvroDataSupplier> suppClass = conf.getClass(
-        AVRO_DATA_SUPPLIER, SpecificDataSupplier.class, 
AvroDataSupplier.class);
-    return ReflectionUtils.newInstance(suppClass, conf).get();
+      AVRO_DATA_SUPPLIER, SpecificDataSupplier.class, AvroDataSupplier.class);
+
+      return ReflectionUtils.newInstance(suppClass, conf).get();

Review Comment:
   Could you please revert these format changes?



##########
parquet-avro/src/main/java/org/apache/parquet/avro/AvroRecordConverter.java:
##########
@@ -169,6 +172,46 @@ public void add(Object value) {
     }
   }
 
+  /**
+   * Returns the specific data model for a given SpecificRecord schema by 
reflecting the underlying
+   * Avro class's `MODEL$` field, or Null if the class is not on the classpath 
or reflection fails.
+   */
+  static SpecificData getModelForSchema(Schema schema) {

Review Comment:
   @gszadovszky Could you help review this? I am not that familiar with Avro.





> Improve default SpecificRecord model selection for Avro{Write,Read}Support
> --------------------------------------------------------------------------
>
>                 Key: PARQUET-2292
>                 URL: https://issues.apache.org/jira/browse/PARQUET-2292
>             Project: Parquet
>          Issue Type: Improvement
>            Reporter: Claire McGinty
>            Assignee: Claire McGinty
>            Priority: Major
>
> AvroWriteSupport/AvroReadSupport can improve the precision of their default 
> `model` selection. Currently they default to new 
> SpecificDataSupplier().get()[0]. This means that SpecificRecord classes that 
> contain logical types will fail out-of-the-box unless a specific 
> DATA_SUPPLIER is configured that contains logical type conversions.
> I think we can improve this and make logical types work by default by 
> defaulting to the value of the `MODEL$` field that every SpecificRecordBase 
> implementation contains, which already contains all the logical conversions 
> for that Avro type. It would require reflection, but that's what the Avro 
> library is already doing to fetch models for Specific types[1].
>  
> [0] 
> [https://github.com/apache/parquet-mr/blob/d38044f5395494e1543581a4b763f624305d3022/parquet-avro/src/main/java/org/apache/parquet/avro/AvroWriteSupport.java#L403-L407]
> [1] 
> https://github.com/apache/avro/blob/release-1.11.1/lang/java/avro/src/main/java/org/apache/avro/specific/SpecificData.java#L76-L86



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to