openinx commented on a change in pull request #2873:
URL: https://github.com/apache/iceberg/pull/2873#discussion_r677913738



##########
File path: core/src/main/java/org/apache/iceberg/io/WriterFactory.java
##########
@@ -0,0 +1,62 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *   http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied.  See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+
+package org.apache.iceberg.io;
+
+import org.apache.iceberg.PartitionSpec;
+import org.apache.iceberg.StructLike;
+import org.apache.iceberg.deletes.EqualityDeleteWriter;
+import org.apache.iceberg.deletes.PositionDeleteWriter;
+import org.apache.iceberg.encryption.EncryptedOutputFile;
+
+/**
+ * A factory for creating data and delete writers.
+ */
+public interface WriterFactory<T> {

Review comment:
       > That's why I don't accept FileFormat as an argument for methods 
constructing writes.
   
   Yes,  we don't need the `FileFormat` to construct the data/delete file 
writers.
   
   > Is there a use case when a single txn will write both Avro and Parquet 
data files?
   
   For the current compute-engine + iceberg integration work,  I don't think we 
support writing both avro and parquet data files in the same txn.  If people 
really want to do this,  they will need to generate their own avro and parquet 
files, and finally commit those files by using Iceberg Table API. I think 
that's the correct practice.
   
   > We can use different file formats for data and delete files.
   
   Yes, that's really helpful.   I remember that the JOIN process between data 
files and positional delete files is a MERGE-SORT JOIN  because the both data 
files and delete files have sorted joined keys : <file_name, row_offset>.   If 
we have `N` positional delete files in the JOIN progress,  then we will 
allocate at least `N*128MB` memory for each row group (To read the whole row) 
from the positional delete files (If we use parquet),  that's quite huge memory 
consuming, which will lead to OutOfMemory in the real production.  In this 
case, the idea  file format for positional delete files is `AVRO` I think.




-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]



---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to