LuciferYang edited a comment on pull request #30484:
URL: https://github.com/apache/spark/pull/30484#issuecomment-749542175


   The branch to be deleted in current pr is from SPARK-11787,.
   
   In SPARK-11787`SqlNewHadoopPartition` create `UnsafeRowParquetRecordReader` 
when `spark.parquet.enableUnsafeRowRecordReader` is true and data format is 
`org.apache.parquet.hadoop.ParquetInputFormat` 
   
   The key point is  the `ParquetInputSplit` pass to 
`UnsafeRowParquetRecordReader.init` method, if `rowGroupOffsets` in 
`ParquetInputSplit` is not null will enter the code branch which I want to 
delete in this pr.
   
   `ParquetInputSplit` produce by `SqlNewHadoopRDD#getPartitions` use 
`ParquetInputFormat.getSplits` method.
   
   
   
https://github.com/apache/parquet-mr/blob/32c46643845ea8a705c35d4ec8fc654cc8ff816d/parquet-hadoop/src/main/java/org/apache/parquet/hadoop/ParquetInputFormat.java#L287-L306
   
   
   The `else branch` will create a `ParquetInputSplit` and `rowGroupOffsets` 
not null, further calls are as follows:
   
   ```
   ParquetInputFormat#getSplits(JobContext) 
     -> ParquetInputFormat#getSplits(Configuration,List<Footer>)
       -> ClientSideMetadataSplitStrategy#getSplits
         -> ClientSideMetadataSplitStrategy#generateSplits
           -> ClientSideMetadataSplitStrategy.SplitInfo#getParquetInputSplit
   ```
   
https://github.com/apache/parquet-mr/blob/32c46643845ea8a705c35d4ec8fc654cc8ff816d/parquet-hadoop/src/main/java/org/apache/parquet/hadoop/ParquetInputFormat.java#L635-L649
   


----------------------------------------------------------------
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.

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