[SPARK-21661] SparkSQL can't merge load table from Hadoop

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 2.2.0
• Fix Version/s: 2.3.0
• Component/s: SQL
• Assignee: Yuanjian Li
• Reporter: Dapeng Sun
• Created: 08/Aug/17 03:57
• Updated: 12/Dec/22 18:11
• Resolved: 11/Oct/17 14:34

Description

Here is the original text of external table on HDFS：

Permission	Owner	Group	Size	Last Modified	Replication	Block Size	Name
-rw-r--r--	root	supergroup	0 B	8/6/2017, 11:43:03 PM	3	256 MB	income_band_001.dat
-rw-r--r--	root	supergroup	0 B	8/6/2017, 11:39:31 PM	3	256 MB	income_band_002.dat
...
-rw-r--r--	root	supergroup	327 B	8/6/2017, 11:44:47 PM	3	256 MB	income_band_530.dat

After SparkSQL load, every files have a output file, even the files are 0B. For the load on Hive, the data files would be merged according the data size of original files.

Reproduce:

CREATE EXTERNAL TABLE t1 (a int,b string)  STORED AS TEXTFILE LOCATION "hdfs://xxx:9000/data/t1"
CREATE TABLE t2 STORED AS PARQUET AS SELECT * FROM t1;

The table t2 have many small files without data.

Issue Links

• links to: [Github] Pull Request #18654