[SPARK-22165] Type conflicts between dates, timestamps and date in partition column

• Type: Bug
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 2.1.1, 2.2.0, 2.3.0
• Fix Version/s: 2.3.0
• Component/s: SQL
• Labels: release-notes
• Assignee: Hyukjin Kwon
• Reporter: Hyukjin Kwon
• Created: 29/Sep/17 05:27
• Updated: 12/Dec/22 18:10
• Resolved: 21/Nov/17 19:54

Description

It looks we have some bugs when resolving type conflicts in partition column. I found few corner cases as below:

Case 1: timestamp should be inferred but date type is inferred.

val df = Seq((1, "2015-01-01"), (2, "2016-01-01 00:00:00")).toDF("i", "ts")
df.write.format("parquet").partitionBy("ts").save("/tmp/foo")
spark.read.load("/tmp/foo").printSchema()

root
 |-- i: integer (nullable = true)
 |-- ts: date (nullable = true)

Case 2: decimal should be inferred but integer is inferred.

val df = Seq((1, "1"), (2, "1" * 30)).toDF("i", "decimal")
df.write.format("parquet").partitionBy("decimal").save("/tmp/bar")
spark.read.load("/tmp/bar").printSchema()

root
 |-- i: integer (nullable = true)
 |-- decimal: integer (nullable = true)

Looks we should de-duplicate type resolution logic if possible rather than separate numeric precedence-like comparison alone.

Issue Links

• links to: [Github] Pull Request #19389 (HyukjinKwon)