[SPARK-33184] spark doesn't read data source column if it is used as an index to an array under a struct

• Type: Bug
• Status: Open
• Priority: Minor
• Resolution: Unresolved
• Affects Version/s: 3.0.0
• Component/s: SQL
• Assignee: Unassigned
• Reporter: colin fang
• Created: 19/Oct/20 18:06
• Updated: 20/Oct/20 04:40

Description

df = spark.createDataFrame([[1, [[1, 2]]]], schema='x:int,y:struct<a:array<int>>')
df.write.mode('overwrite').parquet('test')

# This causes an error "Caused by: java.lang.RuntimeException: Couldn't find x#720 in [y#721]"
spark.read.parquet('test').select(F.expr('y.a[x]')).show()

# Explain works fine, note it doesn't read x in ReadSchema
spark.read.parquet('test').select(F.expr('y.a[x]')).explain()

== Physical Plan ==
*(1) !Project [y#713.a[x#712] AS y.a AS `a`[x]#717]
+- FileScan parquet [y#713] Batched: false, DataFilters: [], Format: Parquet, Location: InMemoryFileIndex, PartitionFilters: [], PushedFilters: [], ReadSchema: struct<y:struct<a:array<int>>>

The code works well if I

# manually select the column it misses
spark.read.parquet('test').select(F.expr('y.a[x]'), F.col('x')).show()

# use element_at function
spark.read.parquet('test').select(F.element_at('y.a', F.col('x') + 1)).show()