Closed zml1206 closed 1 month ago
It cannot be reproduced without union or filter. I didn't find out what the problem is, can you help take a look @zhztheplayer @rui-mo thank you.
@yma11 Any insights?
spark.read.parquet("tmp/t1") .union(spark.read.parquet("tmp/t1")) .filter("a.aA > 3") .collect()
will it work if you use a.aa
?
spark.read.parquet("tmp/t1")
.union(spark.read.parquet("tmp/t1"))
.filter("a.aa > 3")
.collect()
I think the column name becomes lower case after some operation so it can't be matched any more.
will it work if you use
a.aa
?
Yes, it worked with a.aa
.
Fallback scan or remove union, it also worked.
Then it should be problem of native parquet scan together with union, do you prefer to fix it by yourself?
Backend
VL (Velox)
Bug description
Found in read delta lake with checkpoint in version 3.0.0rc1. Simplify recurrence logic:
Spark version
None
Spark configurations
No response
System information
No response
Relevant logs