Issues

perf: coalesce small shuffle-read batches before native sort

requires-triage

#5686

· peterxcli opened

on Sep 4, 2026

Comet's explode is slower than Spark when the element is a struct, worst on nested rows

#5685

· andygrove opened

on Sep 4, 2026

Investigate: why Comet shuffle files can be larger than Spark's, and whether byte-based block sizing would fix it

enhancement

requires-triage

#5678

· andygrove opened

on Sep 4, 2026

Plain `posexplode` evaluates its array argument twice per batch

#5676

· andygrove opened

on Sep 4, 2026

concat_ws with array<string> arguments fails natively instead of flattening like Spark

requires-triage

#5675

· peterxcli opened

on Sep 4, 2026

String-to-timestamp cast diverges from Spark's parseTimestampString: rejects 1-digit segments and empty fractions, accepts zone suffixes on date-only strings and 7-digit years

requires-triage

#5674

· peterxcli opened

on Sep 4, 2026

ANSI cast of double/float to int/long falsely reports CAST_OVERFLOW for exactly-representable boundary values (2147483647.0, -2147483648.0, ±2^63)

requires-triage

#5673

· peterxcli opened

on Sep 4, 2026

rpad/lpad panic (Option::unwrap on None) when the length column contains NULL

requires-triage

#5672

· peterxcli opened

on Sep 4, 2026

Nested (struct/list/map) Parquet schema-evolution conversions bypass Spark's type-conversion rules: silent NULLs, string parsing, and a native panic

requires-triage

#5671

· peterxcli opened

on Sep 4, 2026

Decimal to double/float cast is not correctly rounded: last-digit-wrong results for ordinary DECIMAL(38,18) values

requires-triage

#5670

· peterxcli opened

on Sep 4, 2026

JVM exit hangs when the application returns from main without SparkContext.stop()

bug

requires-triage

#5664

· zhangfengcdt opened

on Sep 3, 2026

Regular native scan forwards S3 options without checking divergence from Hadoop's resolution

requires-triage

#5662

· dwsmith1983 opened

on Sep 3, 2026

Navigation Menu

Uh oh!

All issues

perf: coalesce small shuffle-read batches before native sort

Comet's explode is slower than Spark when the element is a struct, worst on nested rows

Investigate: why Comet shuffle files can be larger than Spark's, and whether byte-based block sizing would fix it

Plain `posexplode` evaluates its array argument twice per batch

concat_ws with array<string> arguments fails natively instead of flattening like Spark

String-to-timestamp cast diverges from Spark's parseTimestampString: rejects 1-digit segments and empty fractions, accepts zone suffixes on date-only strings and 7-digit years

ANSI cast of double/float to int/long falsely reports CAST_OVERFLOW for exactly-representable boundary values (2147483647.0, -2147483648.0, ±2^63)

rpad/lpad panic (Option::unwrap on None) when the length column contains NULL

Nested (struct/list/map) Parquet schema-evolution conversions bypass Spark's type-conversion rules: silent NULLs, string parsing, and a native panic

Decimal to double/float cast is not correctly rounded: last-digit-wrong results for ordinary DECIMAL(38,18) values

JVM exit hangs when the application returns from main without SparkContext.stop()

Regular native scan forwards S3 options without checking divergence from Hadoop's resolution

Uh oh!

Search results