rs-0.55.1

🐛 Bug Fixes

• Fixed IO plugins receiving predicates they cannot evaluate.
• Fixed hanging or crashing on `sqrt`/`cbrt`/`pct_change`/`ewm_*` with a Struct column.
• Raised on overflow in Decimal sum instead of silently wrapping.
• Preserved nulls when importing Arrow maps.
• Propagated null `by` column values in `rolling_*_by`.
• Fixed self-referencing `field` in `struct.with_fields` with `over`.
• Fixed Arrow buffer offset for `Utf8` and `Binary`.
• Clamped group-by slice offset.
• Ensured categorical `fill_null(strategy="min"/"max")` uses lexical min/max.
• Released the GIL while resolving the schema in `__arrow_c_stream__`.
• Fixed data corruption on Series from nested Arrow `LargeList`.
• Properly propagated enum metadata for iceberg column mapping.
• Correctly wrote parquet field IDs for enums and categoricals.
• Fixed arrow C interop binview offset.
• Fixed duplicate hive values in hive pre-partition rewrite.
• Corrected double-counted offset for Arrow export of sliced `Series(Array)`.
• Fixed Struct/List/Array equality handling in sort and join.
• Used try_new in StructArray construction in polars-json.
• Raised ComputeError instead of panicking in repeat_by when output exceeds IdxSize::MAX.
• Ran type coercion pass on pivot's internally generated group_by.
• Improved slice pushdown into joins.
• Preserved ordering in sliced unions.
• Dropped input sortedness when casting to a string.
• Cleared sortedness flags in `StringChunked` substring kernels.
• Fixed bad mask handling when reading optional parquet column.
• Ensured any operation on `Unknown(Int)` and `Unknown(Float)` results in `Unknown(Float)`.
• Flipped `nulls_last` after `Expr.reverse()`.
• Fixed high blocking thread use in sink_parquet with async local path.
• Serialized LazyFrames backed by bytes.
• Released GIL in `SQLContext.execute()`.
• Propagated `nulls_last` in `function_expr_sortedness`.
• Fixed incorrect slicing when a join requires sorting.
• Fixed panic on self join of scan_delta / scan_iceberg.
• Decremented ParkGroup worker count when worker exits.
• Added missing Send bounds in Connector.
• Fixed UB on first/last_non_null with empty chunk.
• Did not CSE non-column height expr on streaming engine.
• Propagated adjusted sortedness through Gather.
• Deduplicated the expression equality checks.
• Addressed 3VL issue with SQL `NOT IN` interaction with `NULL` values and joins.
• Used total comparison in Expr literal values.
• Ensured SQL `SUM` and `CORR` aggregates return NULL for all-null inputs, added `TOTAL`.
• Shared `null_count_dtype` helper between Delta and Iceberg, fixing `SchemaError`.
• Removed non-output columns from the equi-join and semi/anti-join operators.
• Fixed dropped slice on multiple unions.
• Fixed in-memory engine incorrect slice on maintain order join.
• Checked join schema by position.
• Blocked predicate pushdown past overwritten window keys.
• Avoided panic when union slice skips all rows.
• Solved panic in `dt.replace` when there were multiple chunks.
• Propagated `is_scalar` from the input to the output of `.sort()` and `.sort_by()`.
• Inserted missing coercions from `Unknown(_)` in list/array arithmetic.
• Resolved CSV column names overwrite in DSL->IR conversion.
• Avoided IEJoin rewrite for Categorical comparisons.
• Did not rewrite `sort().reverse()` to `sort(descending=True)` when `maintain_order=True`.
• Propagated bigidx to `polars-plan` from `polars`.
• Ensured BinaryView offset+len does not exceed i32::MAX where possible.
• Fixed invalid offset in strptime.
• Fixed incorrect schema type for decimal <-> primitive division.
• Fixed panic in in-memory CSEE handling.