vortex-data
vortex
Blog
Docs
Changelog
Blog
Docs
Changelog
Overview
Branches
Benchmarks
Runs
Performance History
Latest Results
Allow caching LayoutReader in VortexFile (#8244) Not-reinitializing LayoutReader saves us some time. As a related change, make root() static throughout the program. Before ScanBuilder::new() created a new root() which didn't compare with the old one pointer-eq-wise which was an issue Signed-off-by: Mikhail Kot <mikhail@spiraldb.com>
develop
38 minutes ago
remove Signed-off-by: Robert Kruszewski <github@robertk.io>
refactor/parent-ref-stack-allocated
12 hours ago
Unify Matcher on a single ParentRef-based entry point Replace the dual Matcher entry points (matches/try_match over ParentRef plus matches_ref/try_match_ref over ArrayRef) with one generic pair over a new sealed AsParent trait, implemented by ArrayRef and ParentRef. Each matcher now has a single Match associated type and one implementation. - Merge CanonicalView/ParentCanonicalView and ColumnarView/ ParentColumnarView into single ParentView-backed enums; drop the ScalarFnArrayView inner-view generic and ParentScalarFnArrayView alias. - Remove the unused ParentRef::optimize; ArrayParts::optimize remains the construction-side entry point. - Add ParentView::materialize_view and Array::as_parent_view to bridge call sites that need heap-only ArrayView APIs (free for heap-backed parents). - Heap-only compute kernels downcast via as_typed to keep ArrayView ergonomics; matcher-driven call sites use explicit materialization. Signed-off-by: Robert Kruszewski <github@robertk.io>
refactor/parent-ref-stack-allocated
12 hours ago
fixes Signed-off-by: Robert Kruszewski <github@robertk.io>
rk/sparkstats
12 hours ago
Optimize `fsst_compress_offsets_overflow_i32` by 2X locally (#8324) ## Summary Optimize the infamous `fsst_compress_offsets_overflow_i32` test by dropping the symbol table - instead compressing with an empty symbol table to get constant escapes. Locally this change drops the test's runtime from 136s to 57s (in normal test profile), which should have a real impact on our CI runtimes. This change is provided by the new fable-5 model. Signed-off-by: Adam Gutglick <adam@spiraldb.com>
develop
12 hours ago
Fix generation of stat pruning expression for unsupported dtypes Signed-off-by: Robert Kruszewski <github@robertk.io>
rk/statfixes
12 hours ago
Allow running reduce_parent operations on stack allocated parents Add indirection logic to support stack allocated array parents in reduce rules. This lets us avoid heap allocation for commonly used operations that are often immediately optimised away. Signed-off-by: Robert Kruszewski <github@robertk.io>
refactor/parent-ref-stack-allocated
12 hours ago
ArrowImportTable impl Signed-off-by: Andrew Duffy <andrew@a10y.dev>
aduffy/df-geo
12 hours ago
Latest Branches
CodSpeed Performance Gauge
-21%
Allow running reduce_parent operations on stack allocated parents
#7751
12 hours ago
5b8d3e9
refactor/parent-ref-stack-allocated
CodSpeed Performance Gauge
+29%
Support reporting statistics in spark datasource
#8057
12 hours ago
deb1dc0
rk/sparkstats
CodSpeed Performance Gauge
+20%
Fix generation of stat pruning expression for unsupported dtypes
#8326
12 hours ago
0c6b621
rk/statfixes
© 2026 CodSpeed Technology
Home
Terms
Privacy
Docs