Eventual-Inc/Daft - CodSpeed

Daft

Blog Docs Changelog

Performance History

Latest Results

GitHub Actions run

14 minutes ago

GitHub Actions run

17 minutes ago

fix(udf): honor per-method max_retries/on_error overrides

BABTUNA:feat/method-retry-onerror-overrides

24 minutes ago

test(fixed_size_list): update to_arrow assertion to expect 'item' Address rohitkulshreshtha review on #6733: the prior assertion captured the pre-fix behavior where to_arrow passed flat_child's name through verbatim. Now that to_arrow normalizes the inner field to "item", update the assertion to match.

veinkr-bot:fix-chunk-return-type

32 minutes ago

fix at push_down_filter.rs

gavin9402:fix_predicate_pushdown

38 minutes ago

Merge remote-tracking branch 'origin/claude/mock-flotilla-tasks-tab-Wt30L' into claude/integrate-task-events-xazrE # Conflicts: # src/daft-context/src/lib.rs # src/daft-context/src/subscribers/dashboard.rs # src/daft-context/src/subscribers/events.rs # src/daft-dashboard/frontend/src/app/query/mock-tasks.ts # src/daft-dashboard/frontend/src/app/query/page.tsx # src/daft-dashboard/frontend/src/app/query/tasks-sidebar.tsx # src/daft-distributed/src/python/dashboard.rs # src/daft-distributed/src/scheduling/dispatcher.rs # src/daft-distributed/src/statistics/mod.rs # src/daft-distributed/src/statistics/task_lifecycle.rs

claude/integrate-task-events-xazrE

48 minutes ago

fix(cli): avoid stderr noise for invalid query-id env values

BABTUNA:feat/show-query-id-cli

48 minutes ago

feat(distributed): enable two-stage HLL aggregation for approx_count_distinct (#6597) The distributed aggregation pipeline special-cased `ApproxCountDistinct` to skip two-stage pre-aggregation, forcing all raw data to shuffle to a single node. The comment cited an inability to recursively merge HLL sketches. However, the real issue was that `MergeSketch(HyperLogLog)` collapsed sketches directly to a `UInt64` count (via `hll_merge` calling `final_hll.count()`), making the intermediate sketch type unusable across distributed stages. The local `AggregateSink` would further decompose `ApproxSketch` into `ApproxSketch` (sink) + `MergeSketch` (finalize), producing `UInt64` where the second distributed stage expected `FixedSizeBinary`. To fix this, we align with the Spark/Trino/DuckDB convention where the merge step always produces a sketch, and count extraction happens only at the final evaluate step: - `hll_merge`/`grouped_hll_merge` now return `FixedSizeBinaryArray` (merged sketch registers) instead of `UInt64Array` - `MergeSketch(HyperLogLog)::to_field` returns `HLL_SKETCH_DTYPE` instead of `UInt64` - New `hll_cardinality` scalar function (`FixedSizeBinary` -> `UInt64`) extracts the count, following the same pattern as `sketch_percentile` for DDSketch - `ApproxCountDistinct` decomposition in `agg.rs` uses `hll_cardinality` as its final projection - The `ApproxCountDistinct` bypass in the distributed `AggregateNode` is removed --------- Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

main

1 hour ago

Latest Branches

0%

fix: add missing serde feature to uuid workspace dependency#6773

13 hours ago

1019e7b

chenghuichen:cargo-fix

-1%

feat: support Ray 2.55.0#6767

16 hours ago

3cedc6e

XuQianJin-Stars:support-ray-2.55.0

-1%

feat(functions): add seq expression for per-row integer sequences#6772

17 hours ago

4627e23

yuchen-ecnu:support_seq

© 2026 CodSpeed Technology

Home Terms Privacy Docs