Eventual-Inc/Daft - CodSpeed

Daft

Blog Docs Changelog

Performance History

Latest Results

address comment

slade/cleanup-plan-viz

58 seconds ago

feat(dashboard): add interactive tree visualizer for query plans (#6295) ## Summary <img width="1130" height="1502" alt="image (2)" src="https://github.com/user-attachments/assets/8050bcfd-9dfa-4d0d-828c-3fe90788eedc" /> - Replace raw JSON `<pre>` blocks in the Optimized Plan and Unoptimized Plan tabs with an interactive tree visualizer - Nodes are color-coded by operator type: Source (green), Filter (amber), Project (blue), Sort (violet), Join (rose), Aggregate (fuchsia) - Click any node to expand/collapse its properties (predicates, projections, sort keys, join conditions, etc.) - CSS connector lines show parent-child relationships with branching for multi-child nodes (e.g. joins) - Toggle between "Tree View" and "JSON" modes to switch back to the raw formatted JSON ## Test plan - [x] Verified with `next build` (TypeScript compilation passes) - [x] Tested with `next dev` against live dashboard data - [x] Tested with `daft dashboard start` (production build via `make build-release`) - [x] Verified simple queries (filter + sort) and complex queries (join + aggregate + filter pushdown) - [x] Verified JSON toggle switches back to raw formatted view - [x] Verified graceful fallback for unparseable JSON 🤖 Generated with [Claude Code](https://claude.com/claude-code) --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

main

21 minutes ago

add back query end state

colin/fix-native-executor-closing

23 minutes ago

refactor(csv): migrate daft-csv and daft-decoding from arrow2 to arrow-rs Replace daft-arrow (arrow2) dependencies with native arrow-rs crates (arrow-schema, arrow-array, arrow-buffer) and the sync csv crate. This removes the arrow2 CSV I/O layer, inlining row reading helpers and rewriting deserialization to use arrow-rs builders and primitive types instead of arrow2's TrustedLen-based APIs.

universalmind303/arrow-rs-csv

39 minutes ago

Merge branch 'main' into colin/fix-native-executor-closing

colin/fix-native-executor-closing

3 hours ago

Merge branch 'universalmind303/json-refactor' of https://github.com/Eventual-Inc/Daft into universalmind303/json-refactor

universalmind303/json-refactor

4 hours ago

Merge branch 'main' into universalmind303/json-refactor

universalmind303/json-refactor

4 hours ago

refactor(arrow2): move the auto conversion logic into array init (#6289) ## Changes Made In all cases when we receive an arrow array, We were implicitly correcting the type into a daft type where possible. Such as arrow::utf8 -> daft::utf8 _(arrow::largeutf8)_. The old `cast_array_for_daft_if_needed` approach is a caller side obligation... every place that ingests an arrow array has to remember to call it before constructing a Daft array. This led to the casting logic being duplicated across multiple callsites. By moving the casting into the array level `from_arrow` constructors (`DataArray::from_arrow`, `ListArray::from_arrow`), the invariant is self-enforcing. it's impossible to construct a Daft array with the wrong arrow type regardless of how you got there. Callers no longer need to know about Daft's internal type coercions at all. This inadvertently eliminates the duplication in `PySeries::from_arrow` and `record_batch_from_arrow` and makes every future ingestion path correct by default. It also co-locates the coercion logic with the type information that defines it (DataType::to_arrow()), so there's a single source of truth rather than a parallel `coerce_to_daft_compatible_type` function that has to manually mirror the same mappings. ## Related Issues

main

4 hours ago

Active Branches

chore: Move EXPLAIN ANALYZE to Python

last run

58 seconds ago

0%

refactor(arrow2): daft-csv crate

last run

39 minutes ago

0%

refactor(arrow2): refactor daft-json to arrow-rs

last run

4 hours ago

+27%

© 2026 CodSpeed Technology

Home Terms Privacy Docs