Eventual-Inc/Daft - CodSpeed

Daft

Blog Docs Changelog

Performance History

Latest Results

test(udf): cover combined large-stderr + no-newline use_process case (#6825) ## Summary Follow-up test to #6793. The two existing tests in `test_use_process_deadlocks.py` cover each deadlock pathology in isolation: - Large stderr **with** newline → exercises the `mp.connection.wait()`-based drain. - Small stderr **without** newline → exercises the prepended-`\n` divider fix. Per @ohbh's review on #6793: even with the drain in place, `readline()` in `trace_output()` could in principle still block on a no-newline buffer while the child is stalled on a full OS pipe. This test combines both pathologies — a UDF that writes >64 KiB to stderr without ever emitting a newline — and confirms the current implementation handles the combined case correctly. ## Test plan - [x] `DAFT_RUNNER=native pytest tests/udf/test_use_process_deadlocks.py` — all 4 tests pass locally (macOS). 🤖 Generated with [Claude Code](https://claude.com/claude-code)

main

17 seconds ago

style: rustfmt assert_eq! line break in task_store_tests

claude/integrate-task-events-xazrE

51 minutes ago

fix: fix style and implement clippy suggestions

mytsvy:mytsvy/list-flatten

1 hour ago

fix(test): update test_local_full_ls to expect canonical file:// URIs (#6824) ## Summary Follow-up to #6817. After that PR landed, `iter_dir` emits canonical `file:///C:/...` (three-slash) URIs on Windows. But `test_local_full_ls` in `daft-io/src/local.rs` was still building its expected values with `format!(\"file://{path}\")` directly, which produces the old non-canonical `file://C:/...` (two-slash) form for Windows drive-letter paths. Result: the test passed before #6817 and breaks on Windows after. CI on `main` after #6817 merged showed: \`\`\` left: file:///C:/Users/RUNNER~1/AppData/Local/Temp/.tmpgPPMMn/.tmpB03N0L right: file://C:/Users/RUNNER~1/AppData/Local/Temp/.tmpgPPMMn/.tmpB03N0L \`\`\` ## Fix Switch the test to build expected values via `local_path_to_file_uri` (the same helper the production code uses), so the test stays in sync with `iter_dir`'s output. POSIX behavior is unchanged — `local_path_to_file_uri` only adds the third slash for Windows drive-letter paths; POSIX absolute paths still produce `file:///tmp/...`. ## Test plan - [x] `cargo test -p daft-io test_local_full_ls` passes locally (macOS). - [x] Windows rust-tests on `main` should go green after merge. 🤖 Generated with [Claude Code](https://claude.com/claude-code)

main

1 hour ago

feat: add pmod function for PySpark parity (#6801) ## Changes Made Implements `pmod(a, b)` (positive modulo) for PySpark parity. - Returns `((a % b) + b) % b` — result sign follows divisor (e.g. `pmod(-7, 3) == 2`). - Returns NULL when `b == 0` (instead of panicking like `%`). - Supports integer and float inputs; promotes operands to a common supertype, then normalizes to one of `{Int64, UInt64, Float32, Float64}` so the kernel only has four code paths. - Uses `wrapping_rem`/`wrapping_add` on the signed kernel to avoid the `i64::MIN % -1` overflow panic in debug builds. ## Related Issues Part of #3793

main

2 hours ago

ci: revert TEMP Windows rust-tests enablement on PRs Windows verification on this PR is complete (run 25080590564): rust-tests-platform (Windows, false) passed in 40m. Restore the matrix exclude so PR runs don't pay the 50+ min Windows tax going forward — Windows continues to gate `main` only.

rohit/bugfix/test-local-full-ls-windows-uri

2 hours ago

test(udf): cover combined large-stderr + no-newline use_process case Follow-up to #6793. The two existing tests in test_use_process_deadlocks.py cover each deadlock pathology in isolation: - Large stderr WITH newline (mp.connection.wait drain). - Small stderr WITHOUT newline (prepended-newline divider). Per a review concern, even with the drain in place readline() in trace_output() could in principle still block on a no-newline buffer while the child is stalled on a full OS pipe. Add a regression test that exercises both pathologies together — a UDF that writes >64 KiB to stderr without ever emitting a newline. Confirms the current implementation handles the combined case correctly.

rohit/test/use-process-large-stderr-no-newline

2 hours ago

Revert "refactor(events): rename TaskInfo.origin_node_id to last_node_id" Per @samstokes review: the field is conceptually the same as LocalNodeContext::origin_node_id — both are "the distributed pipeline node ID associated with [thing]". Also matches the dashboard server's TaskGroupKey::origin_node_id, which uses it for the same UI grouping purpose. The numeric coincidence with `node_ids.last()` is a construction detail, not the field's semantic role. Updated the doc-comment to spell out the relationship (= last_node_id = node_ids.last(); same kind of value as LocalNodeContext.origin_node_id).

claude/integrate-task-events-xazrE

2 hours ago

Latest Branches

0%

feat(dashboard): Draft tasks view for Flotilla#6783

1 hour ago

3b7d86a

claude/integrate-task-events-xazrE

+1%

feat(list): add list_flatten expression#6812

2 hours ago

355ac60

mytsvy:mytsvy/list-flatten

0%

fix(test): update test_local_full_ls to expect canonical file:// URIs#6824

21 hours ago

38af98f

rohit/bugfix/test-local-full-ls-windows-uri

© 2026 CodSpeed Technology

Home Terms Privacy Docs