Eventual-Inc/Daft - CodSpeed

Daft

Blog Docs Changelog

Performance History

Latest Results

Merge branch 'main' into everettVT/hf-storage-buckets

everettVT/hf-storage-buckets

1 hour ago

fix: fix column not found when using count_rows() for sparse data (#6703) ## Changes Made followup https://github.com/Eventual-Inc/Daft/pull/5681  Previously, the issue of missing fields in sparse data was resolved at the plan level for APIs such as collect() and show(). However, the count_rows() API still suffered from the same problem during the physical execution path for local files. Specifically, when calling count_rows() on a DataFrame that includes a filter on a missing field (e.g., where("sound is null")), the operation failed with a FieldNotFound error because the schema did not contain the required column. ``` daft.read_json('/mnt/test/vedio.audio-00002.jsonl') .where("sound is null") .count_rows() ``` ``` DaftError::FieldNotFound Column "sound" not found in schema: [Field { name: "video_meta", dtype: Struct(...), metadata: {} }] ``` The current modification extends the field completion logic to the physical execution path of count_rows(), ensuring that missing fields are properly filled before counting rows. This aligns the behavior of count_rows() with that of collect() and show(). ## Related Issues

main

1 hour ago

fix(cli): scope query id output to interactive sessions

BABTUNA:feat/show-query-id-cli

2 hours ago

introduce task resource

gavin9402:introduce_file_resource

4 hours ago

test(list_chunk): move groupby regression into tests/dataframe

veinkr-bot:fix-chunk-return-type

5 hours ago

feat(observability): enable event logs by default

BABTUNA:feat/enable-event-log-default

11 hours ago

feat(cli): print query id at query start

BABTUNA:feat/show-query-id-cli

12 hours ago

ci: propagate LLVM_PROFILE_FILE to ray actors via runtime_env The absolute-path fix in 54b25f0c5 did nothing for actor-side coverage. Per-file lcov inspection shows driver-side functions like `PyDistributedPhysicalPlan::from_logical_plan_builder` hit 20 times (once per TPC-H query) while every actor-side function including `PyDistributedPhysicalPlanRunner::new` and `run_plan` remain at zero hits. The driver's instrumentation output is fine; actor profraw is simply never produced in the workspace target directory. The root cause is that Ray actor workers start as fresh subprocesses and do not inherit arbitrary env vars from the driver. Ray explicitly scrubs the spawn env and only forwards a curated set plus whatever is specified via `runtime_env.env_vars`. `LLVM_PROFILE_FILE` was not on that list, so the LLVM profile runtime in every actor fell back to `default.profraw` in the actor's cwd (`/tmp/ray/session_*/...`) where `cargo llvm-cov report` never scans. Forward `LLVM_PROFILE_FILE` explicitly via `runtime_env.env_vars` on both `RaySwordfishActor.options()` and `RemoteFlotillaRunner.options()`. In non-coverage builds the env var is absent, the helper returns an empty dict, and the `runtime_env` stays `None` - so there is no behavior change outside of coverage-instrumented CI. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

desmond/coverage-integration-ray

12 hours ago

Latest Branches

0%

feat(io): add Hugging Face Storage Bucket support#6731

1 hour ago

c78e42d

everettVT/hf-storage-buckets

+11%

feat(cli): show query id in CLI output#6755

2 hours ago

1ba17f4

BABTUNA:feat/show-query-id-cli

0%

feat: Introduce file resource management#6637

14 days ago

b37ec10

gavin9402:introduce_file_resource

© 2026 CodSpeed Technology

Home Terms Privacy Docs