Eventual-Inc/Daft - CodSpeed

Daft

Blog Docs Changelog

Performance History

Latest Results

chore: drop stray sqlite file committed by mistake my_database.db is produced by the write_sql doctest in the repo root and was picked up by a git add -A. It is not part of this change.

atovk:feat/iceberg-overwrite-filter

6 hours ago

Merge branch 'main' into feat/iceberg-overwrite-filter

atovk:feat/iceberg-overwrite-filter

8 hours ago

fix(distributed): preserve notify tokens when fusing task builders combine_with built the merged builder with `notify_tokens: vec![]` and `cancel_token: None`, dropping both inputs' tokens. A dropped oneshot resolves to a cancellation, so an upstream node waiting on the fused task saw it as finished before it ever ran. LimitNode does exactly that: it tears down its Ray limit-counter actor once every forwarded task has reported, so the actor was killed underneath the fused task, which then failed with ActorDiedError on every retry. Consume both inputs and merge their tokens instead. CrossJoinNode fans one builder out against many partners and a oneshot cannot report several completions, so it keeps the borrowed form, now named combine_fanout.

hello-peter-tang:push-down-limit-outer-join

8 hours ago

fix(sql): make the io_config argument reachable The config constructors were registered under mixed-case names while lookup lowercases the parsed function name first, so s3config, httpconfig, azureconfig, gcsconfig and tosconfig were all unreachable and the io_config argument could not be constructed at all. Registering them lowercase makes every casing resolve, matching the other 72 SQL functions. Also restores the explicit-io_config test that had to be dropped while the constructors were unreachable, and lifts the test module's imports to the top of the file. Closes #7447

jackylee-ch:sql-reader-default-io-config

8 hours ago

refactor(iceberg): make overwrite_filter a Daft-only API Per review: hoist the pyiceberg imports to module scope, and drop pyiceberg's BooleanExpression from the accepted filter types so the public surface stays in Daft's own vocabulary. A predicate now comes in as a Daft expression or an Iceberg predicate string. The module-level imports keep pyiceberg out of `import daft`: nothing on the import or Parquet-write paths loads iceberg_write.

atovk:feat/iceberg-overwrite-filter

11 hours ago

fix: remove trailing blank line in test_context.py

Lucas61000:issue-7407

11 hours ago

chore: remove deprecated min_cpu_per_task from ExecutionContext Remove the deprecated min_cpu_per_task configuration option from DaftExecutionConfig for the v0.8.0 release. This was a configurable option in the old Ray runner and has not been useful since the Flotilla release. - Remove min_cpu_per_task field, default value, and env var parsing from DaftExecutionConfig in Rust (lib.rs, python.rs) - Remove min_cpu_per_task parameter and deprecation warning from daft_execution_config() in context.py - Remove min_cpu_per_task from PyDaftExecutionConfig type stubs - Remove test_min_cpu_per_task_is_deprecated test Closes #7407

Lucas61000:issue-7407

14 hours ago

Merge branch 'main' into fix/escape-sql-string-literals

chuqCTC:fix/escape-sql-string-literals

16 hours ago

Latest Branches

0%

feat(iceberg): add overwrite_filter for static partition overwrite#7419

6 hours ago

dac1ae6

atovk:feat/iceberg-overwrite-filter

0%

feat(optimizer): push limit through row-preserving join sides#7439

11 hours ago

8c7cef1

hello-peter-tang:push-down-limit-outer-join

0%

fix(sql): make the io_config argument work in SQL readers#7449

8 hours ago

b9e1913

jackylee-ch:sql-reader-default-io-config

© 2026 CodSpeed Technology

Home Terms Privacy Docs