Eventual-Inc/Daft - CodSpeed

Daft

Blog Docs Changelog

Performance History

Latest Results

fix(sql): remove unused noqa comment to fix ruff RUF100 Signed-off-by: jiangxt2 <jiangxt2@vip.qq.com>

jiangxt2:fix/doris-connectorx-fallback

5 hours ago

fix(sql): add missing type annotations to fix mypy errors - Add Any, Engine imports for type annotations - Annotate __getstate__/__setstate__ with dict[str, Any] - Add return type Engine | None to _get_or_create_engine - Add type: ignore[import-untyped] for pymysql import - Add assert engine is not None guard for type narrowing Signed-off-by: jiangxt2 <jiangxt2@vip.qq.com>

jiangxt2:fix/sql-engine-caching

5 hours ago

perf(sql): cache SQLAlchemy engine with thread-safe double-checked locking Signed-off-by: jiangxt2 <jiangxt2@vip.qq.com>

jiangxt2:fix/sql-engine-caching

6 hours ago

fix(sql): fall back to SQLAlchemy on ConnectorX COM_STMT_PREPARE failure Signed-off-by: jiangxt2 <jiangxt2@vip.qq.com>

jiangxt2:fix/doris-connectorx-fallback

6 hours ago

Merge remote-tracking branch 'origin/main' into fix/doris-connectorx-fallback # Conflicts: # daft/datasets/lerobot.py

jiangxt2:fix/doris-connectorx-fallback

6 hours ago

Merge remote-tracking branch 'origin/main' into fix/sql-engine-caching # Conflicts: # daft/datasets/lerobot.py

jiangxt2:fix/sql-engine-caching

6 hours ago

fix(sql): use quantileExact for ClickHouse partition percentile computation ClickHouse does not support the standard SQL PERCENTILE_DISC function. Use quantileExact() instead for exact percentile computation when calculating partition boundaries in read_sql(). quantileExact is preferred over quantile() (which uses reservoir sampling) because quantile(0)/quantile(1) are not guaranteed to equal the true min/max — this would cause silent row loss at partition boundaries (col >= bound_0 AND col <= bound_last). Other dialects continue to use the standard PERCENTILE_DISC WITHIN GROUP syntax. Signed-off-by: jiangxt2 <jiangxt2@vip.qq.com>

jiangxt2:fix/clickhouse-percentile

7 hours ago

feat(iceberg): inject DataFile column statistics into ScanTask Pass PyIceberg DataFile.lower_bounds/upper_bounds into ScanTask.catalog_scan_task(stats=...) as a 2-row Arrow RecordBatch (row 0=min, row 1=max), matching the format used by Delta Lake and Hudi. This enables MicroPartition.filter() to short-circuit on TruthValue::False after a file has been loaded. Implementation details: - Decode binary-encoded bounds via pyiceberg.conversions.from_bytes() - Map field_id to column name via SchemaFieldIdMappingVisitor - Limit to top-level primitive fields; nested fields out of scope - Precompute top_level_primitive_ids in __init__ to avoid per-file overhead - Handle missing bounds, schema evolution, and decode failures gracefully Tests: 19 Python unit tests (all types + edge cases), 7 integration tests (stats injection, value correctness, query consistency), 5 Rust tests (TableStatistics eval + MicroPartition.filter short-circuit). Closes #7299 Signed-off-by: jiangxt2 <jiangxt2@vip.qq.com>

jiangxt2:feat/iceberg-scan-stats

8 hours ago

Latest Branches

0%

fix(sql): fall back to SQLAlchemy on ConnectorX COM_STMT_PREPARE failure#7303

6 hours ago

003874c

jiangxt2:fix/doris-connectorx-fallback

0%

perf(sql): cache SQLAlchemy engine with thread-safe double-checked locking#7305

6 hours ago

a0b46c2

jiangxt2:fix/sql-engine-caching

0%

fix(sql): use quantileExact for ClickHouse partition percentile computation#7304

10 hours ago

61f9488

jiangxt2:fix/clickhouse-percentile

© 2026 CodSpeed Technology

Home Terms Privacy Docs