Data This Week #35
·
10 min read
Why sorted Parquet files don't give DataFusion an ordered scan, point-in-time leak-free graphs on Iceberg, rank-space Hilbert clustering in DuckDB, Netflix's ontology-driven observability, Uber's dependency analysis, Apache Polaris disaster recovery, Redshift data sharing and Iceberg writes, JEV classification, duckxlsx, and debugging data mismatches.