duckdb

· opensource

duckdb, the analytical database in a process · 42,000 stars.

what i did

piping READ_JSON of nested data straight to COPY TO parquet reordered the rows: the file came out in an order the input never had. a data-correctness bug in a path many pipelines take, reported with the reproduction, and closed by the project.

sources

  • issue #18252, READ_JSON improperly orders rows of nested data when piped to copy to parquet, july 2025.