Currently, Parquet files can be exported from QuestDB by converting a table's partitions to Parquet format, and copying them out. This goes via the WAL, so can impact ingestion performance.

Instead, we will add two more means to export data in parquet format:

• bulk export: COPY ... TO ... WITH FORMAT PARQUET;
  • This supports exporting a table or arbitrary query to parquet
  • The parquet format can be configured (row group sizes, compression codecs)
  • The data can be newly partitioned
• ad-hoc export: /exp?query=...&format=parquet
  • This is similar to the above option, but works via the REST API
  • This allows you to download smaller datasets in parquet format, instead of CSV.

Chronologically-ordered parquet files can be ingested very quickly into QuestDB. This feature will make it much simpler to get data in and out of the database, and hook it up to other tools such as Pandas/Polars/DuckDB.