Convert Parquet to CSV or JSON

Spreadsheets still speak CSV. APIs still speak JSON. Parquet is what data teams ship. Converting the whole file is often a mistake — a 200-column extract becomes a 2 GB CSV nobody can open. Convert the slice you need.

In ParquetView

1. Open the viewer and drop the Parquet file.
2. Hide columns you do not want (the column panel at the bottom of the grid). Exports include visible columns only.
3. Filter a header, search, or run SQL — for example SELECT user_id, event_time FROM events WHERE dt = '2026-08-01'.
4. Open Export and choose CSV or JSON. Parquet is available if you want a smaller typed file back out.

Very large exports are capped at 100,000 rows so the browser stays stable. If you need more, filter harder or use DuckDB/Spark locally.

Why not convert everything?

CSV drops types (timestamps become strings, integers become floats in Excel). Nested Parquet structs flatten poorly. JSON explodes in size. Use CSV when a person will look at the file. Use JSON when a program expects records. Stay in Parquet when the next consumer is a warehouse.

Command line alternative

If you already have DuckDB installed: duckdb -c "COPY (SELECT * FROM 'data.parquet') TO 'data.csv' (HEADER, DELIMITER ',')". ParquetView is the same idea without installing DuckDB — useful on a locked-down laptop.

Related

• Open a Parquet file
• Filter with SQL before you export