How to convert Parquet to CSV or JSON without code

By Mykola-Bohdan Vynnytskyi · Updated November 7, 2026

Open the file in Parquet Reader, click Export above the table and pick CSV, JSON or Markdown. To export only part of the file, run a SQL query first: Export then writes the query result instead of the whole file. There is no script to write, and the file stays on your Mac.

Export the whole file

1. Open the file in Parquet Reader.
2. Click Export in the table header.
3. Choose CSV, JSON or Markdown.
4. Pick a name and a location. The name of the source file is suggested.

Here is what each format gives you:

• CSV: comma-separated with a header row. A NULL becomes an empty field.
• JSON: one array with an object per row.
• Markdown: a table you can paste into a README or a pull request.

Export writes the whole dataset, not only the page on screen. For a quick paste, the Copy button puts the current table on the clipboard as Markdown.

Filter with SQL before exporting

Most of the time you want some of the rows and a few of the columns. Write that as a query in the SQL Query panel. The file is available as parquet_table:

SELECT order_id, country, amount
FROM parquet_table
WHERE status = 'paid'
  AND order_date >= DATE '2026-01-01'

Run it, then click Export. The export contains the rows the query returned. Show Original or ⌘⇧R takes you back to the full file.

Right-click filters, sorting and hidden columns only change the view. They do not change the export. SQL does.

Other ways to convert it

pandas

import pandas as pd

df = pd.read_parquet("orders.parquet")
df.to_csv("orders.csv", index=False)
df.to_json("orders.json", orient="records", date_format="iso")

# only some rows
df[df["status"] == "paid"].to_csv("paid.csv", index=False)

Watch the dates in JSON. A date column came out as 2026-01-03T00:00:00.000 when I ran this, so a plain date gained a time.

DuckDB command line

duckdb -c "COPY (SELECT * FROM 'orders.parquet' WHERE status = 'paid') TO 'paid.csv' (HEADER, DELIMITER ',')"
duckdb -c "COPY (SELECT * FROM 'orders.parquet') TO 'orders.json' (FORMAT json, ARRAY true)"

Here the same date stays 2026-01-03, and the filter runs while the file is read.

What CSV loses

Parquet stores a type for every column. CSV stores text. After the conversion, whatever reads the file has to guess that 2026-01-03 is a date, and a NULL can no longer be told apart from an empty string. Keep the Parquet file as the source and treat the CSV as a copy.

What the free version does not do

Export is free, including query results. It writes CSV, JSON and Markdown. Free works with files, and a join of several files exports like any other query result. Exporting a whole folder as one dataset needs a folder open, which is Pro. See the guide to partitioned folders.