Parquet Viewer & Converter
View Parquet files online, convert Parquet to CSV or JSON and CSV to Parquet, run SQL on CSV files and profile every column, all in your browser with DuckDB. Open several Parquet, CSV, TSV, JSON and NDJSON files at once, join them with SQL, inspect the schema and Parquet metadata, and download the result. Files are never uploaded.
What does this tool do?
Parquet is the column-oriented file format used by Spark, pandas, Polars, BigQuery and Snowflake exports and data lakes. It is compact and fast, but it is binary, so you cannot open it in a text editor or Excel. This tool runs DuckDB, an analytical SQL database, inside your browser with WebAssembly, and puts four tools in one page:
- Parquet viewer: see the rows, columns, types and internal metadata of a
.parquetfile - Converter: Parquet to CSV, CSV to Parquet, Parquet to JSON, JSON to Parquet, CSV to JSON and more
- SQL on CSV and Parquet files: filter, aggregate and join files with SQL, no database setup needed
- Data profiler: min, max, distinct count, mean, quartiles and null percentage for every column
Your files are read directly from your disk by the browser and are not uploaded anywhere. Only the DuckDB engine itself (about 7 MB) is downloaded from the jsDelivr CDN the first time you use the tool.
How to open a Parquet file online
- Drop a
.parquetfile onto the drop area, or click it to select the file. Click Try sample data if you just want to see how it works. - The file becomes a table named after the file, for example
sales_2024.parquetbecomessales_2024, and its rows are shown right away. - Click Schema for the column names and types, or File metadata, Row groups and Key-value metadata for Parquet internals.
Up to 1,000 rows are shown by default; change Rows shown to see up to 10,000.
How to convert Parquet to CSV (or JSON)
- Drop the Parquet file. The query
SELECT * FROM <table>;is run automatically. - Click Export and choose CSV, TSV, JSON (array) or NDJSON.
Export always writes every row of the query, not only the rows shown in the table. To convert only part of the file, edit the query first, for example SELECT id, name FROM sales WHERE year = 2024;.
How to convert CSV to Parquet
- Drop a
.csvor.tsvfile. The delimiter, header row and column types are detected automatically. - Check the types with Schema. To change one, cast it in SQL, for example
SELECT * REPLACE (CAST(zip AS VARCHAR) AS zip) FROM customers;. - Click Export → Parquet (zstd). The file is written with zstd compression, which is usually several times smaller than the CSV.
JSON and NDJSON (JSON Lines) files are converted to Parquet or CSV the same way. Nested JSON objects and arrays become STRUCT and LIST columns.
How to run SQL on CSV files
Every dropped file is a table, so you can query CSV, Parquet and JSON files together with full DuckDB SQL: joins, GROUP BY, window functions, PIVOT, regular expressions and date functions. Edit the SQL and press Run or Ctrl+Enter (Cmd+Enter on a Mac). Table and column names are suggested as you type.
This is useful for CSV files that are too large or too slow for Excel. DuckDB reads only the columns a query needs, so files of several hundred MB can be explored.
How to profile a dataset
Select a table and click Summarize. For every column you get the type, min, max, approximate distinct count, average, standard deviation, 25th / 50th / 75th percentiles, row count and null percentage. It is a quick way to check a new dataset for missing values, outliers and unexpected types before using it.
Parquet metadata
| Button | What it shows |
|---|---|
| File metadata | Writer (created_by, for example parquet-mr, pyarrow or DuckDB), number of rows and row groups, format version and file size |
| Row groups | For every column chunk: physical type, compression codec (Snappy, zstd, gzip…), encodings, compressed and uncompressed size, and min / max / null count statistics |
| Key-value metadata | Extra metadata stored by the writer, such as the pandas or Arrow schema |
Supported files
| Format | Extensions |
|---|---|
| Parquet | .parquet, .parq, .pq |
| CSV / TSV | .csv, .tsv, .txt |
| JSON | .json (array of objects), .jsonl and .ndjson (one object per line) |
You can export to CSV, TSV, JSON, NDJSON and Parquet (zstd). Dropped files up to 20 MB in total are kept with the tool, so they are still there after a reload.
FAQ
What is a Parquet file?
Apache Parquet is an open, column-oriented file format for tabular data. Values of each column are stored together and compressed, which makes files small and lets tools read only the columns they need. It is the standard format for data lakes and for exchanging data between Spark, pandas, Polars, DuckDB and cloud warehouses.
How do I open a Parquet file without Python?
Drop it onto this page. You do not need Python, pandas, Spark or any installation; the file is read in your browser.
Are my files uploaded?
No. Files are registered with the DuckDB engine running in your browser and read from your disk on demand. Nothing is sent to a server.
Is there a file size limit?
There is no fixed limit. Parquet files are read in parts, so large files work well. CSV and JSON files have to be scanned completely for some queries, and very large files (several GB) may exceed the memory available to the browser tab.
Why does the first run take a few seconds?
The DuckDB engine (WebAssembly, about 7 MB compressed) is downloaded and started the first time. Your browser caches it, so later visits are faster.
Can I open Excel (.xlsx) files?
Not directly. Save the sheet as CSV first and drop the CSV file.
What happens to tables I create with SQL?
Tables and views created with CREATE TABLE or CREATE VIEW live in memory and are lost when the page is reloaded. Dropped files are kept (up to 20 MB in total) and become tables again automatically.