Convert CSV files to Apache Parquet in your browser. Choose a compression codec and download the result locally — files never leave the page.
How the CSV to Parquet conversion works
Pick a .csv file — the header row becomes column names, values are type-detected, and you preview the parsed rows before downloading a typed .parquet with Snappy, GZIP, or no compression. Nothing uploads; it works offline once loaded.
Parquet stores data column-by-column with compression and real types: the same table typically shrinks several-fold versus CSV and loads dramatically faster in pandas, DuckDB, Spark, and warehouses.
How column types are inferred
true/false only → BOOLEAN
integers within ±2,147,483,647 → INT32
larger integers → INT64
any decimals → DOUBLE
ISO 8601 dates/datetimes → TIMESTAMP
anything else or mixed → STRING (UTF-8)
Inference samples the first 200 values per column; mixed columns fall back to STRING rather than corrupting data. Empty cells become nulls.
Large files
Parsing happens in memory; for very large CSVs use pandas (read_csv().to_parquet()) or DuckDB (COPY … TO … FORMAT PARQUET).