Convert Parquet to TSV — Free, Private, In-Browser

Parquet is a columnar, strongly-typed, compressed file built for engines. TSV is flat, untyped, tab-delimited text. You reach for this conversion when the next hop cannot read Parquet and also hates commas: a warehouse COPY that wants TAB, a bulk loader, a genetics or logs pipeline that already splits on a tab, or a colleague whose script has delimiter=tab hardcoded. The in-browser converter reads the Parquet with DuckDB-WASM and writes a headered TSV locally — drop the file, preview the schema, download. Nothing is uploaded.

The cost is the schema. Every value becomes text on the way out: timestamps serialize to ISO-ish strings, DECIMALs are written verbatim (exact in the file, but whatever reads the TSV may parse them as floats and lose digits), and BIGINTs past 2^53 will round if a downstream tool treats numbers as doubles. Nested STRUCT and LIST columns have no TSV equivalent, so they flatten to JSON strings inside a cell. NULLs become empty fields, not the word null. A header row is included. There is no compression, so expect the TSV to be several times larger than the Parquet it came from.

Tabs are the point. Comma-heavy values — addresses, notes, currency — no longer need RFC 4180 quoting, which is why people pick TSV over CSV for this hop. The rare value that does contain a literal tab or a newline is quoted automatically so the file stays parseable. If the receiver is delimiter-agnostic, CSV is the more widely supported target and you can skip this page. Everything runs in your browser via DuckDB-WASM; the Parquet never leaves your machine.