Parquet to CSV trades typed column storage for a plain delimited file that spreadsheet and import tools can open.
From real runs
| Element | Result | What happens |
|---|---|---|
| Column order | Kept | The CSV header preserved the Parquet field order id,created_at,active,amount,note,optional in the verified run. |
| Integer values | Kept | Small integer values such as 101 and 102 were written as CSV numerals without quotes or decimal suffixes. |
| Timestamps | Changed | Timestamp-like values came out as ISO strings with millisecond precision, including 2026-09-10T18:30:00.000Z. |
| Booleans | Changed | Boolean columns became the text tokens true and false, which most CSV readers can re-infer but CSV itself cannot type. |
| Nulls | Changed | Parquet nulls became empty CSV fields, as seen after alpha, quoted in the optional column. |
| Comma-containing text | Kept | The note value alpha, quoted was wrapped in CSV quotes so it stayed one field. |
| Parquet metadata | Lost | Column schema metadata, physical encodings, row groups, and compression details are not represented in the CSV file. |
The CSV header preserved the Parquet field order id,created_at,active,amount,note,optional in the verified run.
Small integer values such as 101 and 102 were written as CSV numerals without quotes or decimal suffixes.
Timestamp-like values came out as ISO strings with millisecond precision, including 2026-09-10T18:30:00.000Z.
Boolean columns became the text tokens true and false, which most CSV readers can re-infer but CSV itself cannot type.
Parquet nulls became empty CSV fields, as seen after alpha, quoted in the optional column.
The note value alpha, quoted was wrapped in CSV quotes so it stayed one field.
Column schema metadata, physical encodings, row groups, and compression details are not represented in the CSV file.
id,created_at,active,amount,note,optional and wrote 101,2026-09-10T18:30:00.000Z,true,1234.5,"alpha, quoted",.102,2026-09-11T00:00:00.000Z,false,,Bob's sample,nullable, leaving amount blank.We converted our Apache Parquet sample (816 B) with Table engine on 2026-09-11. It took 2 ms and produced a 281 B CSV file, 66% smaller than the input. The engine raised no warnings.
Input · sample.parquet · 816 B
Parquet file (PAR1 magic present)
Output · sample.csv · 281 B
4 data rows × 5 columns; header: company,stage,arr,owner,close_date
First lines of the output
company,stage,arr,owner,close_date
Northwind Traders,Proposal,48000,R. Ayers,2026-10-14T00:00:00.000Z
Contoso Ltd,Negotiation,120000,P. Ng,2026-09-30T00:00:00.000Z
Fabrikam,Qualified,22500,R. Ayers,2026-11-02T00:00:00.000Z
Tailspin Toys,Won,64000,M. Okafor,2026-09-03T00:00:00.000ZDuckDB
duckdb -c "COPY (SELECT * FROM 'in.parquet') TO 'out.csv' (HEADER, DELIMITER ',');"DuckDB may format timestamps differently depending on types and settings, so compare date text before importing.
The HTTP API and the MCP server run the exact engine this page uses (Table engine); the response carries the same warnings the web converter shows. Check support first with GET /api/v1/pairs?from=parquet&to=csv.
curl -X POST https://filetypeconverters.com/api/v1/convert \
-H "Authorization: Bearer $FTC_API_KEY" \
-F "file=@input.parquet" -F "from=parquet" -F "to=csv" \
-F 'options={"encoding":"utf-8"}' \
--output output.csv| Key | Applies to | Type | Default |
|---|---|---|---|
| encoding | Reading Apache Parquet | utf-8 · latin1 | utf-8 |
| delimiter | Writing CSV | string | , |
| newline | Writing CSV | · |
/convert/parquet-to-json when downstream code needs booleans, numbers, and nulls represented explicitly.No. CSV is text with delimiters. Numbers, booleans, timestamps, and nulls are serialized into CSV fields and may be re-inferred later.
Null values become empty fields. In the test output, a null amount produced two adjacent commas before the note field.
No. Row groups and encoding metadata belong to the Parquet container. The CSV output contains only headers and rows.