Parquet — Columnar Table
A small employee table as Apache Parquet — the columnar format at the heart of modern data lakes and analytics. For testing Parquet readers (pandas, Spark, DuckDB) and conversion.
| idint64 | namestring | emailstring | departmentstring | activebool | scoredouble | joineddate |
|---|---|---|---|---|---|---|
| 1001 | Ada Lovelace | ada.lovelace@example.com | Engineering | true | 98.5 | 2021-03-01 |
| 1002 | Alan Turing | alan.turing@example.com | Research | true | 95 | 2020-06-15 |
| 1003 | Grace Hopper | grace.hopper@example.com | Engineering | false | 91.2 | 2019-11-20 |
| 1004 | Katherine Johnson | katherine.johnson@example.com | Operations | true | 96.8 | 2022-01-10 |
| 1005 | Edsger Dijkstra | edsger.dijkstra@example.com | Research | false | 89.4 | 2018-09-05 |
Specifications
- Rows
- 5
- Columns
- 7
- Format
- Apache Parquet
- Layout
- columnar
- Compression
- snappy
What is a .parquet file?
Apache Parquet (.parquet) is a binary, columnar storage format for analytical data. It stores each column separately with per-column compression and encoding, embeds a schema and statistics, and is the de-facto standard for data lakes and engines like Spark, DuckDB, and pandas/pyarrow.
How to use this file
Use an example .parquet file to test columnar readers (pyarrow, DuckDB, Spark), schema and predicate-pushdown handling, and Parquet-to-CSV/JSON converters.
How to use this file for testing
“Parquet — Columnar Table” is a deterministic Novus Examples fixture for Conversion testing, Data engineering. The same content exported across many formats and linked as a group, so you can convert one and diff against the expected twin.
Documented properties for this file: 5 rows · 7 columns · Apache Parquet. Compare results against paired or grouped companions on this page when present (clean↔damaged, searchable↔scanned, or format twins) so scores stay reproducible across runs.
Download the file once, keep the path stable in CI or local scripts, and treat the spec table as the contract: dimensions, seeds, field lists, and roles are intentional. Corrupt or invalid samples are labelled as such — expect parsers to fail loudly rather than silently accept them.
Data fixtures document their exact quirks — delimiters, encodings, null handling, schema, and row counts — in the spec table. Point your parser or importer at the file and assert it handles the documented edge cases; clean and deliberately-messy siblings make before/after diffs straightforward.
Code examples
import pandas as pd # pip install pyarrow
df = pd.read_parquet("employees.parquet")
print(df.head())
print(df.dtypes)Related files
- h5HDF5 — Hierarchical Scientific DataA small HDF5 file with a compound 'employees' dataset and a numeric 'readings' grid — the hierarchical format used across science and ML. For testing h5py/HDF5 readers and conversion.
- orcConvert v2 ORC Employee Table SourceBinary orc source for the five-row P8 employee conversion table, preserving ids, names, departments, booleans, and scores. Stable P8 artifact p8-convert-orc-source.
- csvConvert v2 ORC Expected CSVCsv semantic reference for the five-row P8 employee conversion table, preserving ids, names, departments, booleans, and scores. Stable P8 artifact p8-convert-orc-csv.
- jsonConvert v2 ORC Expected JSONJson semantic reference for the five-row P8 employee conversion table, preserving ids, names, departments, booleans, and scores. Stable P8 artifact p8-convert-orc-json.
- csvE-commerce Customers (CSV, 500 rows)A realistic e-commerce customer directory (500 rows) — part of a relational dataset (products, customers, orders) with CSV, JSON, SQL, and Parquet twins for testing joins, imports, and conversion.
- csvE-commerce Orders (CSV, 2000 rows)A realistic e-commerce order lines (customer_id → customers, product_id → products) (2000 rows) — part of a relational dataset (products, customers, orders) with CSV, JSON, SQL, and Parquet twins for testing joins, imports, and conversion.
Generated by generation/data_binary.py. Free for any use, no attribution required — license.