How to Open a Parquet File (Python, pandas, DuckDB & More)
Parquet is a compressed, columnar file format — smaller than CSV and far faster to query. Here's how to open one in every common tool.
pandas (Python)
import pandas as pd
df = pd.read_parquet("realestatedataset-texas-2026-08-01.parquet")
print(df.shape)
print(df[["city","beds","baths","last_sold_price"]].head())pandas reads Parquet natively (via pyarrow). Nested price_history / tax_history come through as lists of structs.
DuckDB — query without loading
DuckDB queries a Parquet file in place — no import step, no database:
SELECT city, COUNT(*) n, AVG(last_sold_price) avg_price
FROM 'realestatedataset-texas-2026-08-01.parquet'
WHERE status = 'active'
GROUP BY city ORDER BY n DESC LIMIT 10;It reads only the columns you touch, so even the multi-gigabyte national file is fast on a laptop.
Polars, R, and the command line
- Polars:
pl.read_parquet("file.parquet") - R (arrow):
arrow::read_parquet("file.parquet") - CLI (DuckDB):
duckdb -c "SELECT * FROM 'file.parquet' LIMIT 5"
Prefer a spreadsheet? Download the CSV version instead — it opens in Excel or Google Sheets directly.
The full schema
Every column, type, and example is in the data dictionary. Parquet includes the nested history columns that CSV omits. Everything is CC0.
Frequently asked
What opens a Parquet file?
pandas (pd.read_parquet), DuckDB, Polars, Apache Spark, and R's arrow package all read Parquet natively. DuckDB can even query it without loading it into memory.
Is Parquet better than CSV?
For analysis, yes — it's columnar and compressed, so files are smaller and queries only read the columns you use. CSV is better only when you need a plain table in a spreadsheet.
How do I open Parquet without Python?
Use the DuckDB command-line tool (duckdb -c "SELECT * FROM 'file.parquet'"), or just download the CSV version of the dataset, which opens in any spreadsheet.
Want to query one property live instead of a whole file?
Explore the API →