What is a Parquet file?

A Parquet file is a compressed, columnar data file. Analytics tools — Spark, DuckDB, pandas, BigQuery, Snowflake — use it because scanning one column does not require reading the whole row. The usual extension is .parquet.

Unlike CSV, you cannot open a Parquet file in Notepad or Excel and expect to read it. The bytes are binary: encodings, dictionaries, and optional compression (Snappy, Gzip, Zstd, Brotil) sit in front of the values. That is why people search for a Parquet viewer.

Columnar vs row-based

CSV stores row 1, then row 2, then row 3. If you only need the fare column, you still parse every other field. Parquet stores all fare values together, then all pickup values, and so on. Readers skip unread columns. For wide event logs that is the difference between a usable laptop query and a Spark job.

What is inside the file

• Schema — column names, types (int, string, timestamp, nested structs), and nullability.
• Row groups — chunks of rows that a reader can skip using min/max stats.
• Column chunks — the actual encoded pages, often dictionary-encoded.
• Footer — metadata so a reader can seek without scanning from byte 0.

ParquetView’s metadata panel surfaces schema, Arrow types, and file-level key-value metadata so you can confirm a file before you trust it in a pipeline.

How do I open one?

You do not need Python. Open ParquetView, drop the file, and the grid shows rows. From there you can run SQL or export CSV. Desktop apps and the VS Code parquet-viewer extension exist too; a browser viewer is enough for a quick look.

Parquet vs CSV vs JSON

CSV is universal and wasteful. JSON is nested and even heavier. Parquet is smaller, typed, and faster to scan — and awkward until you have a viewer. If a vendor hands you a 400 MB .parquet, do not convert the entire thing to CSV first. Open it, select the columns you need, then export.

GeoParquet and nested types

GeoParquet is Parquet with a geometry metadata convention. Nested structs and lists are valid Parquet. A good viewer should show types, not just stringify everything. If a column looks like JSON in the grid, check the schema panel — it may be a struct.