Parquet Viewer

A Linux-first desktop Parquet viewer written in C++.

The app is designed for Fedora and Ubuntu users who want to open Parquet files, inspect them as tables, keep multiple files open, and filter rows without writing SQL.

Current features

• Native Qt 6 desktop application
• VS Code-style file explorer for .parquet / .parq files
• Multiple open files through tabs
• Paged table display for large files
• Global search across all rows and columns
• Optional all-rows display by typing -1 into the row count control
• Quick filters such as target = 1, score > 10, and active = true
• Visual filter builder:
  • equals / does not equal
  • greater than / less than
  • at least / at most
  • contains / starts with / ends with
  • is true / is false
  • is empty / is not empty
• DuckDB-backed Parquet queries under the hood

Users never need to type a WHERE clause. The UI builds safe filter expressions internally from dropdowns and input fields.

Build

Requirements

• CMake 3.24 or newer
• A C++20 compiler
• Qt 6 Widgets development files
• DuckDB CLI available as duckdb

The app uses Qt for the desktop UI and the DuckDB CLI for Parquet scanning, filtering, and paging.

DuckDB

CMake looks for the duckdb executable in PATH.

If DuckDB is installed somewhere custom:

cmake -S . -B build -DDUCKDB_EXECUTABLE=/path/to/duckdb

The app embeds that executable path at build time and runs commands like:

duckdb -no-init -json -c "SELECT ..."

Fedora

Install the Qt/C++ toolchain with your normal Fedora package manager. Install DuckDB so duckdb --version works in a terminal.

Ubuntu

Install the Qt/C++ toolchain with your normal Ubuntu package manager. Install DuckDB so duckdb --version works in a terminal.

Commands

cmake -S . -B build
cmake --build build
./build/parquet-viewer

Open files from the toolbar, by double-clicking .parquet files in the explorer, or by passing file paths:

./build/parquet-viewer data/example.parquet

How filtering works

The UI exposes simple filter rows:

[column] [operator] [value]

For example:

status equals 1
active is true
country contains text de

Internally, those choices are converted to DuckDB expressions and applied to the open Parquet file.

The find box also accepts simple quick filters when the left side is a real column name:

target = 1
target=1
score > 10
active = true
name = Alice

If the text is not recognized as a quick filter, it is treated as a normal search value across all columns.

Roadmap

• Column-specific search from header menus
• Schema/details side panel
• Saved filter presets
• Export filtered rows
• Optional persistent cache/index metadata for repeated searches
• Sorting support

License

Parquet Viewer is free software licensed under the GNU General Public License v3.0 or later. See LICENSE for the full license text.