Parq-Bench

A high-performance desktop application for exploring large Parquet files locally. Built with Tauri, React, and DuckDB. No cloud, no accounts, no telemetry — all processing happens on your machine.

Installation

Download the latest installer for your platform from the Releases page.

Features

Preview Tab

Open any Parquet file and see data instantly. Parq-Bench uses virtual scrolling to handle files with millions of rows — only the visible rows are rendered, so scrolling stays smooth regardless of file size.

• Open a file — Click Open Parquet or press Ctrl+O to pick a .parquet file. You can also drag and drop a file onto the preview pane.
• File metadata — Path, size, row count, and column count are shown at the top.
• Schema at a glance — Column names and DuckDB types appear in the header row.
• Interactive pivot view — Once Perspective.js finishes loading in the background, click Switch to interactive view to pivot, sort, filter, and chart the first 5,000 rows.
• Close file — Click Close File to clear the preview and free memory.

SQL Tab

Register one or more files as named tables, then write SQL to query across them.

• Mount a table — Enter an alias (e.g. sales), browse for a file, and click Mount. The table is immediately queryable. You can also drag and drop files onto the SQL pane to auto-register them.
• Glob patterns — Check the Glob box to mount an entire folder of Parquet files as a single table (e.g. logs/*.parquet).
• CSV / TSV support — Enable Slo-mo mode to register delimited text files (CSV, TSV, TXT) with optional custom delimiter.
• Rename or remove tables — Click an alias to rename it in-place. Click x to remove.
• Schema summary — Column names and types for each mounted table are shown as compact pills below the table list.
• SQL editor — A Monaco-powered editor with SQL syntax highlighting, word wrap, and line numbers. Resize it vertically by adjusting the font size with the A- / A+ buttons.
• Autocomplete — Type . after a table alias to get column suggestions. SQL keywords, table names, and column names from all mounted tables are offered as completions.
• Run a query — Click Run SQL or press Ctrl+Enter. Results appear in a scrollable table below the editor with row count, execution time, and column types.
• Export results — Click Export Query CSV or Export Query Parquet to save query output to a file. Press Ctrl+Shift+E to export as CSV directly.
• Table statistics — Click Stats on any mounted table to view column-level statistics. By default, stats are read from Parquet metadata (instant). Enable Deep Stats in Settings to run a full table scan for exact values.
• EXPLAIN ANALYZE — Press Ctrl+Shift+Enter to see the query execution plan instead of result rows.
• Schema diff — Compare column schemas between two mounted tables to spot added, removed, or type-changed columns.
• Charting — After running a query, pick a chart plugin (Y Bar, X Bar, Y Line, Treemap), select X/Y columns and an aggregation (sum, avg, count, min, max), then click Chart in Perspective to visualize results interactively.

Keyboard Shortcuts

Shortcut	Action
Ctrl+O	Open Parquet file
Ctrl+1	Switch to Preview tab
Ctrl+2	Switch to SQL tab
Ctrl+Enter	Run SQL query
Ctrl+Shift+Enter	Explain Analyze
Ctrl+Shift+E	Export query as CSV
Ctrl+,	Open Settings
Escape	Close About dialog or clear error

Theme

Choose System, Light, or Dark from the dropdown in the header. The choice persists across sessions.

Settings

Press Ctrl+, to open the Settings panel. Configurable options:

Setting	Default	Description
SQL Row Limit	200	Maximum rows returned by workspace queries
Perspective Max Rows	5,000	Maximum rows loaded into the Perspective viewer
Editor Font Size	13 px	Monaco editor font size
Expand Mode	Fullscreen	Fullscreen overlay or drag-to-resize panels
Deep Stats	Off	Full table scan for exact stats (may be slow on large files)
Show Visualization	On	Toggle chart controls in workspace
Show Perspective Configure	On	Toggle Perspective's built-in settings button

Memory Guard

Parq-Bench monitors system memory. If usage gets too high, the memory guard trips and blocks file opens and queries until pressure drops. A banner explains what happened and how to recover.

Building from Source

Prerequisites

• Node.js 20+
• Rust (stable)
• Platform build tools for Tauri v2

Development

cd app
npm install
npm run tauri dev

Production Build

cd app
npm run tauri build

Installers are output to app/src-tauri/target/release/bundle/.

Architecture

Layer	Technology	Role
Frontend	React 19, TypeScript, Vite	UI, virtual-scroll table, Perspective viewer
Editor	Monaco Editor	SQL workspace with autocomplete
Visualization	Perspective.js (FINOS)	Interactive grid, charts, pivoting
Backend	Rust, Tauri 2	File I/O, DuckDB queries, Arrow IPC transport
Query Engine	DuckDB 1.4.4 (bundled)	Parquet reading, SQL execution, exports
Transport	Arrow IPC / WebSocket	< 1 MB via Tauri IPC, >= 1 MB via localhost socket

Project Structure

parq_bench/
  app/
    src/            # React frontend (TypeScript)
    src-tauri/
      src/
        lib.rs      # Tauri commands, DuckDB integration, transport
        main.rs     # App entry point
        bin/        # CLI tools (fixture generator, compat matrix)
  docs/             # Architecture docs, upgrade policies
  testdata/         # Synthetic Parquet test fixtures

Release Gate

Validates the full pipeline before a release:

cd app
npm run release:gate

This runs:

• TypeScript + Vite production build
• cargo check and cargo test (13 integration tests)
• Parquet fixture regeneration (5 synthetic files)
• DuckDB compatibility matrix (5/5 must pass)

DuckDB Version Policy

DuckDB is pinned to 1.4.4 for reproducibility. See docs/duckdb-upgrade-policy.md for the upgrade process.

Contributing

Bug reports and feature requests are welcome — please open an issue.

License

GNU General Public License v3.0 — KISA-Keep it Simple Analytics LLC