CDV — Columnar Data Viewer

A desktop app for viewing and querying columnar data files — Parquet, Avro, Arrow IPC — alongside CSV, JSON, and Excel. Drop a file in, get a schema and a SQL editor. No server, no setup.

Under the hood it's DuckDB doing the work: files are registered as views (or optionally materialized into tables), and results travel to the frontend as Arrow IPC batches. CDV also connects to PostgreSQL, Snowflake, SQLite, other DuckDB databases, and DuckLake catalogs, so you can pull data from live systems into the same workspace.

Tauri v2 (Rust) backend, React + TypeScript frontend.

Supported Data Formats

Input

Format	Extensions	Engine
Parquet	.parquet	read_parquet()
Avro	.avro	read_avro()
Arrow IPC	.arrow, .ipc	Direct reference
CSV	.csv	read_csv_auto()
TSV	.tsv	read_csv_auto(delim='\t')
JSON	.json	read_json_auto()
JSONL	.jsonl	read_json_auto()
Excel	.xlsx	read_xlsx()

Export

Format	Output
CSV	FORMAT csv
Parquet	FORMAT parquet
JSON	FORMAT json

Cloud Storage

Provider	Scheme	Auth
Amazon S3 (+ compatible)	s3://	Access Key + Secret Key, optional endpoint
Google Cloud Storage	gcs://	Key ID + Secret
Cloudflare R2	s3://	Account ID + Access Key + Secret Key

Database & Lakehouse Connectors

Connector	How it works
PostgreSQL	Attaches via ADBC, supports multiple databases
Snowflake	ADBC driver auto-downloaded on first connect
SQLite	Direct attach
DuckDB	Direct attach
DuckLake	Attaches as a DuckLake catalog

Features

Data Source Management

• Add sources from local files (native file picker) or cloud connections
• Drag-and-drop file import
• Auto-format detection by extension with manual override
• Schema preview with column names, types, nullable, and primary key selection
• Materialize option — store as a DuckDB table for faster repeated access, or keep as a lazy view
• Update, reimport, or remove data sources
• Properties dialog with name, view name, path, format, row count, and full schema

SQL Query Editor

• Monaco Editor with SQL language mode and dark theme
• Context-aware autocomplete — table names after FROM/JOIN, columns after SELECT/WHERE, dot-notation support
• Ctrl+Enter to execute
• Per-tab query state with debounced persistence

Data Viewing

• Virtualized table (@tanstack/react-table + @tanstack/react-virtual) for smooth scrolling
• Column sorting and resizing
• Row numbering
• Type-aware rendering — NULL shown italic, booleans colored, numbers locale-formatted
• View modes: First 100, Last 100, All Rows, Filtered Rows

Cloud Connections

• Create connections to S3, GCS, or Cloudflare R2 with credentials
• Browse remote files via glob() queries
• Import remote files directly as data sources

Export

• Export query results to local file (CSV, Parquet, JSON)
• Native save dialog with customizable query before export

Workspace & Tabs

• Multi-tab interface — data tabs (one per source) and query tabs (unlimited)
• Tab state persisted and restored across restarts

Tech Stack

• Tauri (React, Tailwind, Rust)
• DuckDB
• Arrow

Application Data (com.cdv.desktop)

File	Purpose
cdv.duckdb	Materialized tables persist across sessions
catalog.json	Data source metadata and connection records
settings.json	App preferences
workspace.json	Open tabs and active tab state

All persisted in the OS app data directory. Falls back to in-memory DuckDB if the persistent database fails to open.

Development

Prerequisites

• Rust
• Bun (or Node.js)
• Tauri CLI

Setup

bun install
bun tauri dev

Build

bun tauri build

Contributing

See CONTRIBUTING.md.

License

MIT — see LICENSE.