Quick Look for Apache Parquet on macOS. Select a .parquet file in Finder, press Space, see your data — schema, types, and rows in a native table. No Python, no Java, nothing else to install.
- Native table view — looks and feels like macOS's built-in CSV preview, in both light and dark mode
- Schema at a glance — column names with their types (
BIGINT,VARCHAR,TIMESTAMP, …) in the header - File facts in the status bar — total rows × columns, file size, compression codec, row group count
- Select & copy — click a cell (or its
#row number for the whole row) and hit ⌘C; rows copy as CSV with full cell contents, even past the display truncation - Inspect long values — hover a truncated cell for a tooltip with more of its content
- Fast on big files — powered by an embedded DuckDB; only the first 500 rows are read, row counts come from parquet metadata
- Handles the ugly stuff — NULLs shown distinctly, 300+ column files, megabyte-sized cells, embedded newlines/tabs shown escaped (
\n,\t), corrupt files get a readable error instead of a blank panel - Signed & notarized — no Gatekeeper warnings, no quarantine dance
brew install --cask inanalper/tap/kolonDownload Kolon-<version>.dmg from Releases, open it, drag the column into place, then launch Kolon once so macOS registers the Quick Look extension.
Requires Xcode and XcodeGen (brew install xcodegen).
git clone https://github.com/inanalper/Kolon && cd Kolon
./Scripts/fetch-duckdb.sh # downloads the pinned libduckdb release
# (or ./Scripts/build-duckdb.sh to compile the slim parquet-only dylib we ship — takes ~30 min)
xcodegen generate
xcodebuild -project Kolon.xcodeproj -scheme Kolon -configuration Release \
-derivedDataPath build build
cp -R build/Build/Products/Release/Kolon.app /Applications/
open /Applications/Kolon.app # registers the extension, then you can close itIf a preview doesn't appear right away, restart Quick Look with killall QuickLookUIService or log out and back in.
Kolon is a sandboxed Quick Look app extension with a slim, parquet-only build of DuckDB embedded — parquet decoding, type inference, and metadata all come from DuckDB's parquet reader, rendered into an AppKit NSTableView. Previews are read-only and never leave your machine.
Preview limits (the status bar tells you when they kick in): first 500 rows, first 200 columns, cells truncated at 300 characters.
- More formats on the same engine: Arrow/Feather, Avro, ORC, compressed JSONL/CSV

