parqcat

parqcat is a lightweight command-line utility for inspecting local Parquet files.

Parquet is a good storage format, but quick inspection often means starting Python, a JVM tool, a SQL shell, or a larger data system. parqcat is meant to cover the everyday terminal workflow: show rows, preview the beginning or end of a file, inspect the schema, and pipe rows into tools such as jq.

Install

From crates.io:

cargo install parqcat

From this repository:

cargo install --path .

For development:

cargo build
cargo test

Features

• Native Rust CLI with no Python or JVM runtime dependency.
• cat, head, and tail commands for row inspection.
• schema command for logical schema inspection without decoding row data.
• Table output for quick terminal viewing.
• JSONL output for scripts and pipelines.
• Transparent support for .gz, .zst, and .zstd outer-compressed Parquet files.
• Support for common Parquet primitive, nested, dictionary, temporal, decimal, and binary values.

Usage

parqcat cat <FILE>
parqcat head [-n <N>] <FILE>
parqcat tail [-n <N>] <FILE>
parqcat schema <FILE>

Examples: Author: Robert Kende

parqcat cat data.parquet
parqcat head -n 20 events.parquet
parqcat tail --lines 5 archive/events.parquet.zstd
parqcat schema data.parquet

When stdout is a terminal, row commands default to compact table output. When stdout is piped or redirected, row commands default to compact newline-delimited JSON.

parqcat cat events.parquet | jq -r '.event_type'
parqcat head -n 100 events.parquet | jq 'select(.status == "failed")'

Force JSONL with -j or --jsonl:

parqcat cat -j events.parquet

Force table output with -t or --table:

parqcat head -t -n 20 events.parquet

Output

JSONL output is one compact JSON object per logical row. It writes only data to stdout; diagnostics go to stderr.

Table output is intended for humans. It is compact, deterministic, uncolored, and may truncate very wide values. Date, time, and timestamp columns reserve enough width for common rendered values.

Schema output is a compact table:

name      type     nullable
--------  -------  --------
id        Int32    no
name      Utf8     yes

Nested schema fields are indented below their parent field.

Supported Input

parqcat targets mainstream, unencrypted local Parquet files.

Supported:

• local file paths
• common primitive types
• nullable fields
• dictionary encoding
• nested lists, structs, and maps
• common Parquet internal compression codecs
• outer .gz, .zst, and .zstd wrappers

Not supported in v1:

• stdin
• remote object stores
• multiple input files
• SQL, filtering, projection, or file mutation
• CSV output
• encrypted Parquet files requiring keys

Development

Run the local checks before opening a change:

cargo fmt --check
cargo test
cargo clippy --all-targets -- -D warnings

Tests generate compact Parquet fixtures in Rust; no Python, JVM, or external data engine is required.

License

Apache-2.0. See LICENSE.