data-glance

Fast data profiling CLI for parquet and CSV files. Powered by ydata-profiling and Polars.

Installation

Install from PyPI:

# Run with uvx (cached)
uvx data-glance profile data.parquet

# Install globally
uv tool install data-glance

# Install with pip
pip install data-glance

Or run directly from GitHub:

# Run from GitHub (always latest)
uvx --from git+https://github.com/bswrundquist/data-glance data-glance profile data.parquet

Quick Start

# Profile a file
data-glance profile data.parquet

# Quick preview
data-glance head data.csv

# Check data quality
data-glance diagnose data.parquet

# View schema
data-glance schema data.parquet

Commands

Command	Description
profile	Generate HTML profile report
diagnose	Check data quality issues
head	Preview first N rows
tail	Preview last N rows
schema	Display column types
stats	Quick statistics
count	Count rows (fast)
columns	List column names
unique	Show unique values
filter	Filter data by expression
sample	Extract random sample
convert	Convert between formats
compare	Compare two files
validate	Validate data rules
info	File metadata
generate	Create test data

Profile Command

Basic Usage

data-glance profile data.parquet
data-glance profile data.csv --preset quick
data-glance profile data.parquet --preset full
data-glance profile huge.parquet --sample 10000

Column Filtering

data-glance profile data.csv --include "user_*,order_*"
data-glance profile data.csv --exclude "*_id,*_hash"

Null Handling

data-glance profile data.csv --nulls drop-cols
data-glance profile data.csv --drop-null-threshold 0.5
data-glance profile data.csv --drop-constant

Output Options

data-glance profile data.csv -o report.html
data-glance profile data.csv --json report.json
data-glance profile data.csv --no-browser
data-glance profile data.csv --dry-run

CSV Options

data-glance profile data.tsv --delimiter tab
data-glance profile data.csv --encoding latin-1
data-glance profile messy.csv --ignore-errors

Data Inspection

head / tail - Preview Data

data-glance head data.parquet --rows 20
data-glance tail data.csv --rows 10

schema - View Structure

data-glance schema data.parquet

stats - Quick Statistics

data-glance stats data.parquet

count - Row Count

data-glance count data.parquet          # Single file
data-glance count *.csv                 # Multiple files
data-glance count *.parquet --total     # Just the number

columns - List Columns

data-glance columns data.parquet
data-glance columns data.csv --one       # One per line (for piping)
data-glance columns data.csv --types     # With data types
data-glance columns data.csv --one | grep user  # Filter columns

unique - Value Distribution

data-glance unique data.csv status
data-glance unique data.parquet category --counts --sort
data-glance unique data.csv user_id --limit 50

info - File Metadata

data-glance info data.parquet

Shows file size, modification time, and for parquet: row count, columns, row groups.

Data Operations

filter - Query Data

# Filter by condition
data-glance filter data.csv "col('status') == 'active'"
data-glance filter data.parquet "col('age') > 30" -o filtered.parquet
data-glance filter data.csv "col('name').str.contains('test')" --limit 100

# Expression syntax (Polars)
col('column') == 'value'
col('column') > 100
col('column').is_in(['a', 'b'])
col('column').is_null()
col('column').str.contains('pattern')

sample - Extract Sample

data-glance sample data.parquet sample.parquet -n 1000
data-glance sample big.csv small.csv --fraction 0.1
data-glance sample data.parquet sample.csv  # Convert while sampling

convert - Format Conversion

data-glance convert data.csv data.parquet
data-glance convert data.parquet data.csv
data-glance convert data.csv data.parquet --compression zstd

Data Quality

diagnose - Quality Check

data-glance diagnose data.csv

Shows: schema, null percentages, quality issues, suggested fixes.

compare - Diff Files

data-glance compare data_v1.parquet data_v2.parquet

Shows: row/column differences, schema changes, null changes.

validate - Check Rules

# Check for nulls
data-glance validate data.csv --no-nulls "id,email"

# Check uniqueness
data-glance validate data.parquet --unique "id"

# Check row count
data-glance validate data.csv --min-rows 1000

# Check null percentage
data-glance validate data.csv --max-null-pct 0.1

# Check required columns
data-glance validate data.csv --required-cols "id,name,email"

# Combine rules
data-glance validate data.parquet \
    --unique "id" \
    --no-nulls "id,email" \
    --min-rows 1000

Returns exit code 1 if validation fails (useful in CI/CD).

Global Options

data-glance -q profile data.csv    # Quiet mode
data-glance -v profile data.csv    # Verbose mode

Test Data

data-glance generate test.parquet --rows 5000
data-glance generate test.csv --edge-cases --nulls 0.1

Presets

Preset	Speed	Detail	Use Case
quick	Fast	Minimal	Large files, quick checks
default	Medium	Standard	Most use cases
full	Slow	Detailed	Deep analysis

Tips

• Use --preset quick or --sample for large files
• Use diagnose before profile to understand data quality
• Use --dry-run to preview what will be profiled
• Use validate in CI/CD pipelines
• Use count --total for scripting
• Use columns --one to pipe to other tools
• Use filter to extract subsets before profiling

Development

# Clone and install
git clone https://github.com/bswrundquist/data-glance
cd data-glance
make install-dev

# Run tests
make test

# Lint and format
make lint
make format

# Build
make build

# Release
make release-patch  # 0.1.0 -> 0.1.1
make release-minor  # 0.1.0 -> 0.2.0
make release-major  # 0.1.0 -> 1.0.0