feat: Command Line Interface

[fdosani]

Summary

DataComPy is currently a Python library only — users must write a script to compare two datasets. A command-line interface would make it immediately accessible for ad-hoc data validation, CI pipelines, and use by non-Python practitioners without any boilerplate.

Motivation

CICD pipelines often run as shell tasks (Airflow BashOperator, GitHub Actions, GitLab CI) where Python entry points are awkward. A native CLI that exits non-zero on mismatch and emits machine-readable JSON is the highest-leverage single CICD improvement DataComPy could add.

Other common friction points:

• Ops / data engineers want to diff two CSV or Parquet files without opening a Python REPL
• Analysts want a one-liner to spot-check a file diff without writing a throw-away script every time
• Automation wants to shell out to datacompy compare ... and capture a structured result

Proposed interface

# Basic comparison
datacompy compare \
    --left s3://bucket/before.parquet \
    --right s3://bucket/after.parquet \
    --on id,date \
    --backend pandas \
    --max-unequal-rows 0 \
    --json
# exit 0 on match, 1 on mismatch, 2 on error

# Local files, HTML report written to file
datacompy compare \
    --left df1.csv \
    --right df2.csv \
    --on id \
    --output report.html

# Tolerances
datacompy compare \
    --left before.parquet \
    --right after.parquet \
    --on account_id \
    --abs-tol 0.01 \
    --rel-tol 0.001

# Ignore whitespace and case
datacompy compare \
    --left df1.csv \
    --right df2.csv \
    --on id \
    --ignore-spaces \
    --ignore-case

# Control report verbosity
datacompy compare \
    --left df1.csv \
    --right df2.csv \
    --on id \
    --sample-count 20 \
    --column-count 5