Build and Test the Tool

Software Development with Claude Code

Dani Zysman

AI Curriculum Manager @ DataCamp

From Plan to Code

recraft: half: Developer at desk coding a CLI tool with terminal showing argparse output

Our specification from Lesson 4.1:
- --artist — artist name to query
- --format — JSON or CSV output
- --output — file path for results
Claude will generate:
- argparse argument parsing
- Database query via MCP
- File output logic

CLI Argument Patterns

Pattern	Example	When to Use
Required	`--artist "AC/DC"`	Must be provided
Optional	`--format json`	Has sensible default
Flag	`--verbose`	Boolean on/off
Positional	`reporter.py data.csv`	Primary input

Best practices:
- Add --help descriptions for every argument
- Use defaults to reduce required inputs
- Validate early, fail fast with clear messages

Building the Reporter

> Use @PLAN.md to build reporter.py

Done! Created reporter.py following PLAN.md

Core functions:
- get_db_connection()     Read-only SQLite with validation
- fetch_artist_stats()    4 SQL queries for aggregates
- format_json/csv()       Output formatting
- write_output()          File writing
- parse_args() / main()   CLI orchestration

Features:
- Case-insensitive artist lookup
- Per-album breakdown with revenue
- All error messages as specified
- --db flag for test injection

The Generated Code

import argparse, sqlite3, json, csv

def main():
    parser = argparse.ArgumentParser()
    ...
    args = parser.parse_args()

    conn = sqlite3.connect("chinook.db")
    # Query artist's albums and tracks...

    if args.format == "json":
        with open(args.output, "w") as f:
            json.dump(data, f, indent=2)

Test the Tool

flowchart: reporter.py, pytest, Verify Output

Tests verify:
- Valid artist returns data
- Output file is created
- JSON/CSV format is correct

Testing CLI Tools

Unit Tests
- Test individual functions
- Mock database calls
- Fast, isolated, repeatable
Integration Tests
- Run the actual CLI command
- Use real database
- Slower but more realistic

What to test:
- Happy path — valid artist, correct output
- Error cases — artist not found, bad format
- Edge cases — empty name, special characters
- File handling — write permissions, overwrite

Testing in Action

> Use @PLAN.md to build test_reporter.py

Created test_reporter.py: 31 tests across 7 classes

Coverage:
- Connection     Valid DB, missing file, invalid file
- Stats          Known/unknown artist, case-insensitive, multi-genre
- JSON           Format, indentation, round-trip serialization
- CSV            Two sections, pipe-delimited genres, row counts
- File I/O       Write, overwrite, unwritable path
- CLI args       Required args, choices validation
- Integration    JSON/CSV output, artist/DB not found, cleanup

Uses minimal seeded SQLite DB: independent of chinook.db

All 31 tests passing!

Let's Practice!

Software Development with Claude Code