bash-cli-refactoring

name

bash-cli-refactoring

description

Refactor bash CLI tools systematically using BDD methodology. Covers helper extraction (debug-aware wrappers, interactive pickers, two-tier resolution), god function decomposition (orchestrator pattern, extraction by concern), bash-specific gotchas (set -e traps, local masking, ANSI-C quoting, pipefail), BDD testing for bash (pytest-bdd with subprocess runner, testing without external dependencies), standard command patterns (entry point architecture, two-path flow, argument parsing), and a 9-item anti-pattern catalog with detection signals and fix recipes. Synthesized from real refactoring of a 1500-line kubectl wrapper CLI. Trigger phrases: "refactor bash", "bash CLI", "shell script refactoring", "bash refactor", "refactor shell script", "bash best practices", "shell script quality", "bash anti-patterns", "bash god function", "bash helper extraction", "BDD bash", "test bash CLI", "bash cleanup", "bash set -e", "bash debugging", "refactoring CLI tool", "bash code quality", "shell script patterns", "bash code review"

Bash CLI Refactoring

Systematic methodology for refactoring bash CLI tools, synthesized from real refactoring of a 1500-line Kubernetes CLI wrapper.

Core Philosophy

1. Extract by duplication, decompose by concern — Helper extraction eliminates repeated code. God function decomposition separates concerns. Different triggers, different techniques.
2. Two-tier resolution — Pure resolve functions return values or empty. Ensure functions add interactive fallback. Never mix side effects with logic.
3. BDD before refactoring — Write tests for existing behavior first. Then refactor with confidence. pytest-bdd + subprocess = zero mocks.
4. Bash has landmines — local masks exit codes, set -o pipefail breaks grep, single-quoted ANSI codes are literal strings. Know the gotchas.
5. Fix anti-patterns, not symptoms — A 9-item catalog with detection signals. Grep for the pattern, apply the recipe.

Quick-Start: Common Problems

"This 200-line function does everything"

1. Identify concerns: arg parsing, resolution, data transform, output, tool logic
2. Name each concern as _module_action()
3. Extract one at a time, test after each extraction
4. Keep orchestrator thin (~50 lines) -> See references/god-function-decomposition.md

"Same 7 lines copy-pasted in every command"

1. Identify the repeated pattern (selection, resolution, cleanup)
2. Extract to a helper with edge case handling
3. Choose naming: _fzf_select (infra), select_resource (domain), ensure_pod (resolution)
4. Replace all occurrences, test each one -> See references/helper-extraction.md

"How do I test a CLI that needs a Kubernetes cluster?"

1. Set up .bdd/ directory with pytest-bdd
2. Write subprocess-based CLI runner with ANSI stripping
3. Test what does NOT need the cluster: help, validation, routing, error messages
4. Test command routing via expected output headers -> See references/bdd-for-bash.md

"I'm refactoring and things keep breaking silently"

1. Check for local var=$(cmd) — masks exit codes
2. Check ANSI quoting: use $'\033[1m' not '\033[1m'
3. Check set -e is in ALL sourced library files
4. Do NOT add set -o pipefail if using grep in pipelines -> See references/bash-gotchas.md

"Where do I even start with this messy CLI?"

1. Run the anti-pattern detection grep commands
2. Prioritize: blind stderr suppression > duplication > god functions
3. Add BDD tests for existing behavior first
4. Refactor one anti-pattern at a time -> See references/anti-pattern-catalog.md

Decision Trees

What to Refactor First

Signal	Priority	Action
2>/dev/null on data-fetching calls	High	Debug-aware wrapper
Same 5+ lines in 3+ commands	High	Helper extraction
Function >100 lines	Medium	God function decomposition
No set -e in library files	Medium	Add to all files
No tests	Medium	BDD test suite first
Hardcoded credentials	Low	Env var override
Dead code	Low	Remove + update routing

Extraction Type Selection

Pattern	Technique	Reference
Same code in multiple commands	Helper extraction	references/helper-extraction.md
One huge function, many concerns	God function decomposition	references/god-function-decomposition.md
Repeated interactive selection	Generic picker + domain layer	references/helper-extraction.md
Repeated resource resolution	Two-tier resolve/ensure	references/helper-extraction.md
Missing validation	Regex guard pattern	references/anti-pattern-catalog.md

Communication Between Bash Functions

Data Type	Mechanism	Example
Return values	stdout	local ns=$(resolve_namespace "$app")
Configuration	Global variables	_LOGS_FOLLOW="-f" set by orchestrator
Input parameters	Function arguments	_logs_stream_stern "$namespace" "${apps[@]}"
Error signaling	Return code + stderr	print_error "msg" >&2; return 1
Status messages	stderr	print_warning "msg" >&2

Reference Files

File	Contents
references/helper-extraction.md	Debug-aware wrappers, interactive pickers, two-tier resolution, cleanup patterns — when and how to extract reusable helpers
references/god-function-decomposition.md	Orchestrator pattern, extraction by concern, naming conventions, step-by-step decomposition procedure
references/bash-gotchas.md	set -e traps, local masking exit codes, pipefail vs grep, ANSI-C quoting, heredoc escaping, array handling, subshell scope
references/bdd-for-bash.md	pytest-bdd setup for CLIs, subprocess runner with ANSI stripping, testing without external dependencies, feature file categories
references/command-patterns.md	Entry point architecture, standard cmd_* pattern, two-path flow, argument parsing, library module organization, help system, debug mode
references/anti-pattern-catalog.md	9 anti-patterns with detection signals, fix recipes, before/after code, and grep-based detection commands