go-ansible/CLAUDE.md

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

Project Overview

core/go-ansible is a pure Go Ansible playbook engine. It parses YAML playbooks, inventories, and roles, then executes tasks on remote hosts via SSH. 41 module handler implementations (plus 3 community modules), Jinja2-compatible templating, privilege escalation (become), and event-driven callbacks. This is a library — there is no standalone binary. The CLI integration lives in cmd/ansible/ and is compiled as part of the core CLI binary.

Build & Test

go build ./...                                    # verify compilation
go test ./...                                     # run all tests
go test -race ./...                               # with race detection
go test -run TestParsePlaybook_Good_SimplePlay     # single test
go test -v ./...                                  # verbose output
go work sync                                      # if in a Go workspace

No SSH access is needed for tests — the suite uses a mock SSH client (mock_ssh_test.go).

Architecture

Single flat ansible package with four layers:

Playbook YAML ──► Parser ──► []Play ──► Executor ──► Module Handlers ──► SSH Client ──► Remote Host
                               │                         │
Inventory YAML ──► Parser ──► Inventory        Callbacks (OnPlayStart, OnTaskEnd, ...)

• types.go — Core structs (Playbook, Play, Task, TaskResult, Inventory, Host, Facts) and KnownModules registry (80 entries: both FQCN ansible.builtin.* and short forms).
• parser.go — YAML parsing for playbooks, inventories, tasks, and roles. Custom Task.UnmarshalYAML scans map keys against KnownModules to extract the module name and args (since Ansible embeds the module name as a YAML key, not a fixed field). Free-form syntax (shell: echo hello) is stored as Args["_raw_params"]. Iterator variants (ParsePlaybookIter, ParseTasksIter, etc.) return iter.Seq values.
• executor.go — Orchestration engine: host resolution from inventory, play execution order (gather facts → pre_tasks → roles → tasks → post_tasks → notified handlers), when: condition evaluation, {{ }} Jinja2-style templating with filter support, loop execution, block/rescue/always, handler notification.
• modules.go — 41 module handler implementations dispatched via a switch on the normalised module name. Each handler extracts args via getStringArg/getBoolArg, constructs shell commands, runs them via SSH, and returns a TaskResult.
• ssh.go — SSH client with lazy connection, auth chain (key file → default keys → password), known_hosts verification, become/sudo wrapping, file transfer via cat > piped through stdin.
• cmd/ansible/ — CLI command registration via core/cli. Provides ansible <playbook> and ansible test <host> subcommands with flags for inventory, limit, tags, extra-vars, verbosity, and check mode.

Adding a New Module

1. Add both FQCN and short form to KnownModules in types.go
2. Add the dispatch case in executeModule switch in modules.go
3. Implement module{Name}(ctx, client, args) method on Executor in modules.go
4. Write tests in the appropriate modules_*_test.go file using mock SSH infrastructure

If adding new YAML keys to Task, update the knownKeys map in Task.UnmarshalYAML (parser.go) to prevent them being mistaken for module names.

Test Organisation

File	Coverage
types_test.go	YAML unmarshalling for Task, RoleRef, Inventory, Facts
parser_test.go	Playbook, inventory, and task file parsing
executor_test.go	Executor lifecycle, conditions, templating, loops, tags
ssh_test.go	SSH client construction and defaults
modules_cmd_test.go	Command modules: shell, command, raw, script
modules_file_test.go	File modules: copy, template, file, lineinfile, stat, slurp, fetch, get_url
modules_svc_test.go	Service modules: service, systemd, user, group
modules_infra_test.go	Infrastructure modules: apt, pip, git, unarchive, ufw, docker_compose
modules_adv_test.go	Advanced modules: debug, fail, assert, set_fact, pause, wait_for, uri, blockinfile, cron, hostname, sysctl, reboot

Coding Standards

• UK English in comments and documentation (colour, organisation, centre)
• Test naming: _Good (happy path), _Bad (expected errors), _Ugly (edge cases/panics)
• Use coreerr.E(scope, message, err) from go-log for all errors in production code (never fmt.Errorf)
• Tests use testify/assert (soft) and testify/require (hard)
• Licence: EUPL-1.2