Skip to content

TODO.md

Latest commit

 

History

History
105 lines (75 loc) · 4.06 KB

TODO.md

File metadata and controls

105 lines (75 loc) · 4.06 KB
Raw

sumd — TODO

Auto-derived from metrics — CC hotspots from ast analysis, coverage gaps from coverage.json. Last updated: 2026-04-19 (v0.3.3) Manual items in P0 and Manual backlog sections survive regeneration.

P0 — Stable release blockers (manual, v0.2.0)

  • Raise test coverage to ≥ 50% before tagging 0.2.0 stable
  • Add sumd/__main__.py to support python -m sumd
  • Real testql CLI scenario file (sumd-cli.testql.toon.yaml) instead of smoke placeholder
  • Validate rc1 on CI before promoting to 0.2.0 stable

CC hotspots (auto — functions with CC ≥ 10)

CC File Function Action
15 sumd/renderer.py _parse_calls_hubs Split hub detection from formatting
14 sumd/renderer.py _collect_pkg_sources Extract _scan_sources() helper
13 sumd/parser.py validate_codeblocks Split: _check_fence_format(), _check_markpact_kind()
12 sumd/renderer.py _render_test_contracts Split by contract type (smoke / crud / api)
12 sumd/extractor.py extract_openapi Extract _parse_paths(), _build_endpoint()
12 sumd/extractor.py extract_dockerfile Separate stage detection from content rendering
10 sumd/renderer.py _render_extras Extract per-format renderers
10 sumd/renderer.py _render_deployment_docker Split multi-compose logic
10 sumd/extractor.py generate_map_toon Split: _collect_map_files(), _render_map_detail()
10 sumd/extractor.py extract_taskfile Extract task-type classifiers
10 sumd/extractor.py _collect_map_files Already a sub-function — revisit call sites
10 sumd/extractor.py _analyse_py_top_classes Split parsing from formatting
10 sumd/cli.py _scan_one_project Extract profile dispatch

Target: CC ≤ 10 for all functions. Current average (production code): ~4.6.

Coverage gaps (auto — modules < 60% covered)

Coverage Missing lines Module
0% 103 sumd/mcp_server.py
0% 486 sumd/cli.py
14% 91 sumd/toon_parser.py
36% 303 sumd/extractor.py
36% 417 sumd/renderer.py

Action plan

  • cli.py (0%) — add tests/test_dogfood.py integration tests (see P1) + CLI unit tests via click.testing.CliRunner
  • mcp_server.py (0%) — add tests/test_mcp_server.py; at minimum test tool registration and that main() doesn't crash
  • toon_parser.py (14%) — add tests/test_toon_parser.py covering each _parse_toon_block_* function
  • extractor.py (36%) — extend test_extractor.py with edge-case fixtures (missing files, empty files, malformed YAML)
  • renderer.py (36%) — add tests/test_renderer.py parametrised over section fixtures

P1 — Integration / dogfood tests (high ROI)

Add tests/test_dogfood.py — tests that scan the sumd project itself:

def test_sumd_scans_itself():
    # sumd scan . --fix --profile rich must exit 0 and produce valid SUMD.md
    ...

def test_sumd_scans_all_profiles():
    for profile in ['minimal', 'light', 'rich']:
        # each profile must exit 0
        ...

def test_sumr_generates_sumr_md():
    # sumr . must produce SUMR.md with expected sections
    ...

These catch full-pipeline regressions not visible via unit tests.

P2 — mcp_server.py tests

sumd/mcp_server.py — 103 lines, 0% coverage, no tests anywhere.

# tests/test_mcp_server.py
def test_mcp_tools_registered():
    ...

def test_mcp_main_no_crash():
    ...

P3 — Autonomization backlog (manual)

These items improve autonomy of the dev process; not blocking release.

  • Coverage ratchet: sumd scan . --update-gates — only raises coverage_min, never lowers it
  • Pre-commit hook: task install:hooks → .git/hooks/pre-commit runs sumd scan . --fix --profile light on every commit touching .py files
  • TODO auto-regeneration: sumd scan . --todo generates this file from live metrics (CC + coverage + duplication)