Repository navigation
Build: repo hygiene sweep (template leftovers, dead code) #59
Description
Question
Build: repo-hygiene sweep — delete template leftovers and dead code.
Root-level leftovers (all tracked, all from the python_template origin or superseded):
PROMPT.md,ralph.yml,@fix_plan.md— Ralph-agent scaffolding, content still says "python_template"@AGENT.md— duplicatesAGENTS.md(the oneCLAUDE.mdlinks to)WORKTREE_BACKEND_PLAN.md,WORKTREE_BACKEND_README.md— superseded bydocs/worktree-backend-spec.md; backend shippedcoverage.xml— stale (Jan 2026, predates the wholeworktree/package, reports 32.5%); confirm gitignored-and-untracked or remove from tracking
Dead code (zero references from src and tests): dl.get_git_branches (dl.py:563), dl.workspace_status (dl.py:907), config.save_config (config.py:104). Test-only-referenced candidates (BranchManager.checkout_branch, create_remote_branch_via_ssh — docstring says "legacy method from dl.py", dl.get_remote_head_sha) — delete function+tests together where the docstring or history confirms legacy; keep anything genuinely load-bearing.
Also the trivially dead duplicate assignment at workspace_clone.py:217/219 (is_new_workspace = False twice).
Deletion-only PR — no behavior change; CI green is the acceptance signal.
Note added by #53 (Rust rewrite: go, with a final Python release first): both halves of this ticket still land, for different reasons.
- The root-level leftovers are repo-level and outlive the rewrite — the repo persists and will host the
devlaunch-core/dlcargo workspace. - The dead code deletions now do double duty: they shrink the port surface. Deleting a function is an unambiguous "do not port this", where leaving it in makes that a judgment call during the port.
Activity
blooop commentedon Aug 7, 2026
Answer
Landed in #66 — deletion-only,
git diffadds zero lines.Leftovers:
PROMPT.md,ralph.yml,@fix_plan.md,@AGENT.md,WORKTREE_BACKEND_PLAN.md,WORKTREE_BACKEND_README.md, plus the Ralph state block in.gitignore.@AGENT.mdwas verbatim python_template boilerplate with nothingAGENTS.mdlacks.WORKTREE_BACKEND_README.mdwas wrong, not merely redundant — it documented~/.devlaunch/repos, a[worktree] enabledkey and a--backendflag, none of which exist.Dead code: all six functions confirmed dead by grep across src, tests, completions and docs, with no dynamic dispatch anywhere in the package (sole entry point is
dl = "devlaunch.dl:main"). Two needed history to judge:get_remote_head_shahad a real caller until0d64544rewrotecreate_remote_branchto a local push and left its tests orphaned;checkout_branchwas grepped across every revision ofbranch_manager.pyand was never called from production in any of them. 12 tests removed with them, 409 -> 397, and every branch of the surviving_git_ls_remotehelper is still covered.coverage.xml: already untracked and ignored (.gitignore:47). No action, as the ticket allowed.Beyond the list, one more deletion earned its place: the README "Backend Selection" section documented
dl --backend devpodandDEVLAUNCH_BACKEND, neither of which appears anywhere indevlaunch/. User-facing docs for flags that never shipped.Fog this sharpens
Two template leftovers survive, both needing a change rather than a deletion, so neither fit this PR:
pyproject.tomldeclarestomli_w, now unused aftersave_configwent. Dropping it forces a lock regeneration, which on current pixi also bumpspixi.lockv6 to v7 — that belongs in #62, where the lock is touched anyway.pyproject.toml'sdev/dev-restarttasks stillssh pythontemplate.devpod, which cannot resolve for this repo.scripts/update_from_template.shstill points atblooop/python_template.Also noted while grepping, unrelated to this ticket:
conda.recipe/recipe.yamlrun:deps list onlypython,iterfzf,devpod—tomlihas never been declared, so the conda package has been shipping without it.