Publish cw 0.1.0

[thorwhalen]

Cut the first real release, under the policy ADR-0005 defines.

Pre-flight

• python -m cw.testing parity green on Linux, macOS, Windows.
• Both acceptance-gate repos green; the falsifying gate green.
• pip install cw in a clean venv pulls zero third-party packages
  (pip list diffed against a bare venv, asserted in CI).
• import cw pulls no i2 (the regression test from the M0 issue).
• rg -n 'argh' --glob '!misc/docs/*' . returns nothing outside goldens metadata.
• Every module has a top-level docstring (pyproject.toml selects ruff D100; they are
  auto-extracted for generated docs).

Docs the release must carry

• README, progressive disclosure, easy examples first:

  cw.dispatch(f)                                       # 0. zero ceremony
  cw.dispatch([f, g], prog='x')                        # 1. subcommands
  cw.dispatch(COMMANDS, config={'cmd': {'p': {...}}})  # 2. this call's particulars
  dispatch = functools.partial(cw.dispatch, convention=cw.MODERN, prog='priv')
  dispatch(COMMANDS)                                   # 3. configs shrink to nothing

• The one sentence that carries the whole config story: config has the same shape as
  obj — keyed the way you address the thing on the command line. A single callable has no
  command word to type, so it has no command key to write.
• The ARGH vs MODERN table and what each switch does.
• The migration section, with the ADR-0005 dependency spelling verbatim.
• The v1 non-goals (ADR-0006 cut list), so nobody re-proposes them.
• Documented deliberate divergences from argh: coroutine → TypeError; the
  function-named-parameter case; priv's one renamed command.

Release

• Version 0.1.0; CHANGELOG written.
• Published to PyPI (a merge to the default branch on this fleet is the release —
  note it in the PR).
• The grammar-freeze test from ADR-0005 runs on every release.
• t/theremin bumped to the released cw and its holding argh declaration removed once
  it is on cw.compat or the real API.

Depends on

• The falsifying gate: priv --help and wads pack go --help byte-identical #25 — The falsifying gate: priv --help and wads pack go --help byte-identical
• ADR-0005: cw's release, pinning and rollback policy for the fleet #12 — ADR-0005: cw's release, pinning and rollback policy for the fleet
• Drop the argh dependency from cw's pyproject #2 — Drop the argh dependency from cw's pyproject
• Make import cw stdlib-only: move the i2 import into resource_inputs #3 — Make import cw stdlib-only: move the i2 import into resource_inputs
• cw.__main__ and the cw[completion] extra #24 — cw.__main__ and the cw[completion] extra

---

Research & rationale: the canonical spec and the full fleet audit live in the private research ledger (https://github.com/thorwhalen/priv/discussions/65). Effort: medium.

[thorwhalen]

Published: cw 0.1.0 — https://pypi.org/project/cw/0.1.0/

Landed by #33 (squash 54369a0, [bump minor] in the subject so isee bumped 0.0.15 → 0.1.0 rather than 0.0.16). The wads publish job ran green on master (33337281492).

Verified from PyPI, in a fresh venv, outside the source tree:

$ pip install cw==0.1.0
$ pip list
cw      0.1.0
pip     25.0.1          <- zero dependencies, as the README promises
$ python -m cw.testing parity
8 shapes / 137 cases: identical      exit 0

The gate ships inside the wheel (fixtures + eight goldens), so it runs from an installed cw and not only from a checkout.

dependencies = ["cw>=0.1,<0.2"] is the pin the migration notes give the fleet, and ADR-0005's rollback drill was re-executed against it and still reproduces byte for byte.