Release v0.11.0

🧱 TopMark 0.11.0 — Policy & STDIN Architecture Hardening

This release delivers a set of internal architectural improvements that strengthen
policy correctness, STDIN handling, and CLI/API parity.

➡️ User-facing behavior remains compatible with the 0.10.x series, but there is an
intentional internal breaking change for integrators or tests that rely on TopMark
pipeline internals.

---

⚠️ Internal Breaking Change

Mandatory PolicyRegistry at Pipeline Bootstrap

• ProcessingContext.bootstrap() now requires a PolicyRegistry argument
• Internal callers must:
  1. Resolve a Config
  2. Construct a PolicyRegistry from it
  3. Pass the registry explicitly when bootstrapping the context
• This change:
  • Eliminates ad-hoc and repeated policy merging
  • Guarantees deterministic effective policy resolution
  • Reduces per-file and per-step overhead

Note: This affects internal and test code only.
The public API (topmark.api.check, topmark.api.strip) and all CLI commands remain
source-compatible.

---

What Changed

Policy Evaluation

• Introduced PolicyRegistry to precompute effective policies (global + per-file-type) once per run
• Centralized policy access via ctx.get_effective_policy()
• Removed per-context caching and None-guarded fallbacks
• Ensured strip eligibility and guidance are based on:
  • the strip axis
  • the effective policy
  • independent of comparison status

Outcome Bucketing

• Refactored bucketing into a precedence-ordered classifier (first match wins)
• Added stable debug tags (bucket[...]) for easier diagnosis
• Normalized bucket labels and removed duplication across header / comparison / strip reasons

CLI / Config / Runtime Alignment

• Unified write-mode semantics across CLI, config, and writer:
  • CLI write_mode now maps cleanly to OutputTarget + FileWriteStrategy
  • STDIN content mode forces STDOUT output and clears file write strategies
  • Explicit diagnostics are emitted when config or CLI values are overridden due to STDIN
• Renamed stdin → stdin_mode throughout CLI, config, API, and runtime
• Propagated stdin_filename so STDIN-backed runs participate correctly in:
  • config discovery
  • policy evaluation
  • file resolution
• Centralized all STDIN-related normalization in MutableConfig.sanitize(), enforcing invariants
  before freezing and removing duplicated CLI-only logic

Enums & Config Parsing

• Introduced a stable, machine-keyed enum pattern (KeyedStrEnum) for config-facing enums:
  • OutputTarget
  • FileWriteStrategy
• Decoupled machine identifiers from human-readable labels
• Removed brittle string / Literal-based parsing
• Standardized .parse() helpers for both config and CLI normalization

Writer Behavior

• Updated WriterStep.may_proceed() to consistently respect:
  • apply intent
  • output target
  • STDIN mode
• Ensured writer eligibility rules behave identically across CLI and API runs

API Runtime Parity

• Aligned API execution with CLI behavior by applying the same:
  • config discovery
  • normalization rules
  • policy overlays

---

Fixes

CLI Guidance Correctness

• Made check and strip per-file guidance both policy-aware and feasibility-aware
• Prevented misleading “run --apply …” suggestions when changes are blocked by:
  • policy
  • feasibility
  • empty-file rules
  • strip-only scenarios

---

Tests

• Updated pipeline and API tests to bootstrap contexts via PolicyRegistry
• Added shared helpers to keep test setup DRY and consistent across:
  • pipeline tests
  • API tests
  • CLI tests

---

ℹ️ Notes

• There are no user-facing breaking changes relative to 0.10.x
• The public API surface remains stable
• Internal consumers, extensions, and test harnesses must be updated to construct and pass a
  PolicyRegistry when bootstrapping pipeline contexts

---

This release makes policy behavior more predictable, easier to reason about, and safer to extend 🚀

Full Changelog: v0.10.1...v0.11.0