django-wireview 0.2.1__tar.gz

django_wireview-0.2.1/.gitignore

@@ -0,0 +1,40 @@
+.DS_Store
+*.egg-info/
+*.eggs/
+*.log
+*.map
+*.min.js
+*.sqlite3
+.envrc
+dist/
+__pycache__/
+.pytest_cache/
+htmlcov/
+node_modules/*
+.coverage
+site/
+.vscode/
+pyrightconfig.json
+*.sublime-*
+
+# Collected static files
+tests/static/
+.idea/
+
+# IDE LSP metadata cache
+.wireview/
+
+# Auto-generated type stubs
+*.pyi
+
+# Claude Code personal overrides
+.claude/settings.local.json
+
+# Benchmarks (results are versioned; the sqlite scratch DB is not)
+bench/.data/
+
+# Redis snapshot, if a redis-server ever writes one into the working directory
+dump.rdb
+
+# Claude Code worktrees
+.claude/worktrees/

django_wireview-0.2.1/CHANGELOG.md

@@ -0,0 +1,125 @@
+# Changelog
+
+All notable changes to django-wireview are documented here. The format follows
+[Keep a Changelog](https://keepachangelog.com/en/1.1.0/). Git tags `v*` are the
+version source of truth; `pyproject.toml` is bumped in the release commit.
+Feature-level history is tracked by GAP number in `docs/FEATURE-GAP.md`.
+
+The django-reactor era changelog (2.x) is preserved in
+[docs/legacy/CHANGELOG-reactor.md](./docs/legacy/CHANGELOG-reactor.md).
+
+## [Unreleased]
+
+## [0.2.1] - 2026-09-08
+
+### Added
+
+- Published on PyPI: `pip install django-wireview` works. The release workflow now builds through
+  `make ci-build` (which fails if the wheel is missing `wireview.min.js`), checks that the tag
+  matches `pyproject.toml`, and publishes the sdist and wheel with PyPI trusted publishing (OIDC,
+  environment `pypi`). No API token is stored anywhere. Releases up to v0.2.0 were GitHub-only
+
+## [0.2.0] - 2026-09-08
+
+The release that makes the SQLite-plus-Windows deployment premise real: a channel
+layer that needs no Redis, partial HTML diffs that actually fire, a transport seam,
+and reproducible benchmarks on macOS, Linux and Windows to back the claims.
+
+### Added
+
+- `make bench ARGS="--layer redis"` benchmarks channels_redis next to channels-nats, so the channel-layer
+  choice can be made on measurements (`docs/design/transport-abstraction.md` §5-3). The benchmark starts
+  the broker itself on a free port; `REDIS_URL` / `NATS_URL` point it at an existing one instead
+- Windows benchmark lane and results: `make bench ARGS="--server uvicorn"` (also `uvicorn-wsproto`) picks the
+  ASGI server, `bench/windows/` runs the same benchmark inside a Parallels Windows guest, and
+  `bench/results/win11-parlab-*.json` hold the numbers next to the macOS control group `a993181-*.json`.
+  daphne dies at about 500 connections per process on Windows (`select()` limit); a single-process uvicorn
+  does not. `docs/DEPLOYMENT.md` gained the Windows single-server recipe (uvicorn × N behind Caddy,
+  channels-nats, SQLite WAL) and `docs/design/transport-abstraction.md` §5-2 the measurements
+- NATS channel layer support in the test project and the benchmark: `tests/testproj/settings_nats.py`,
+  `tests/test_nats_layer.py` (cross-process broadcast reaches a consumer through channels-nats),
+  `make bench ARGS="--layer nats --processes N"` with a broadcast fan-out measurement
+- `bench/`: reproducible benchmarks (`make bench`) for render payload sizes, per-event cost,
+  memory and WebSocket connection density, plus `make bench-compare BASE=<ref>` to benchmark a
+  past commit in a throwaway worktree and print a side-by-side table
+- Comprehensions and blocks: `{% for %}` output is one dynamic slot holding the item template's
+  statics once and per-item dynamics, and `{% if %}` output is a nested block with its own
+  statics. Adding, removing or changing items and switching branches now produce partial diffs
+  instead of full renders (GAP-025). Client side lives in `wireview/static/wireview/rendered.mjs`
+  and is unit-tested with `npm test`
+- `wireview.core.transport`: `Outbound` and `Broker` interfaces with Channels implementations.
+  Every channel-layer call now goes through them, so another connection layer only has to
+  implement the two interfaces (GAP-026)
+- `Rendered.to_dict()` / `from_dict()` so render snapshots can live outside the process
+- `docs/implementation/wire-protocol.md` (message shapes in all four directions) and
+  `docs/design/transport-abstraction.md` (measurements, options, remaining steps)
+- `on_mount` hooks and `attach_hook()` / `detach_hook()` for lifecycle interception (GAP-021)
+- `.claude/settings.json` with a permission allowlist and a PostToolUse hook that runs `ruff format` on edited Python files
+
+### Changed
+
+- NATS is now the layer this project targets. `make test-e2e` runs the browser suite on
+  channels-nats and `tests/e2e.sh` starts a throwaway nats-server for it, so no broker has to be
+  running first;
+  `WIREVIEW_TEST_LAYER` (memory by default, nats or redis) picks the layer for the
+  test project, and README and `docs/DEPLOYMENT.md` recommend it first. CI runs the E2E suite on
+  NATS too, and channels-nats is a dev dependency now that it is on PyPI. channels_redis stays
+  fully supported
+- The event transpiler cache is a small pure-Python LRU instead of `lru-dict`, so wireview installs
+  without a C compiler on platforms that have no `lru-dict` wheel (Windows ARM64). The dependency is gone
+- `WireviewMeta` accepts a `broker=` argument; `channel_layer=` still works and builds a
+  `ChannelsBroker`. Tests that patched `get_channel_layer` should patch
+  `wireview.core.transport.get_channel_layer` instead
+- `data-state` now carries a compact, zlib-compressed `Signer.sign_object` payload
+  (`wireview.core.state`). The legacy `Signer().sign(json)` format is still accepted on join
+- Rewrote `CLAUDE.md` as a compact agent guide (repository map, commands, conventions, gotchas) that links to `docs/` instead of duplicating API docs
+- `make lint` now runs `ruff format --check` and djlint, and CI's lint job reuses it
+- Release workflow attaches only the wheel (sdist removed)
+- Package version unified to the latest tag (`pyproject.toml`, `package.json`)
+- README and tutorials state the actual Python requirement (>=3.12)
+
+### Fixed
+
+- The published wheel and sdist now contain `wireview.min.js`. hatchling honours `.gitignore`,
+  which ignores `*.min.js` as a build artifact, so every release up to v0.1.1 shipped a package
+  whose `{% wireview_header %}` pointed at a file that was not there and no JavaScript loaded.
+  `make ci-build` now fails if the wheel is missing it
+- HTTP renders no longer leak diff markers into attributes (`value="<!--$0-->…"`); markers are
+  stripped for non-live renders
+- Variables inside `{% if %}` branches were never marked, so any change inside a conditional
+  was a full render; branches are now wrapped like the rest of the template
+- Empty variable output is now marked too, so a value toggling between "" and text no longer
+  shifts every following index into a full render
+- Partial HTML diffs never fired for live components: `{% tag_header %}` embedded the freshly
+  signed `data-state` in the static parts, so the fingerprint changed on every render and every
+  event shipped a full render. The signed state is now a dynamic part (GAP-024)
+- `Component._lifecycle_hooks` type annotation so `pyright` passes
+- djlint formatting of the livecomp and slots test templates
+- Duplicate test component class names (`ChildComponent`, `SimpleComponent`) that triggered registration warnings on every test run
+
+### Removed
+
+- Stale `MANIFEST.in` (referenced `reactor/` paths; hatchling is the build backend) and placeholder `docs/index.md`
+
+## [0.1.1] - 2025-12-09
+
+### Added
+
+- Wheel build in the release workflow
+
+## [0.1.0] - 2025-12-09
+
+First tagged release of django-wireview, the Pydantic v2 / Django Channels
+successor to django-reactor. Highlights over reactor: Streams, Presence, chunked
+and external file uploads, AsyncResult and `assign_async`, `JS()` client
+commands, JavaScript Hooks, Slots, Function Components, LiveComponent,
+`temporary_assigns`, `params_changed`, flash messages, `push_title`, form
+auto-recovery, viewport bindings, optimistic UI attributes, type stub
+generation, and `mount()` testing utilities. See `docs/FEATURE-GAP.md` for the
+Phoenix LiveView parity table.
+
+[Unreleased]: https://github.com/itda-work/django-wireview/compare/v0.2.1...HEAD
+[0.2.1]: https://github.com/itda-work/django-wireview/compare/v0.2.0...v0.2.1
+[0.2.0]: https://github.com/itda-work/django-wireview/compare/v0.1.1...v0.2.0
+[0.1.1]: https://github.com/itda-work/django-wireview/compare/v0.1.0...v0.1.1
+[0.1.0]: https://github.com/itda-work/django-wireview/releases/tag/v0.1.0