2 Commits

Include renames

Author	SHA1	Message	Date
ChuckandClaude Opus 5.5	342e9164b8	fix: settings the display ignored, a memory leak, and the plugin card handler (#605) ... * fix(errors): stop affected_plugins growing without bound Each repeat of an error pattern appended every plugin in the time window to the pattern's list again, so a plugin failing in a loop grew the display process's memory without limit: 3,000 errors from three plugins reached 2.5 million entries. Keep the list unique. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(fonts): load a BDF font at its native size instead of PIL's default FreeType rejects any size but a BDF strike's own, and FontManager answered that with ImageFont.load_default() -- a different typeface -- so 5x7.bdf requested at 8 or 10px rendered as PIL's default font. Retry at the native strike, as element_style already does. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(web): plugin toggle failures no longer claim "operation in progress" Every exception in POST /plugins/toggle was mapped to PLUGIN_OPERATION_CONFLICT, so any failure told the user "A plugin operation is already in progress". Report the failure as what it is, and record the plugin id in the operation history for form posts too (it read a `data` variable that only the JSON path set). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(web): route plugin card clicks through handlePluginAction The document-level delegation checked `typeof handlePluginAction`, which is scoped inside the plugin-manager IIFE and so never visible to it. Every card click took a copied fallback that stopped propagation (the grid's own listener never ran), confirmed an uninstall twice, and sent Starlark app uninstalls to POST /plugins/uninstall instead of DELETE /starlark/apps/<id>. Expose the handler on window and delegate to it. Also run every test/js/unit suite under pytest: they need only node, but CI ran one of the eight. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(display): apply Rotation durations, WiFi messages and Vegas settings Three settings the web UI saves never reached the display: - Rotation & Durations: display.display_durations was never read. Every plugin inherits get_display_duration() and the plugin was asked first. A saved value now wins. The page shows unsaved screens blank with the plugin's own duration as a placeholder, and saving a blank removes the override, so one save no longer pins every screen. - WiFi status overlay: the controller looked for wifi_status.json one directory above the repo. Both sides now use wifi_manager.get_wifi_status_path(). The message is written by rename so the display never reads it half-written, and the resumed plugin redraws the whole panel afterwards. - Vegas: nothing called coordinator.update_config(), so saved Vegas settings never reached a running scroll. They are now queued when display.vegas_scroll changes, and applied while Vegas is stopped too, so a disable then re-enable works. The follower's scroll-speed default (75) now matches VegasModeConfig's (50). Also throttles Vegas's per-frame live-priority scan to 4Hz. It cost 139us per frame on a Pi 4 with two scoreboards (1.7% of a 125fps frame) and grows with each plugin. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix: keep affected_plugins order when serialized; guard non-Element targets ErrorPattern.to_dict() ran the now-ordered list through set(), so get_error_summary() listed plugins in an unstable order. The document-level card-action listener called event.target.closest() without checking the target is an Element. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>	2026-09-23 10:29:16 -04:00
3 people	eedf680a8c	perf: display pipeline optimizations — caching, logging, scroll, text width (#358) ... * docs(core): add module and class docstrings to the 5 undocumented core files Fills the only significant documentation gaps found during a codebase audit. All other core files (plugin_system/, logging_config.py, etc.) already have complete module, class, and function docstrings. Files changed (documentation only — zero logic changes): display_controller.py — module doc explaining orchestration role; DisplayController class doc; main() docstring display_manager.py — module doc; DisplayManager class doc with typical-usage snippet for plugin authors cache_manager.py — module doc explaining two-tier cache; DateTimeEncoder class and default() docstrings config_manager.py — module doc explaining file ownership and atomic-write / hot-reload design; ConfigManager class doc; get_config_path() / get_secrets_path() docstrings font_manager.py — module doc (class docstring already existed) Also noted (but not changed to avoid behaviour risk): display_manager.py and font_manager.py use logging.getLogger() directly instead of the project's get_logger() wrapper. display_manager.py also calls setLevel(logging.INFO) immediately after, which would be lost if switched to get_logger(). Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * perf(display_controller): three targeted hot-path optimizations Opt 1 — cache inspect.signature() per plugin_id inspect.signature() is called at most once per plugin_id; the result (bool: accepts display_mode param) is stored in _plugin_accepts_display_mode and reused on every subsequent display() call. Eliminates all reflection from the display path at runtime. Cache is invalidated when a plugin instance is replaced in plugin_modes. Opt 2 — pre-cache config values that never change during a run _normal_brightness and _scroll_speed are resolved from the config dict once in __init__ and stored as typed instance attributes. - Removes 2+ chained dict.get() calls with temporary {} default objects from the 60fps follower loop (vegas_speed) and from every _check_dim_schedule call. - current_brightness init now uses _normal_brightness directly. Opt 3 — schedule minute-gate: re-evaluate at most once per clock minute _check_schedule and _check_dim_schedule both performed pytz.timezone(), datetime.now(), strftime(), and datetime.strptime() on every outer loop call. Schedule state can only change on a minute boundary, so both methods now: - lazily build self._tz once and reuse it - skip the full re-parse when (hour, minute) matches the last evaluated key (_schedule_checked_minute / _dim_checked_minute) - _check_dim_schedule stores its return value in _cached_target_brightness for the gate fast-path Tests: 23 new tests in test_display_controller_optimizations.py covering all three optimisation invariants (cache init, hit, miss, invalidation). All pre-existing test failures are unrelated to these changes (confirmed by stash+run on main). Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * fix: resolve 22 pre-existing test failures across 6 groups Test fixes (tests were asserting wrong values or patching wrong objects): basketball scoreboard — update display mode assertions from generic basketball_live/recent/upcoming to league-prefixed nba_live/recent/upcoming to match the current manifest display_controller schedule — inject schedule directly into controller.config (what _check_schedule actually reads) instead of patching config_service.get_config; also reset minute-gate state so the optimisation doesn't interfere git cache (3 tests) — production code refactored from 4 subprocess calls (rev-parse + abbrev-ref + config + log) to a single git log --format=%H%n%cI that returns SHA and date on two lines; update fake and call-count assertions web_api dotted-key (2 tests) — validate_config_against_schema mock returned [] (empty list); endpoint unpacks as is_valid, errors = ... causing ValueError; fix: return_value = (True, []) state reconciliation — test expected save_config() to be called with enabled=False (treating state as source of truth); production code correctly syncs the state manager to match config instead; fix: assert set_plugin_enabled('plugin1', True) Production fixes (production code had bugs or missing features): reconcile endpoint — add force parameter parsing with isinstance(payload, dict) guard for non-object bodies; route through _coerce_to_bool; pass force= to reconcile_state() (8 tests) transactional uninstall — add _do_transactional_uninstall() helper that: (1) snapshots config before touching anything; (2) calls cleanup_plugin_config first and aborts on failure; (3) rolls back config + reloads plugin on uninstall failure; (4) propagates unexpected errors (TypeError etc.) instead of swallowing them (6 tests) fix_array_structures / ensure_array_defaults — recursive calls passed the full ancestor prefix into calls where config_dict is already navigated, so dotted property keys like eng.1 caused parent_parts.split('.') to mis-navigate; fix: drop prefix on recursive calls; also add _fix_none_arrays pass after merge_with_defaults so None arrays in JSON requests are replaced with schema defaults (2 tests) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * perf: four targeted optimizations across the display pipeline Opt 1 — cache data-fetch interval per plugin (plugin_manager.py) _get_plugin_update_interval fell back to config_manager.get_config() (a full dict copy) when the manifest lacked an interval. Called for every plugin on every run_scheduled_updates() tick (~30fps), this was up to 300 dict copies/sec with 10 plugins. Fix: cache the resolved interval in _update_interval_cache[plugin_id] on first call; return the cached value on subsequent calls. Cache is cleared on load_plugin and unload_plugin. Opt 2 — demote noisy per-cycle INFO logs to DEBUG (display_controller.py) Four logger.info calls fired on every mode cycle or every FPS-loop entry, including one that called list(self.plugin_modes.keys()) unconditionally (allocating a list every outer loop iteration). - "Processing mode" kept at INFO but reformatted to %s (lazy) and the plugin_modes key dump moved to logger.debug - "Attempting/Got cycle duration" → logger.debug - "Entering high/normal FPS loop" → logger.debug Mode name at INFO is preserved for black-screen troubleshooting. Opt 3 — use Image.frombytes instead of Image.fromarray in scroll hot path (scroll_helper.py) Image.fromarray on a non-contiguous numpy slice goes through numpy's array protocol. Image.frombytes on an ascontiguousarray is ~50% faster for the 128×32 display-sized frames used here. Applied to all three code paths in _get_visible_portion_integer (simple, wrap- around, and edge cases). Opt 5 — cache get_text_width per (text, font) pair (display_manager.py) FreeType fonts require one load_char() per character per call; PIL fonts call textbbox(). Plugins that measure the same text every frame (centering a score, ticker label, etc.) were re-measuring from scratch on every display() call. Fix: _text_width_cache[(text, id(font))] stores results; cleared automatically in _load_fonts() when fonts are reloaded so stale entries from old font objects are evicted. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * fix(scroll_helper): fix edge-case bug exposed by frombytes switch The previous commit replaced Image.fromarray with Image.frombytes in _get_visible_portion_integer. This surfaced a pre-existing bug in the edge-case branch (start_x >= image_width): the original code returned a wrong-size Image silently (Image.fromarray accepts a too-short array); Image.frombytes raises ValueError instead. Fix: consolidate all non-simple-slice paths to use the pre-allocated _frame_buffer, which is always display_width wide. The edge-case path now clamps the source to available columns and zero-pads the remainder. Verified pixel-identical output vs original across: - normal case (single slice, multiple start positions) - wrap-around case (tail + head of scroll image) - edge case (start_x at or past image end) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * fix: address CodeRabbit review comments on PR #358 1. display_controller — add _refresh_config_cache() and wire it into a controller-level ConfigService subscriber so _normal_brightness, _scroll_speed, _tz, and the schedule minute-gates stay in sync with the live config after a hot-reload (was using stale init-time values) 2. display_manager — narrow bare except Exception in get_text_width to (AttributeError, TypeError, ValueError, OSError) to avoid masking unrelated bugs 3. plugin_manager — import ConfigError; narrow except Exception in _get_plugin_update_interval to (ConfigError, OSError, ValueError, TypeError) — fixes Ruff BLE001 4. api_v3 _do_transactional_uninstall — snapshot and restore secrets in addition to main config; previously a failed uninstall_plugin() would leave the plugin's secrets deleted even after rollback 5. api_v3 uninstall endpoint — queued path now delegates to _do_transactional_uninstall instead of using the old ad-hoc flow, so rollback/state behaviour is consistent whether or not an operation queue is in use Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * fix(display_controller): move _plugin_accepts_display_mode init before plugin loop Codacy HIGH: 'access to member before its definition' — the dict was initialised at line 441 but accessed at line 364 inside the plugin- loading loop, both within __init__. Fix: move the initialisation to line 194 (before the plugin loop), remove the now-unnecessary hasattr guard, and delete the duplicate initialisation that remained at the old location. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> --------- Co-authored-by: Chuck <chuck@example.com> Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>	2026-06-01 11:58:21 -04:00