:config and :conv were not in the prefix lists, so typing them
showed 'Plugin' mode but didn't route to the config/converter
providers. Also added :settings, :converter aliases.
Register ConfigProvider as built-in dynamic provider. Extend
execute_plugin_action to dispatch CONFIG:* commands via the
DynamicProvider::execute_action trait method.
Calculator and converter registered as built-in dynamic providers.
System registered as built-in static provider. All gated by config
toggles (calculator, converter, system — default true).
Foundation for built-in calculator, converter, and system providers.
DynamicProvider trait for per-keystroke providers. ProviderManager
iterates builtin_dynamic alongside native dynamic plugins in search.
The 5-second timer emitted 'changed' on the search entry in daemon
mode, triggering a full re-query that rebuilt the result list and
selected row 0 — jumping the user back to the top while browsing.
Widget refresh is a daemon-side concern; the UI gets updated data
on the next user-initiated search. Only keep the timer for local
(dmenu) mode where the UI owns the providers directly.
Calculator and converter results now get a 10k grouping bonus so all
their results stay together above websearch/filesearch. Previously
websearch (priority 9000) would interleave with converter results
(9000, 8999, 8998...) since they had the same base priority.
Websearch is a generic fallback — it always shows a result, so
highlighting it adds no signal. Filesearch returns fuzzy matches,
not auto-detected conversions. Only calc and conv produce direct
answers that deserve highlighting.
Highlighting:
- Dynamic plugin results (calculator, converter, websearch, filesearch)
get a subtle accent left-border + background tint when auto-detected
- Exact name matches (case-insensitive) are highlighted the same way
- Exact match on apps gets a higher score boost (50k) than other
providers (30k), so apps rank first when names match exactly
Shadow:
- Removed hardcoded box-shadow from all theme CSS files
- Added --owlry-shadow variable in base.css (defaults to none)
- Themes can opt into shadow via --owlry-shadow if desired
CSS class: .owlry-result-highlight on ResultRow
has_icon() returns true even for broken themes since it checks all
search paths. Instead, verify the theme directory actually exists
in the search path. Falls back to Adwaita only when the configured
theme is genuinely missing from disk.
If the configured icon theme (e.g. Sweet-Blue) doesn't exist on disk,
GTK falls back to hicolor which has almost no icons. Detect this by
probing for a standard icon, and set Adwaita as the theme — it's
guaranteed to exist as a GTK4 dependency.
This replaces the broken add_search_path("/usr/share/icons/Adwaita")
approach which doesn't work because search paths are scoped to the
active theme name, not the directory name.
The full results Vec was cloned into lazy_state.all_results and then
separately consumed for current_results. Now we slice for current_results
and move the original into lazy_state, avoiding one full Vec allocation
per query.
Search queries in daemon mode now run on a background thread via
DaemonHandle::query_async(). Results are posted back to the main
thread via glib::spawn_future_local + futures_channel::oneshot.
The GTK event loop is never blocked by IPC, eliminating perceived
input lag.
Local mode (dmenu) continues to use synchronous search since it
has no IPC overhead.
update_results('') was called inside MainWindow::new(), blocking the
window from appearing until the daemon responded. Move it to a
glib::idle_add_local_once callback scheduled after present() so the
window renders immediately.
scroll_to_row walked all GTK children via first_child/next_sibling
to count rows. The count is already available in LazyLoadState, so
use that directly. Eliminates O(n) widget traversal per arrow key.
get_score() called Utc::now() inside calculate_frecency() for every
item in the search loop. Added get_score_at() that accepts a pre-sampled
timestamp. Eliminates hundreds of unnecessary clock_gettime syscalls
per keystroke.
detect_terminal() was spawning up to 17 'which' subprocesses sequentially
on every startup. Replace with std::env::split_paths + is_file() check.
Eliminates 200-500ms of fork+exec overhead on cold cache.
