alexion/dotfiles
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

dotcli: add shortcut identifiers to dot kde save completion

Browse Source 
dot kde save's tab-completion only ever enumerated schema-backed
identifiers; kglobalshortcutsrc identifiers never showed up as
candidates despite being a fully supported mechanism. Enumerate them
live via kglobalaccel (allMainComponents/allActionsForComponent),
printed as a separate block after the schema-backed one, degrading
silently if the D-Bus session is unavailable.
This commit is contained in:
alexion
2026-07-05 22:02:40 -04:00
parent ec43fb2e14
commit 537724989a
2 changed files with 38 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

19
.config/dot/.claude/tasks/0009-kde-shortcut-completion.md
View File

@@ -36,9 +36,16 @@ here.
## Acceptance criteria
- [ ] `python3 kde.py complete` includes every currently-registered `kglobalshortcutsrc.<componentUnique>.<actionUnique>` identifier, sourced live via `allMainComponents`/`allActionsForComponent`
- [ ] Schema-backed identifiers print first, followed by shortcut identifiers, as two distinct blocks — not interleaved into one merged sorted list
- [ ] Shortcut identifiers print plain, with no friendly-name description text
- [ ] If the D-Bus walk raises `RuntimeError` or `OSError`, the shortcuts block is omitted, the schema block still prints normally, and nothing is written to stderr
- [ ] Freeform identifiers remain unlisted by `cmd_complete` (unchanged, confirmed not a regression)
- [ ] Verified manually against a live session — no new automated tests, consistent with the existing shortcuts-mechanism test carve-out (spec's testing decisions, 0005's Implementation Notes)
- [x] `python3 kde.py complete` includes every currently-registered `kglobalshortcutsrc.<componentUnique>.<actionUnique>` identifier, sourced live via `allMainComponents`/`allActionsForComponent`
- [x] Schema-backed identifiers print first, followed by shortcut identifiers, as two distinct blocks — not interleaved into one merged sorted list
- [x] Shortcut identifiers print plain, with no friendly-name description text
- [x] If the D-Bus walk raises `RuntimeError` or `OSError`, the shortcuts block is omitted, the schema block still prints normally, and nothing is written to stderr
- [x] Freeform identifiers remain unlisted by `cmd_complete` (unchanged, confirmed not a regression)
- [x] Verified manually against a live session — no new automated tests, consistent with the existing shortcuts-mechanism test carve-out (spec's testing decisions, 0005's Implementation Notes)
## Implementation Notes
- `iter_shortcut_identifiers` (new, `commands/kde/kde.py`) walks `allMainComponents()` then `allActionsForComponent()` per component, yielding `kglobalshortcutsrc.<componentUnique>.<actionUnique>`. `cmd_complete` wraps that walk in `sorted(set(...))` and appends it as a second print loop after the existing schema-backed one, inside a `try/except (RuntimeError, OSError)` that falls back to an empty list on any failure — so a missing `busctl` or an unreachable D-Bus session degrades completion instead of breaking it.
- Manually verified both paths: live run on this machine prints 278 shortcut identifiers after 322 schema-backed ones; with `busctl` removed from `PATH` (simulating a non-KDE/minimal shell), `cmd_complete` still exits 0, prints only the 322 schema identifiers, and writes nothing to stderr.
- `/review-uncommitted`'s Standards pass flagged two judgement-call smells: (1) the D-Bus call/unpack idiom for `allActionsForComponent` was duplicated between the new function and `_resolve_shortcut_action_id`; (2) the silent `except` swallow had no comment explaining why. Fixed both: extracted a shared `_actions_for_component(component_unique)` helper used by both call sites, and added a comment on the `try` explaining that fish invokes this on every TAB press in shells that may lack a live KDE session, so a broken shortcuts source must never cost the already-printed schema candidates. Re-ran the full test suite (101/101 pass) and both manual checks after the fix.
- No automated tests added, per the task's own acceptance criterion and the shortcuts mechanism's existing test carve-out (0005's Implementation Notes: a live D-Bus session isn't practically substitutable without disproportionate mock infrastructure).