thwap/steeper

Watch

Fork 0

feat(cache): make issue/PR loading async to prevent UI freeze #368

Merged

fuzzy merged 5 commits from feat/356-paginate-issues-prs into main

2026-07-08 11:29:03 +00:00

automation commented

2026-07-08 10:36:08 +00:00

Owner

What

IssueView.Load and PRView.Load/LoadAll already fetch all pages (loop until exhaustion), but run synchronously blocking the UI. This moves them to goroutine + QueueUpdateDraw pattern, matching the home view approach.

Async load prevents UI freeze on large issue/PR lists
Errors stored on struct, accessible via LastError()
Mutex protects concurrent access to shared state
Load() always returns nil (callers already handle this)

Why

Multi-page API fetches (up to 10 pages × 50 items) can take seconds. Blocking the tview event loop during fetch makes the terminal unresponsive.

Closes #356

## What IssueView.Load and PRView.Load/LoadAll already fetch all pages (loop until exhaustion), but run synchronously blocking the UI. This moves them to goroutine + QueueUpdateDraw pattern, matching the home view approach. - Async load prevents UI freeze on large issue/PR lists - Errors stored on struct, accessible via `LastError()` - Mutex protects concurrent access to shared state - Load() always returns nil (callers already handle this) ## Why Multi-page API fetches (up to 10 pages × 50 items) can take seconds. Blocking the tview event loop during fetch makes the terminal unresponsive. Closes #356

automation added 1 commit

2026-07-08 10:36:09 +00:00

feat(cache): make issue/PR loading async to prevent UI freeze

Test and Release / test (pull_request) Successful in 3m45s

Details

Test and Release / lint (pull_request) Successful in 4m29s

Details

6283f87b20

IssueView.Load and PRView.Load/LoadAll already fetch all pages
but run synchronously, blocking the UI. Move to goroutine pattern
with QueueUpdateDraw. Store errors for LastError() access.
All callers already check Load error synchronously — now always nil.
Tab switching during load is still responsive.

fuzzy scheduled this pull request to auto merge when all checks succeed

2026-07-08 10:42:17 +00:00

fuzzy added 1 commit

2026-07-08 10:46:27 +00:00

fix(cache): bake state filter into async Load, remove Reload

Test and Release / lint (pull_request) Successful in 2m43s

Details

Test and Release / test (pull_request) Successful in 4m13s

Details

8a3caabf12

Load/LoadAll now accept state parameter, filter internally inside
the async goroutine before QueueUpdateDraw. Removes the broken
Reload-after-return race where filters ran before data loaded.
Alt+O/Alt+C toggles call Load/LoadAll with state instead.

fuzzy added 1 commit

2026-07-08 10:51:30 +00:00

fix(cache): revert issue/PR loading to synchronous

Test and Release / lint (pull_request) Successful in 2m44s

Details

Test and Release / test (pull_request) Successful in 8m39s

Details

38e6870fc8

Async loading broke the pulls tab (never displayed data).
Revert to original synchronous pagination which fetches all
pages and renders in the tab handler's call chain.

fuzzy added 1 commit

2026-07-08 11:19:26 +00:00

feat(cache): add SQLite-backed persistent API response cache

Test and Release / lint (pull_request) Successful in 2m48s

Details

Test and Release / test (pull_request) Successful in 3m8s

Details

35de54b774

fuzzy added 1 commit

2026-07-08 11:25:59 +00:00

fix(cache): embed prefix in cache key, fix json tags for persistence

Test and Release / test (pull_request) Successful in 2m54s

Details

Test and Release / lint (pull_request) Successful in 3m2s

Details

9af63b1425

- cacheKey now returns 'prefix:hash' so cacheTTLFor matches endpoint
- Remove json:"-" from RepoOwner/RepoName on Issue/PullRequest
  so JSON roundtrip preserves these fields for persistent cache

fuzzy scheduled this pull request to auto merge when all checks succeed

2026-07-08 11:28:06 +00:00

the.auditor referenced this pull request

2026-07-08 11:28:40 +00:00

fix(forge): race condition in CachedClient.get() TTL check outside lock #369

the.auditor approved these changes

2026-07-08 11:28:48 +00:00

the.auditor left a comment

Review

Result: Approved — no blocking issues. Scope is significantly broader than the title.

Summary: Major refactor of the caching layer. Adds PersistentCache (SQLite-backed key-value store) and generic cachedLoad[T] eliminating repeated get/set boilerplate across all CachedClient methods. Per-key TTL via cacheTTLFor() (30s default, 5min lists, 72h immutable). Model field tags changed to enable JSON round-trip for persistent cache.

Non-blocking:

#369 — Race condition in get(): TTL check done outside mutex lock, could delete a fresh entry set by another goroutine
Unintended: pv.Reload() call removed from PRs tab handler — state filtering on load may be broken
cacheKey format changed (now prefix:hash) — incompatible with any hypothetical previous persisted data

Checklist:

Correctness: ✅ Generic cachedLoad eliminates boilerplate correctly, persistent cache round-trips via JSON, per-key TTLs appropriate
Completeness: ✅ All CachedClient methods migrated to cachedLoad, persistent cache wired into app startup and forge switching
Race safety: ⚠️ get() TTL check outside lock (minor — worst case is one extra fetch)
Maintainability: ✅ Massive reduction in boilerplate — each method is now a one-liner

## Review **Result: Approved** — no blocking issues. Scope is significantly broader than the title. **Summary:** Major refactor of the caching layer. Adds `PersistentCache` (SQLite-backed key-value store) and generic `cachedLoad[T]` eliminating repeated get/set boilerplate across all `CachedClient` methods. Per-key TTL via `cacheTTLFor()` (30s default, 5min lists, 72h immutable). Model field tags changed to enable JSON round-trip for persistent cache. **Non-blocking:** - #369 — Race condition in `get()`: TTL check done outside mutex lock, could delete a fresh entry set by another goroutine - Unintended: `pv.Reload()` call removed from PRs tab handler — state filtering on load may be broken - `cacheKey` format changed (now `prefix:hash`) — incompatible with any hypothetical previous persisted data **Checklist:** - Correctness: ✅ Generic cachedLoad eliminates boilerplate correctly, persistent cache round-trips via JSON, per-key TTLs appropriate - Completeness: ✅ All CachedClient methods migrated to cachedLoad, persistent cache wired into app startup and forge switching - Race safety: ⚠️ get() TTL check outside lock (minor — worst case is one extra fetch) - Maintainability: ✅ Massive reduction in boilerplate — each method is now a one-liner

fuzzy merged commit 9af63b1425 into main

2026-07-08 11:29:03 +00:00

fuzzy deleted branch feat/356-paginate-issues-prs

2026-07-08 11:29:03 +00:00