mpiton/vortex - CodSpeed

vortex

Performance History

Latest Results

fix(link): dedupe probe inputs without indexing on originalUrl The intermediate Map<originalUrl, getDuplicateKey> in dispatchDuplicateDetection would silently drop a row if two ResolvedLinks shared the same originalUrl, leaving that row with duplicate === undefined and blocking it from bulk start when skipDuplicates was on. Compute canonical keys per-row, then collapse through a Set.

feat/task-33-detect-duplicates

8 minutes ago

fix(link): collapse bulk starts by canonical URL dispatchDuplicateDetection already collapses probes to unique canonical URLs, but the bulk start handlers iterated per row and called startDownload once for each. Two rows resolving to the same getDuplicateKey (mirror sites, redirects to the same file) both passed the duplicate gate when neither was in active/history yet, queueing the download twice from a single paste batch. New startLinks(links) helper gates each link via isStartable and, when skipDuplicates is on, dedupes by canonical URL via a Set so each unique target is started at most once. handleStartSelected and handleStartAllOnline now both delegate to it.

feat/task-33-detect-duplicates

34 minutes ago

fix(link): unify URL identity for dedupe and start Duplicate detection keyed on link.originalUrl while startLink used link.resolvedUrl ?? link.originalUrl. A redirected link whose canonical form already existed in active/history could pass the duplicate gate (originalUrl was unique) and then re-queue the same underlying URL via download_start. New getDuplicateKey(link) helper returns link.resolvedUrl ?? link.originalUrl. dispatchDuplicateDetection now takes the full ResolvedLink[] and derives canonical keys for the IPC and the inFlight set; startLink calls the same helper. Both sides operate on identical URL identity.

feat/task-33-detect-duplicates

57 minutes ago

fix(link): unblock bulk start when duplicate detection IPC fails When link_detect_duplicates rejected, link.duplicate stayed undefined and isStartable silently rejected every online row whenever skipDuplicates was enabled (the default). dispatchDuplicateDetection now wires onError to mark each in-flight URL with duplicate: null so the bulk-start gate falls back to optimistic behavior instead of freezing the UI on a transient IPC failure.

feat/task-33-detect-duplicates

1 hour ago

fix(link): address PR #145 review comments - url_normalizer: replace byte-index slicing in detect_scheme with str::get(..) accessors so non-ASCII-leading URLs no longer panic on a multi-byte UTF-8 boundary - url_normalizer: lowercase only host portion of authority; userinfo (user:pass@) is preserved byte-for-byte per RFC 3986 to avoid merging distinct credentials - LinkGrabberView: add monotonic batchRef so a stale link_detect_duplicates response cannot clobber a newer resolve batch - LinkGrabberView: while Skip duplicates is on, isStartable returns false for rows whose duplicate probe is still pending - LinkRow: route duplicate badge labels + tooltip through react-i18next (en + fr translations added) - Add two normalizer unit tests for the UTF-8 panic + userinfo paths

feat/task-33-detect-duplicates

2 hours ago

refactor(link): tighten task-33 duplicate detection per code review Round-1 cleanup off the simplify pass: - Extract `application/services/history_paginate.rs` with `for_each_history_page` (callback per page) + `list_full_history` (collect-all wrapper). `export_history` and the new `detect_duplicates` query both consume it; the bounded-page pagination loop is no longer copy-pasted. - Hoist `InMemoryDownloadReadRepo` and a new `make_history_and_downloads_query_bus` builder into `application/test_support`. Drops ~120 lines of stub fixtures from the `detect_duplicates` test module. - Replace `DuplicateCheckDto.source: Option<&'static str>` (stringly- typed IPC) with a proper serde enum `DuplicateSourceDto` lined up with the rest of the IPC layer. - `detect_duplicates` handler: store the existing-id as `u64` in the index, stringify only on a confirmed hit; pre-check `contains_key` on the history loop so shadowed-by-active entries no longer allocate a `(source, id, file_name)` tuple. - `url_normalizer`: detect the scheme via `eq_ignore_ascii_case` on the prefix instead of allocating a lowercase copy of the whole URL, and trim the WHAT-style block-list / behaviour comments to the load-bearing WHY notes (path-case rationale, batch-size rationale). - `LinkGrabberView`: extract `isStartable` + `startLink` helpers so `handleStartSelected` and `handleStartAllOnline` no longer carry the duplicated status / duplicate gates. Wrap `duplicateCount` in `useMemo`. The `detectDuplicates` `onSuccess` now compares against the previous `duplicate` reference and returns `prev` when nothing moved, so an all-unique batch does not retrigger downstream re-renders. - `LinkRow`: replace the duplicate-source ternary chain with a `duplicateLabelMap` lookup and switch the `data-duplicate` sentinel from `"no"` to `"none"` so unique rows surface a meaningful tag. Test file gains a `makeDuplicate` helper that halves the fixture boilerplate. Tests: cargo test --workspace 1438 passed, vitest 668 passed.

feat/task-33-detect-duplicates

13 hours ago

feat(link): auto-group split archives into one package (task 31) (#144) * feat(link): auto-group split archives into one package (task 31) Cluster Link-Grabber URLs by detected archive base name (`*.partNN.rar`, `*.rNN`, `*.7z.NNN`, `*.zip.NNN`, `*.tar.{gz,bz2,xz}.NNN`) and create one `Package` per cluster with `source_type = SplitArchive` and `external_id = "split-archive:{base}"` for idempotent re-resolve. New `GroupSplitArchivesCommand` handler + `link_group_split_archives` Tauri IPC mirror the playlist grouper flow. Gaps in the part numbering emit a new `DomainEvent::SplitArchiveIncomplete { package_id, base_name, missing_parts }` event (forwarded to the frontend as `split-archive-incomplete`) so the UI can warn the user before extraction blocks. Frontend `SplitArchiveLinkInput` / `SplitArchiveGroupResult` types added in `src/types/media.ts`. 25 service unit tests (matcher fixtures + grouping integration) + 3 handler tests cover the contract. Refs PRD-v2 §P1.12, PRD §6.3. * refactor(link): harden split-archive grouper from /simplify review Code-review pass on the task-31 split-archive grouper: - Format-namespaced `external_id` (`split-archive:{format_tag}:{base}`) so a RAR set and a ZIP set sharing a base name produce two distinct packages instead of silently collapsing under one row. Replaces the previous base-only key documented as "acceptable" in the test that has now been rewritten to assert the correct behaviour. - New `application/services/group_lock` module shared by `PlaylistGrouper` and `SplitArchiveGrouper`, removing a duplicated OnceLock + poisoned-mutex-recovery pair. - `group_one_base` now releases the grouper lock before publishing `PackageCreated` / `SplitArchiveIncomplete` events so synchronous subscribers cannot block other concurrent grouping calls. Same refactor also collapses the duplicated reuse-vs-created branching into a single tail block. - `MAX_LINKS = 500` cap on `group_all` mirrors `MAX_URLS` in `resolve_links` so a runaway IPC payload cannot allocate unbounded cluster state. New `test_group_all_caps_link_count_to_avoid_dos` test covers the path. - `SplitArchiveFormat`, `DetectedPart`, `detect_from_filename` demoted to `pub(crate)`: they were exported but only consumed inside the module. Reduces the public API surface. * fix(link): detect legacy RAR header + release playlist lock before publish Legacy multi-volume RAR sets ship the first volume as `name.rar` followed by `name.r00`, `name.r01`, … The previous matcher only captured `.rNN`, so a valid `name.rar + name.r00` set produced a single-part cluster and was dropped by MIN_PARTS_TO_GROUP. Add `match_legacy_rar_header` (part 0), order it after the other matchers so the more specific patterns still win, and extend `compute_missing_parts` to walk from 0 for LegacyRar so a missing header is reported as `rar` instead of silently dropped. PlaylistGrouper used to hold the shared grouper lock across `event_bus.publish`, which can block re-entrant grouping paths and matches the bug already fixed in SplitArchiveGrouper. Scope the lock to the find-then-save window and publish after the guard drops. Addresses CodeRabbit, Codex, and Cubic review feedback on PR #144. * fix(link): bound part indices + gate on distinct part numbers Three converging review findings on PR #144: 1. `MIN_PARTS_TO_GROUP` was checked against raw link count, so two mirrors of `movie.part01.rar` slipped through the singleton guard and produced a misleading "complete" group with `missing_parts` empty. Replace `parts.len()` with the distinct `part_num` count. 2. The modern RAR (`\d+`) and legacy RAR (`\d{2,}`) regexes accepted unbounded suffixes, letting a hostile `name.part1000000000.rar` force `compute_missing_parts` into a multi-billion-step iteration. Add `MAX_PART_INDEX = 10_000` and reject anything above it inside `detect_from_filename` (after the matcher cascade so a rejected index does not silently cascade into `match_legacy_rar_header`). 3. Defensive `checked_add` on `raw_num + 1` in `match_legacy_rar` so the (now-impossible after the cap) `u32` overflow path is closed. Addresses CodeRabbit, Codex, and Cubic feedback. * docs(changelog): correct split-archive test count (31, not 35) --------- Co-authored-by: Mathieu Piton <27002047+mpiton@users.noreply.github.com>

main

14 hours ago

docs(changelog): correct split-archive test count (31, not 35)

feat/task-31-split-archives

14 hours ago

Latest Branches

-23%

refactor(link): tighten task-33 duplicate detection per code review#145

29 minutes ago

c216cff

feat/task-33-detect-duplicates

+19%

feat(link): auto-group split archives into one package (task 31)#144

14 hours ago

144c2ba

feat/task-31-split-archives

N/A

Add CodSpeed performance benchmarks for Rust domain layer#143

17 hours ago

c3a7733

codspeed-wizard-1777816068872

Home Terms Privacy Docs