Skip to content

Add playlists and a one-feature-per-PR Jellyfin parity backlog - #38

Merged
vitobotta merged 4 commits into
vitobotta:mainfrom
phantomic12:jellyfin-parity
Nov 15, 2026
Merged

vitobotta merged 4 commits into
vitobotta:mainfrom
phantomic12:jellyfin-parity

Conversation

@phantomic12

Copy link
Copy Markdown
Contributor

Summary

This is a third, separate PR. It adds user playlists and records the Jellyfin
parity work as a one-feature-per-PR backlog so the remaining items can land
independently instead of being bundled into the two existing PRs.

Playlists (the only feature implemented here)

  • Per-account playlists with name and optional description.
  • Add a movie or show from its detail page, or create a playlist and add the current title in one step.
  • Reorder items, remove items, rename, and delete.
  • Child accounts cannot add a title above their age cap, and titles above the cap are hidden from the playlist view.
  • Playlists hold titles, not individual episodes yet.
  • Playback is item-by-item; a continuous play queue is a follow-up.

Jellyfin parity backlog, one feature per PR

The full table with status and remaining work is in
docs/jellyfin-parity.md. Each item below is
intended to be its own PR, in this order:

  1. Playlists — this PR
  2. Library — Movies and TV metadata — done
  3. Library — Local media scanning and file management — missing
  4. Library — Music library — missing
  5. Library — Books and comics — missing
  6. Library — Photos and home videos — missing
  7. Library — Collections — missing
  8. Library — Favourites — partial (wishlist covers save-for-later)
  9. Library — Mark played / unplayed — missing
  10. Library — Watched state and progress — done
  11. Library — Next Up — done
  12. Library — Genres, studios, people pages — partial
  13. Library — Trailers and theme media — missing
  14. Discovery — Search — done
  15. Discovery — Search filters and sorting — partial
  16. Discovery — Recommendations — done
  17. Discovery — Popular and trending rows — done
  18. Playback — Direct play — done
  19. Playback — Remux — done
  20. Playback — Transcoding — done
  21. Playback — Hardware-accelerated transcoding — missing
  22. Playback — Tone mapping — missing
  23. Playback — Quality and bitrate selection — missing
  24. Playback — Audio track selection — done
  25. Playback — Subtitle track selection — done
  26. Playback — External subtitle search and download — partial
  27. Playback — Subtitle burn-in — done
  28. Playback — Chapter markers — partial
  29. Playback — Trickplay thumbnails — missing
  30. Playback — Chapter images — missing
  31. Playback — Intro and credits detection — partial
  32. Playback — Playback speed — done
  33. Playback — Keyboard shortcuts — done
  34. Playback — Picture-in-picture — missing
  35. Playback — Casting and remote control — missing
  36. Playback — SyncPlay (watch together) — missing
  37. Live TV — Tuners and channel guide — missing
  38. Live TV — DVR and recordings — missing
  39. Users — Multiple profiles per account — missing
  40. Users — Avatars and display identities — partial
  41. Users — Per-library permissions — missing
  42. Users — Parental controls — done
  43. Users — Quick Connect — missing
  44. Users — OIDC / LDAP sign-in — missing
  45. Users — API keys — missing
  46. Users — Active sessions and devices — missing
  47. Users — Remote sign-out — missing
  48. Platform — Webhooks — missing
  49. Platform — Notifications — missing
  50. Platform — Scheduled tasks UI — partial
  51. Platform — Plugin system — missing
  52. Platform — Backup and restore — missing
  53. Platform — Metrics and monitoring — missing
  54. Platform — DLNA — missing
  55. Platform — Mobile apps — partial (installable PWA)
  56. Platform — Themes and branding — partial
  57. Platform — Localisation and RTL — missing
  58. Platform — Accessibility audit — partial

Server-side parity items (local media scanning, Live TV/DVR, plugins, hardware
transcoding) are called out as separate product decisions, not smuggled into a
playback PR.

Testing

  • Parsed every changed Ruby file and the ERB views with Ruby's Prism parser; all clean.
  • Verified playlist routes, model ordering/positions, ownership scoping, and child-account filtering against the new specs.
  • RSpec could not be run in this environment because no Ruby/Rails toolchain is installed here.

Accounts can create named playlists, add films and shows from their
detail pages, reorder or remove items, and keep child-account rating
caps enforced. The parity document lists each remaining Jellyfin
feature with its own scope and proposed PR order.

🤖 Generated with Codebuff
Co-Authored-By: Codebuff <noreply@codebuff.com>
The README rewrite moved features and setup into new sections, and this branch
had added its feature block inside what upstream now uses for the services
table. The merge keeps upstream's structure, folds the playlists feature into
the features list, and links the parity backlog from Contributing.

🤖 Generated with Codebuff
Co-Authored-By: Codebuff <noreply@codebuff.com>
@phantomic12

Copy link
Copy Markdown
Contributor Author

@vitobotta ready for a first review and mergeable against current main: playlists plus the Jellyfin parity backlog, one feature per PR as the backlog describes.

@vitobotta vitobotta left a comment

Copy link
Copy Markdown
Owner

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

First review at c1e5e69, with main merged. The suite is green: 515 examples, 0 failures, and RuboCop is clean on the twelve touched Ruby files, so the note in the description about a missing toolchain is the only reason these were not run before opening. I also checked what the specs do not cover: a child account sees the allowed items and the hidden count, another account gets a redirect with no playlist data, a metadata failure still renders the content page, and both pages hold at 1280 and at phone width.

One blocking problem: adding a show from its detail page cannot succeed, because of the year format. Details on the line below.

The minor items are on the lines below as well: two rows in the parity table are stale, the index count includes hidden items, the reorder arrows are computed from the filtered list, and an unknown move direction is treated as down. The feature is otherwise sound: ownership scoping, the duplicate rule and position ordering all behave.

Comment thread app/views/content/show.html.erb Resolved
Comment thread docs/jellyfin-parity.md Resolved
Comment thread app/views/playlists/index.html.erb Resolved
Comment thread app/views/playlists/show.html.erb Resolved
…lists

A show could never be added from its detail page because Cinemeta reports a year range and numericality validates the raw value; the index also promised titles the playlist page hides, arrows were indexed against the filtered list, and any direction string reordered the playlist.
@phantomic12

Copy link
Copy Markdown
Contributor Author

@vitobotta all five are addressed on the branch.

  • Blocking — show cannot be added: PlaylistItem#normalize_year now takes the first four digits from a raw value before validation, so the 2008–2013 range Cinemeta returns for tt0903747 stores as 2008. Pinned by a request example that posts the range the page sends.
  • Parity table: row 34 (PiP) is now ✅ — the player from the 11 November commits has the button and state handling; row 58 no longer claims there is no reduced-motion mode, since application.css honours prefers-reduced-motion.
  • Index count: the controller computes visible and hidden per playlist through the parental policy, and the card reads 1 title · 1 hidden for this account, matching the page it opens.
  • Reorder arrows: indexes now come from the full ordered id list that move! swaps within, not the filtered list.
  • Move direction: only up and down are accepted; anything else redirects with Invalid move direction. and leaves the order untouched, covered by an example that posts sideways.

playlists_spec + playlist_spec 19/0, content_spec 11/0, RuboCop clean on the four Ruby files.

@vitobotta vitobotta left a comment

Copy link
Copy Markdown
Owner

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Re-reviewed at 6ab56894. The five findings from the first pass are addressed:

  • Series year: normalize_year keeps the first four-digit year before numericality; the new request example posts 2008-2013 and stores 2008.
  • Hidden counts: the index now shows the visible count plus the hidden count for the account; the new example asserts both.
  • Arrows: @ordered_ids comes from the full ordered playlist, so the move boundaries match PlaylistItem#move!.
  • Invalid direction: move accepts only up and down and redirects with an alert; the new example checks the order is unchanged.
  • Docs: rows 34 and 58 now match the code.

One new blocking regression on the create failure path, plus one minor note, both on the lines below.

Gates at the head: playlists 14/0, full suite 518/0 (excluding my throwaway probe), JS 528/0, RuboCop clean on the four touched Ruby files.

Comment thread app/views/playlists/index.html.erb Resolved
Comment thread app/models/playlist_item.rb Resolved
…eal years

The index view ran with nil counts when create re-rendered it for an account that already had playlists, and the year helper turned 20100 into 2010 instead of leaving malformed input to validation.
@phantomic12

Copy link
Copy Markdown
Contributor Author

@vitobotta both fixed at 3bfc659.

  • Blocking — nil counts on the create failure path: the count setup moved into build_item_counts, which index and the rescue ActiveRecord::RecordInvalid branch both call, so re-rendering the index for an account that already has playlists no longer fetches from nil. New example creates a playlist first, posts a blank name, and asserts 422 with the existing playlist still on the page — it fails against the previous head with undefined method fetch for nil.
  • Year normalisation: the helper now matches only \A(\d{4})(?:\s*[–—-]\s*\d{4})?\z, so a single year or a Cinemeta range is accepted and everything else is left to numericality. Examples: 2008–2013 is valid and stores 2008, and 20100 stays 20100 instead of being silently truncated to 2010.

playlists_spec + playlist_item_spec + content_spec: 33/0, RuboCop clean on the four touched files.

@vitobotta vitobotta left a comment

Copy link
Copy Markdown
Owner

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Both open findings are fixed at 3bfc659.

The create failure path now builds the same counts as index through build_item_counts, and the new request example covers an account with an existing playlist. The year normalizer now accepts one four-digit year or a range of two and leaves other input to numericality. A probe confirmed 20100 stays 20100 and fails validation, while a spaced two-year range stores 2008.

The branch merges cleanly with main at 97f9ddf. Gates on the merged tree: 554 RSpec examples, 0 failures; 578 JavaScript tests, 0 failures; RuboCop 185 files, no offenses; Brakeman clean.

No new findings.

@vitobotta
vitobotta merged commit a749a52 into vitobotta:main Nov 15, 2026
3 checks passed
@vitobotta

Copy link
Copy Markdown
Owner

Merged. Thanks @phantomic12 for the playlists and the parity backlog. A new release will be made soon.

@phantomic12
phantomic12 deleted the jellyfin-parity branch November 15, 2026 22:34
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants