ajslater/codex v2.3.0

on GitHub

Upgrading

• Comicbox 5.0.0 reads some tags differently and adds others. Existing
  comics keep the tags they were imported with until they are read again:
  select comics and Force Update Tags, or let a scan pick up files you
  change. Until then the new fields are empty, and age ratings and credit
  roles keep their old spellings, so a filter can list both Everyone 10+
  and Everyone, or both CoverArtist and Cover. Age-restricted access
  is unaffected either way.

Features

• Sort by Reprints, the other editions an issue was published as, in the
  table and in cover view. Filter by a reprint first to sort by that one.
• Read a reprint series as a reading order from the reader's reading-order
  menu, which makes reprint tags usable as durable reading lists.
• "Alternate Series" is now "Reprints" everywhere: it covers MetronInfo
  Reprints and AlternativeNames, ComicInfo AlternateSeries and CoMet
  isVersionOf. A series' other names are listed separately in the metadata
  panel.
• Each top collection remembers the sort it was last browsed with, and
  clearing a search puts back the sort it replaced.
• Manga is its own tag now, separate from reading direction, along with
  MetronInfo's Manga Volume.
• A comic's own web links are shown and editable.
• Credits mark the primary holder of each role and list them first. Codex
  cannot write it back: neither ComicInfo nor MetronInfo stores it.
• Online tagging takes an Effort setting, per scan and as an admin default.
  Thorough restores Comic Vine's unbounded search, Minimal trades matches
  for speed, and Metron ignores it. Auto, the default, drops a large
  unattended scan to Minimal so Comic Vine's hourly limit doesn't stretch it
  into hours.
• The tagging status strip shows Comic Vine's remaining hourly budget.

Fixes

• Reading Bottom to Top jumped to the wrong page: page one showed the last
  page.
• Saving browser settings sent an empty request. The next page request saved
  them anyway, so nothing was lost.
• Sorting by a tag column outside the table view no longer errors.
• Browsing a group whose parents have gone missing shows a button back to
  the top rather than an empty crumb trail and no way out.
• The Admin Tagging Status table shows what is being looked up right now.
• Identifier links are built from the id instead of stored beside it, so
  they can no longer disagree with it. A comic can carry an id for its
  series, and a reprint's id links to the issue it reprints. The tag editor
  writes identifier types the way the formats spell them.
• A Comic Vine key the site rejects now fails validation instead of passing.
• Comicbox's online caches stay under the config directory instead of a
  location that recreating a container discards.
• An online scan's time estimate is keyed on Effort rather than Match Mode,
  which never changed how many requests a comic costs. It reads higher, and
  a real scan beats it.
• Covers are cached by the browser and revalidated instead of re-sent, so
  revisiting a page does not download them again. They also survive a scan,
  which used to discard every cached cover whether or not it had changed.
• Every web request leaked a database connection, and its open files, until
  garbage collection.

Performance

• Web requests share a pool of database connections instead of opening one
  each. db_workers under [server] sizes it, or turns it off with 0.

Dev

• Comicbox 5.0.0, and its metadata schema v3.0.
• Simyan, mokkari and requests-cache are declared directly rather than
  relied on through comicbox.
• Asgiref is declared directly; the connection pool uses its executor
  registry.