Repository navigation
v4.4.1 - Select all means all of them, the page size is yours, and a deep queue keeps its history
Compressatorium 4.4.1
Released 2026-08-15. Select all means all of them, the page size is yours, and a deep queue keeps its history.
This release supersedes v4.4.0. That tag was cut but never published as a release; everything in it is rolled up here, plus the job-history fix below. If you're coming from 4.3.0, this is the only release you need.
4.3.0 was about what Compressatorium can convert. 4.4.x is about what happens when you have two thousand of them. No new tools, no new formats — a file-list release aimed squarely at the moment you open a platform folder holding 2,000+ discs and realise the header checkbox only ticks the fifty you can see.
That checkbox has always been per-page. A Redump platform set at 50 rows a page is 40-odd pages, so queuing the whole folder meant ticking it once per page, all the way down. Now it takes one click. And once one click can queue two thousand jobs, the job history has to survive a queue that deep — which it didn't, until 4.4.1.
The short version
- Select all spans every page. Tick the header checkbox in a multi-page folder and a banner offers the rest: "Select all 2140 items in this folder (all 43 pages)".
- Shift-click the header checkbox does the same without the banner, mirroring the shift-click range select already on the rows.
- Deselect all appears once everything is selected, and drops only what's in the current view — Clear still drops the entire selection.
- It follows the tool, the mode, and the filter. Files under a file mode, convertible folders under makeps3iso
folder_to_iso, archives underromz_extract; with.isopicked in a mixed folder the offer reads "Select all 2100.isoitems" and leaves the rest alone. - It works everywhere the file list does — plain folders, inside an archive, and across recursive Search all results.
- Rows per page is a setting: 25 / 50 / 100 / 250 / 500, remembered between sessions, and changing it keeps your place instead of throwing you back to page 1.
- Fixed: a queue deeper than
MAX_JOB_HISTORYemptied the Completed and Failed tabs while the batch ran. - Fixed: shift-clicking after the page got shorter threw an error and swallowed the click.
Select all, across every page
Once anything is selected in a view spanning more than one page, a banner appears under the selection bar:
All 50 on this page are selected. Select all 2140 items in this folder (all 43 pages)
One click takes the lot. Once everything is selected the banner confirms the scope and offers the inverse:
All 2140 in this folder are selected, across all 43 pages. Deselect all
Two different "unselect" actions, and they aren't redundant. Deselect all drops everything in the current view, so with a filter active it clears just the filtered rows and leaves your other picks intact. Clear, on the selection bar, drops the whole selection regardless of view.
It selects exactly what the header checkbox would. Both run through the same mode-aware predicate, so ticking every page by hand and clicking Select all once produce identical sets — the offer never quietly includes rows the checkbox wouldn't, or omits rows it would. Conversion eligibility is still enforced separately at submit time, as before.
It only appears when there's something to gain. Filter a folder down to a single page and no banner shows at all.
No extra loading. Pagination was already client-side slicing of the listing your browser holds, so "all pages" is the set already in hand — the banner costs one click, not one request.
Rows per page
A Rows picker in the file-list footer offers 25 / 50 / 100 / 250 / 500, replacing a fixed 50. A 2,140-file folder goes from 43 pages to 5.
- Remembered across reloads, so you don't re-pick it for every folder.
- Keeps your place. Going 50 → 100 while on page 8 lands you on page 4 still looking at the same files, rather than the top of the folder.
- Selection survives it. Selection is keyed by path, not by page, so re-paginating mid-selection — including after a cross-page Select all — keeps every ticked row.
- Always reachable. The picker sits outside the pager, which hides itself at a single page, so a size big enough to collapse the pagination can still be undone.
Why it stops at 500. Every row on the visible page gets hydrated — archive summaries, CHD metadata, and a background DAT hash job per file when DATs are loaded — so page size is what bounds that work. Collapsing a 2,000-file folder into one page would kick off a hash of the entire folder just from opening it. Selecting everything no longer needs a giant page anyway, now that Select all spans pages.
Fixed
-
A deep queue no longer wipes your job history mid-run.
MAX_JOB_HISTORY(default 500) caps how many finished jobs are kept, but it was gating on the total job count — pending work included. Queue more than 500 files at once, which cross-page Select all makes a single click, and the total sat permanently over the cap: every sweep then deleted every completed, failed and cancelled job it could find and still couldn't get under it, because the rest were queued and not removable. Since the sweep runs on each completion, the Completed and Failed tabs emptied continuously while the batch ran, and a finished 2,000-file conversion left no record of which files had failed. The cap now counts terminal jobs only. Queued and processing jobs were never at risk of deletion — they simply aren't removable — but they were being counted, and that is what evicted the history. Nothing else changes: history past the cap is still trimmed oldest first, and the just-finished job is still preserved. -
Shift-click no longer breaks after the page shrinks. The shift-click range anchor was an index into the page you were on, but nothing kept it inside the page actually on screen. Walk to a shorter last page — or narrow the extension filter, or re-sort — and the next shift-click ran off the end of the row list and threw, swallowing the click entirely. The anchor is now clamped to the current page, so the range simply stops at its end. Changing rows-per-page additionally starts a fresh range, since repagination re-cuts the page and extending from a row picked under a different layout is surprising even when it happens to be in bounds. This one predates 4.4.x — the short-last-page route was reachable in 4.3.0.
Upgrade notes
No database migration. The schema is unchanged from 4.2.0 (0003 remains the head).
No new settings. No environment variables were added, removed, or renamed.
One behavior change worth knowing. MAX_JOB_HISTORY now means what its name says — it bounds finished jobs, not total jobs. If you raised it as a workaround for history vanishing during large batches, you can put it back to the default; the default 500 finished jobs are now actually retained while a deep queue drains.
MAX_QUEUE_DEPTH is untouched and still defaults to 0 (disabled). That one is a deliberate operator-set ceiling on queued+processing jobs — if you've set it, it still rejects submissions past its limit. Only the history cap changed.
No API changes. Mode strings, endpoint paths, and JSON shapes are unchanged, so anything scripting against the API is unaffected.
The image is unchanged in size. No new binaries or runtimes ship with this release.
One new browser-local key. Your rows-per-page choice is stored in localStorage under compressatorium_file_list_page_size. Nothing is written server-side, and an unset or unrecognised value falls back to the previous default of 50 — so a first run after upgrading looks exactly like 4.3.0 until you change it.
For contributors: shots.yml matched the file-list header checkbox by its aria-label, which is state-dependent and changed here. It now matches structurally (thead input[type="checkbox"]), so the screenshot automation no longer breaks when button wording changes.
Full commit list since v4.3.0 (6 commits, oldest first)
57b8f85chore: sync package.json to v4.3.0 [skip ci]9cf46f3feat(ui): select all across pages, and a rows-per-page picker (#259)f16416cUpdate Web UI screenshots — Sat Aug 15 18:23:20 UTC 2026e821bf7docs: cut 4.3.0 release notes and open 4.4.0 (#260)185d587chore: sync package.json to v4.4.0 [skip ci]d7d925afix: cap job history by finished jobs, not by total queue depth (4.4.1) (#261)