Every Settings change now persists on its own (debounced ~500ms so a slider drag
is one PUT), with a small transient indicator; the draft/dirty-save UI and the
navigation leave-guard are gone. Also unblocks the navigation extraction — there
is no longer a cross-page "unsaved settings" guard to thread through routing.
Locks the five flows the refactor must not break (feed, channel, player,
settings, nav) so later sprints can prove they changed nothing. Adds a thin
data-testid layer and a dedicated seeded non-demo account; the suite runs on
the baked :8080 stack (authoritative) or :5173. No product-behaviour change.
Five view modes behind a reusable ViewSwitcher (cards · cardsSmall · rows ·
rowsCompact · tiles), and a subdued edge-fade on the page scroller and every
capped list. UAT-accepted; not shipped to prod (user's call — the Platform
refactor is next). Version bump + releaseNotes deferred to ship time.
Same regression the S3 edge-fade caused for the Plex player, and the one my
"swept the class" pass missed: PlayerModal is `fixed inset-0 z-50` but rendered
inside the page scroller, so once that scroller got the mask it became a
stacking context and scoped the modal's z-50 inside it — the nav rail (z-40) and
header (z-30) painted over the player, and the fade dimmed its top edge (worst
at browser zoom, where the centered card is smaller against the rail/panel).
Portaled to <body>, where every other full-screen overlay already lives.
Verified: modal is body-parented, full-viewport, and hit-tests on top at the
nav / filter / header / corner points with the main mask active.
The player is `fixed left-0 top-0 z-50` but rendered from inside the page
scroller, so it only ever covered the viewport because that scroller happened
not to be a stacking context. Giving it the edge-fade made it one, and the mask
scoped the player's z-50 inside it: the rail, the side panel and the header
painted straight over the player, and its edges faded. Measured both ways, then
fixed where every other full-screen overlay here already sits — <body>.
Same for the Suspense fallback that stands in while the player chunk loads, and
for the floating draft save bar on the Configuration page. Swept the rest: every
other `fixed` element is either App-level or already portaled.
The sweep's own three (the thread, the Plex collections filter, the tag manager)
plus the seven the spec missed — the playlist picker, the row tag menu, the Plex
playlist/collection dialogs, the player's track menu, the video description and
the table's multi-select filter. Same idiom throughout: a height-capped list that
silently clipped its content now says so, and hides its scrollbar like the rail
and the panels already do.
The table's option list is its own module-level component because it owns a hook:
FilterPopover is redeclared per DataTable render, so React remounts it.
The subdued edge-fade the rail and side panels use now hints at more content on
the page scroller too, which the feed, the channel page and Plex all scroll
inside. Both <main>s move into one PageScroller so the invariants they share —
the load-bearing scrollbar gutter, and the bar staying visible because hiding it
would zero that gutter — are stated once.
Two fixes the hook needed first: it set fresh state on every scroll event, which
above a page tree would re-render it per frame; and it watched a single child
captured at mount, so a lazy page swapping out its Suspense fallback left the
observer on a detached node and the fade frozen at its pre-load value (measured
on the Plex page: no bottom fade over a 3119px list).
My last commit made ROW_COL the literal source of the widths — and froze
them, because inline pixels can't follow the root font that Settings'
text slider drives (0.9-1.3). Measured at 130%: the channel box stuck at
128px while its text wanted 152 and the meta cell, being a rem grid, grew
to 156 — one row in two unit systems.
The classes come back and ROW_COL documents them, saying plainly that the
two are matched by hand and why they have to be. Chasing a comment's
literal truth cost a property I'd already verified.
The comment claimed the boxes were generated from ROW_COL. They weren't —
still w-32/w-16/w-48 by hand — so a comment promised a link that didn't
exist, which is worse than not claiming it. The boxes take the numbers
directly now (inline, since Tailwind can't read a runtime value), and the
thresholds add up in plain arithmetic rather than an IIFE whose result
depended on key order.
They were two hand-written sets of numbers describing the same boxes —
change a w-32 and the thresholds are silently wrong, the symptom being a
column appearing just before it fits. Now computed from one table (which
reproduces the old values exactly).
Width also stops reaching the four views that ignore it: it changes every
frame of a resize, so passing it everywhere pushed all of them through
VideoCard's memo for nothing.
The viewport was the wrong signal: the nav rail and filter panel move the
row's width by ~430px between them, so at a 1280 window with both pinned
every column still qualified while the row had 762px for 766px of them —
the title squeezed back to nothing, the exact bug the breakpoints were
meant to fix. My own comment predicted it and I didn't follow it through.
VirtualFeed already measures the container for colCount, so it now passes
the item's real width down. Measured in the failing config: title -4 ->
359 with the date column standing down.
It sat above the one-line row's early return but only textBlock consumes
it, so every compact row allocated it and threw it away — the same shape
I'd already fixed for textBlock and meta, crept back. Also merges a
rationale comment I'd written twice.
The xl breakpoint was a fact about the FEED's toolbar, hard-coded into
the switcher that E4's managers are meant to reuse — their rows are far
emptier and would have hidden the label for no reason. It's a prop now,
defaulting to icon-only.
The two views spans collapse into one builder, which also fixes a
regression I'd just added: splitting them put the exact-count tooltip on
the number alone, so hovering the word gave nothing.
Three from UAT: the row's cap goes 1280 -> 1472 (+15%), since with the
rail and filters hidden there was a lot of page left over and the fixed
columns mean the extra width can only go to the title; the views column
drops the word and keeps it in the tooltip (a table doesn't repeat its
unit every line), taking that column 144 -> 64; and the view switcher
names the current mode where the toolbar has room, falling back to the
icon alone below xl. Title: 431 -> 511 at 1440, 706 at 1900.
Every column but the title is shrink-0, so once the fixed widths outgrow
the row the title gives all of it: measured clientWidth 0 on every row at
a 1000px window — a table of channel, duration, views and date with no
titles. Columns now drop from the right as space runs out, least
important first, in the order the user named: title, then channel, then
the rest. Measured at 1000px: title 0 -> 351.
I'd written this off as "E15's problem, bites below ~640px" — wrong by
600px, because I only ever measured at 1440.
I set the attribute unconditionally because the one-line row truncates
hard, and never looked at the other four views — where the name fits and
the tooltip just repeated what was already on screen. It now shares the
title's clamped check.
Dropping the sr-only twin last round swapped a double announcement for
none: a bare span's title isn't dependably surfaced, so a screen reader
got no hint a row was live, a stream, or saved. role="img" + aria-label
names each marker once and keeps the hover.
The saved pip borrowed the bookmark button's label, so it read "Saved —
click to remove" on a plain span: hover, be told to click, click, nothing.
It gets its own "Saved" string. The status markers drop their sr-only
twin too — with a title already carrying the meaning, a screen reader
that surfaces title text said it twice.
The cell was a flex row, so whatever sat left pushed the rest: a 36:36
and a 3:31:09 put the marker beside them at different x, and a live video
— which has no duration at all — pulled it left again. The column edges
lined up, which is all I had checked; the data inside them did not.
Each datum now has a fixed slot in a grid, the duration right-aligned as
a number should be. The chip becomes an icon, about a letter wide: it is
the least important thing in the row, and three states have to stay apart
by SHAPE — in the youtube scheme --accent is red, so a live dot and a
stream dot would be the same dot.
It looked <main> up when it bound, which the lazy pages won. On a first
visit to the channel page or Plex the chunk is still loading, so the
Suspense fallback is on screen, there is no <main> to find, and the effect
gave up — its dep never changes again, so the button stayed dead for that
whole visit and only worked once the chunk was cached.
It now reads the scroller off the scroll event instead, captured on the
document: scroll doesn't bubble but it does capture, so there is nothing
to look up and nothing to race. Also 50% bigger, as asked (44 -> 64).
Every width was measured in the English UI, but "101.7K megtekintés" is
139px against "101.7K views" at 79 — so a Hungarian user's view count
ellipsed itself in a column whose job is to show it. The channel column
gives back the room: its names truncate at any width we can afford (12
of 102 at 128 vs 5 at 160), and a clipped name still reads where a
clipped number does not.
The channel name also carries a tooltip now — it truncates hardest in
the fixed column and had no way to be read in full.
A 60-video sample topped out at 2:00:58, so the column was 8px short of
an 11-hour stream that is also saved — and the library holds one. That row
would wrap to two lines in the one layout that promises single ones.
The stacked layouts' textBlock now builds after the one-line row returns,
since only they use it.
Actions lead the row: the title is what you read, so that is the shortest
pointer trip to them. Every column after it is a fixed width, which is
what puts the same datum on the same vertical line in every row — a CSS
grid cannot do it, since VirtualFeed renders each row as its own subtree.
Widths come from measuring the real data, not from taste.
Metadata rises to the channel name's size, and the duration drops its
chip: over a thumbnail that background carries contrast against the
artwork, but in a text column it only made one data column look unlike
its neighbours.
Six buttons need 188px but the small card's text column bottoms out near
160 (its 180px column less padding), so the reset button — the only one
that is conditional, hence the only one anyone noticed — hung outside the
card by up to 15px. The buttons go tighter there (154px), and the row can
wrap as a backstop rather than clip. AddToPlaylist and DownloadButton now
take the button style as a prop so all six stay one size.
Tile actions also pin to the tile's bottom edge, so they land in the same
place whether the title above ran to one line or three.
Its column bottoms out at 180px, so its image is narrower than a row's
(measured 170px) while still carrying the 194px labelled pair — the same
overflow the row fix was for. The rule is the thumbnail's size class, not
the family: only the large card has room for labels.
The feed takes its column count from main's clientWidth and its
thumbnails are aspect-ratio boxes, so the two feed each other once the
content height sits within a scrollbar's width of the viewport: the bar
appears, 15px goes, thumbs shorten, the content fits, the bar leaves,
and round again — every frame. Measured: scrollbar width alternating
0/15 and clientWidth 987/972 on consecutive frames. Reserving the gutter
holds clientWidth constant and breaks it.
The bare row becomes a single line — title truncating into the space
left, then channel, badges, meta and the actions each holding a fixed
column — and it drops the 896px reading cap, since with the actions
right-aligned there is no gap for the extra width to stretch. The
actions get a reserved slot: the reset button only exists for videos
with watch state, so without one the meta sat 32px further right on
some rows than others.
Rows and tiles get a permanent container instead of a hover-only fill.
They were transparent at rest and painted an opaque bg on hover, so
running the pointer down the list kept covering and re-revealing the
ambient backdrop — which reads as the background flickering.
Card actions now sit on the bottom edge rather than wherever the title
ends, and the small thumbnails carry icon-only controls: the labelled
Continue / Restart are sized for a card's image and were clipped at
176px.
Dropping Thumb dropped everything it draws ON the image — duration, the
live/stream badges, the saved marker and the resume bar. Those are
metadata, not decoration, so they move into the meta line and the bar
onto the row. Splits the spec's overloaded compact flag into thumb/dense,
which meant one thing per family and nothing on its own.
Tiles answer the row view's wasted space: the actions move under the meta
into vertical slack the thumbnail already reserves, which frees ~168px of
width, so the block survives in a narrow column and the list can go
multi-column. Rows stay single-column on purpose — reading down one edge
is why that view exists.
The column count now comes from a per-view minimum width, so every
multi-column view reflows on the mechanism the card grid already used.
Seeding the roving index from an effect only scheduled the update, so the
focus effect still read the previous index and focused the wrong item for
one render — a screen reader announced both. Seed it where the menu opens
instead. The modes are one exclusive choice, so they also need a group to
scope their checked-ness to.
Arrows/Home/End move a roving focus and Escape hands focus back to the
trigger — the menu unmounts under the focused item, which stranded focus
on <body>. Outside-click deliberately does not refocus: that would steal
focus from whatever was just clicked.
The view mode moves out of Settings into a switcher next to sort, on the
feed and channel pages alike. It leaves the Settings prefs draft for its
own eager save: a toolbar click must not mark that page dirty and trip
its leave-page guard.
lib/feedView.ts owns the vocabulary and maps the old grid/list prefs
forward on read, so nobody's saved choice is lost.
Sprint 5 made the channel filter multi-select, but this handler still wrote a
single-element array — so arriving from a second channel page silently dropped
the first. Union the id in (deduped, so repeating it for an already-picked
channel is a no-op).
It called closeChannel() before setPage("feed"). closeChannel runs
history.back(), and the popstate that restores the page fires AFTER the
synchronous setPage — so the restore won, dropping you back on whatever page the
channel was opened from. Opening from the feed hid it: the page it restored
happened to be the feed.
setPage already closes an open channel page and pushes a clean history entry, so
the closeChannel() call was both redundant and the bug.
Pick any number of channels; they OR together and every other filter applies
across the lot — "the last week's unwatched from these three".
`channelId`/`channelName` are replaced outright by `channelIds: string[]`, no
legacy reader (standing rule: no back-compat unless asked — a saved view written
against the old field simply loses its channel filter). The wire and the share
URL move with it: repeated `channel_ids` params, `?channel=a,b`. Backend ORs via
`Video.channel_id.in_()`; the count endpoint shares the same params object.
Two things the migration exposed:
- SavedViewsWidget fed DB blobs straight into serializers that now index an
array — the FeedFilters type says nothing about what an older build wrote, so
the blob is coerced where it enters. It white-screened the app before this.
- Chips carried channelName to label themselves. Without it they resolve names
from the channel cache, so they now fall back to the id, not a shared
"This channel" that would render N identical chips while the cache loads.
Channel page: the actions sat in the header's top-right corner, a page-width
mouse trip from anything else you'd click. There are only ever two tabs, so
they ride the tab row now, right next to About.
Channel picker: drop the 50-row cap — 320 plain buttons cost nothing to render
(the feed virtualises thousands), so the whole list scrolls in the same box. The
cap was caution, not a measurement. The box gets the app's usual edge-fade to
hint at more, and a line saying "Priority first, then A-Z" — the backend orders
by `priority DESC, lower(title)`, which reads as random to anyone who hasn't set
a priority.
The picker's list is a flex column with a max height, so its fifty buttons were
flex children free to shrink: each collapsed to ~12px and clipped its own text
into an unreadable smudge instead of the box scrolling. shrink-0 on the rows.
Colour, contrast (9.03) and opacity all measured fine — the height was the
broken part, and no colour check was ever going to see that.
`filters.channelId` was settable by exactly one thing: a "?channel=" URL from a
shared view. The panel still rendered a CHANNEL card whose only control was
"remove" — a remove button for a filter the UI could not apply.
Two ways in now. The panel gets a real Channel group (searchable picker,
first-class alongside date/tags/language, so the chip row surfaces it for free);
normalizeLayout appends it to saved layouts, so nobody loses their arrangement.
And a subscribed channel's page gets "Show in my feed", which applies the filter
and keeps the rest of your filters — the point is "this channel, the way I
normally read", which the channel page itself can't answer (it runs show=all,
scope=all over the whole catalog).
The picker owns its channel query so a few-hundred-entry list loads when the
panel is actually opened, not on every feed visit, and says "showing 50 of N"
rather than quietly stopping.
The chip is about YouTube subscription syncing, so it only earns its place on
the feed those subscriptions fill and the manager where you curate them. On
Plex — a different source entirely — plus playlists, downloads and stats it was
just noise.
Its tooltips were written for a roomy sidebar block and read like documentation
on a pill: the counts explanation went 120 chars -> 48, the full-history one
142 -> 46, and pause/resume drop "background".
Pinning it to the header's right edge read fine on a narrow window and stranded
it on a wide one — it's a status, but it's still part of the pill row. It now
follows the search bar.
Its tooltip was the long explanation of the counts, which is redundant next to a
popover that spells them out and overflowed the viewport from the right edge.
The chip gets the counts themselves; the explanation moves onto the popover's
counts row, where there's room for it.
The rail spent a two-row block plus a divider on per-user sync status. It's
global, not per-module, so it now rides the right end of the floating header as
a pill: the headline count at a glance, an accent dot when something wants you
(paused / missing history), and the counts, state, the channel-manager link and
pause in a popover.
Frees the rail's whole top block. Also drops SyncStatus's two now-dead layouts:
"bar" was a leftover of the top bar removed in 0.42.0, and "rail" lost its only
caller here — the component is chip-only, ~60 lines lighter, and the popover
reuses useDismiss.
A channel page renders the Feed from its own baseline — channel-scoped,
whole-catalog, show-all, live included — so measuring it against the feed's
defaults labelled all five as "filters you applied" when they are just the page
being itself. It also has no filter sidebar (App hides it), and revealing what
the collapsed panel hides is the chip row's only job.
With the filter panel now collapsed to a tab, there was no way to see what was
narrowing the feed without opening it. Every applied filter is now a chip you
can drop on its own, with a Clear all when more than one is on.
The panel's count badge and the chip row read the SAME list (lib/useActiveFilters),
so they can't disagree about what's applied; that list also owns the reset both
"Clear all"s share. Replaces the hand-summed activeCount in Sidebar.
peter