Deferring was the one action that changed your list and said nothing about
it. The task disappeared, no message, and Cmd/Ctrl+Z would not bring it
back because the undo stack only recorded deletes and clear-completed.
That was worse than it sounds. The app had just taught people, via the bar
that appears when something is deleted, that things which vanish come with
a way back. Deferring quietly broke that promise, and anyone who misread
the arrow would reasonably believe they had destroyed the task.
It now uses the same bar, saying which way the task went, and undo sends
it back to the row it left rather than the bottom of the list. That needed
an insert position on moveTaskToDay, which is tested along with the round
trip, clamping and the default append.
Undo for a move is handled separately from applyUndo, which edits a single
day's array and cannot express something that spans two.
Two pieces of copy that were making the app harder to read than it needs
to be.
Carry-over happened in total silence. Open the app after a weekend and
unfamiliar tasks were sitting in Today with nothing explaining where they
came from, which reads as a bug rather than the feature it is. A line now
says how many were brought forward, on the day it happened, and it can be
dismissed.
The day control was a single button labelled with the day you were already
on, which then took you somewhere else. Clicking it was the only way to
find out what it did, and the date underneath already said where you were.
It is two options now with the current one marked, so the control states
both where you are and what the alternative is without being touched.
Also drops labelFor from dates.js along with its tests. It existed to
label that one button and nothing uses it now.
Undo existed but nothing said so. The only route was Cmd/Ctrl+Z, which was
documented in a README nobody reads mid-task, and which does not exist on
a phone at all. On touch there was simply no way to get a deleted task
back.
Deleting a task or clearing completed ones now brings up a short-lived bar
with a real Undo button, so touch has a route for the first time. It also
shows the keyboard shortcut, which is the point: someone learns it once,
here, at the only moment they actually care.
It uses the existing tokens, sits out of the way at the bottom, clears
itself after seven seconds and goes as soon as it is used. The shortcut
badge is hidden where no keyboard exists rather than advertising a chord
nobody can press.
Fixing a typo meant deleting the task and typing it again, which also cost
you its position in the list. Click the text to edit it. Enter saves,
Escape cancels, clicking away saves.
The rename rule lives in tasks.js and is tested. It refuses blank input
rather than treating it as a delete, since selecting all and hitting enter
by accident should give you your task back, and it returns the original
array untouched when nothing changed so an idle edit does not write to
storage.
The interesting part was what editing collides with. Keystrokes are
stopped from reaching the row, or Delete would remove the task you are
typing into and Alt+Arrow would reorder it mid-edit. A drag cannot start
from inside the field, and the click a drag leaves behind is swallowed, or
dropping a row would open it for editing. Switching day, deferring a task
and the midnight rollover all close an open editor, so it cannot be left
hanging over a row that is no longer there.
The app is built around two days and had no way to move anything between
them. "I won't get to this today" is the most obvious thing someone wants
from a two-day list, and the only route was deleting the task and typing
it again somewhere else.
Each row gets an arrow next to delete. On Today it sends the task forward,
on Tomorrow it points back, so one control covers both directions and the
label says which one you are getting. Alt+Right and Alt+Left do the same
from the keyboard, and each only acts in the direction that makes sense
from the day you are on.
The move itself is in src/lib/defer.js with tests, since it touches two
date keys at once and the failure modes are worth pinning down: a missing
id, the same day twice, an empty destination, and a task that somehow
already exists on the other side.
A pass over desktop, tablet and phone widths, measuring rather than
eyeballing.
The one that matters: a task could not be deleted on a phone. The delete
button is revealed on hover, touch devices report hover: none, so it sat
at zero opacity with nothing to reveal it. The README told people to hover
over a task and click it. It is now always visible where hover does not
exist.
Touch targets were under the 44px minimum: the checkbox at 24, delete at
32, the footer links at 23 tall. They now get 44px hit areas from overlays
that leave the drawn size alone, behind a pointer: coarse query so a mouse
keeps small precise targets and the full row stays draggable.
Two contrast failures in light mode. The date was the brand blue on the
near-white background at 3.54:1, so text now uses a darker --accent-text
and reads 5.14:1. Completed rows carried opacity 0.8 over an already
mid-contrast blue, which came to 3.34:1; the tint, the strikethrough and
the colour say "done" well enough without it, and dropping it gives 4.78:1.
Both themes now pass AA everywhere measured.
The checkbox sat vertically centred, so on a task wrapping to five lines
it floated in the middle of the block. Rows align to the top and task text
gets a 24px line box, matching the checkbox exactly, which also spaces
wrapped lines better.
Reduced motion was only honoured by the empty-state drawing. Svelte's
transitions run in JavaScript and never saw the media query, so tasks
still flew in on a stagger. The component reads the preference directly
and collapses its durations, and CSS covers the rest. The lift on a
dragged card stays, since it tracks the pointer rather than playing at you.
Long unbroken words already wrapped and no breakpoint scrolled sideways.
Two things that were deliberately left out of 2.0.0.
Export writes every stored day to a JSON file. Import reads one back, and
only ever adds: a task whose id is already there is left alone, so
importing the same file twice does nothing and importing into a list
you're using can't lose work. That is also why it needs no confirmation
dialog. The trade is that import restores rather than reverts.
Both sit in a quiet line under the task list rather than the header, since
they get used about twice a year and the header is what you look at all
day.
For the PWA half, the service worker is about fifty lines with no
dependency, because the strategy falls out of how Vite builds. Documents
go network first and fall back to cache, so a deploy is picked up as soon
as you're online and nobody ends up stuck on an old build. Fingerprinted
assets go cache first and are kept, since their names change when their
contents do. No build-time asset manifest needed.
Icons are SVG in the manifest, which stays sharp at any size and costs
about a kilobyte, plus one 180px PNG because iOS wants a raster
apple-touch-icon. Two theme-color metas so the phone status bar follows
the theme, and safe-area padding on the header, without which the header
sits under the clock once installed on an iPhone.
Verified against the production build: worker registers and claims the
page, shell and assets land in cache, and with the server stopped the app
still loads, adds a task and persists it.
Removes docs/superpowers, which held a design note and an implementation
plan. Those were working notes rather than anything the project needs, and
they don't belong in the repo.
Everything else here is wording: em dashes swapped for ordinary
punctuation across the README and the code comments, and a few sentences
straightened out.
The README described migration that never ran, and listed Backspace as a
delete key.
It now says what actually happens: unfinished work carries across gaps,
completed tasks are cleared with their day, storage prunes itself, and old
keys convert on first open. Adds Cmd/Ctrl+Z and Alt+Arrow, the touch long
press, the screenshot's new path, and how to run the tests.
Aculix Technologies