The staging deploy runs on push to a non-main branch and tries to comment
the staging URL on the branch's PR. When the branch is pushed before the PR
exists (the common case), the comment is skipped and never reappears once the
PR is opened.
Add a pull_request [opened, reopened] trigger and move the comment into its
own job that runs both after a successful push-deploy and on PR open/reopen.
The branch is taken from github.head_ref on PR events and github.ref_name on
push; the existing dedupe-by-body keeps the two paths from double-posting.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RF5L4HtbcykTfY9YUYGds3
Add filter_url(), a reverse()-style helper that builds a URL to a filtered
list view from a filter object (target inferred from the filter type).
Add OperatorFilter.where(**lookups), a Django-.filter()-style ergonomic
constructor that resolves each field's criterion class from its annotation
(shared with from_json via _criterion_class_for, removing duplication).
Make SessionFilter.timestamp_start/timestamp_end DateCriterion applied via
the __date lookup, so date ranges over the timestamp columns are expressible.
Wire the navbar 'today' / 'last 7 days' totals as links to the matching
filtered session lists, and align the 'last 7 days' total to the same
calendar-day window so the number matches the list it links to.
Stats-table and game-detail links remain a follow-up (see spec).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RF5L4HtbcykTfY9YUYGds3
The flip-up branch cleared inline `top` to "", which let the menu's
`top-[105%]` utility class reassert top:105% on the now-fixed element —
collapsing the menu to a 2px sliver below the viewport, so toggles near the
viewport bottom appeared not to open. Set the unused anchor to "auto" so the
inline value wins over the class. Add an e2e regression for the flip-up path.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The device/status dropdown menu is absolutely positioned inside the session
list's overflow-x-auto wrapper. Because overflow-x:auto forces overflow-y:auto,
a menu taller than a short table was clipped (issue #39). Open the menu with
position:fixed anchored to its toggle so it escapes the clipping ancestor,
bound it to the viewport with an internal scroll, flip it up when there is more
room above, and reposition on scroll/resize while open.
Fixes#39.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
base.css is a Tailwind build artifact already listed in .gitignore, but it
had been tracked since it was first committed (gitignore can't untrack an
already-tracked file), so it kept getting re-committed against intent. Untrack
it to match js/dist (the TS artifact, also gitignored + untracked).
Because nothing in the test path rebuilt it, e2e/static serving relied on the
committed copy. Add 'css' (and 'ts') as prerequisites of the test/test-e2e
make targets and a Build CSS step in CI so the stylesheet is generated before
tests run.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Flowbite re-initialises popovers on every htmx swap. A popover hidden via
Tailwind `invisible` (visibility:hidden) still occupies layout, so once
Popper parks it with a transform offset it expands the table's
overflow-x-auto wrapper and a spurious scrollbar appears (horizontal here,
vertical in #40). Add `[&.invisible]:hidden` so the popover is removed from
layout while hidden; Flowbite drops `invisible` on show, restoring display.
Relates to #40. e2e regression covers no-overflow-after-swap plus
popover-still-shows-on-hover.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Delete stale _session_row_fragment; end_session and reset_session_start
return the canonical row plus an OOB navbar-playtime fragment. Clone keeps
HX-Refresh since it changes row count. Fixes#53.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Adds a confirm-gated button on running sessions in the session list that
sets timestamp_start to now (issue #33). The htmx path returns HX-Refresh;
ButtonGroup gains optional hx_confirm/hx_swap keys.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The comment described what the code does (find wrapper by name attr),
not why. The locator is self-explanatory.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The id (e.g. id_related_game) sat on the <search-select> wrapper, a
non-labelable custom element. Consequences:
- <label for="id_X"> focused nothing (a11y gap)
- .disabled / .focus() on #id_X silently no-oped
- add_purchase.ts needed a [data-search-select-search] descendant
workaround to gate related_game on the type field
id is now on the [data-search-select-search] <input>, making it a real
labelable, disableable control. add_purchase.ts drops the workaround
and gates via #id_related_game directly. E2e tests updated; new test
asserts label-click focuses the search box.
Closes#30
Finish the behavioural refactor from #28: no first-party JS lives on the
global object solely to be reachable from a server-rendered inline on*
attribute, and no inline Alpine blobs remain in the filter bar / year picker.
- Filter-bar collapse: drop the inline onclick for a delegated click listener
on the persistent <filter-bar> custom element (data-filter-bar-toggle). The
inner #filter-bar body is htmx-swapped while connectedCallback does not re-run,
so delegation on the host preserves the swap-survival the inline handler had.
- YearPicker: convert the Alpine x-data/x-on/x-ref/_pickerInstance f-string into
a <year-picker> custom element with typed props (YearPickerProps). Behavior
moves to ts/elements/year-picker.ts; ts/year_picker.ts and _YEAR_PICKER_MEDIA
are removed. The builder lives in primitives.py (next to YearPicker) to avoid a
circular import; registration stays in custom_elements.py for codegen.
- Add bindPopupDismiss (ts/utils.ts): shared Escape + outside-click dismiss with
a cleanup return and an extraInside hook for popups mounted on document.body.
Adopted by date-range-picker.ts (1:1) and year-picker.ts (Datepicker popup is
body-mounted, passed as an extra inside root).
Follow-up #49 tracks unifying popup/dismiss/positioning across the remaining
dropdown/search-select/Flowbite cases.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Staging stores sessions in machine-local SQLite with no shared volume.
Fly's default deploy provisions two machines (HA), so requests after
login could land on the machine that never wrote the session row,
bouncing logged-in users straight back to the login page.
Deploy with --ha=false and scale count 1 so each per-branch staging app
runs on exactly one machine.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YWAvjVEAibhwbeVgbYmg94
Replaces the four onSwap-based widgets with TypeScript custom elements
following the pattern from PR #16. Each widget gets a class extending
HTMLElement with connectedCallback/disconnectedCallback, typed props via
register_element + gen_element_types codegen, and lives in ts/elements/.
- range-slider: RangeSliderElement; Python uses _RangeSlider builder
- date-range-picker: DateRangePickerElement; Python uses _DateRangePicker builder
- search-select: SearchSelectElement; Python uses _SearchSelect builder;
data-* attrs become plain attrs (data-name -> name, data-search-url -> search-url, etc.)
- filter-bar: FilterBarElement; props carry preset URLs; onclick/onsubmit
attrs replaced with data-filter-bar-* sentinel attrs; all window.* globals removed
Deletes ts/range_slider.ts, ts/search_select.ts, ts/date_range_picker.ts,
ts/filter_bar.ts. Updates all tests and e2e pages to use the new element
selectors and script paths (dist/elements/<tag>.js).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Inputs rendered at different sizes: the SearchSelect was 54px tall with a 16px
font (the wrapper's p-2 and the inner box's forms-plugin padding stacked, and
the wrapper had no text-sm), and the string-criterion filter input had no
text-sm (16px) with p-2. Native inputs are 42px / 14px / px-3 py-2.5.
Align both to the native size:
- SearchSelect wrapper supplies px-3 py-2.5 + text-sm and the inner search box
zeroes its own padding (p-0), so the field is one 42px row like an input.
- String filter input gets text-sm + px-3 py-2.5.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The string-criterion filter input (e.g. Playevent Note) had transition-all, so
on focus its border-color animated from the near-white default to the focus
blue — a visible white "blink". It also lacked an explicit border width and
focus:border-brand/ring-brand, so its focus came from the forms-plugin default.
Drop transition-all (the other inputs snap to focus), add border +
focus:border-brand focus:ring-brand, and use rounded-base — so it focuses
instantly to the same brand border as every other input.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The SearchSelect wrapper had no border and no focus styling, so it looked unlike
a native input (which has border-default-medium → border-brand on focus). Add
border-default-medium + focus-within:border-brand/ring to the shared container
class — focus-within because the focusable element is the inner search box. This
also makes filter-bar comboboxes consistent with the other filter inputs, which
already have borders.
e2e asserts the wrapper border matches a native input's at rest and turns brand
on focus.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Disabled controls looked inconsistent: the SearchSelect faded (opacity-50)
while native inputs used a solid strong surface. Standardize on the faded look
(opacity-50) the user preferred, via shared constants so every form element
matches:
- DISABLED_CONTROL_CLASS (disabled:opacity-50 disabled:cursor-not-allowed) on
the control — native inputs/select/textarea via PrimitiveWidgetsMixin, plus
the Checkbox component (previously had no disabled style).
- DISABLED_WITHIN_CLASS (has-[:disabled]: wrapper variant) for composite
controls like SearchSelect whose disabled state lives on an inner element.
e2e asserts a disabled SearchSelect and the Name input fade identically
(opacity 0.5) and return to 1 when enabled. CLAUDE.md documents the shared
disabled constants.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Form controls were styled "at a distance": Django renders bare
<input>/<select>/<textarea>/<label>, so input.css reached in with ID-scoped
#add-form descendant rules plus a global form *:disabled rule and .errorlist.
The #add-form ID specificity forced state rules to climb, needed
:not([data-search-select-search]) carve-outs, and broke on markup changes — it
surfaced as the add_purchase Name/related_game fields not reading as disabled.
Components now own all form styling via utilities on the elements themselves:
- PrimitiveWidgetsMixin stamps INPUT/SELECT/TEXTAREA_CLASS (incl. disabled:
variants) onto native widgets by type, skipping SearchSelect (self-styled)
and checkboxes.
- New FormFields(form, *, extras=...) renders label + control + errors + row
layout with their own classes (replaces form.as_div()); the <form> owns its
flex layout. extras appends a node into a named field's row (session
timestamp buttons).
- AddForm/purchase/session render via FormFields; login too — a new
LoginForm(PrimitiveWidgetsMixin, AuthenticationForm) styles its inputs and
auth.py renders it via FormFields + a StyledButton (was as_table).
- input.css loses the entire #add-form block, the global :disabled rule, and
.errorlist. State (disabled:) now lives on the element — no specificity wars,
no carve-outs, robust to markup edits.
Tests: error rendering uses the component class (not .errorlist); add-form
labels/inputs carry their own classes; e2e login fixtures click the Login
button by text (submit is now a <button>); Name disabled cursor asserted.
CLAUDE.md documents the no-styling-at-a-distance + FormFields conventions.
513 passed; lint/format/ts-check clean.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Excluding the inner search box from the global disabled rule also dropped its
cursor: not-allowed, so the pointer flickered between not-allowed (wrapper) and
the text I-beam (input) when moving across the disabled widget. Add
disabled:cursor-not-allowed to the search input so the cursor stays consistent.
e2e: assert the disabled inner input computes cursor: not-allowed.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The disabled widget showed two clashing surfaces in dark mode: the wrapper
(faded via has-[:disabled]) plus the inner search input, which picked up the
global disabled-input fill from common/input.css
(`form input:disabled { background: neutral-secondary-strong }`). That rule is
unlayered, so it beat any utility override on the input.
Exclude the SearchSelect's inner search box from that global rule
(`:not([data-search-select-search])`) so it stays transparent — the wrapper is
then the single faded surface. Standalone inputs (e.g. the Name field) keep
their distinct disabled surface, unchanged.
e2e: assert the disabled inner input computes transparent background (one
element), alongside the existing wrapper-opacity check.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Follow-ups on the add-form fixes:
- syncSelectInputUntilChanged now actually stops mirroring once the user edits
the target (the "UntilChanged" contract). The old focus-based stop was a
no-op (wrong removeEventListener reference), so live sync kept clobbering a
manually-edited Sort name. Track dirty targets in a Set keyed by syncData
index; programmatic writes don't fire "input", so only real user edits mark a
target dirty. Drops the dead focus listener.
- SearchSelect now greys itself when disabled, via has-[:disabled]: utilities on
its container class — the visible "box" is the wrapper <div>, so disabling the
transparent inner input alone left it looking active. The component owns its
disabled appearance; callers only toggle the inner control's `disabled`.
- Document the composite-widget disabling approach in CLAUDE.md and the
SearchSelect docstring.
Extends the e2e tests: sync drops after a manual Sort name edit; disabled
related-game wrapper computes opacity 0.5 (and 1 when re-enabled).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
lukas
Claude