Cleanup of hacky leftovers from the node-tree migration (no behaviour
change):
- Return annotations: the component builders return Node subtrees, not
SafeText strings, but ~40 functions still declared `-> SafeText`. Correct
them to `-> Node` across filters / search_select / date_range_picker /
domain. The genuine string returners keep `-> SafeText`: the Alpine
selectors (GameStatusSelector / SessionDeviceSelector, which build f-string
markup) and the script-tag helpers (CsrfInput / ModuleScript /
ExternalScript / StaticScript).
- layout.render_page / layout.Page / AddForm now accept `Node` in their
`content` / `scripts` / `fields` parameters (TYPE_CHECKING import in
layout to avoid the components import cycle), matching what views already
pass.
- session._session_fields builds a `Fragment(*rows, separator="\n")` instead
of `mark_safe("\n".join(str(row) ...))` — keeps the tree intact so media
could bubble, per the Fragment convention.
- Inline SVG icon children use `Safe(...)` nodes instead of `mark_safe(...)`
strings (filters mode-toggle + collapse icons, date_range_picker calendar
icon).
- _filter_field reads the widget's own id from its node `.attributes`
(`_widget_id`) for the label's `for`, dropping the superfluous `for_widget`
argument that always rendered `for="None"`. Removes the two TODOs whose
premise ("the Component function can't expose the id") the class/node
refactor retired, plus RangeSlider's dead commented-out Label block.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Page() now calls collect_media(content) and emits the ModuleScript /
StaticScript tags itself, so views no longer thread scripts= for
component-owned JS. The list views (game/session/purchase/device/
platform/playevent) compose with Fragment(filter_bar, content) instead of
mark_safe(str(filter_bar) + str(content)) — keeping the node tree intact
so the filter bar's media (filter_bar.js + search_select.js +
range_slider.js, and date_range_picker.js on purchases) reaches Page().
The stats views drop _STATS_SCRIPTS; YearPicker's datepicker.umd.js is
collected from its declared media.
The scripts= argument remains for page-specific glue not owned by a
component (the add-form helpers add_game.js / add_purchase.js /
add_session.js, alongside search_select.js for their form widgets).
Adds regression tests asserting the list and stats pages auto-load their
widget scripts with no scripts= in the view, and documents the node/
media model in CLAUDE.md.
https://claude.ai/code/session_01BKurBhE3Qj25p7Bfsg7EeK
Generic leaf builders (Div, Span, Td, Tr, Th, Ul, Li, Strong, Label,
Template, P) are now generated from one _html_element factory over the
single Element class — the tag name is data, not a per-tag body. Only
elements that add classes/behaviour (Button, Pill, Checkbox, Radio,
Input, A, SearchField, H1, Modal, AddForm, tables) stay hand-written.
All primitives now return Node objects; string-built widgets (Icon,
SimpleTable, YearPicker) return Safe, and YearPicker declares its
datepicker media. Raw concatenation (_popover_html, Popover slot) uses
Fragment.
Node.__str__/__html__ now return a SafeString: a node's rendered output
is safe HTML by construction, so str(node) stays safe when fed back into
a child list or template (matching the old SafeText behaviour and
preventing double-escaping).
Consumers adapted: the form widgets (SearchSelectWidget,
PrimitiveCheckboxWidget) return render(component) so Django gets a safe
string; the session form's manual field markup joins via str(row).
Component tests render nodes to HTML before asserting.
https://claude.ai/code/session_01BKurBhE3Qj25p7Bfsg7EeK
Port FastHTML's proc_htmx as onSwap(selector, initializeElement) in
utils.js, built on htmx.onLoad: it runs an initializer once per matching
element, on initial page load and inside every htmx-swapped fragment.
Migrate search_select.js, range_slider.js, filter_bar.js and
add_purchase.js to it, removing the hand-rolled DOMContentLoaded +
htmx:afterSwap listeners and per-element guard flags. This also fixes a
latent bug: both events passed the Event object as range_slider's
"force" parameter, so every htmx swap force-re-initialized all sliders
and stacked duplicate listeners. The collapse button's
window.initRangeSliders() call was a no-op (handles are positioned in
percentages, so hidden-init is safe) and is removed with the global.
Add e2e/test_widgets_e2e.py covering the onSwap lifecycle (initial-load
init, htmx-swap init, single-fire toggles) plus FilterSelect pills and
the add-purchase type toggle. The synthetic page in
test_search_select_e2e.py now loads htmx and search_select.js as a
module, matching the new initialization path.
https://claude.ai/code/session_01BKurBhE3Qj25p7Bfsg7EeK
Serve alpinejs 3.15.12, @alpinejs/mask 3.15.12, flowbite 2.4.1 and
flowbite-datepicker 2.0.0 from games/static/js/ instead of jsdelivr, so
pages (and browser tests) work without network access. Adds the
StaticScript primitive for vendored UMD bundles, which cannot be loaded
as ES modules.
https://claude.ai/code/session_01BKurBhE3Qj25p7Bfsg7EeK
Implements the DateRangePicker design: a DateRangeField that looks like a
single input but splits each date into DD/MM/YYYY part inputs (ordered by
the new common.time.dateformat_hyphenated), and a DateRangeCalendar popup
with a preset column (today, yesterday, last 7/30 days, this/last month,
this year), anchor-style range picking with an outlined/filled/muted range
track, and a Cancel / Clear / Select footer.
Typing fills each part's placeholder from the right (YYYY -> YY19 -> 1987),
auto-advances between parts, and Backspace/Delete reverts the active part.
The committed value lives in hidden ISO {prefix}-min/{prefix}-max inputs --
the same contract as DateRangeFilter, so filter_bar.js needs no changes.
As a tryout, the Purchased filter in PurchaseFilterBar now uses the
DateRangePicker; Refunded keeps the native-date DateRangeFilter, and the
native-path e2e tests were repointed at it.
Includes unit tests for the component family and the filter-bar
integration, plus Playwright e2e tests for segment entry, calendar
picking, presets, and footer actions.
https://claude.ai/code/session_017b75KJAu4kNNpZPu9NAPBM
Closes#10.
Backend (common/criteria.py):
- Treat `excludes` as an always-orthogonal AND'd negative across both
MultiCriterion and ChoiceCriterion; the modifier now governs only the
`value` (include) set. This removes the prior divergence where
MultiCriterion.EXCLUDES dropped the excludes list and ChoiceCriterion.EXCLUDES
swapped include/exclude into a positive.
- Fold INCLUDES / INCLUDES_ALL / EXCLUDES (+ EQUALS/NOT_EQUALS aliases) into the
shared _SetCriterion base so the two subclasses cannot drift; remove _extra_q.
M2M "has all" (games/filters.py):
- PurchaseFilter._games_to_q builds a pk__in subquery with one join per value so
INCLUDES_ALL on the many-to-many games field works in a single .filter()
(a naive Q(games=a) & Q(games=b) collapses to one join and matches nothing).
UI (FilterSelect + filter_bar.js):
- Add an optional any/all/none match-mode <select> (INCLUDES/INCLUDES_ALL/
EXCLUDES) rendered before the pills via a new `leading` slot on the shared
combobox shell. A native control so its value is its state. readSearchSelect
serialises it to data-match; filter_bar folds it into the criterion modifier.
Orthogonal to the (Any)/(None) presence pseudo-options and the exclude channel.
- Enable it for the M2M Purchase.games field (INCLUDES_ALL is only meaningful
for multi-valued relations). Styled with already-compiled utilities.
Tests: harmonized EXCLUDES + INCLUDES_ALL for both criterion types, a DB-backed
INCLUDES_ALL vs INCLUDES contrast on Purchase.games, and FilterSelect /
PurchaseFilterBar rendering + round-trip of the match mode.
https://claude.ai/code/session_01KwVrGFbq13mZdhDL9G6zhg
- Move {id,label} stripping into _SetCriterion.from_json() so both
MultiCriterion and ChoiceCriterion normalise at the parse boundary;
the querying layer stays typed (list[int] / list[str]) and clean.
- Revert MultiCriterion to a thin _extra_q() override; _SetCriterion.to_q()
is no longer duplicated.
- JS: readSearchSelect always emits {id, label} objects — no conditional
mixed-type arrays. filter_bar.js stores them as-is for all fields,
removing the fragile isIdField hardcoded list.
- Update tests to use the {id, label} filter format.
https://claude.ai/code/session_01EyAJcMoDktLrY9tSbdHViA
Store {id, label} objects instead of bare IDs in MultiCriterion value/excludes.
FilterSelect pills now render directly from the embedded labels — no DB round-trip
to _resolve_game/device/platform_options. The filter URL and saved presets are
self-describing. MultiCriterion.to_q() extracts ids for querying; bare ints are
still accepted for backward compatibility.
Closes#9https://claude.ai/code/session_01EyAJcMoDktLrY9tSbdHViA
Spell out the abbreviated data-ss-* hook attributes (data-search-select-option,
-label, -mode, -template, -action, -type, -modifier, -modifier-option, -pills,
-search, -options, -no-results) and the JS expando properties (_searchSelectInit,
_searchSelectLabel, _searchSelectDirty, _searchSelectOption) across components,
JS, and tests — no abbreviations left in the widget's hooks.
https://claude.ai/code/session_01XzhXvMvw42CQGc9kmin3GS
Rename abbreviated identifiers in the PR's code to full words: tpl→template,
e→event, el→element, removeBtn→removeButton, and single-letter loop variables
(o→option, g/d/p→game/device/platform, v→value/modifier_value). Add a
'name variables with complete words' convention to CLAUDE.md.
https://claude.ai/code/session_01XzhXvMvw42CQGc9kmin3GS
Eliminate the Python/JS class-string duplication: the server renders hidden
<template> prototypes (row, pill, include/exclude/modifier pills) using the same
component functions, and search_select.js clones them, filling only the
[data-ss-label] slot, value, and data-* attrs. All Tailwind class strings and DOM
structure now live solely in the Python components — the JS no longer hardcodes
any class. Pill gains an opt-in label_slot; the shell takes a templates list.
Companion issue #8 tracks the further HTMX-idiomatic step of returning rendered
row HTML from the search endpoint.
https://claude.ai/code/session_01XzhXvMvw42CQGc9kmin3GS
- Wire the long-defined-but-unused _FIELD_CLASS into the container so pills and
the search input form a single padded flex row; the flex-1 input now fills the
widget instead of looking unclickable inside a larger box (affects both
SearchSelect and FilterSelect via the shared shell).
- Filter option labels get text-body so they're readable on dark backgrounds.
- Filter +/- buttons get text-body (readable at rest) and hover:border-brand-strong
so the border stays visible against the brand hover fill.
- Mirror the filter class changes in search_select.js and rebuild base.css.
https://claude.ai/code/session_01XzhXvMvw42CQGc9kmin3GS
FilterSelect fully replaces it: delete SelectableFilter and its _selectable_*
helpers, the now-unused _get_filter_options, selectable_filter.js, and the .sf-*
rules in input.css (rebuilt base.css). The three list views load search_select.js
instead of selectable_filter.js. Drop the SelectableFilter export and refresh
docs/comments that referenced it.
https://claude.ai/code/session_01XzhXvMvw42CQGc9kmin3GS
Replace the bespoke SelectableFilter in all three bars with FilterSelect: enum
fields (status, type, ownership) pre-render their fixed options; model-backed
fields (game(s), platform, device) use the search endpoints with prefetch and
resolve only the selected ids to pill labels — dropping the per-page queries that
fetched every game/platform/device. filter_bar.js now reads filter-mode
SearchSelect widgets via readSearchSelect (data-included/excluded/modifier),
preserving the {value, excludes, modifier} JSON and id Number() coercion; the
redundant session game/device blocks are gone. Drop FilterBar's now-unused
platform_options param. Rebuild base.css for the inline filter-pill utilities and
update the bar tests to the new markup.
https://claude.ai/code/session_01XzhXvMvw42CQGc9kmin3GS
Dispatch on data-ss-mode: in filter mode, value rows (server-rendered or fetched
via buildRow) carry +/- buttons that add include/exclude pills, and pinned
modifier pseudo-options set a lone, mutually-exclusive modifier pill. Pill removal
handles the modifier flag; filter pills carry no hidden inputs. Extend
readSearchSelect to serialise filter widgets into data-included / data-excluded /
data-modifier (the shape the filter bar consumes), leaving form widgets'
data-values path unchanged. JS class strings mirror the FilterSelect constants.
https://claude.ai/code/session_01XzhXvMvw42CQGc9kmin3GS
Introduce a general 'prefetch' option (rows to load on first open, default 0 =
unchanged) carried as data-prefetch. Rework the JS search so a search_url widget
filters its loaded window instantly on every keystroke while issuing a debounced
server request for the rest, with an AbortController so a slower earlier response
can never overwrite a newer one. No-results stays hidden until the server
response decides it, avoiding a flash over an incomplete window. On first focus a
prefetch-enabled widget seeds its window immediately. Rename single-letter locals
to full words while reworking these functions.
https://claude.ai/code/session_01XzhXvMvw42CQGc9kmin3GS
- Parameterize SearchSelectWidget with a required options_resolver so
each widget explicitly names its resolver instead of implicitly using
_game_options
- Add autofocus support: SearchSelect forwards it to the search input,
and SearchSelectWidget extracts it from Django's attrs dict
- Add _device_options and _platform_options resolvers (single pk__in
queries, same pattern as _game_options)
- Add /api/devices/search and /api/platforms/search endpoints
- Switch PlayEventForm.game from plain Select to SearchSelectWidget
(preserving autofocus), and use SingleGameChoiceField for correct labels
- Switch SessionForm.device to SearchSelectWidget
- Switch PurchaseForm.platform and GameForm.platform to SearchSelectWidget
- Wire ModuleScript("search_select.js") into add/edit playevent and
add/edit game views
https://claude.ai/code/session_013fpJD54HxRgxRv2xzwXGNo
lukas
Claude