{% extends "vantage/base.html" %}
{% load static vantage vantage_inline vantage_slots i18n %}
{% block content %}
{% comment %}
Resolved once, then handed to every tag that needs it. The allowlist behind it walks
the model's fields and re-derives what the change form would allow, so recomputing it
per cell on a fifty-row page would be a quiet cost on exactly the models that are
already the widest. Inside the block on purpose: a tag outside one never runs in a
template that extends another.
{% endcomment %}
{% inline_config panel as inline %}
{{ panel.get_label_plural }}
{% if panel.description %}
{{ panel.description }}
{% endif %}
{% if list_prefs %}{% include "vantage/partials/list_prefs.html" %}{% endif %}
{% comment %}
Every link carries the active search, filters and sort, so you export what you
are looking at rather than the whole table.
{% endcomment %}
{% translate "Export" %}
{% slot "before_list" %}
{% comment %}
One element wraps the whole list: the control band, the filter rail and the table.
It carries data-grid, and that is why it is out here rather than around the table
alone. vantage-grid.js resolves every control it owns with root.querySelector, so
the density and column buttons have to be inside this element — widening it is what
allowed four stacked control strips to collapse into one band *above* the two
columns. That is also what puts the rail and the data on one baseline: neither
column opens with a line of chrome that the other one lacks.
The page links are handed over as data rather than scraped from the pagination
block below, so `[` and `]` carry the active search, filters and sort — the server
already built those URLs correctly once.
{% endcomment %}
{% if searchable or rows or saved_view_write_url or timeline_field or patterns.soft_delete or has_active_filters %}
{% include "vantage/partials/list_toolbar.html" %}
{% endif %}
{% comment %}
The rail sits beside the table on a wide screen and stacks above it on a narrow one,
which is the grid doing it — no second markup path and no breakpoint-specific DOM.
{% endcomment %}
{% if filters %}
{% endif %}
{% if analytics.stats or analytics.charts or analytics.deferred %}
{% include "vantage/partials/analytics.html" %}
{% endif %}
{% if confirming %}
{{ confirming.label }}?
{{ confirm_text }}
{% blocktranslate %}This will apply to {{ confirm_scope }}.{% endblocktranslate %}
{% endif %}
{% comment %}
The keyboard handlers bind to the table inside the grid element, never to the
document, which is how the single-character shortcuts stay compliant with WCAG
2.1.4 (exemption 3, "active only when the component has focus").
{% endcomment %}
{% if inline.enabled and rows %}
{% comment %}
Loaded here rather than from base.html, the same arrangement the grid controls and
the command palette use: a page with nothing editable on it pays for neither file.
{% endcomment %}
{% comment %}
An empty form, and it is load-bearing. Every inline control carries
form="v-inline-form", which re-parents it away from the bulk-action form it is
physically nested inside. Without that, pressing Enter in a cell implicitly submits
a bulk action, and every cell on the page rides along in the POST body.
It must sit outside #v-action-form: HTML forbids nesting forms, and the parser
silently drops the inner one, which would put every control back in the action form
without anything looking wrong.
{% endcomment %}
{% comment %}
The changeset bar. Persistent rather than a toast: it is the review step, and a
review step that fades after four seconds is not one. Hidden until something is
staged, and the count goes in the button's own label so nobody commits twelve edits
believing they are committing one.
{% endcomment %}
{% comment %}
Announcements for cell mode and for the outcome of a save. A region of its own
rather than the grid's, so an edit never overwrites "row 4 of 55" mid-sentence.
{% endcomment %}
{% include "vantage/partials/conflict.html" %}
{% comment %}
After vantage-grid.js in document order, which is what deferred scripts guarantee.
That ordering is not cosmetic: the grid's table-scoped keydown listener is
registered first, so a cell control that stops an arrow key from propagating is
stopping something that already exists rather than racing it.
{% endcomment %}
{% endif %}
{% if rows %}
{% comment %}
The quick-peek drawer, non-modal on purpose: the list behind it stays live and
focus is never trapped, so Tab walks out of it the way it walks out of any other
region. A modal peek would be a worse detail page, not a faster one.
It is empty until something is peeked; the content is fetched from the record's
own detail URL, which means there is exactly one server-side renderer for a
record and the drawer can never drift from the page it previews.
{% endcomment %}
{% endif %}
{% if page.paginator.num_pages > 1 %}
{% blocktranslate with start=page.start_index end=page.end_index total=paginator.count %}{{ start }}–{{ end }} of {{ total }}{% endblocktranslate %}