Skip to main content

makeover_webview/
lib.rs

1//! The webview renderer for [`makeover_layout`].
2//!
3//! <!-- wiki: makeover-webview -->
4//!
5//! # The renderer that needs no palette
6//!
7//! `makeover-immediate` and `makeover-tui` both take a `Palette`, because egui
8//! and a terminal need an actual colour before they can put anything on
9//! screen. A webview does not: `var(--surface-raised)` *is* the late binding,
10//! and the browser resolves it against whatever `themes.js` last wrote onto
11//! `:root`.
12//!
13//! So this crate emits text naming intents, and never learns a colour. It is
14//! the deferral rule with no adapter in the way, and it is why the webview was
15//! always the wrong renderer to derive a vocabulary from: it can express
16//! anything, so it never pushes back.
17//!
18//! # Phase A: the stylesheet
19//!
20//! This module emits component CSS and no markup, deliberately. GoingsOn has
21//! 145 `innerHTML` sites and Balanced Breakfast 175 `createElement` sites, so
22//! moving markup is a migration where adopting a generated stylesheet is not.
23//! The apps keep every line of their markup and gain the classes.
24//!
25//! It is not a deletion either, which this header claimed until the measurement
26//! came in. Adoption across goingson removed 49 declarations net and *added* 25
27//! lines: a rule loses its depth declarations and gains a variant selector next
28//! to it, so the file stays the same size. What phase A moves is where depth is
29//! defined, not how much CSS exists. Numbers and method in the wiki note under
30//! "The deletion test, run".
31//!
32//! The bevel properties are byte-identical to what both apps already
33//! hand-write, which is asserted below.
34//!
35//! Some of what phase A emits is not a look but the withdrawal of one. A
36//! renderer that picks its element from the description inherits that element's
37//! user-agent chrome, and [`reset`] is where a primitive says which parts of it
38//! were never asked for.
39//!
40//! # Phase B: the markup, one description at a time
41//!
42//! [`form`] renders [`makeover_layout::Field`], which is the half of phase B
43//! whose description is settled. It emits strings, because both apps
44//! interpolate their fields into larger string-built forms and returning nodes
45//! would rewrite those too. It owns its own escaping, on the reasoning in that
46//! module: a Rust encoder can cover element text and attribute values with one
47//! function, where the apps need four and have to choose correctly at every
48//! call site.
49//!
50//! [`facet`] renders `makeover_layout::Facet`: a dimension a set is narrowed by,
51//! and the one phase-B emitter whose markup an app is not keeping, because the
52//! markup it replaces was two mechanisms rather than one. A tag's selection and
53//! a tag's browse position were separate state on MNW's discover page, which is
54//! why every filter row there carries a tick box *and* a chevron; one gesture
55//! doing both is what lets the second one go.
56//!
57//! [`list`] is the other half: column tracks, the narrowing rules, and the cell
58//! containers a row is made of. It stops at the cell boundary and does not
59//! render what goes inside one, on the reasoning in that module. So phase B is
60//! now the frame around content in both directions, and what an app still owns
61//! is the content itself.
62//!
63//! # What phase A settled, and what it costs
64//!
65//! Decided 2026-07-29 against goingson's `styles.css` rather than against a
66//! component list. The useful finding there was that `.btn` (line 644),
67//! `.card` (768) and `.tag, .badge` (882) each hand-write the same
68//! composition, so three quarters of phase A is one rule with several names.
69//!
70//! Two of the four decisions change how goingson looks, and adoption should
71//! not be described as a pure deletion:
72//!
73//! - **Pressed carries its fill.** [`interactive_rules`] emits
74//!   [`Depth::pressed`] whole. goingson presses to `--surface-sunken` today and
75//!   will press to `--surface-well`, and hovers to `--surface-overlay` today
76//!   and will hover to `--hover-surface`. Since `surface-well` inverts by theme
77//!   where `surface-sunken` does not, a dark theme presses *lighter* than it
78//!   hovers. That falls out of `makeover`'s own derivation, which says outright
79//!   that `surface-sunken` cannot serve as a well, so if it reads wrong the
80//!   answer is there and not here.
81//! - **Badges go flat.** See [`token_rules`].
82//!
83//! The other two: the progress trough is renderer-local and the scrollbar
84//! track was dropped ([`component_rules`]), and no class prefix ships by
85//! default, so adoption means deleting the app's hand-written rule in the same
86//! commit that adds the generated one. `.card`, `.badge` and the tab classes
87//! all already exist in goingson, and while both rules exist the cascade order
88//! decides which wins. That is the one real risk in adopting this, and it is
89//! why the migration lands per component rather than in one commit.
90//!
91//! # 0.10.0: the states this crate used to leave to its consumers
92//!
93//! [`interactive_rules`] emitted hover and pressed and stopped, because
94//! `makeover-layout` modelled no interaction state. Focus and disabled were
95//! therefore unsayable, and every app completed the primitive from outside the
96//! only way that works: by out-specifying a rule it does not own. goingson
97//! carries 19 such rules and the MNW server 21, and the three focus rings do
98//! not match each other.
99//!
100//! That also blocked the cascade-layer work outright. An app that declares
101//! `@layer` puts its own rules in a named layer, and unlayered declarations
102//! outrank every named layer regardless of specificity, so all of those
103//! overrides lose in the commit that adopts layers. They cannot simply be
104//! deleted, because they are the only thing supplying the missing states.
105//! Emitting the states here is what turns that adoption into a deletion.
106//!
107//! Four states now, in emission order, and the order is load-bearing: they are
108//! all specificity (0,2,0), so disabled beats hover by coming last and by
109//! nothing else. Nothing here reaches for `:not(:disabled)`, which would raise
110//! a selector this crate will shortly be wrapping in its own layer.
111//!
112//! Hover additionally sits inside a capability query now. `makeover-touch`
113//! answers whether a fingertip has hover and `makeover-geometry` spells the
114//! condition; this crate asks and does not decide. goingson's section 60 exists
115//! solely to take the hover state back on touch, which is a fight it should
116//! never have been handed.
117//!
118//! # 0.11.0: the layer contract
119//!
120//! [`stylesheet`] emits into the `makeover` cascade layer ([`CSS_LAYER`], which
121//! lives in `makeover-geometry` because that is the one crate every CSS emitter
122//! in the family already depends on). `makeover-geometry` 0.6.0 does the same
123//! for `geometry.css`.
124//!
125//! The cascade resolves origin and importance, then layer, then specificity,
126//! then source order, and **unlayered normal declarations outrank every named
127//! layer**. So before this, an app that declared `@layer base, components,
128//! responsive` put every rule it owns into a named layer and lost all of them to
129//! this unlayered file, regardless of specificity and regardless of loading
130//! last. Nothing errors when that happens: the CSS is valid, the minifier is
131//! happy, and buttons and badges look subtly wrong.
132//!
133//! That is why the layer belongs here rather than in each app. An app cannot fix
134//! it from its own stylesheet, because the fix is to layer the file it does not
135//! own.
136//!
137//! **What it flips**, and the reason each app wants a look when it bumps the
138//! pin: a generated rule that currently beats an app rule by being more specific
139//! stops beating it. The direction is always "the app wins", which is what the
140//! apps already assume, but a hand-written rule an app thought was dead can come
141//! back to life.
142//!
143//! An app should declare the order once, or the layer's position is decided by
144//! whichever generated file the browser happens to see first:
145//!
146//! ```css
147//! @layer makeover, base, components, responsive;
148//! ```
149//!
150//! [`in_css_layer`] is re-exported for an app that assembles its own stylesheet
151//! from this crate's pieces. goingson builds `tables.css` in its own `build.rs`
152//! out of [`list::narrowing_css`] and [`list::grid_template_columns`], and those
153//! rules are as generated as the ones here, so they belong in the same layer and
154//! this crate cannot put them there on the app's behalf.
155//!
156//! # 0.12.0: the ring gets its own width
157//!
158//! [`focus_rule`] reused [`Emit::border_width`] and emitted a 1px ring. That was
159//! an implementation convenience dressed as consistency with the invalid-field
160//! ring: a bevel and a focus indicator answer different questions, and only one
161//! of them has to be noticed from across a desk.
162//!
163//! Caught while adopting 0.11.0 into goingson, by the check the adoption tasks
164//! ask for. Every consumer had already written its own ring and all three chose
165//! at least 2px: the MNW server 2px across 10 rules, Balanced Breakfast 2px,
166//! goingson 2px on three rules and 3px on the one covering twelve selectors. The
167//! design system was the only thing in the tree saying 1px, so deleting the app
168//! rules in favour of it would have thinned the focus indicator everywhere.
169//!
170//! [`Emit::focus_width`] now carries it, defaulting to `2px`, and the offset is
171//! the same magnitude with its sign off the depth. Both values are the measured
172//! consensus rather than a new opinion.
173//!
174//! # 0.57.0: a suggestion's second line, in place of a reason it cannot be picked
175//!
176//! `1fcf2e9b`, ruled by Max 2026-08-21. `Outcome::Suggestions` now carries
177//! `Candidate` rather than `Choice`, and a candidate has no `unavailable`: a
178//! suggestion that cannot be picked is a row a route should not have offered.
179//! What it has instead is a `detail`, the line that tells it from a row reading
180//! the same -- "Format" under audio, software, writing and video.
181//!
182//! So `.form-suggestion-why` is gone and `.form-suggestion-detail` is here, and
183//! the difference is not only the name. `why` was drawn in the disabled token,
184//! which said the row was refusing. A detail orients rather than refuses, so it
185//! is `--content-muted`, which is what every other secondary line in this crate
186//! reads as.
187//!
188//! No consumer is affected: the only emitter is `quasi-webview`, which moves in
189//! the same cascade.
190//!
191//! # 0.56.0: an interval, as the group MNW already wrote by hand
192//!
193//! [`makeover_layout::FieldKind::Interval`] emits a `role="group"` named by the
194//! field's label, holding one `<input type="number">` per end. That is not a
195//! shape invented here: MNW's discover sidebar has a `role="group"` with
196//! `aria-labelledby` over `min_price` and `max_price` today, written by hand
197//! because nothing in the description could say the two boxes were one
198//! question. The markup is what the measurement found, and this emits it.
199//!
200//! - **The group carries the error and the descriptions**, on the split
201//!   [`makeover_layout::FieldKind::Radio`] already uses here: what is wrong is
202//!   the answer, and a crossed interval is not the fault of either end.
203//! - **Both boxes take the whole extent.** `min`, `max` and `step` describe the
204//!   axis, so they are written twice. The crossing rule is not emitted, because
205//!   HTML has no attribute for it and the description does not carry it: it
206//!   comes back as an error on the group, like every other refusal.
207//! - **Which end is which is `aria-label` and nothing more.** The description
208//!   states direction structurally, by which member holds which name, and never
209//!   in words. Visible Min and Max captions are a page's own and reach the
210//!   group through [`form::Filling::trailing`].
211//!
212//! [`form::Value::Between`] is the second value. A separator inside one string
213//! would have made this crate the owner of a delimiter that either end could
214//! contain.
215//!
216//! # 0.55.0: a number's unit, as adjacent text
217//!
218//! `makeover-layout` 0.33.0's `Field::unit`. HTML has no unit attribute and
219//! inventing one would be markup nothing reads, so it is a `<span>` after the
220//! control -- and it is named in `aria-describedby` rather than left as
221//! decoration, because a number and what it is measured in are one fact and
222//! reading the first without the second is reading it wrong.
223//!
224//! This is the renderer the member costs something. egui already had somewhere
225//! better than the label to put a unit; here the two placements are equally
226//! adjacent, and what is bought is that the unit is a value a consumer can read
227//! back rather than a suffix on a label it would have to parse.
228//!
229//! # 0.54.0: a curve this renderer can carry, and one it declines
230//!
231//! `makeover-layout` 0.32.0's `Curve`. A range takes its granularity from the
232//! curve now (`Field::curve.step()`), every other kind keeps `Field::step`, and
233//! `Curve::Linear` emits exactly what it emitted before.
234//!
235//! **A constant-ratio curve emits a linear track, and that is the answer, not a
236//! debt.** HTML has no logarithmic range input, so a described screen asking
237//! for one is asking the browser for something it does not have, the same class
238//! of request as [`makeover_layout::FieldKind::Date`] on a host with no
239//! calendar. The renderer answers with the nearest control the host really
240//! offers and keeps every fact that survives the translation: the extent, the
241//! granularity, and the value's own units. What does not survive is resolution
242//! at the small end. The value submitted is still a value in the field's own
243//! units, which is what every handler on this path reads.
244//!
245//! The alternatives are worse in the specific way this stack exists to avoid.
246//! Shipping JS that maps thumb position to value puts app code back in the
247//! renderer. Changing what the control submits from a value to a fraction moves
248//! the mapping to whoever reads the form, and the MNW server reads these forms
249//! with its own handlers, so a fraction arriving where a value is expected
250//! would be silent.
251//!
252//! When this reopens: the day a described screen on the webview path asks for a
253//! non-linear range. The answer then is mapping in `quasi-router`, where one
254//! implementation serves every host, not JS here. Nothing asks today. No
255//! described screen in the tree uses a non-linear curve; the one real consumer
256//! is audiofiles' ADSR (`instrument_panel.rs:643`), which is egui and stays
257//! egui because audiofiles takes only `quasi-immediate` and
258//! `makeover-immediate`; and the two webview-served consumers, goingson and the
259//! MNW server, use no [`makeover_layout::FieldKind::Range`] at all.
260//!
261//! # 0.52.0: a markdown field gets the preview it was permitted
262//!
263//! 0.50.0 marked a [`makeover_layout::FieldKind::Rich`] field with
264//! `data-format="markdown"` and stopped there, and nothing read the mark: no
265//! rule, no markup, no selector anywhere in the tree. So a described editor was
266//! a plain textarea carrying an attribute, and MNW converting its four
267//! hand-written section editors onto the member would have *lost* the
268//! Write/Preview pair one of them already has. A permission taken and not spent
269//! turns every conversion into a regression.
270//!
271//! [`form::editor_rules`] and the chrome beside it are the spending. The pair is
272//! a segmented control, so it takes the depth, the focus ring and the chosen
273//! state from rules that already existed; the preview pane is a well, because it
274//! stands where the control stood. Both are gated on the attribute rather than
275//! on a class, which is what the attribute was for.
276//!
277//! **This crate still renders no markdown.** The pane arrives empty and is
278//! filled by whatever binds the editor, which is where the host's sanitiser
279//! already is. A converter here would have moved that guarantee into a crate
280//! with no view of the host's content-security posture.
281//!
282//! # 0.47.0: a range, a chooser's ghost text, and an option that cannot be
283//! picked yet
284//!
285//! `makeover-layout` 0.28.0's three form findings, all of them cheap here and
286//! none of them cheap in the app that found them.
287//!
288//! - `FieldKind::Range` emits `<input type="range">`, and `Field::step` emits
289//!   `step`. The step is emitted only when the description carries one: the
290//!   browser's own default is `step="1"`, which is what a description means by
291//!   saying nothing, and is also what turns a 0-to-1 threshold into a
292//!   two-position control.
293//! - A select with nothing chosen emits a disabled, selected, valueless first
294//!   option carrying `Field::placeholder`. HTML has no placeholder attribute on
295//!   `<select>`; this is the idiom, and `required` keeps working through it
296//!   because the option's value is empty.
297//! - `Choice::unavailable` emits `disabled` plus the reason. Where it goes
298//!   differs by control and the difference is forced: a radio group gets a
299//!   `.form-option-reason` span beside the label, and a `<select>` option has
300//!   room for no element at all, so the reason runs into its text.
301//! - `Choice::detail` takes the same split for the same reason, one release
302//!   later (makeover-layout 0.39.0): a `.form-option-detail` span in a radio
303//!   group, run into the text of a `<select>`'s option. An option carrying both
304//!   reads what it is before why it cannot be picked.
305//!
306//! # 0.25.0: a cell says what it holds
307//!
308//! 0.23.0 gave a table its layout and left every cell the same. One `.cell`
309//! carried the whole thing, so a cell holding text and a cell holding a button
310//! were one class and one content colour, and a control in a cell was painted
311//! as text. That is the drift [`RowPart::intent`](makeover_layout::RowPart)
312//! has prevented for list rows since 0.2.0 and prevented for nothing here.
313//!
314//! makeover-layout 0.14.0's [`CellPart`](makeover_layout::CellPart) names the
315//! four things a cell holds, and [`table_rules`] turns them into
316//! `.cell-value`, `.cell-tokens`, `.cell-actions` and `.cell-link`. Only the
317//! first takes a colour: a token carries its own tone, an action is a control
318//! rather than text, and a link takes the action colour from the anchor it is.
319//!
320//! The colour going on `.cell-value` rather than on `.cell` is the fix rather
321//! than an implementation detail. On the container it cascades into the parts
322//! that are not text, which is the bug said in one rule.
323//!
324//! [`list::Cell::part`] is `Option<CellPart>` here, where it was
325//! `Option<RowPart>`. A table cell borrowing the list row's vocabulary was the
326//! drift with a type on it: the two answer different questions, and only one of
327//! them was ever about a cell.
328//!
329//! # 0.23.0: a table lays itself out, and a row shows its controls
330//!
331//! Three things a description could say and this renderer had no rule for,
332//! found together by rendering the MNW server's SSH-keys settings tab through
333//! `quasi` and preferring the hand-written Askama original.
334//!
335//! **A table had no layout at all.** [`list::narrowing_css`] emits the track
336//! list, and it has to be called with the columns, so it works where the
337//! columns are known at build time: goingson builds `tables.css` in its own
338//! `build.rs` and is untouched. A table a description produced knows its
339//! columns at render time, and the rules would have had to travel with the
340//! markup: a `<style>` element per table, which needs `style-src
341//! 'unsafe-inline'` that the MNW server is working to drop, or the head, which
342//! an htmx fragment swap does not carry. [`table_rules`] lays a table out with
343//! `display: table` instead, which aligns columns across rows knowing nothing
344//! about how many there are. [`Priority`](makeover_layout::Priority) hiding
345//! moves from a generated rule per dropped column to one rule per drop class,
346//! and [`list::column_classes`] is what puts those classes on a cell. A header
347//! row emitted by a renderer's own code should call it too, or the header and
348//! the body disagree about which column just dropped.
349//!
350//! **A destructive button had nowhere for its tone to land.** `.button` carried
351//! no tone, on the reading that a control's colour is its surface. Every
352//! consumer had written the danger rule itself. It joins the badge in taking
353//! the four tones as colour, off `data-tone`.
354//!
355//! **A list rendered as a bulleted list**, because nothing here reset the `ul`
356//! a renderer emits for one.
357//!
358//! `RowPart::revealed_on_hover` also stops being honoured, which its own doc
359//! sanctioned: a renderer decides. It was retired outright in makeover-layout
360//! 0.13.0, once this had been its only consumer for a release. The rule hid a
361//! row's actions until hover,
362//! and every escape it grew was a report that hiding was wrong for somebody:
363//! `focus-within` for the keyboard, the capability gate for a fingertip with no
364//! way to unhide. What survived hid the controls from pointer users alone, who
365//! are the ones scanning a list to learn what can be done to a row.
366//!
367//! # 0.17.0: the depth classes stop being controls
368//!
369//! [`depth_rules`] gave `.raised` the whole interactive set. A depth is a
370//! statement about shape, so that left the vocabulary with no raised surface
371//! that is merely an object, and an app wanting one had two moves: write its
372//! own class from tokens, or take a control class and cancel the control half.
373//! goingson took the second, in three variants over sixteen elements
374//! (`.card--static` at 14 call sites, `.card--muted` at 2, `.card--shell` at 1),
375//! each re-asserting the resting fill and bevel on `:hover` and `:active`.
376//!
377//! Measured before changing it: `.raised` is emitted into goingson, Balanced
378//! Breakfast and the MNW server, and none of the three has a single call site.
379//! The states were unasked-for everywhere at once, and dropping them costs no
380//! migration anywhere.
381//!
382//! `.card` and `.button` are unchanged. They are the same depth *and* controls,
383//! and they take their states from [`surface_rules`], which is where a state
384//! belongs: on the thing that claims to answer a pointer.
385//!
386//! # Substitution, three ways
387//!
388//! `Fill::Well` has no colour on makeover before 2.3.0, and each renderer
389//! answers that differently, which is the evidence that dropping
390//! `Fill::fallback` from the description was right:
391//!
392//! - `makeover-immediate` substitutes the page in Rust.
393//! - `makeover-tui` refuses to substitute and draws an edge instead, because a
394//!   terminal would quantise the two together.
395//! - here, CSS already has the mechanism: `var(--surface-well,
396//!   var(--surface-page))` falls back in the browser, and nothing in Rust
397//!   decides anything.
398
399#![forbid(unsafe_code)]
400
401pub mod facet;
402pub mod figure;
403pub mod form;
404pub mod list;
405pub mod meter;
406pub mod placeholder;
407pub mod reset;
408pub mod vocabulary;
409
410/// A render of every emitter, scraped for the classes it wrote.
411///
412/// Test-only, and the guard behind [`vocabulary::names`]. See the module's own
413/// header for why the check renders rather than reads the source.
414#[cfg(test)]
415mod corpus;
416
417use crate::list::{cell_part_class, part_class};
418use crate::reset::Reset;
419use makeover_geometry::{Density, SizeClass};
420// Re-exported rather than redefined. An app assembling its own stylesheet out
421// of this crate's pieces needs the same layer name, and most such apps depend
422// on this crate and not on `makeover-geometry` directly: goingson builds
423// `tables.css` in its own build.rs from [`list::narrowing_css`], and those
424// rules are as generated as the ones here.
425pub use makeover_geometry::{CSS_LAYER, in_css_layer};
426
427/// This crate's version, as the generated stylesheet reports it.
428///
429/// A consumer whose lockfile still pins an old `makeover-webview` gets a
430/// well-formed sheet with components missing and no error anywhere, so the
431/// emitter has to name itself in what it writes. Read by
432/// `makeover_build::layout_css` through [`stylesheet`].
433pub const VERSION: &str = env!("CARGO_PKG_VERSION");
434use makeover_layout::{
435    Bevel, CellPart, Depth, Fallback, Fill, Flow, Intent, RowPart, Selector, Sort, State, Token,
436    Tone,
437};
438use makeover_touch::Affordance;
439use std::fmt::Write as _;
440
441/// How the emitted CSS is shaped.
442#[derive(Debug, Clone, Copy, PartialEq, Eq)]
443pub struct Emit {
444    /// Bevel thickness, as a CSS length.
445    ///
446    /// A value, so it arrives from the caller: border widths belong to
447    /// `makeover-geometry` and will come from there once it carries them.
448    pub border_width: &'static str,
449    /// Focus ring thickness, as a CSS length.
450    ///
451    /// Separate from [`border_width`](Self::border_width), which it reused
452    /// until 0.12.0. That reuse was an implementation convenience dressed as
453    /// consistency, and it emitted a 1px ring: a bevel and a focus indicator
454    /// are answering different questions, and only one of them has to be
455    /// noticed from across a desk.
456    ///
457    /// The default is the measured consensus rather than a new opinion. Every
458    /// consumer had already written its own ring and all three chose at least
459    /// 2px: the MNW server 2px across 10 rules, Balanced Breakfast 2px,
460    /// goingson 2px on three rules and 3px on the one covering twelve
461    /// selectors. The design system was the only thing in the tree saying 1px.
462    pub focus_width: &'static str,
463    /// Prefix for emitted class names, without the leading dot.
464    pub class_prefix: &'static str,
465}
466
467impl Default for Emit {
468    fn default() -> Self {
469        Self {
470            border_width: "1px",
471            focus_width: "2px",
472            class_prefix: "",
473        }
474    }
475}
476
477/// A string as CSS escapes, for a `content` value.
478///
479/// `\u{25B2}` becomes `\25B2`. Emitted escaped rather than literally so the
480/// stylesheet is ASCII whatever the description spells: a `content` string is
481/// read by whatever encoding the consumer serves the file as, and a caret that
482/// depends on that is a caret that works on one machine.
483///
484/// Terminated by the closing quote at every site here. A CSS hex escape takes
485/// up to six digits and ends at the first character that cannot be one, so an
486/// escape followed by more text would need a space that these do not.
487fn css_escape(text: &str) -> String {
488    text.chars().fold(String::new(), |mut out, c| {
489        let _ = write!(out, "\\{:X}", c as u32);
490        out
491    })
492}
493
494/// The CSS custom property holding a bevel's composition.
495#[must_use]
496pub fn bevel_var(bevel: Bevel) -> &'static str {
497    match bevel {
498        Bevel::Raised => "--bevel-raised",
499        Bevel::Inset => "--bevel-inset",
500    }
501}
502
503/// A `var()` reference to a fill intent, with the browser's own fallback where
504/// the intent may be absent.
505///
506/// The fallback is CSS syntax, not a decision made here. That is the whole
507/// difference between this renderer and the other two.
508#[must_use]
509pub fn fill_var(fill: Fill) -> String {
510    match fill {
511        Fill::Well => format!("var(--{}, var(--{}))", fill.token(), Fill::Page.token()),
512        other => format!("var(--{})", other.token()),
513    }
514}
515
516/// The two-tone edge as a `box-shadow` value.
517///
518/// Two inset shadows, one per corner pair: the light one offset down and
519/// right so it lands on the top and left edges, the dark one the other way.
520/// The same assignment `makeover-immediate` draws with polylines and
521/// `makeover-tui` draws with box-drawing characters.
522#[must_use]
523pub fn bevel_shadow(bevel: Bevel, opts: &Emit) -> String {
524    let (top_left, bottom_right) = bevel.edges();
525    let w = opts.border_width;
526    format!(
527        "inset {w} {w} 0 var(--{}), inset -{w} -{w} 0 var(--{})",
528        top_left.token(),
529        bottom_right.token()
530    )
531}
532
533/// The custom properties both bevels resolve through.
534///
535/// Emitted as properties rather than inlined into every rule because that is
536/// what the apps already do, and because a consumer that wants the edge
537/// without the fill reads the property directly.
538#[must_use]
539pub fn bevel_properties(opts: &Emit) -> String {
540    let mut css = String::new();
541    for bevel in [Bevel::Raised, Bevel::Inset] {
542        let _ = writeln!(
543            css,
544            "    {}: {};",
545            bevel_var(bevel),
546            bevel_shadow(bevel, opts)
547        );
548    }
549    css.push_str(ELEVATION_PROPERTY);
550    css
551}
552
553/// The cast shadow of a surface that floats over the page.
554///
555/// Composed here for the reason the bevel pair is: `makeover` derives the tone,
556/// this crate owns the geometry, and neither has to know the other's numbers.
557///
558/// **Only for a surface that overlays the page.** A menu, a toast, a popover, a
559/// dropdown. A surface *in* the page takes `.raised` and its bevel, and a rule
560/// that reaches for this on a card or a plate has renamed a literal rather than
561/// replaced it.
562///
563/// Two lengths rather than one, because a single blur reads as a smudge at
564/// plate size and as a halo at menu size. The offset is small and downward: a
565/// Platinum-era menu sits just off the page rather than hovering above it.
566const ELEVATION_PROPERTY: &str =
567    "    --elevation-overlay: 0 2px 4px var(--elevation), 0 8px 24px var(--elevation);\n";
568
569/// The class name for a depth.
570#[must_use]
571pub fn depth_class(depth: Depth, opts: &Emit) -> Option<String> {
572    let name = match depth {
573        Depth::Flat => return None,
574        Depth::Raised => "raised",
575        Depth::Well => "well",
576        Depth::Sunken => "sunken",
577        // A depth added to the description since this renderer was last
578        // built. No class, on the same footing as Flat: emitting a name
579        // whose rule body we cannot write would put a class in the markup
580        // that the stylesheet never defines.
581        _ => return None,
582    };
583    Some(format!("{}{name}", opts.class_prefix))
584}
585
586/// A prefixed class name.
587///
588/// Public since 0.27.0, for the renderers that emit markup this crate does not.
589/// A screen renderer writing `class="row"` has to prefix it the way the
590/// stylesheet half does or a prefixed app gets rules matching everything except
591/// the elements that renderer wrote, and the failure is invisible: the CSS
592/// stays valid and one element is unstyled. quasi-webview carried a byte
593/// identical copy of this function until it could call this one.
594#[must_use]
595pub fn class(name: &str, opts: &Emit) -> String {
596    let mut out = String::with_capacity(opts.class_prefix.len() + name.len());
597    push_class(&mut out, name, opts);
598    out
599}
600
601/// A prefixed class name, written into a buffer the caller already has.
602///
603/// The form the emitters use, and the reason it exists is [`escape_into`]'s:
604/// every class on every element went through a `format!` before 0.40.0,
605/// including the default case where the prefix is empty and the answer is the
606/// argument. A described table row carried roughly eighty transient
607/// allocations, and this and the escaper were most of them.
608///
609/// [`class`] stays for callers holding a name rather than a buffer.
610///
611/// [`escape_into`]: crate::form::escape_into
612pub fn push_class(out: &mut String, name: &str, opts: &Emit) {
613    out.push_str(opts.class_prefix);
614    out.push_str(name);
615}
616
617/// The class an option of a selector carries, which is what the rules key off.
618///
619/// Named for the option and not for the group: [`selector_rules`] styles the
620/// thing that gets picked, so `Selector::Tabs` is `tab` and not `tabs`. The
621/// distinction is not pedantry. quasi-webview spelled these `tabs`, `segmented`
622/// and `option`, put `toggle` on the wrapping element rather than on the
623/// buttons inside it, and every described selector in that renderer came out
624/// with no depth, no focus ring and no chosen state, while the toggle group got
625/// a bevel meant for its buttons.
626///
627/// The chosen option additionally carries `chosen`, the same way a latched chip
628/// carries `latched`. That name is this crate's too; there is no reason for a
629/// caller to spell it, and [`selector_rules`] is where it is written down.
630#[must_use]
631pub fn option_class(selector: Selector) -> &'static str {
632    match selector {
633        Selector::Tabs => "tab",
634        Selector::Segmented => "segment",
635        Selector::Toggle => "toggle",
636    }
637}
638
639/// The fill and edge declarations for a depth, as a rule body.
640///
641/// Empty for [`Depth::Flat`], which has neither and inherits what it sits on.
642/// Callers lean on the emptiness to skip the rule rather than emit a class that
643/// sets nothing: a class that sets no properties is a class that means "I
644/// thought about this", which is what comments are for.
645///
646/// The two halves are emitted independently because [`Depth::Sunken`] has a
647/// fill and no bevel. Requiring both, which this did before makeover-layout
648/// 0.3.0, silently dropped the fill for exactly that case. Independent does not
649/// mean unpaired: both halves still come off one `Depth`, so they cannot
650/// disagree about what the region is.
651#[must_use]
652pub fn depth_declarations(depth: Depth) -> String {
653    let mut css = String::new();
654    if let Some(fill) = depth.fill() {
655        let _ = writeln!(css, "    background: {};", fill_var(fill));
656    }
657    if let Some(bevel) = depth.bevel() {
658        let _ = writeln!(css, "    box-shadow: var({});", bevel_var(bevel));
659    }
660    css
661}
662
663/// One rule giving a selector a depth, or nothing when the depth declares
664/// nothing.
665#[must_use]
666pub fn depth_rule(selector: &str, depth: Depth) -> String {
667    let body = depth_declarations(depth);
668    if body.is_empty() {
669        return String::new();
670    }
671    format!(".{selector} {{\n{body}}}\n")
672}
673
674/// The media condition a hover rule has to sit inside, or `None` if hover is
675/// unconditional.
676///
677/// Two crates answer this and neither answer is made here. `makeover-touch`
678/// owns *whether* hover exists at a density, and `makeover-geometry` owns how
679/// that capability is spelled as a media condition. Asking both is what stops
680/// this renderer minting a third opinion, which is what all three apps did:
681/// goingson sniffed the user agent, Balanced Breakfast used `(hover: none)`
682/// alone, and the MNW server had no gate at all.
683///
684/// [`SizeClass`] is required by [`Affordance::available`] and ignored by this
685/// member, which reports as much through `reads_size`. Passing Compact is not
686/// a claim about width; the test below pins that every class agrees.
687fn hover_condition() -> Option<&'static str> {
688    if Affordance::Hover.available(Density::Touch, SizeClass::Compact) {
689        // A fingertip grew a hover state. Nothing to gate, and this renderer
690        // should not invent a reason to gate anyway.
691        None
692    } else {
693        Some(Density::Pointer.media_condition())
694    }
695}
696
697/// Put a rule inside a media query, or leave it alone.
698fn gated(condition: Option<&str>, rule: &str) -> String {
699    let Some(condition) = condition else {
700        return rule.to_string();
701    };
702    let mut css = format!("@media {condition} {{\n");
703    for line in rule.lines() {
704        // Blank lines stay blank. Indenting one leaves trailing whitespace,
705        // which is the sort of thing a formatter later reverts and calls a diff.
706        if line.is_empty() {
707            css.push('\n');
708        } else {
709            let _ = writeln!(css, "    {line}");
710        }
711    }
712    css.push_str("}\n");
713    css
714}
715
716/// The keyboard focus ring, placed by the depth it lands on.
717///
718/// This is the webview's **focus ring** and nothing more. **Reach** and
719/// **focus** are both the browser's — the document decides what is reachable
720/// and `:focus-visible` decides which reached thing wears the ring — and no
721/// description states either. The three terms are defined once in
722/// `makeover_layout`'s crate header, "Reach, focus and the focus ring".
723///
724/// One ring for the whole system, because a focus ring's job is to be
725/// recognised and three apps having three of them is the failure. What varies
726/// is where it sits, and that comes off [`Depth`] rather than off a per-
727/// component choice: a well takes the ring inside its own edge, and anything
728/// standing proud of the page takes it outside.
729///
730/// `outline` rather than the composed `box-shadow` the invalid-field ring at
731/// [`field_rules`] uses, and deliberately the one place the two rings are built
732/// differently. A `box-shadow` ring has to restate the bevel beside it, because
733/// `box-shadow` is not additive and a lone ring silently drops the well out
734/// from under the element. That restatement is a second copy of the depth,
735/// living in a different function from the first, and it is exactly the
736/// duplication `Depth` exists to prevent. `outline` occupies its own property,
737/// so the bevel survives untouched and there is nothing to keep in agreement.
738/// They render the same: both are a flush ring one border-width wide.
739#[must_use]
740pub fn focus_rule(selector: &str, depth: Depth, opts: &Emit) -> String {
741    let w = opts.focus_width;
742    // Same magnitude either way, and only the sign comes off the depth. Both
743    // values are what the consumers had already converged on independently:
744    // 2px out is what all three wrote, and 2px in is the MNW server's own
745    // answer for the one inset ring it had.
746    let offset = match depth.bevel() {
747        // Inside the well, clear of its edge rather than painted over it.
748        Some(Bevel::Inset) => format!("calc(-1 * {w})"),
749        // Raised, or no edge at all. Outside, standing off by its own width.
750        _ => w.to_string(),
751    };
752    // The token by name. It is `makeover`'s, derived from the action colour,
753    // and reaching it through a description member was a second path to the
754    // same variable for as long as one existed.
755    format!(
756        ".{selector}:focus-visible {{\n    outline: {w} solid var(--focus-ring);\n    outline-offset: {offset};\n}}\n"
757    )
758}
759
760/// Present, visible, and not answering.
761///
762/// Matches the ARIA attribute as well as the pseudo-class, because `:disabled`
763/// only matches form elements and half the things this crate emits are not
764/// one: a `div` carrying `.chip` or `.tab` can never be `:disabled`. Keying on
765/// the accessible state is the pattern [`field_rules`] already establishes for
766/// `aria-invalid`, on the reasoning that one fact read by both the styling and
767/// the accessibility tree cannot drift from itself.
768///
769/// The rest depth is re-asserted rather than assumed, because this rule has to
770/// beat the hover and pressed rules above it. It does that on source order at
771/// equal specificity, not by out-specifying them: every rule this function's
772/// caller emits is (0,2,0), and adding a `:not(:disabled)` anywhere would raise
773/// one of them and have to be unpicked when this output moves inside its own
774/// cascade layer.
775#[must_use]
776pub fn disabled_rule(selector: &str, depth: Depth) -> String {
777    format!(
778        ".{selector}:disabled,\n.{selector}[aria-disabled=\"true\"] {{\n{}    color: var(--{});\n    cursor: not-allowed;\n}}\n",
779        depth_declarations(depth),
780        State::Disabled.token()
781    )
782}
783
784/// Every state a selector that answers a click implies: hover, pressed, focus
785/// and disabled, in that order.
786///
787/// Order is the whole cascade mechanism here. All four selectors are
788/// specificity (0,2,0), so disabled wins over hover and pressed by coming last
789/// and by nothing else.
790///
791/// Pressed emits [`Depth::pressed`] in full, fill and edge together. Emitting
792/// only the edge is what left goingson hand-writing `background:
793/// var(--surface-sunken)` on three separate rules, and a fill that does not
794/// travel with its edge is precisely the disagreement `Depth` exists to make
795/// unrepresentable. So the pressed fill comes from the description
796/// (`--surface-well`) rather than from whatever each app reached for.
797///
798/// Hover has no member in the description and is renderer policy: a terminal
799/// and an immediate-mode painter have no hover to express. It resolves against
800/// `--hover-surface`, which `makeover` already derives and which nothing
801/// consumed until now. What it *is* gated on is capability, via
802/// [`hover_condition`]. Before that gate existed the apps each wrote their own:
803/// goingson's section 60 exists solely to take back the hover state this
804/// function had just handed it, by out-specifying a rule it does not own.
805///
806/// `depth` is the selector's **rest** depth, used to place the focus ring and
807/// to restore the surface under a disabled control. The pressed rule keeps
808/// inverting from [`Depth::Raised`] regardless, which is what every caller got
809/// before this parameter existed: a tab's unchosen depth is
810/// [`Depth::Sunken`], and `Sunken.pressed()` is `Sunken`, so deriving the press
811/// from the rest depth would leave a tab with no press at all.
812#[must_use]
813pub fn interactive_rules(selector: &str, depth: Depth, opts: &Emit) -> String {
814    let mut css = gated(
815        hover_condition(),
816        &format!(".{selector}:hover {{\n    background: var(--hover-surface);\n}}\n"),
817    );
818    css.push_str(&depth_rule(
819        &format!("{selector}:active"),
820        Depth::Raised.pressed(),
821    ));
822    css.push_str(&focus_rule(selector, depth, opts));
823    css.push_str(&disabled_rule(selector, depth));
824    css
825}
826
827/// One rule per depth: its fill and its edge, together.
828///
829/// A depth and nothing else. `.raised` says a surface sits on what is behind
830/// it, which is a statement about the shape and not about what happens when a
831/// pointer arrives, so it emits no hover, press, focus or disabled rule. The
832/// named surfaces are where interaction lives: `.card` and `.button` are the
833/// same depth *and* controls, and they get their states from
834/// [`surface_rules`].
835///
836/// This class carried the interactive set until 0.17.0, which left the
837/// vocabulary with no raised surface that is merely an object. Consumers that
838/// needed one took a control class and cancelled half of it instead: sixteen
839/// elements in goingson across three `.card--*` variants, each re-asserting the
840/// resting fill and bevel on `:hover` and `:active`. Nothing anywhere used
841/// `.raised` itself, so the states were unasked-for in every consumer at once.
842#[must_use]
843pub fn depth_rules(opts: &Emit) -> String {
844    let mut css = String::new();
845    for depth in [Depth::Raised, Depth::Well] {
846        let Some(class) = depth_class(depth, opts) else {
847            continue;
848        };
849        css.push_str(&depth_rule(&class, depth));
850    }
851    css
852}
853
854/// The three surfaces that are a depth with a name.
855///
856/// `button` and `card` are both [`Depth::Raised`], and `field` is a
857/// [`Depth::Well`] because that is the reading `Depth`'s own documentation
858/// gives a text field. Their bodies come out identical by construction rather
859/// than by hand: three hand-written copies in goingson's stylesheet is what
860/// phase A deletes, and generating them from one call is what stops them
861/// drifting apart again.
862fn surface_rules(opts: &Emit) -> String {
863    let mut css = String::new();
864    for name in ["button", "card"] {
865        let c = class(name, opts);
866        css.push_str(&depth_rule(&c, Depth::Raised));
867        css.push_str(&interactive_rules(&c, Depth::Raised, opts));
868    }
869
870    let field = class("field", opts);
871    css.push_str(&depth_rule(&field, Depth::Well));
872
873    // A field takes focus and refuses input like everything else here, and got
874    // neither until now, which is why all three apps hand-write a focus ring
875    // for it and no two of them match. No hover or pressed: a text field does
876    // not light up under the pointer and does not invert when clicked, so the
877    // two states `interactive_rules` would add are the two it does not have.
878    css.push_str(&focus_rule(&field, Depth::Well, opts));
879    css.push_str(&disabled_rule(&field, Depth::Well));
880
881    // Keyed on the ARIA attribute rather than on a class, so the visual state
882    // and the accessible state cannot drift apart: there is one fact and both
883    // read it. goingson already drove its invalid styling this way and was
884    // right to; the `.invalid` class this emitted before 0.5.0 was a second
885    // place to forget.
886    //
887    // The ring composes *after* the bevel rather than replacing it. box-shadow
888    // is not additive, so a lone ring silently dropped the well out from under
889    // an invalid field. Flat and unlit: this edge is saying "wrong", and
890    // lighting one side would have it say "raised" at the same time.
891    let _ = writeln!(
892        css,
893        ".{field}[aria-invalid=\"true\"] {{\n    box-shadow: var({}), 0 0 0 {} var(--danger);\n}}",
894        bevel_var(Bevel::Inset),
895        opts.border_width
896    );
897    css
898}
899
900/// Badges and chips.
901///
902/// The one place phase A changes how goingson looks rather than only where its
903/// rules live. [`Token::Badge`] is [`Depth::Flat`], so a badge emits no fill
904/// and no edge at all, where goingson ships `.tag, .badge` as a single rule
905/// carrying the raised bevel. Splitting that means reading every call site to
906/// decide which of the two it always was.
907///
908/// What a badge does carry is a [`Tone`], the intent family it shares with
909/// notices and nothing else. Neutral is the bare class rather than a variant,
910/// because it is the absence of a status and not a status called "none".
911/// Text that goes somewhere.
912///
913/// The one inline control, and the vocabulary had no word for it until a
914/// description layer needed one. A table cell has carried `cell-link` since
915/// 0.14.0's part list, deliberately unruled because the cell's own rule covers
916/// it; what was missing is the same thing outside a table, which is what a
917/// described run holds when a sentence contains a link.
918///
919/// Colour and underline only. Whether a link is inline in a sentence or sitting
920/// on its own line is the app's layout, and how much room it takes is
921/// `makeover-geometry`'s. What is here is the pair of signals that say "this
922/// goes somewhere" and nothing that says where it sits.
923///
924/// The visited arm is deliberately absent. A link inside an app points at the
925/// app's own screens, which the user is expected to have been to, so painting
926/// them differently marks almost everything and distinguishes nothing.
927fn link_rules(opts: &Emit) -> String {
928    let mut css = String::new();
929    let link = class("link", opts);
930
931    let _ = writeln!(
932        css,
933        ".{link} {{\n    color: var(--action);\n    \
934         text-decoration: underline;\n}}"
935    );
936    // The hover step is the same one every other control takes, and it is a
937    // colour rather than a surface: a link has no box to raise.
938    let _ = writeln!(
939        css,
940        "@media (hover: hover) and (pointer: fine) {{\n    .{link}:hover \
941         {{\n        color: var(--action-hover);\n    }}\n}}"
942    );
943    let _ = writeln!(
944        css,
945        ".{link}:focus-visible {{\n    outline: {} solid var(--focus-ring);\n    \
946         outline-offset: 2px;\n}}",
947        opts.focus_width
948    );
949    // A link is often a `<button>` rather than an `<a>`: a renderer picks the
950    // element from the method, so a link that writes is a button that has to
951    // stop looking like one. What that costs is named in [`reset`].
952    css.push_str(&Reset::TEXT_BUTTON.rule(&format!("button.{link}")));
953    css
954}
955
956fn token_rules(opts: &Emit) -> String {
957    let mut css = String::new();
958
959    // No `depth_rule` call here, deliberately: `Token::Badge.depth(_)` is Flat,
960    // and a label with an edge says it can be pressed.
961    let badge = class("badge", opts);
962    // `content-muted` literally, not `Tone::Neutral.token()`. What makes a
963    // badge quiet is `Token::Badge` answering no click, which this crate holds
964    // and `Tone` genuinely does not know. Routing it through Neutral put the
965    // claim where the evidence was not, and the bill arrived on the figure
966    // value: it took the same muting from the same call and read as its own
967    // caption. Neutral answers `content` from makeover-layout 0.36.0.
968    let _ = writeln!(css, ".{badge} {{\n    color: var(--content-muted);\n}}");
969    for tone in [Tone::Info, Tone::Success, Tone::Warning, Tone::Danger] {
970        let _ = writeln!(
971            css,
972            ".{badge}[data-tone=\"{0}\"] {{\n    color: var(--{0});\n}}",
973            tone.token()
974        );
975    }
976
977    // A button carries the four tones a badge does. It had none, on the reading
978    // that a control's colour is its surface rather than its text, and that
979    // reading has one hole big enough to matter: the button that destroys
980    // something. Every consumer had written that rule itself, and a description
981    // that says `Tone::Danger` on an act had nowhere for it to land.
982    //
983    // Colour and not a fill, matching the badge. A red surface is a decision
984    // about emphasis that belongs to an app's own layer, and two of them
985    // fighting is worse than neither.
986    let button = class("button", opts);
987    for tone in [Tone::Info, Tone::Success, Tone::Warning, Tone::Danger] {
988        let _ = writeln!(
989            css,
990            ".{button}[data-tone=\"{0}\"] {{\n    color: var(--{0});\n}}",
991            tone.token()
992        );
993    }
994
995    // A chip holds itself down, which is `Depth::pressed` arrived at
996    // independently by two apps. `removable` is a remove affordance, so it is
997    // markup and waits for phase B.
998    let chip = class("chip", opts);
999    let unlatched = Token::Chip { removable: false };
1000    css.push_str(&depth_rule(&chip, unlatched.depth(false)));
1001    css.push_str(&interactive_rules(&chip, unlatched.depth(false), opts));
1002    css.push_str(&depth_rule(
1003        &format!("{chip}.latched"),
1004        unlatched.depth(true),
1005    ));
1006    css
1007}
1008
1009/// The three selectors, each named by what it picks.
1010///
1011/// A tab comes *forward* to join the pane it opens, which is why
1012/// [`Selector::Tabs`] chooses [`Depth::Raised`] where a segment and a toggle
1013/// are held in. That is the folder semantic, and it is the whole reason the
1014/// three are not one member with a flag.
1015///
1016/// [`Selector::abutting`] is not emitted: whether the options touch is
1017/// spacing, and spacing is `makeover-geometry`'s question to answer.
1018///
1019/// Both states emit as of makeover-layout 0.3.0. Before it the description
1020/// named only the chosen option, so an unchosen one fell through to
1021/// [`Depth::Flat`] and nothing was drawn for it, which left goingson's tab
1022/// strip hand-writing the recess that makes its chosen tab read as forward.
1023fn selector_rules(opts: &Emit) -> String {
1024    let mut css = String::new();
1025    for selector in [Selector::Tabs, Selector::Segmented, Selector::Toggle] {
1026        let c = class(option_class(selector), opts);
1027        css.push_str(&depth_rule(&c, selector.unchosen()));
1028        css.push_str(&interactive_rules(&c, selector.unchosen(), opts));
1029        css.push_str(&depth_rule(&format!("{c}.chosen"), selector.chosen()));
1030    }
1031    css
1032}
1033
1034/// The parts of a list row.
1035///
1036/// The list is written out rather than derived because `RowPart` is
1037/// `#[non_exhaustive]` as of makeover-layout 0.9.0, so there is nothing to
1038/// iterate. A member added upstream emits no rule until it is named here, which
1039/// is the trade `non_exhaustive` makes: a silent gap instead of a build break.
1040/// [`part_class`] carries the same list and the same obligation.
1041fn row_rules(opts: &Emit) -> String {
1042    let mut css = String::new();
1043
1044    // The container the rows sit in, giving back what a `<ul>` brought. Not a
1045    // size: there is no magnitude in it, which is the line this crate holds.
1046    css.push_str(&Reset::BULLETS.rule(&format!(".{}", class("list", opts))));
1047
1048    for part in [
1049        RowPart::Primary,
1050        RowPart::Secondary,
1051        RowPart::Meta,
1052        RowPart::Actions,
1053        RowPart::Tokens,
1054        RowPart::Proportion,
1055    ] {
1056        let c = class(part_class(part), opts);
1057
1058        // Actions carry controls rather than text, and `RowPart::intent` says
1059        // so by returning the same intent inheriting already gives. Pinning it
1060        // would be louder than saying nothing. Tokens answer alike, for their
1061        // own reason: each token carries its own tone, and a colour on the
1062        // strip would fight the things sitting in it. A proportion is the same
1063        // case again: the meter inside carries the tone.
1064        if !matches!(
1065            part,
1066            RowPart::Actions | RowPart::Tokens | RowPart::Proportion
1067        ) {
1068            let _ = writeln!(css, ".{c} {{\n    color: var(--{});\n}}", part.intent());
1069        }
1070
1071        // A row's actions are shown at rest. `RowPart::revealed_on_hover` said
1072        // otherwise and was not honoured here from 0.23.0; makeover-layout
1073        // 0.13.0 retired the method, so there is no longer a description saying
1074        // one thing and a renderer doing another.
1075        //
1076        // The rule was `opacity: 0` gated to pointer devices, revealed on
1077        // `:hover` and on `:focus-within`. Each escape it needed was a report
1078        // that hiding was wrong for somebody: `focus-within` because tabbing
1079        // could never reach an action; the gate because a fingertip had no way
1080        // to unhide, which both webview apps had already hand-written
1081        // `opacity: 1` to undo. What was left was a control hidden from
1082        // exactly one group: people using a pointer, who are also the group
1083        // scanning a list to find out what can be done to a row.
1084        //
1085        // A settings screen is where that reads worst: the whole reason to be
1086        // on it is to remove a key, and the button doing so was invisible
1087        // until pointed at. A table cell's actions were never hidden, so the
1088        // two arrangements now agree.
1089    }
1090
1091    // A part that may take two lines. `Flow::Tight` gets no rule: one line is
1092    // what a run already does, and restating it here would put a declaration on
1093    // every part in every row to say nothing.
1094    //
1095    // This is the shape both webview apps had already written by hand and
1096    // commented -- Balanced Breakfast on a feed row's title, goingson on a
1097    // problem's body -- which is the whole argument for the description
1098    // carrying it. `-webkit-` prefixed and unprefixed together: the prefixed
1099    // trio is what every engine actually implements, and `line-clamp` is the
1100    // standard property landing behind it.
1101    let _ = writeln!(
1102        css,
1103        ".{} {{\n    display: -webkit-box;\n    -webkit-box-orient: vertical;\n    -webkit-line-clamp: {lines};\n    line-clamp: {lines};\n    overflow: hidden;\n}}",
1104        class("row-relaxed", opts),
1105        lines = Flow::Relaxed.lines()
1106    );
1107
1108    css
1109}
1110
1111/// A row of things that share their space, and what each fallback gets here.
1112///
1113/// Ruling: wiki `layout-room-and-fallback`, Max 2026-08-18. The bug it answers
1114/// is goingson's `styles.css:702`, which pinned every `.page-header` over the
1115/// pill strip with `position: absolute` and so took the toolbar out of flow:
1116/// zero width contributed to the row it shared, nothing able to collide with
1117/// it, and therefore nothing preventing the collision. Rule 1 of the ruling is
1118/// that every described member is in flow, and these rules are how that is kept
1119/// rather than asked for.
1120///
1121/// # The floor, which is most of the fix
1122///
1123/// `.run > *` gets `min-width: min-content`. That is the derived minimum the
1124/// ruling asks for, in this renderer's own unit and stated by the browser
1125/// rather than by anybody: a member cannot be squeezed narrower than what is
1126/// in it, so members in one flow push each other instead of overlapping. It
1127/// costs no query and no number, and it is what fixes all four measured widths
1128/// whichever fallback the group declared.
1129///
1130/// # What each fallback gets, exactly
1131///
1132/// [`Fallback::Wrap`] is `flex-wrap: wrap`, which is exact. The browser wraps
1133/// the run when the members no longer fit, deciding that from their own
1134/// intrinsic widths, which is the derived minimum doing the whole job.
1135///
1136/// [`Fallback::Stack`] is wrap plus `flex: 1 1 max-content` on the members, so
1137/// a member that cannot sit beside its sibling takes a line of its own and
1138/// fills it. For the two-member run this was ruled on -- a tab strip and a
1139/// band -- that is precisely "a row becomes a column".
1140///
1141/// [`Fallback::Shed`] and [`Fallback::Menu`] get wrap, and this renderer is
1142/// honouring less than the description says. **CSS cannot express either one
1143/// without breaking the ruling's own first constraint.** Both need to know that
1144/// the run is out of room in order to take a member out of it, a container
1145/// query is the only construct that can ask, and `@container` compares against
1146/// a `<length>` -- there is no `@container (inline-size < min-content)`. So
1147/// every honest spelling of Shed here needs an authored breakpoint, which is
1148/// the thing the ruling exists to forbid, and the dishonest ones are worse: a
1149/// clamped height clips by document order rather than by [`Priority`], and
1150/// `display: none` under a viewport `@media` is the `nth-child(n+5)` bug the
1151/// vocabulary replaced.
1152///
1153/// Wrapping is the right thing to do instead. It keeps every member reachable,
1154/// which is the property that was actually broken -- goingson's new-contact
1155/// button left the viewport entirely at 560 -- and it keeps rule 1. A renderer
1156/// answering with less than was described is precedented and deliberate here:
1157/// [`makeover_layout::Region::Columns`] says a terminal stacking a board's
1158/// columns is honouring the description rather than degrading it.
1159///
1160/// The real mechanism needs the shed members to have somewhere to go, which is
1161/// markup and belongs to quasi-webview: an overflow control is a member of the
1162/// run, and the description does not yet say that a member *is* one.
1163///
1164/// # Menu, once a script is measuring
1165///
1166/// That is now built, in `quasi-webview`'s `menu.js`, and this crate's half of
1167/// it is two classes and one override. A script that has taken a menu run over
1168/// marks it `data-menu`, and a marked run goes back to `nowrap`: wrapping is
1169/// what hides the overflow condition the script is trying to measure. The
1170/// unmarked rule above is untouched, so a page that ships no script still
1171/// wraps, which is rule 1 — every member reachable — rather than a strip with
1172/// tabs squeezed off the end.
1173///
1174/// `.run-overflow` is the control the shed members move into and
1175/// `.run-overflow-items` is where they land. The geometry is this crate's the
1176/// way every other surface's is; what is *in* it is the script's, because which
1177/// members no longer fit is a measurement and not a description.
1178fn run_rules(opts: &Emit) -> String {
1179    let run = class("run", opts);
1180    let mut css = String::new();
1181
1182    // `flex-wrap: nowrap` is stated rather than left to the default, because
1183    // the fallbacks below are read as overrides of this line and a reader
1184    // should not have to know which way flexbox leans to see that.
1185    //
1186    // No gap. Spacing between members is the app's, the same way this crate
1187    // states no margins anywhere else; a gap here would be a size, and the one
1188    // hardcoded size in the mechanism is makeover-geometry's contact patch.
1189    let _ = writeln!(
1190        css,
1191        ".{run} {{\n    display: flex;\n    flex-wrap: nowrap;\n    align-items: center;\n}}"
1192    );
1193
1194    // The derived minimum, and the whole reason a member can no longer be
1195    // overlapped. `min-width: auto` is flexbox's default for a flex item and is
1196    // *not* the same thing: auto lets an item be compressed below its content
1197    // in a nowrap run, which is how a toolbar ends up drawn over a tab strip
1198    // even without anything leaving the flow.
1199    let _ = writeln!(css, ".{run} > * {{\n    min-width: min-content;\n}}");
1200
1201    for fallback in [
1202        Fallback::Wrap,
1203        Fallback::Stack,
1204        Fallback::Shed,
1205        Fallback::Menu,
1206    ] {
1207        let name = fallback_class(fallback);
1208        let c = class(name, opts);
1209        let _ = writeln!(css, ".{c} {{\n    flex-wrap: wrap;\n}}");
1210        if matches!(fallback, Fallback::Stack) {
1211            let _ = writeln!(css, ".{c} > * {{\n    flex: 1 1 max-content;\n}}");
1212        }
1213    }
1214
1215    // A menu run a script has taken over. The mark is the script's and this is
1216    // the only rule that reads it: wrapping is what a run does when nothing is
1217    // measuring, and it is also what makes the overflow unmeasurable, since a
1218    // wrapped run always fits. The two cannot both be on.
1219    let menu = class(fallback_class(Fallback::Menu), opts);
1220    let _ = writeln!(css, ".{menu}[data-menu] {{\n    flex-wrap: nowrap;\n}}");
1221
1222    // The overflow control, and it is a member of the run like any other: in
1223    // flow, at the end, taking the width of what is in it. `relative` is what
1224    // the items hang off.
1225    let overflow = class("run-overflow", opts);
1226    let items = class("run-overflow-items", opts);
1227    let _ = writeln!(css, ".{overflow} {{\n    position: relative;\n}}");
1228
1229    // Overlaid rather than in flow, for the reason every menu is: a control
1230    // that pushed the page down when it opened would change the layout it was
1231    // opened to escape. `inset-inline-end: 0` rather than a left, so the panel
1232    // stays on the page in both writing directions.
1233    //
1234    // No width, no padding and no border. All three are sizes and sizes are
1235    // makeover-geometry's; what is stated here is placement, the surface and
1236    // the shadow that separates it from the page, which is the same division
1237    // `figure_rules` and the timeline entry make. The elevation shadow is what
1238    // an overlaid surface takes instead of an edge -- see `ELEVATION_PROPERTY`.
1239    let _ = writeln!(
1240        css,
1241        ".{items} {{\n    \
1242         position: absolute;\n    \
1243         inset-block-start: 100%;\n    \
1244         inset-inline-end: 0;\n    \
1245         z-index: 1;\n    \
1246         display: flex;\n    \
1247         flex-direction: column;\n    \
1248         align-items: stretch;\n    \
1249         background: var(--surface-raised);\n    \
1250         box-shadow: var(--elevation-overlay);\n\
1251         }}"
1252    );
1253
1254    // `hidden` is how the script closes it, and a flex display would otherwise
1255    // beat the attribute's own `display: none`.
1256    let _ = writeln!(css, ".{items}[hidden] {{\n    display: none;\n}}");
1257
1258    css
1259}
1260
1261/// Every class [`fallback_class`] can return, plus the run itself.
1262///
1263/// [`ROW_PART_CLASSES`](crate::list::ROW_PART_CLASSES)'s reasoning and the same
1264/// obligation: a `match` over a `#[non_exhaustive]` enum cannot be enumerated
1265/// from outside, so the list sits beside it and a test holds the two together.
1266/// `run` is in it because it is emitted in its own right rather than only as a
1267/// fallback's fallback.
1268pub const RUN_CLASSES: &[&str] = &[
1269    "run",
1270    "run-wrap",
1271    "run-stack",
1272    "run-shed",
1273    "run-menu",
1274    // Not returned by `fallback_class`: these two are the overflow control a
1275    // measuring renderer builds, and they are in the list because the list is
1276    // what a host seals its vocabulary against. A class emitted by a script and
1277    // missing from here is a control with no surface and no edge.
1278    "run-overflow",
1279    "run-overflow-items",
1280];
1281
1282/// The class a run carries for what it does when it is tight.
1283///
1284/// A run always carries `.run` as well, so an unrecognised fallback -- the enum
1285/// is `#[non_exhaustive]` -- lands as a plain nowrap row with the min-content
1286/// floor still under it. That is the safe failure: every member in flow and
1287/// none overlapped, which is the property, with only the rearrangement missing.
1288#[must_use]
1289pub fn fallback_class(fallback: Fallback) -> &'static str {
1290    match fallback {
1291        Fallback::Wrap => "run-wrap",
1292        Fallback::Stack => "run-stack",
1293        Fallback::Shed => "run-shed",
1294        Fallback::Menu => "run-menu",
1295        _ => "run",
1296    }
1297}
1298
1299/// The progress trough these rules fill.
1300///
1301/// This was renderer-local chrome with nothing behind it until makeover-layout
1302/// 0.10.0, which is the unusual order: the tones below were emitted for every
1303/// bar in the tree while the only way to describe one was to concatenate the
1304/// numbers into a heading. `Meter` is the word that arrived late, and
1305/// [`meter::meter_html`](crate::meter::meter_html) is what now fills these.
1306///
1307/// The rules stay a superset of what a description can ask for. An app drawing
1308/// its own bar keeps these classes, which is what the four goingson grew
1309/// independently were adopted onto.
1310///
1311/// The trough is a [`Depth::Well`], the same reading a text field gets:
1312/// something with its content down inside it.
1313fn progress_rules(opts: &Emit) -> String {
1314    let progress = class("progress", opts);
1315    // `progress-fill` rather than a bare `fill`: an unprefixed build claims
1316    // these names in the app's own stylesheet, and `.fill` is grabby enough to
1317    // catch things that have nothing to do with progress. goingson already
1318    // calls it `.progress-fill`, so this is also the name that deletes.
1319    let fill = class("progress-fill", opts);
1320    let mut css = depth_rule(&progress, Depth::Well);
1321
1322    // The untoned bar is `--action`, not [`Tone::Neutral`]. That is the one
1323    // place this differs from the badge rules, and deliberately: a badge with
1324    // no status is a muted label, while a bar with no status is still
1325    // reporting progress, and `content-muted` would read as disabled.
1326    let _ = writeln!(
1327        css,
1328        ".{progress} > .{fill} {{\n    background: var(--action);\n}}"
1329    );
1330
1331    // A bar can be saying something, same as a badge: goingson colours subtask
1332    // progress as success and an over-estimate as danger, which is real
1333    // information rather than decoration. Emitting the tones is what lets that
1334    // survive adoption instead of staying hand-written.
1335    for tone in [Tone::Info, Tone::Success, Tone::Warning, Tone::Danger] {
1336        let _ = writeln!(
1337            css,
1338            ".{progress} > .{fill}[data-tone=\"{0}\"] {{\n    background: var(--{0});\n}}",
1339            tone.token()
1340        );
1341    }
1342    css
1343}
1344
1345/// What a wait looks like, for the attribute that has been describing one to
1346/// nobody.
1347///
1348/// Wiki `loading-and-progress-standard`, phase 2. `data-awaiting` has been
1349/// emitted since the vocabulary gained `Awaiting`, with
1350/// `data-awaiting="determinate"` and a `data-awaiting-amount` beside it or
1351/// `data-awaiting="indeterminate"` alone, and until 0.60.0 **nothing in the tree
1352/// styled either**. Determinate and indeterminate rendered identically, which is
1353/// to say not at all. This is the half that was missing.
1354///
1355/// # Why it is keyed on `aria-busy` and not on the attribute alone
1356///
1357/// `data-awaiting` is a fact about the control: pressing this waits. It is true
1358/// when the page is painted and it stays true. Whether a wait is *running* is
1359/// true only between two events, so it is the binder's to set, and `aria-busy`
1360/// is the standard spelling of it — announced as well as drawn, which a class
1361/// of our own would not be.
1362///
1363/// That also keeps this crate out of any one client library's vocabulary.
1364/// `quasi-webview` sets `aria-busy` from htmx's request events; a host driving
1365/// the same markup another way sets it the same way and gets the same drawing.
1366///
1367/// # The two drawings
1368///
1369/// One pseudo-element either way, so no renderer has to emit an extra node.
1370///
1371/// Indeterminate is the activity mark of rule 2: a small square that blinks.
1372/// Determinate is a trough with a fill, drawn as a single gradient whose stop is
1373/// `--awaiting-share`, a plain number from 0 to 1 that the binder sets from
1374/// bytes it has actually watched land. A determinate control with nothing
1375/// setting the share draws an empty trough rather than a full one, which is the
1376/// honest reading: the size is known and the delivery is not.
1377///
1378/// **What the bar may not do**, from rule 1 and from `Awaiting`'s own docs:
1379/// what is done over what there is, and never a remaining time, an arrival time
1380/// or a rate extrapolated forward. Nothing here can express one, which is
1381/// deliberate — the only input is a share of a measured payload.
1382///
1383/// # Sizes, and the deferral rule
1384///
1385/// `progress_rules` emits the tones and never the width, because the width is
1386/// not this crate's to know. A pseudo-element has no intrinsic size at all, so
1387/// the same treatment would render nothing anywhere. Both sizes are therefore
1388/// custom properties with defaults: an app that wants a different mark sets
1389/// `--awaiting-mark` and `--awaiting-bar` once, and one that says nothing gets a
1390/// mark that is visible.
1391///
1392/// # The cadence, and what happens without it
1393///
1394/// `--cadence-activity` comes from `makeover-timing` through `makeover-build`,
1395/// and is a half-period, so a full cycle is twice it. It is used bare rather
1396/// than with a fallback: a number written here would be a second heartbeat for
1397/// a mark three renderers draw.
1398///
1399/// A sheet assembled without the time axis leaves `animation-duration` invalid,
1400/// which resolves to `0s`, which runs no animation and leaves the base style —
1401/// a lit, still mark. That is also exactly what the reduced-motion block does,
1402/// since it sets the cadence to `0ms`. Both fall out of one rule because the
1403/// base style is lit and the keyframes do the dimming, which is the ordering
1404/// `makeover_timing::reduced_motion_css` asks its consumers for by name.
1405fn awaiting_rules(opts: &Emit) -> String {
1406    let mut css = String::new();
1407
1408    // Dimming rather than lighting, so a zero-length animation leaves a lit
1409    // mark rather than a blank one. See `makeover_timing::reduced_motion_css`.
1410    css.push_str(
1411        "@keyframes makeover-activity {\n    \
1412         0%, 49.99% {\n        background: var(--action);\n    }\n    \
1413         50%, 100% {\n        background: var(--surface-sunken);\n    }\n\
1414         }\n",
1415    );
1416
1417    // Nothing is drawn until something is waiting. `content` on the base rule
1418    // rather than on the busy one keeps the box the same box across the
1419    // transition, so a mark appearing does not reflow the line it is in.
1420    let _ = writeln!(
1421        css,
1422        "[data-awaiting]::after {{\n    \
1423         content: \"\";\n    \
1424         display: none;\n    \
1425         margin-inline-start: var(--awaiting-gap, 0.5ch);\n    \
1426         vertical-align: baseline;\n\
1427         }}"
1428    );
1429
1430    let _ = writeln!(
1431        css,
1432        "[data-awaiting][aria-busy=\"true\"]::after {{\n    \
1433         display: inline-block;\n    \
1434         inline-size: var(--awaiting-mark, 0.5em);\n    \
1435         block-size: var(--awaiting-mark, 0.5em);\n    \
1436         background: var(--action);\n    \
1437         opacity: 1;\n    \
1438         animation: makeover-activity calc(var(--cadence-activity) * 2) \
1439         step-end infinite;\n\
1440         }}"
1441    );
1442
1443    // The measured half. A wider box, no blink, and a gradient whose stop is
1444    // the share: the fill and the trough in one paint, so the markup stays one
1445    // pseudo-element on both branches.
1446    //
1447    // `--awaiting-share` unset is an empty trough, not a full one. A bar that
1448    // read full because nobody was counting would be the confidently-wrong
1449    // drawing rule 1 exists to forbid.
1450    //
1451    // The trough takes an edge for the reason a well does: a bar at zero share
1452    // is otherwise a rectangle of the surface it sits on, which is nothing at
1453    // all. `border_width` rather than a literal, the way `depth_rule` and
1454    // `track_rules` write theirs.
1455    let _ = writeln!(
1456        css,
1457        "[data-awaiting=\"determinate\"][aria-busy=\"true\"]::after {{\n    \
1458         inline-size: var(--awaiting-bar, 6em);\n    \
1459         animation: none;\n    \
1460         outline: {} solid var(--border);\n    \
1461         outline-offset: -{};\n    \
1462         background: linear-gradient(\n        \
1463         to inline-end,\n        \
1464         var(--action) 0 calc(var(--awaiting-share, 0) * 100%),\n        \
1465         var(--surface-sunken) 0\n    \
1466         );\n\
1467         }}",
1468        opts.border_width, opts.border_width
1469    );
1470
1471    css
1472}
1473
1474/// A strip of figures, and the two spans inside each one.
1475///
1476/// Colour only, which is the deferral rule applied to a component that badly
1477/// wants to break it. A figure reads as a figure because the value is set large
1478/// over a small caption, and that is a size: `makeover-geometry` answers how
1479/// much space and this crate answers what the thing is. Emitting `font-size`
1480/// here would be this crate naming a value, which is the one thing it is defined
1481/// by not doing, and `progress_rules` is the precedent — it emits the tones and
1482/// never the width, because the width is not its to know.
1483///
1484/// So the arrangement and the type scale are the app's, and what is generated is
1485/// the part an app cannot get right by itself: which of the two spans carries
1486/// the tone.
1487fn figure_rules(opts: &Emit) -> String {
1488    let figure = class("figure", opts);
1489    let value = class("figure-value", opts);
1490    let caption = class("figure-caption", opts);
1491    let change = class("figure-change", opts);
1492    let mut css = String::new();
1493
1494    // `content` literally. A figure's value is the thing itself, at full
1495    // weight -- wiki `three-tone-convention` classes it "active, emphasised",
1496    // and both other renderers already draw it that way (makeover-tui
1497    // `piece.rs:391` bold, makeover-immediate `widget.rs:244`). It reached
1498    // here through `Tone::Neutral.token()` and came out `content-muted`, so
1499    // the headline number sat at the colour of its own caption.
1500    let _ = writeln!(
1501        css,
1502        ".{figure} > .{value} {{\n    color: var(--content);\n}}"
1503    );
1504    let _ = writeln!(
1505        css,
1506        ".{figure} > .{caption} {{\n    color: var(--content-muted);\n}}"
1507    );
1508    let _ = writeln!(
1509        css,
1510        ".{figure} > .{change} {{\n    color: var(--content-muted);\n}}"
1511    );
1512
1513    // A toned figure tones one part and never the caption. The caption is the
1514    // noun and stays muted.
1515    //
1516    // Which part depends on whether there is a change, and that is the whole of
1517    // what 0.13.0 changed here. A figure with a delta is an ordinary number that
1518    // has moved in a direction worth reading, so the delta takes the colour and
1519    // the number stays plain; a figure without one has nowhere else to put it.
1520    // `:has` is what lets one attribute mean both, and the alternative was the
1521    // emitter deciding by writing the attribute onto a different element, which
1522    // leaves two elements able to disagree about a figure's one meaning.
1523    for tone in [Tone::Info, Tone::Success, Tone::Warning, Tone::Danger] {
1524        let _ = writeln!(
1525            css,
1526            ".{figure}[data-tone=\"{0}\"] > .{change} {{\n    color: var(--{0});\n}}",
1527            tone.token()
1528        );
1529        let _ = writeln!(
1530            css,
1531            ".{figure}[data-tone=\"{0}\"]:not(:has(> .{change})) > .{value} \
1532             {{\n    color: var(--{0});\n}}",
1533            tone.token()
1534        );
1535    }
1536    css
1537}
1538
1539/// A picture, its frame and its caption.
1540///
1541/// `makeover_layout::Image` arrived at 0.21.0, found by trying to describe MNW's
1542/// carousel and discovering nothing named a picture.
1543///
1544/// # The frame is a border, and this is the one place a bevel is wrong
1545///
1546/// 0.36.0 emitted [`Depth::Raised`] here through [`depth_rule`], the same call
1547/// `button` and `card` make. It was wrong, and MNW's landing page is where it
1548/// showed: **the frames had no visible edge at all.**
1549///
1550/// `Depth::Raised`'s edge is `--bevel-raised`, which is an *inset* shadow — a
1551/// 1px light run at the top-left and a dark one at the bottom-right, drawn
1552/// **inside** the element's box. On a button or a card that box is a surface
1553/// this crate owns, so an inset edge reads as the surface catching the light.
1554/// On a picture it is drawn on top of the picture, over whatever pixels the
1555/// image happens to have at its border. MNW's screenshots are light-on-light
1556/// parchment, so the light half landed on a light image and the frame
1557/// disappeared.
1558///
1559/// **You cannot bevel a surface you do not own.** A picture's content is the
1560/// app's, arrives at request time, and can be any colour, so its edge has to
1561/// sit *outside* the content rather than on it. That is a border.
1562///
1563/// The fill stays, and it is not decoration: it is what shows through a
1564/// transparent PNG and what stands in the frame's place while the image is
1565/// still loading.
1566///
1567/// This is a real limit on [`Depth`] rather than a special case. Every other
1568/// consumer of a depth draws its own surface; a picture is the first member
1569/// whose surface belongs to someone else.
1570///
1571/// # No size
1572///
1573/// `width: 100%` and nothing else. How large a picture is depends on the box it
1574/// was put in, which is the app's arrangement and `makeover-geometry`'s scales,
1575/// and a renderer that picked one would be answering for every consumer at
1576/// once. This is where `figure_rules` landed for the same reason.
1577fn picture_rules(opts: &Emit) -> String {
1578    let picture = class("picture", opts);
1579    let img = class("picture-img", opts);
1580    let caption = class("picture-caption", opts);
1581    let mut css = String::new();
1582
1583    // Block, or an inline image sits on the text baseline and carries a
1584    // descender's worth of space under it that no app ever wants and every app
1585    // deletes by hand. The border is the frame; see the type docs for why it is
1586    // not the bevel every other surface here gets.
1587    let _ = writeln!(
1588        css,
1589        ".{img} {{\n    display: block;\n    width: 100%;\n    height: auto;\n    \
1590         background: var(--{});\n    border: {} solid var(--border);\n}}",
1591        Fill::Raised.token(),
1592        opts.border_width
1593    );
1594
1595    // The two fits that need a rule. `Fit::Natural` emits no attribute at all,
1596    // so it is the bare rule above and needs nothing here.
1597    let _ = writeln!(
1598        css,
1599        ".{img}[data-fit=\"cover\"] {{\n    height: 100%;\n    object-fit: cover;\n}}"
1600    );
1601    let _ = writeln!(
1602        css,
1603        ".{img}[data-fit=\"contain\"] {{\n    height: 100%;\n    object-fit: contain;\n}}"
1604    );
1605
1606    // A caption reads back one step, which is `figure-caption`'s answer and the
1607    // same claim: it says what the thing above it is, and it is not the thing.
1608    let _ = writeln!(
1609        css,
1610        ".{picture} > .{caption} {{\n    color: var(--content-muted);\n}}"
1611    );
1612
1613    css
1614}
1615
1616/// A region showing one child at a time, and the chrome that moves between them.
1617///
1618/// [`makeover_layout::Showing`] lets a description say that a region holds
1619/// several children and shows some of them. A renderer derives its own chrome
1620/// from that, which is what stops every renderer growing a `match` on a widget
1621/// name; these are the rules the derived chrome needs.
1622///
1623/// # Why the default is every child, and the enhancement takes them away
1624///
1625/// The controls are a lie until something binds them. A prev button rendered
1626/// into a document with no script is a control that looks live and answers
1627/// nothing, and the reader it lies to is exactly the one who cannot see the
1628/// other children either — the collapsing and the moving are the same half.
1629///
1630/// So the rules run in the direction the enhancement does. Nothing here hides a
1631/// child and nothing here shows a control. Whatever binds the region sets
1632/// `data-ready` on it, and that is what collapses the stack to one and reveals
1633/// the row that moves it. A reader with no script gets every child in order and
1634/// no controls, which is more content rather than less, and a reader with
1635/// script gets a settled page rather than a stack that jumps to one frame after
1636/// load.
1637///
1638/// MNW proved this shape by hand — a `<noscript>` stylesheet opening its
1639/// carousel back out — and it is here rather than there because the property is
1640/// the description's, not one app's.
1641///
1642/// # Not spacing
1643///
1644/// The row's gaps are `makeover-geometry`'s question and are absent for
1645/// [`row_rules`]'s reason. What is here is `display`, which carries no
1646/// magnitude, and the muted readout, which is the same claim `picture-caption`
1647/// makes: it says where you are among the children and it is not one of them.
1648fn showing_rules(opts: &Emit) -> String {
1649    let controls = class("showing", opts);
1650    let position = class("showing-position", opts);
1651    let frame = class("showing-frame", opts);
1652    let mut css = String::new();
1653
1654    // Hidden until something binds it, which is the whole argument above.
1655    let _ = writeln!(css, ".{controls} {{\n    display: none;\n}}");
1656    // Block, and nothing about how the three sit in it. A button and a span are
1657    // inline already, so they make a row without this crate saying so, and
1658    // saying so is where `align-items` and a gap would follow -- both spacing,
1659    // both `makeover-geometry`'s, and `row_rules` refuses them for the same
1660    // reason.
1661    let _ = writeln!(
1662        css,
1663        "[data-ready] > .{controls} {{\n    display: block;\n}}"
1664    );
1665
1666    // A child is in flow until the region is bound, and then only the current
1667    // one is. `.current` is a modifier for the reason `.chosen` and `.latched`
1668    // are: one name for the state, set by whoever knows it.
1669    let _ = writeln!(
1670        css,
1671        "[data-ready] > .{frame}:not(.current) {{\n    display: none;\n}}"
1672    );
1673
1674    // Reads back one step. `picture-caption`'s rule and its reason.
1675    let _ = writeln!(css, ".{position} {{\n    color: var(--content-muted);\n}}");
1676
1677    css
1678}
1679
1680/// A time axis: the container, its gridlines and ruler, and the placed things.
1681///
1682/// `makeover-layout` 0.24.0's [`Track`](makeover_layout::Track), and the first
1683/// member here whose whole point is *position*. That makes the magnitude line
1684/// this crate keeps worth restating rather than assuming.
1685///
1686/// # Where the numbers come from
1687///
1688/// Every value that varies per item is a custom property the caller sets
1689/// inline, and every rule here reads one. `--track-at` and `--track-for` are
1690/// percentages of the span, which
1691/// [`Track::fraction`](makeover_layout::Track::fraction) computes once so three
1692/// renderers cannot disagree about it. Nothing here knows a pixel.
1693///
1694/// That is what lets the stylesheet stay static while the items move: an entry
1695/// carries `style="--track-at: 37.5%; --track-for: 4.166%"` and the rule below
1696/// turns it into a box. The alternative was emitting a rule per item, which is
1697/// a stylesheet that grows with the data.
1698///
1699/// **The height of the track is the app's**, not this crate's. 96 quarter-hour
1700/// slots at some slot height is a size, and a size is `makeover-geometry`'s
1701/// question -- the same refusal `figure_rules` and `placeholder_rules` make.
1702/// Percentages need a resolved height above them, so `.track` gets
1703/// `position: relative` and nothing else; the app says how tall a day is.
1704///
1705/// # Overlap
1706///
1707/// Two things at the same time is intrinsic to an axis and has no analogue in a
1708/// list. The description does not declare it -- `Placement::overlaps` derives it
1709/// from the times -- so what arrives here is a lane index and a lane count, and
1710/// the rule divides the width. A renderer that would rather stack them ignores
1711/// both properties and the defaults give it one full-width lane.
1712fn track_rules(opts: &Emit) -> String {
1713    let track = class("track", opts);
1714    let slot = class("track-slot", opts);
1715    let tick = class("track-tick", opts);
1716    let entry = class("track-entry", opts);
1717    let mut css = String::new();
1718
1719    // The positioning context every entry resolves against, and the whole of
1720    // what this crate says about the container. No height: see the doc above.
1721    let _ = writeln!(css, ".{track} {{\n    position: relative;\n}}");
1722
1723    // Gridlines and the ruler are the axis reading itself back, which is
1724    // `picture-caption` and `showing-position`'s claim: about the thing rather
1725    // than one of the things.
1726    //
1727    // `border_width` rather than a literal, the way `depth_rule` writes its
1728    // edge. A gridline is the one place a timeline would most naturally reach
1729    // for a hardcoded 1px, and a hardcoded 1px is this crate naming a size.
1730    let _ = writeln!(
1731        css,
1732        ".{slot} {{\n    border-top: {} solid var(--border);\n}}",
1733        opts.border_width
1734    );
1735    let _ = writeln!(css, ".{tick} {{\n    color: var(--content-muted);\n}}");
1736
1737    // The one rule that does real work. Top and height are the placement;
1738    // left and width are the lane. Both lane properties default so an entry
1739    // that names neither is full width, which is the common case and the one a
1740    // renderer gets for free.
1741    let _ = writeln!(
1742        css,
1743        ".{entry} {{\n    \
1744         position: absolute;\n    \
1745         top: var(--track-at, 0%);\n    \
1746         height: var(--track-for, 100%);\n    \
1747         left: calc(var(--track-lane, 0) / var(--track-lanes, 1) * 100%);\n    \
1748         width: calc(100% / var(--track-lanes, 1));\n\
1749         }}"
1750    );
1751
1752    css
1753}
1754
1755/// A region's stand-in, and the header of a table that can be reordered.
1756///
1757/// Both are 0.12.0 members and both are colour and affordance only, which is
1758/// where `figure_rules` landed after trying to emit a type scale. How much room
1759/// a stand-in gets is a size — goingson has the same one at three, as
1760/// `--compact`, `--dashboard` and `--padded` — and a size is
1761/// `makeover-geometry`'s question.
1762///
1763/// The caret is the one thing here that is neither colour nor affordance, and it
1764/// is a renderer's own expression rather than a value the description named:
1765/// `aria-sort` is what the table actually says, and this turns it into something
1766/// visible for everyone not using a screen reader. A terminal draws its own; an
1767/// immediate-mode painter draws its own.
1768fn state_rules(opts: &Emit) -> String {
1769    let placeholder = class("placeholder", opts);
1770    let text = class("placeholder-text", opts);
1771    let heading = class("table-heading", opts);
1772    let mut css = String::new();
1773
1774    let _ = writeln!(
1775        css,
1776        ".{placeholder} > .{text} {{\n    color: var(--content-muted);\n}}"
1777    );
1778    // Only the failure is toned. An empty list is the normal state of a new
1779    // install, and `Readiness::tone` is what says so.
1780    let _ = writeln!(
1781        css,
1782        ".{placeholder}[data-tone=\"{0}\"] > .{text} {{\n    color: var(--{0});\n}}",
1783        Tone::Danger.token()
1784    );
1785
1786    // A header that reorders the table is a control, and the pointer is the
1787    // only part of saying so that is not the app's own type and spacing.
1788    let _ = writeln!(
1789        css,
1790        ".{heading}[data-sortable] {{\n    cursor: pointer;\n}}"
1791    );
1792    // The caret carries its own leading space, the way `makeover-tui` and
1793    // `makeover-immediate` both write `" \u{25B2}"`. It used to be emitted bare,
1794    // and both apps that adopted the vocabulary had to put the gap back in their
1795    // own stylesheets on the same afternoon -- each having to work out first that
1796    // app CSS outranks this crate's cascade layer, so adding the space the
1797    // obvious way, as `content`, silently wins over the glyph and leaves the
1798    // heading with no caret at all. A consumer should not have to know that, and
1799    // with the space emitted here there is nothing left for one to add.
1800    //
1801    // Three states, three tones, on the convention in wiki
1802    // `three-tone-convention`. A column in force is `content`; a column offering
1803    // to reorder and not doing it now is `content-secondary`, because it still
1804    // answers a press; a column that is not sortable emits no caret at all and
1805    // takes nothing. The idle arm used to hide its glyph and reserve the box,
1806    // which cost a reflow-free press and said nothing. It draws now, and the
1807    // reservation stops being a thing to get right.
1808    //
1809    // The glyph is `Sort::glyph`, escaped rather than written: a CSS `content`
1810    // string cannot carry the character literally through this file's own
1811    // escaping, and spelling it here as well would put the third copy back that
1812    // `makeover-layout` 0.27.5 exists to remove.
1813    let _ = writeln!(
1814        css,
1815        ".{heading}[data-sortable]::after \
1816         {{\n    content: \" {}\";\n    color: var(--content-secondary);\n}}",
1817        css_escape(Sort::Ascending.glyph())
1818    );
1819    for direction in [Sort::Ascending, Sort::Descending] {
1820        let _ = writeln!(
1821            css,
1822            ".{heading}[aria-sort=\"{}\"]::after \
1823             {{\n    content: \" {}\";\n    color: var(--content);\n}}",
1824            direction.as_str(),
1825            css_escape(direction.glyph())
1826        );
1827    }
1828    css
1829}
1830
1831/// The frame a table sits in.
1832///
1833/// # Why this is a CSS table and not the grid the rest of the module assumes
1834///
1835/// A grid row needs `grid-template-columns`, which has to name every column in
1836/// order, so it cannot be written without knowing the columns. That is what
1837/// [`list::narrowing_css`] is for, and it works: goingson builds `tables.css` in
1838/// its own `build.rs` out of it, and nothing here changes that.
1839///
1840/// It does not work for a table a *description* produced. Those columns are
1841/// known at render time rather than at build time, and the rules would have to
1842/// travel with the markup: a `<style>` element per table, which needs
1843/// `style-src 'unsafe-inline'` and so blocks the MNW server's standing plan to
1844/// drop it. The head is not an escape: a table swapped in by htmx after a
1845/// delete arrives as a fragment with no head at all.
1846///
1847/// A CSS table aligns its columns across rows knowing nothing about how many
1848/// there are, so there is no track list to emit and nothing per-table to carry.
1849/// The cost is that a described table cannot take a per-column fixed length,
1850/// which costs nothing today: [`Sizing`](list::Sizing) is looked up by column
1851/// name and a description carries no lengths to put in it, so every track a
1852/// described table could ask for is already content, fill or auto.
1853///
1854/// [`Priority`] hiding is unchanged in kind. It moves from a generated
1855/// `display: none` per dropped column to one rule per drop class, which is the
1856/// same fact addressed by class rather than by cutoff, and still never by
1857/// position.
1858fn table_rules(opts: &Emit) -> String {
1859    let table = class("table", opts);
1860    let head = class("table-head", opts);
1861    let row = class("table-row", opts);
1862    let heading = class("table-heading", opts);
1863    let cell = class("cell", opts);
1864    let mut css = String::new();
1865
1866    let _ = writeln!(
1867        css,
1868        ".{table} {{\n    display: table;\n    width: 100%;\n}}"
1869    );
1870    let _ = writeln!(css, ".{head},\n.{row} {{\n    display: table-row;\n}}");
1871    // Scoped under `.{table}` rather than keyed on the classes alone. The
1872    // table model is what a cell takes *by being in a table*, and only there:
1873    // goingson's task headings are `.table-heading` inside a CSS grid, so the
1874    // unscoped rule reached them, and the sort caret and the sortable cursor
1875    // -- which `state_rules` keys on `.table-heading` and rightly still does
1876    // -- came with a `display` the grid had to blockify away.
1877    //
1878    // It costs no markup anywhere: quasi-webview always nests the heading
1879    // inside the table (`quasi-webview/src/node.rs:1591`, `:1605`, `:1615`).
1880    let _ = writeln!(
1881        css,
1882        ".{table} .{heading},\n.{table} .{cell} {{\n    display: table-cell;\n}}"
1883    );
1884
1885    // A content column shrinks to what is in it. `width: 1%` is how a CSS table
1886    // is told that: auto layout hands the slack to the columns that asked for
1887    // room, and a column asking for almost none gets what it needs and no more.
1888    // The `nowrap` is what stops it being given less by wrapping.
1889    let _ = writeln!(
1890        css,
1891        ".{} {{\n    white-space: nowrap;\n    width: 1%;\n}}",
1892        class("cell-content", opts)
1893    );
1894
1895    // A fixed column has no length to be fixed to. The description carries none
1896    // and `Sizing` is not reachable from here, so it behaves as content: the
1897    // honest answer to a width nobody supplied, and the same one `Sizing::track`
1898    // gives it.
1899    let _ = writeln!(
1900        css,
1901        ".{} {{\n    white-space: nowrap;\n}}",
1902        class("cell-fixed", opts)
1903    );
1904
1905    // Optional columns go at the narrowest class, secondary ones go with them,
1906    // which is the cutoff walk `kept_at` describes said as two media queries.
1907    // Essential columns have no rule at all, because never dropping is what not
1908    // being mentioned already means.
1909    for (size, drops) in [
1910        (
1911            SizeClass::Compact,
1912            &["cell-drops-first", "cell-drops-next"][..],
1913        ),
1914        (SizeClass::Medium, &["cell-drops-first"][..]),
1915    ] {
1916        let selectors: Vec<String> = drops
1917            .iter()
1918            .map(|drop| format!(".{}", class(drop, opts)))
1919            .collect();
1920        css.push_str(&gated(
1921            Some(&size.media_condition()),
1922            &format!("{} {{\n    display: none;\n}}\n", selectors.join(",\n")),
1923        ));
1924    }
1925
1926    // What is inside a cell, which the table side could not say until
1927    // makeover-layout 0.14.0. Every cell was one `.cell` and one content
1928    // colour, so a button in a cell was painted as text -- the drift
1929    // `RowPart::intent` has prevented for list rows since 0.2.0 and prevented
1930    // for nothing here.
1931    //
1932    // The colour goes on `.cell-value` rather than on `.cell`, and that
1933    // placement is the whole fix. On the container it would cascade into the
1934    // tokens and the controls sitting beside the text, which is the bug said
1935    // in one rule; on the part that is text, it reaches text and stops.
1936    for part in [
1937        CellPart::Value,
1938        CellPart::Tokens,
1939        CellPart::Actions,
1940        CellPart::Link,
1941    ] {
1942        // Three of the four inherit, each for its own reason: a token carries
1943        // its own tone, an action is a control rather than text, and a link
1944        // takes the action colour from the anchor it is. `CellPart::intent`
1945        // says so by answering with the intent inheriting already gives, and
1946        // pinning that would be louder than saying nothing.
1947        //
1948        // Written as a skip-list rather than as a match on Value, so a member
1949        // added upstream gets its intent emitted rather than being silently
1950        // dropped. That is the same trade `part_class`'s fallback makes: land
1951        // plainly, never land as nothing.
1952        if !matches!(part, CellPart::Tokens | CellPart::Actions | CellPart::Link) {
1953            let _ = writeln!(
1954                css,
1955                ".{} {{\n    color: var(--{});\n}}",
1956                class(cell_part_class(part), opts),
1957                part.intent()
1958            );
1959        }
1960    }
1961
1962    css
1963}
1964
1965/// The component layer: every named thing phase A emits.
1966///
1967/// No scrollbar track. It was on the phase A list and came off: eight lines of
1968/// `::-webkit-scrollbar` with no shape a terminal or an immediate-mode painter
1969/// would want handed to it, so it stays with the apps.
1970#[must_use]
1971pub fn component_rules(opts: &Emit) -> String {
1972    let mut css = String::new();
1973    css.push_str(&surface_rules(opts));
1974    css.push_str(&link_rules(opts));
1975    css.push_str(&token_rules(opts));
1976    css.push_str(&selector_rules(opts));
1977    css.push_str(&row_rules(opts));
1978    css.push_str(&run_rules(opts));
1979    css.push_str(&progress_rules(opts));
1980    css.push_str(&awaiting_rules(opts));
1981    css.push_str(&figure_rules(opts));
1982    css.push_str(&picture_rules(opts));
1983    css.push_str(&showing_rules(opts));
1984    css.push_str(&track_rules(opts));
1985    css.push_str(&state_rules(opts));
1986    css.push_str(&table_rules(opts));
1987    css.push_str(&facet::facet_rules(opts));
1988    css.push_str(&form::editor_rules(opts));
1989    css.push_str(&form::suggestion_rules(opts));
1990    css.push_str(&form::unit_rules(opts));
1991    css.push_str(&form::option_detail_rules(opts));
1992    css.push_str(&form::note_rules(opts));
1993    css
1994}
1995
1996/// The whole phase-A stylesheet: properties, depth rules and components, in
1997/// [`CSS_LAYER`], under a generated-file banner.
1998///
1999/// The banner sits outside the layer, because a comment participates in no
2000/// cascade and a reader opening the file should see what it is before seeing
2001/// an at-rule.
2002#[must_use]
2003pub fn stylesheet(opts: &Emit) -> String {
2004    let body = in_css_layer(&format!(
2005        ":root {{\n{}}}\n\n{}\n{}",
2006        bevel_properties(opts),
2007        depth_rules(opts),
2008        component_rules(opts)
2009    ));
2010    // Counted from the body rather than through `vocabulary::names`, which
2011    // calls back into here.
2012    let classes = vocabulary::classes_in_css(&body).len();
2013    let version = VERSION;
2014    format!(
2015        "/* Generated by makeover-webview {version} from makeover-layout, \
2016         {classes} classes.\n   \
2017         Do not edit. The version and the count are here because a stale\n   \
2018         lockfile fails silently: an older emitter writes a well-formed sheet\n   \
2019         with components missing, and nothing else in the file says so. If\n   \
2020         this version trails what the manifest asks for, re-resolve.\n\n   \
2021         Depth is a fill and an edge together; naming them apart is what let\n   \
2022         them disagree. See the crate's README and wiki note makeover-layout.\n\n   \
2023         Everything below is in the `{CSS_LAYER}` cascade layer. Declare the\n   \
2024         order once in your own stylesheet, or this layer's position is decided\n   \
2025         by whichever generated file the browser happens to see first:\n\n   \
2026         @layer {CSS_LAYER}, base, components, responsive; */\n{body}"
2027    )
2028}
2029
2030#[cfg(test)]
2031mod tests {
2032    use super::*;
2033    use makeover_layout::Edge;
2034
2035    #[test]
2036    fn every_fallback_class_is_one_a_checker_knows_about() {
2037        // The obligation ROW_PART_CLASSES carries, for the same reason: a class
2038        // this crate can write and the vocabulary list does not carry is
2039        // invisible to the dead-vocabulary seal and to the overlap check both.
2040        for fallback in [
2041            Fallback::Wrap,
2042            Fallback::Stack,
2043            Fallback::Shed,
2044            Fallback::Menu,
2045        ] {
2046            assert!(
2047                RUN_CLASSES.contains(&fallback_class(fallback)),
2048                "{fallback:?} is missing from RUN_CLASSES"
2049            );
2050        }
2051        let names = crate::vocabulary::names(&Emit::default());
2052        for name in RUN_CLASSES {
2053            assert!(names.contains(*name), "{name} is not in the vocabulary");
2054        }
2055    }
2056
2057    #[test]
2058    fn a_run_gives_every_member_a_floor_it_cannot_be_squeezed_below() {
2059        // The whole of what stops the overlap, and it is not a fallback: it
2060        // applies to every run whatever the group declared. flexbox's default
2061        // min-width is auto, which lets an item be compressed below its own
2062        // content in a nowrap row, and that is how a toolbar is drawn over a
2063        // tab strip even with nothing out of flow.
2064        let css = run_rules(&Emit::default());
2065        assert!(css.contains(".run > * {\n    min-width: min-content;\n}"));
2066        // No number anywhere in it. The minimum is derived by the browser from
2067        // what the members contain, which is the ruling's own requirement.
2068        assert!(!css.contains("px"));
2069        assert!(!css.contains("rem"));
2070        assert!(!css.contains("@media"));
2071    }
2072
2073    #[test]
2074    fn room_is_never_asked_of_the_viewport() {
2075        // The 913 case: a window in SizeClass::Expanded holding a group out of
2076        // room. A viewport query answers about the window and would be wrong
2077        // about the group, which is why the table's @media walk is not the
2078        // precedent this follows.
2079        let css = run_rules(&Emit::default());
2080        for size in [SizeClass::Compact, SizeClass::Medium] {
2081            assert!(!css.contains(&size.media_condition()));
2082        }
2083    }
2084
2085    #[test]
2086    fn every_fallback_lands_as_a_class_and_an_unknown_one_lands_plainly() {
2087        let css = run_rules(&Emit::default());
2088        for fallback in [
2089            Fallback::Wrap,
2090            Fallback::Stack,
2091            Fallback::Shed,
2092            Fallback::Menu,
2093        ] {
2094            let class = fallback_class(fallback);
2095            assert!(css.contains(&format!(".{class} {{")), "{class} unemitted");
2096        }
2097        // Stack is the one that also says what a member does with the line it
2098        // took, which is what separates it from wrapping.
2099        assert!(css.contains(".run-stack > * {\n    flex: 1 1 max-content;\n}"));
2100    }
2101
2102    #[test]
2103    fn the_emitted_bevel_matches_what_the_apps_already_hand_write() {
2104        // Balanced Breakfast's styles.css, verbatim. Adoption has to be a
2105        // deletion, not a redesign, or nobody will take it.
2106        let opts = Emit::default();
2107        assert_eq!(
2108            bevel_shadow(Bevel::Raised, &opts),
2109            "inset 1px 1px 0 var(--bevel-light), inset -1px -1px 0 var(--bevel-dark)"
2110        );
2111        assert_eq!(
2112            bevel_shadow(Bevel::Inset, &opts),
2113            "inset 1px 1px 0 var(--bevel-dark), inset -1px -1px 0 var(--bevel-light)"
2114        );
2115    }
2116
2117    #[test]
2118    fn no_colour_ever_reaches_the_output() {
2119        let css = stylesheet(&Emit::default());
2120        assert!(!css.contains('#'), "a hex literal escaped into the CSS");
2121        assert!(
2122            !css.contains("rgb"),
2123            "a colour function escaped into the CSS"
2124        );
2125        // Every colour is named, never resolved.
2126        assert!(css.contains("var(--surface-raised)"));
2127        assert!(css.contains("var(--bevel-light)"));
2128    }
2129
2130    #[test]
2131    fn a_well_falls_back_through_css_rather_than_through_rust() {
2132        assert_eq!(
2133            fill_var(Fill::Well),
2134            "var(--surface-well, var(--surface-page))"
2135        );
2136        // Nothing else needs one.
2137        assert_eq!(fill_var(Fill::Raised), "var(--surface-raised)");
2138        assert_eq!(fill_var(Fill::Page), "var(--surface-page)");
2139    }
2140
2141    #[test]
2142    fn raised_and_well_do_not_collapse_onto_each_other() {
2143        let css = depth_rules(&Emit::default());
2144        assert!(css.contains(".raised {"));
2145        assert!(css.contains(".well {"));
2146        assert!(css.contains("var(--bevel-raised)"));
2147        assert!(css.contains("var(--bevel-inset)"));
2148    }
2149
2150    /// The cast shadow is composed here from the tone `makeover` derives, so
2151    /// neither crate has to hold the other's numbers.
2152    ///
2153    /// It is a `:root` property and deliberately not a depth class. There is no
2154    /// `Depth::Overlay` in the description layer, and adding one would be a
2155    /// claim about what a screen means rather than about how it is painted;
2156    /// until something asks for it, a consumer names the property on the rule
2157    /// for the menu or the toast it already has.
2158    #[test]
2159    fn the_cast_shadow_is_a_root_property_not_a_depth() {
2160        let css = bevel_properties(&Emit::default());
2161        assert!(css.contains("--elevation-overlay:"));
2162        assert!(css.contains("var(--elevation)"));
2163        assert!(
2164            !depth_rules(&Emit::default()).contains("elevation"),
2165            "elevation is not a depth class"
2166        );
2167    }
2168
2169    #[test]
2170    fn the_cascade_carries_the_pressed_state() {
2171        let css = surface_rules(&Emit::default());
2172        // The one thing this renderer gets free that the other two resolve by
2173        // hand, eighteen call sites deep in audiofiles' case. Asserted on a
2174        // named surface: pressing belongs to the control, not to the depth.
2175        assert!(css.contains(".card:active {"));
2176        assert!(css.contains(".button:active {"));
2177    }
2178
2179    #[test]
2180    fn the_depth_class_is_a_surface_and_not_a_control() {
2181        let css = depth_rules(&Emit::default());
2182        // The static surface the vocabulary was missing. Sixteen goingson
2183        // elements wore .card and cancelled its hover and press to get this,
2184        // because a raised object that is not pressable had no other spelling.
2185        for state in [":hover", ":active", ":focus-visible", ":disabled"] {
2186            assert!(
2187                !css.contains(&format!(".raised{state}")),
2188                "the depth class claimed {state}: {css}"
2189            );
2190        }
2191        assert!(css.contains("var(--bevel-raised)"), "still raised: {css}");
2192    }
2193
2194    #[test]
2195    fn pressing_moves_the_fill_and_not_only_the_edge() {
2196        // The decision-1 guard, and the regression that mattered: emitting the
2197        // bevel flip alone is what left goingson hand-writing `background:
2198        // var(--surface-sunken)` on .btn, .card and .tag/.badge alike, so none
2199        // of the three could be deleted.
2200        let pressed = interactive_rules("button", Depth::Raised, &Emit::default());
2201        assert!(pressed.contains(".button:active {"));
2202        assert!(
2203            pressed.contains("background: var(--surface-well, var(--surface-page))"),
2204            "pressed dropped its fill: {pressed}"
2205        );
2206        assert!(pressed.contains("box-shadow: var(--bevel-inset)"));
2207    }
2208
2209    #[test]
2210    fn pressed_takes_its_fill_from_the_description_not_from_the_app() {
2211        // goingson presses to --surface-sunken. The description says a pressed
2212        // raised region reads as a well, and makeover says outright that
2213        // surface-sunken cannot serve as one, so the app is the thing that
2214        // moves.
2215        //
2216        // Scoped to the pressed rules rather than to the whole sheet: since
2217        // makeover-layout 0.3.0 an unchosen tab is legitimately
2218        // --surface-sunken, so the token appearing somewhere in the output no
2219        // longer means the app's choice leaked in.
2220        let css = stylesheet(&Emit::default());
2221        let mut checked = 0;
2222        for rule in css.split("}\n") {
2223            if !rule.contains(":active") {
2224                continue;
2225            }
2226            checked += 1;
2227            assert!(
2228                !rule.contains("surface-sunken"),
2229                "a pressed rule took the app's fill: {rule}"
2230            );
2231        }
2232        assert!(checked > 0, "no pressed rules found to check");
2233        assert_eq!(
2234            Depth::Raised.pressed().fill(),
2235            Some(Fill::Well),
2236            "the description changed under us"
2237        );
2238    }
2239
2240    #[test]
2241    fn the_whole_stylesheet_is_emitted_in_the_family_layer() {
2242        // The point of 0.11.0. Unlayered normal declarations outrank every
2243        // named layer, so an app declaring `@layer base, components` loses
2244        // every rule it owns to this file until this file is layered too.
2245        let css = stylesheet(&Emit::default());
2246        assert!(css.contains(&format!("@layer {CSS_LAYER} {{")));
2247
2248        // Exactly one layer block, and nothing outside it but the banner.
2249        assert_eq!(css.matches("@layer").count(), 2, "banner names it once");
2250        let opened = css.find("@layer makeover {").expect("layer opens");
2251        for (i, line) in css.lines().enumerate() {
2252            let before_layer = css.lines().take(i).map(str::len).sum::<usize>() < opened;
2253            if before_layer || line.is_empty() {
2254                continue;
2255            }
2256            assert!(
2257                line.starts_with("    ") || line == "}" || line.starts_with("   "),
2258                "line outside the layer: {line:?}"
2259            );
2260        }
2261    }
2262
2263    #[test]
2264    fn the_generated_sheet_carries_no_trailing_whitespace() {
2265        // A checked-in generated file that a formatter wants to rewrite is a
2266        // diff every time somebody saves it.
2267        let css = stylesheet(&Emit::default());
2268        for (i, line) in css.lines().enumerate() {
2269            assert_eq!(line, line.trim_end(), "trailing whitespace on line {i}");
2270        }
2271    }
2272
2273    #[test]
2274    fn the_banner_tells_an_app_how_to_order_the_layer() {
2275        // Without a declared order the layer's position depends on which
2276        // generated file the browser sees first, which is not a contract.
2277        let css = stylesheet(&Emit::default());
2278        assert!(css.contains("@layer makeover, base, components, responsive;"));
2279        // And the banner is outside the layer, not a rule inside it.
2280        assert!(css.starts_with("/* Generated by makeover-webview"));
2281    }
2282
2283    #[test]
2284    fn the_banner_names_the_emitter_so_a_stale_pin_is_visible_on_sight() {
2285        // A consumer whose lockfile pins an old version gets a well-formed
2286        // sheet with components missing and no error. balanced_breakfast ran
2287        // on 657 bytes from a 0.1.0 emitter while its manifest asked for
2288        // 0.5.1, and the only way it surfaced was diffing two apps' generated
2289        // files. The version and the count are what the file says instead.
2290        let css = stylesheet(&Emit::default());
2291        let banner = css.lines().next().unwrap();
2292        assert!(
2293            banner.contains(VERSION),
2294            "{banner} does not name the emitter"
2295        );
2296        let classes = vocabulary::classes_in_css(&css).len();
2297        assert!(classes > 0);
2298        assert!(
2299            banner.contains(&format!("{classes} classes")),
2300            "{banner} does not carry the class count"
2301        );
2302    }
2303
2304    #[test]
2305    fn a_primitive_owns_every_state_it_implies() {
2306        // The whole point of 0.10.0. Anything emitting a hover rule owes the
2307        // other three, or the consuming app supplies them by out-specifying a
2308        // rule it does not own: 19 such rules in goingson, 21 in the MNW
2309        // server, and three focus rings that do not match.
2310        let css = stylesheet(&Emit::default());
2311        for selector in ["button", "card", "chip", "tab", "segment", "toggle"] {
2312            assert!(css.contains(&format!(".{selector}:hover {{")), "{selector}");
2313            assert!(
2314                css.contains(&format!(".{selector}:active {{")),
2315                "{selector}"
2316            );
2317            assert!(
2318                css.contains(&format!(".{selector}:focus-visible {{")),
2319                "{selector} has no focus ring"
2320            );
2321            assert!(
2322                css.contains(&format!(".{selector}:disabled,")),
2323                "{selector} has no disabled state"
2324            );
2325        }
2326    }
2327
2328    #[test]
2329    fn a_field_takes_focus_and_refuses_input_without_taking_a_hover() {
2330        // A text field does not light up under the pointer, so it gets the two
2331        // states it has and not the two it does not.
2332        let css = stylesheet(&Emit::default());
2333        assert!(css.contains(".field:focus-visible {"));
2334        assert!(css.contains(".field:disabled,"));
2335        assert!(!css.contains(".field:hover {"));
2336        assert!(!css.contains(".field:active {"));
2337    }
2338
2339    #[test]
2340    fn disabled_is_emitted_after_hover_so_source_order_settles_it() {
2341        // Every one of these selectors is specificity (0,2,0), so nothing but
2342        // order decides which wins. A disabled button taking the hover fill is
2343        // the exact bug goingson's `.button:disabled:hover` was written to fix,
2344        // and the reason it had to reach (0,3,0) to do it.
2345        let css = interactive_rules("button", Depth::Raised, &Emit::default());
2346        let hover = css.find(":hover").expect("hover");
2347        let active = css.find(":active").expect("active");
2348        let focus = css.find(":focus-visible").expect("focus");
2349        let disabled = css.find(":disabled").expect("disabled");
2350        assert!(hover < active && active < focus && focus < disabled);
2351
2352        // And it restores the surface, or the hover fill survives underneath.
2353        let tail = &css[disabled..];
2354        assert!(tail.contains("background: var(--surface-raised)"));
2355    }
2356
2357    #[test]
2358    fn a_disabled_state_reaches_things_that_cannot_be_disabled() {
2359        // `:disabled` matches form elements only, and a chip is a div. Keying
2360        // on the ARIA attribute too is the pattern the invalid field already
2361        // set: one fact, read by the styling and the accessibility tree alike.
2362        let css = disabled_rule("chip", Depth::Raised);
2363        assert!(css.contains(".chip:disabled,"));
2364        assert!(css.contains(".chip[aria-disabled=\"true\"]"));
2365        assert!(css.contains("cursor: not-allowed"));
2366    }
2367
2368    #[test]
2369    fn the_focus_ring_does_not_disturb_the_bevel_it_lands_on() {
2370        // `outline` has its own property, so unlike the invalid ring there is
2371        // no bevel to restate beside it and nothing to keep in agreement.
2372        let opts = Emit::default();
2373        let css = focus_rule("button", Depth::Raised, &opts);
2374        assert!(css.contains("outline: 2px solid var(--focus-ring)"));
2375        assert!(!css.contains("box-shadow"), "the ring restated the bevel");
2376    }
2377
2378    #[test]
2379    fn a_well_takes_the_ring_inside_and_a_raised_surface_outside() {
2380        // One ring, placed by depth. The offset comes off `Depth::bevel` and
2381        // not off a per-component choice, which is what gave three apps three
2382        // different rings.
2383        let opts = Emit::default();
2384        assert!(focus_rule("field", Depth::Well, &opts).contains("outline-offset: calc(-1 * 2px)"));
2385        assert!(focus_rule("button", Depth::Raised, &opts).contains("outline-offset: 2px"));
2386        // Nothing to sit inside of, so it sits outside.
2387        assert!(focus_rule("badge", Depth::Sunken, &opts).contains("outline-offset: 2px"));
2388
2389        // And the ring is not the bevel. Reusing border_width emitted a 1px
2390        // ring that every consumer had already overridden.
2391        assert_ne!(opts.focus_width, opts.border_width);
2392    }
2393
2394    #[test]
2395    fn hover_is_gated_on_capability_and_the_keyboard_path_is_not() {
2396        // goingson's section 60 exists only to take back the hover state this
2397        // crate handed it. Gating at the source is what deletes that section
2398        // in all three apps rather than having each fight for it.
2399        let css = stylesheet(&Emit::default());
2400        let condition = format!("@media {}", Density::Pointer.media_condition());
2401        assert!(css.contains(&condition));
2402
2403        // What is gated is every hover state the surfaces carry. The row's
2404        // actions used to be the other half of this test and are not gated any
2405        // more, because they are not hidden any more: a rule that reveals
2406        // nothing needs no capability answer.
2407        let gated: Vec<&str> = css.lines().filter(|line| line.contains(":hover")).collect();
2408        assert!(!gated.is_empty(), "{css}");
2409        for line in gated {
2410            let indent = line.len() - line.trim_start().len();
2411            assert!(indent > 4, "an ungated hover rule: {line}");
2412        }
2413        assert!(!css.contains(".row:hover"), "{css}");
2414    }
2415
2416    #[test]
2417    fn the_capability_answer_is_asked_for_and_not_assumed() {
2418        // Both halves come from the crates that own them. If `makeover-touch`
2419        // ever says a fingertip has hover, this stops gating on its own.
2420        assert!(!Affordance::Hover.available(Density::Touch, SizeClass::Compact));
2421        assert!(Affordance::Hover.available(Density::Pointer, SizeClass::Compact));
2422        assert_eq!(hover_condition(), Some(Density::Pointer.media_condition()));
2423
2424        // And the size class passed to that call is not a claim about width.
2425        assert!(Affordance::Hover.reads_density());
2426        for size in [SizeClass::Compact, SizeClass::Medium, SizeClass::Expanded] {
2427            assert!(!Affordance::Hover.available(Density::Touch, size));
2428        }
2429    }
2430
2431    #[test]
2432    fn hover_resolves_against_the_token_makeover_already_derives() {
2433        let css = interactive_rules("card", Depth::Raised, &Emit::default());
2434        assert!(css.contains(".card:hover {"));
2435        assert!(css.contains("background: var(--hover-surface)"));
2436        // Not the app's choice, which was --surface-overlay.
2437        assert!(!css.contains("surface-overlay"));
2438    }
2439
2440    #[test]
2441    fn a_badge_gets_no_edge_and_no_fill() {
2442        // Decision 2, and the one visible redesign in phase A. Token::Badge is
2443        // Flat: an edge on a label says it can be pressed.
2444        let css = token_rules(&Emit::default());
2445        let badge = css
2446            .lines()
2447            .skip_while(|l| !l.starts_with(".badge {"))
2448            .take_while(|l| !l.starts_with('}'))
2449            .collect::<Vec<_>>()
2450            .join("\n");
2451        assert!(!badge.contains("box-shadow"), "badge kept an edge: {badge}");
2452        assert!(!badge.contains("background"), "badge kept a fill: {badge}");
2453        assert_eq!(Token::Badge.depth(false), Depth::Flat);
2454        assert_eq!(Token::Badge.depth(true), Depth::Flat);
2455    }
2456
2457    #[test]
2458    fn a_badge_carries_a_tone_and_neutral_is_the_bare_class() {
2459        let css = token_rules(&Emit::default());
2460        // Neutral is the absence of a status, not a status named "none".
2461        assert!(css.contains(".badge {\n    color: var(--content-muted);"));
2462        assert!(!css.contains("data-tone=\"content-muted\""));
2463        for tone in ["info", "success", "warning", "danger"] {
2464            assert!(
2465                css.contains(&format!(".badge[data-tone=\"{tone}\"]")),
2466                "missing tone {tone}"
2467            );
2468            assert!(css.contains(&format!("color: var(--{tone})")));
2469        }
2470    }
2471
2472    #[test]
2473    fn a_chip_is_raised_and_latches_into_a_well() {
2474        let css = token_rules(&Emit::default());
2475        assert!(css.contains(".chip {"));
2476        assert!(css.contains(".chip.latched {"));
2477        assert!(css.contains(".chip:active {"));
2478        // The whole difference from a badge: it answers a click.
2479        assert!(Token::Chip { removable: false }.interactive());
2480        assert!(!Token::Badge.interactive());
2481    }
2482
2483    #[test]
2484    fn only_a_tab_comes_forward_when_chosen() {
2485        // The folder semantic. Collapsing the three selectors would lose it.
2486        let css = selector_rules(&Emit::default());
2487        assert!(css.contains(".tab.chosen {"));
2488        assert!(css.contains(".segment.chosen {"));
2489        assert!(css.contains(".toggle.chosen {"));
2490        assert_eq!(Selector::Tabs.chosen(), Depth::Raised);
2491        assert_eq!(Selector::Segmented.chosen(), Depth::Well);
2492        assert_eq!(Selector::Toggle.chosen(), Depth::Well);
2493
2494        let tab = css
2495            .lines()
2496            .skip_while(|l| !l.starts_with(".tab.chosen {"))
2497            .take_while(|l| !l.starts_with('}'))
2498            .collect::<Vec<_>>()
2499            .join("\n");
2500        assert!(
2501            tab.contains("var(--bevel-raised)"),
2502            "tab was held in: {tab}"
2503        );
2504    }
2505
2506    #[test]
2507    fn an_unchosen_tab_recedes_without_looking_picked() {
2508        let css = selector_rules(&Emit::default());
2509        // Recessed by colour and given no edge. An edge would make every option
2510        // look picked; flat would leave the chosen one nothing to come forward
2511        // from, which is the gap makeover-layout 0.3.0 closed.
2512        assert!(
2513            css.contains(".tab {\n    background: var(--surface-sunken);\n}"),
2514            "unchosen tab is not recessed: {css}"
2515        );
2516        assert_eq!(Selector::Tabs.unchosen(), Depth::Sunken);
2517        assert!(css.contains(".tab:hover {"));
2518    }
2519
2520    #[test]
2521    fn a_segment_stands_up_so_the_chosen_one_can_be_held_in() {
2522        // The inverse of the tab, and why the three selectors are not one
2523        // member with a flag.
2524        let css = selector_rules(&Emit::default());
2525        assert!(css.contains(".segment {\n    background: var(--surface-raised);"));
2526        assert_eq!(Selector::Segmented.unchosen(), Depth::Raised);
2527        assert_eq!(Selector::Segmented.chosen(), Depth::Well);
2528    }
2529
2530    #[test]
2531    fn a_rows_actions_are_shown_at_rest() {
2532        let css = row_rules(&Emit::default());
2533
2534        // The hover reveal is gone, and with it every escape it needed. What
2535        // it hid was hidden from pointer users alone, who are the ones
2536        // scanning a list to learn what can be done to a row.
2537        assert!(!css.contains("opacity"), "{css}");
2538        assert!(!css.contains("pointer-events"), "{css}");
2539        assert!(!css.contains(":hover"), "{css}");
2540        assert!(!css.contains(":focus-within"), "{css}");
2541
2542        // Nor is it hidden any other way. `display: none` would reflow the row
2543        // and `visibility: hidden` would take the actions out of the focus
2544        // order; the point is that neither is reached for.
2545        assert!(!css.contains("display: none"), "{css}");
2546        assert!(!css.contains("visibility:"), "{css}");
2547    }
2548
2549    #[test]
2550    fn a_figures_tone_lands_on_the_delta_when_there_is_one() {
2551        // 0.13.0. The delta is the part that reads as good or bad; the number
2552        // itself is an ordinary fact. A figure with no delta has nowhere else to
2553        // put the colour, so the value takes it, and `:has` is what lets one
2554        // attribute mean both without the emitter choosing an element.
2555        let css = stylesheet(&Emit::default());
2556
2557        assert!(
2558            css.contains(".figure[data-tone=\"success\"] > .figure-change"),
2559            "{css}"
2560        );
2561        assert!(
2562            css.contains(
2563                ".figure[data-tone=\"success\"]:not(:has(> .figure-change)) > .figure-value"
2564            ),
2565            "{css}"
2566        );
2567        // The caption is the noun and never takes the tone.
2568        assert!(
2569            !css.contains("[data-tone=\"success\"] > .figure-caption"),
2570            "{css}"
2571        );
2572    }
2573
2574    #[test]
2575    fn the_three_text_parts_take_their_intents_and_actions_inherits() {
2576        let css = row_rules(&Emit::default());
2577        assert!(css.contains(".row-primary {\n    color: var(--content);"));
2578        assert!(css.contains(".row-secondary {\n    color: var(--content-secondary);"));
2579        assert!(css.contains(".row-meta {\n    color: var(--content-muted);"));
2580        // Actions carry controls, not text. Pinning the colour it would inherit
2581        // anyway is louder than saying nothing.
2582        assert!(!css.contains(".row-actions {\n    color:"));
2583    }
2584
2585    #[test]
2586    fn the_token_strip_takes_no_colour_of_its_own() {
2587        // makeover-layout 0.9.0. A token carries its own tone, so a colour on
2588        // the strip would be a rule fighting the things sitting in it -- the
2589        // same reasoning as actions, reached for a different reason.
2590        let css = row_rules(&Emit::default());
2591        assert!(!css.contains(".row-tokens {\n    color:"));
2592    }
2593
2594    #[test]
2595    fn an_unknown_row_part_renders_plainly_rather_than_failing_to_build() {
2596        // What `#[non_exhaustive]` bought and what it cost. `part_class` can no
2597        // longer be exhaustive, so a member added upstream lands as a bare
2598        // class with no rule instead of stopping the build. Asserting the
2599        // fallback exists is what keeps it from being written as `unreachable!`
2600        // by someone who reads the match as closed.
2601        assert_eq!(part_class(RowPart::Tokens), "row-tokens");
2602        assert_eq!(part_class(RowPart::Meta), "row-meta");
2603    }
2604
2605    #[test]
2606    fn a_link_takes_the_action_colour_the_theme_actually_defines() {
2607        // `--action-primary` shipped here for months and no theme has ever
2608        // defined it, so every `.link` dropped its colour declaration outright
2609        // and fell back to inherited text. Nothing caught it because the sheet
2610        // is valid CSS either way; MNW's no-undefined-token lint is what found
2611        // it, 2026-08-14. The hover arm two lines below was always `--action-hover`,
2612        // which is what makes the typo legible in hindsight.
2613        let css = link_rules(&Emit::default());
2614        assert!(css.contains("color: var(--action);"));
2615        assert!(!css.contains("--action-primary"));
2616        assert!(css.contains("color: var(--action-hover);"));
2617    }
2618
2619    #[test]
2620    fn the_progress_trough_is_a_well() {
2621        let css = progress_rules(&Emit::default());
2622        assert!(css.contains(".progress {"));
2623        assert!(css.contains("box-shadow: var(--bevel-inset)"));
2624        assert!(css.contains(".progress > .progress-fill {"));
2625        assert!(css.contains("background: var(--action)"));
2626        // A bare `.fill` would catch things that have nothing to do with
2627        // progress once the sheet lands unprefixed.
2628        assert!(!css.contains("> .fill "));
2629    }
2630
2631    #[test]
2632    fn a_progress_bar_can_carry_a_tone_and_defaults_to_action() {
2633        let css = progress_rules(&Emit::default());
2634        // Untoned is --action, not Tone::Neutral's content-muted: a bar with no
2635        // status is still reporting progress, and muted would read as disabled.
2636        assert!(css.contains(".progress > .progress-fill {\n    background: var(--action);"));
2637        assert!(!css.contains("progress-fill {\n    color: var(--content-muted)"));
2638        for tone in ["info", "success", "warning", "danger"] {
2639            assert!(
2640                css.contains(&format!(".progress > .progress-fill[data-tone=\"{tone}\"]")),
2641                "missing progress tone {tone}"
2642            );
2643        }
2644        // goingson's two live cases, which is why the tones are emitted at all.
2645        assert!(css.contains("[data-tone=\"success\"] {\n    background: var(--success);"));
2646        assert!(css.contains("[data-tone=\"danger\"] {\n    background: var(--danger);"));
2647    }
2648
2649    #[test]
2650    fn no_scrollbar_track_is_emitted() {
2651        // Decision 3's negative half. It was on the phase A list and came off;
2652        // this is what stops it drifting back in.
2653        let css = stylesheet(&Emit::default());
2654        assert!(!css.contains("scrollbar"));
2655        assert!(!css.contains("::-webkit"));
2656    }
2657
2658    #[test]
2659    fn an_invalid_field_is_ringed_without_being_lit() {
2660        let css = surface_rules(&Emit::default());
2661        assert!(css.contains(".field {"));
2662        // The ARIA attribute, not a class: one fact, read by both the visual
2663        // and the accessible state, so they cannot drift.
2664        assert!(css.contains(".field[aria-invalid=\"true\"] {"));
2665        assert!(!css.contains(".field.invalid"));
2666        // A flat ring: this edge says "wrong", and a two-tone bevel would have
2667        // it say "raised" at the same time.
2668        assert!(css.contains("0 0 0 1px var(--danger)"));
2669    }
2670
2671    #[test]
2672    fn an_invalid_field_keeps_the_well_underneath_it() {
2673        // box-shadow is not additive. A lone ring replaces the bevel and drops
2674        // the well out from under the field, which is what this emitted before
2675        // 0.5.0 and is the whole reason the rule composes.
2676        let css = surface_rules(&Emit::default());
2677        let invalid = css
2678            .lines()
2679            .skip_while(|l| !l.starts_with(".field[aria-invalid"))
2680            .take_while(|l| !l.starts_with('}'))
2681            .collect::<Vec<_>>()
2682            .join("\n");
2683        assert!(
2684            invalid.contains("var(--bevel-inset)"),
2685            "the well was dropped: {invalid}"
2686        );
2687        assert!(invalid.contains("var(--danger)"));
2688    }
2689
2690    #[test]
2691    fn button_and_card_come_out_identical_by_construction() {
2692        // The duplication phase A deletes. They are the same composition, so
2693        // the only honest way to emit both is from one call.
2694        let opts = Emit::default();
2695        let css = surface_rules(&opts);
2696        assert_eq!(
2697            depth_declarations(Depth::Raised),
2698            depth_declarations(Depth::Raised)
2699        );
2700        assert!(css.contains(".button {"));
2701        assert!(css.contains(".card {"));
2702        assert_eq!(
2703            interactive_rules("button", Depth::Raised, &Emit::default()).replace("button", "card"),
2704            interactive_rules("card", Depth::Raised, &Emit::default())
2705        );
2706    }
2707
2708    #[test]
2709    fn a_prefix_reaches_the_component_classes_too() {
2710        let opts = Emit {
2711            class_prefix: "mo-",
2712            ..Emit::default()
2713        };
2714        let css = stylesheet(&opts);
2715        for name in [
2716            "mo-button",
2717            "mo-card",
2718            "mo-field",
2719            "mo-badge",
2720            "mo-chip",
2721            "mo-tab",
2722            "mo-row-primary",
2723            "mo-progress",
2724            "mo-progress-fill",
2725        ] {
2726            assert!(css.contains(&format!(".{name}")), "unprefixed: {name}");
2727        }
2728        // The bare names must be gone entirely, or a prefixed build still
2729        // collides with the app's own stylesheet.
2730        assert!(!css.contains(".button {"));
2731        assert!(!css.contains(".card {"));
2732        assert!(!css.contains(".badge {"));
2733    }
2734
2735    #[test]
2736    fn the_class_a_renderer_puts_on_an_option_is_the_one_the_rules_key_off() {
2737        // `option_class` is the contract a screen renderer writes markup
2738        // against, and the rules below are the other half of it. They come off
2739        // one mapping now, so this asserts the mapping is the one that reaches
2740        // the stylesheet rather than that two lists still agree.
2741        let css = stylesheet(&Emit::default());
2742        for selector in [Selector::Tabs, Selector::Segmented, Selector::Toggle] {
2743            let name = option_class(selector);
2744            assert!(css.contains(&format!(".{name} {{")), "{name}: {css}");
2745            assert!(css.contains(&format!(".{name}.chosen {{")), "{name}: {css}");
2746        }
2747        assert_eq!(option_class(Selector::Tabs), "tab");
2748    }
2749
2750    #[test]
2751    fn the_caret_brings_its_own_gap_and_is_the_glyph_the_description_names() {
2752        let css = stylesheet(&Emit::default());
2753
2754        // The space is inside the glyph, which is what the other two renderers
2755        // write. Emitted bare, every consumer has to add it back, and the
2756        // obvious way to add it -- `content` in an app stylesheet, which is
2757        // unlayered and so outranks this sheet -- deletes the caret instead.
2758        assert!(css.contains("content: \" \\25B2\";"), "{css}");
2759        assert!(css.contains("content: \" \\25BC\";"), "{css}");
2760        assert!(!css.contains("content: \"\\2"), "{css}");
2761
2762        // The arrows this renderer used to draw alone are gone. Composition
2763        // rather than agreement: the glyph comes from `Sort::glyph`, so a
2764        // fourth spelling cannot appear here without appearing everywhere.
2765        assert!(!css.contains("2191") && !css.contains("2193"), "{css}");
2766        assert!(css.contains(&css_escape(Sort::Ascending.glyph())), "{css}");
2767
2768        // Three states, three tones. An idle sortable heading draws its caret
2769        // now rather than reserving a hidden box for it, so there is no
2770        // visibility to order and no reflow left to guard against; what
2771        // separates the states is the colour, and the sorted arms come after
2772        // the idle one because the specificity is the same.
2773        let idle = css
2774            .find(".table-heading[data-sortable]::after")
2775            .expect("the idle caret is emitted");
2776        let sorted = css
2777            .find(".table-heading[aria-sort=\"ascending\"]::after")
2778            .expect("the ascending caret is emitted");
2779        assert!(idle < sorted, "{css}");
2780        assert!(
2781            css[idle..sorted].contains("color: var(--content-secondary);"),
2782            "{css}"
2783        );
2784        assert!(css[sorted..].contains("color: var(--content);"), "{css}");
2785        assert!(!css.contains("visibility: hidden;"), "{css}");
2786    }
2787
2788    #[test]
2789    fn a_destructive_button_has_somewhere_for_its_tone_to_land() {
2790        let css = component_rules(&Emit::default());
2791
2792        for tone in [Tone::Info, Tone::Success, Tone::Warning, Tone::Danger] {
2793            assert!(
2794                css.contains(&format!(".button[data-tone=\"{}\"]", tone.token())),
2795                "{css}"
2796            );
2797        }
2798        // Colour, not a fill. A red surface is an app's decision about emphasis.
2799        assert!(!css.contains(".button[data-tone=\"danger\"] {\n    background"));
2800    }
2801
2802    #[test]
2803    fn the_table_model_is_scoped_to_the_table_but_the_caret_is_not() {
2804        // goingson's task headings are `.table-heading` inside a CSS grid.
2805        // They want the sort caret and the sortable cursor; they do not want
2806        // `display: table-cell`, which a grid item blockifies away anyway.
2807        // Scoping the one and not the other is what separates them.
2808        let css = component_rules(&Emit::default());
2809
2810        assert!(
2811            css.contains(".table .table-heading,\n.table .cell {\n    display: table-cell;"),
2812            "{css}"
2813        );
2814        assert!(
2815            !css.contains(".table-heading,\n.cell {\n    display: table-cell;"),
2816            "the table model is still unscoped: {css}"
2817        );
2818
2819        let states = state_rules(&Emit::default());
2820        assert!(
2821            states.contains(".table-heading[aria-sort"),
2822            "the caret got scoped along with the model: {states}"
2823        );
2824    }
2825
2826    #[test]
2827    fn a_figure_value_is_the_thing_itself_and_a_badge_is_quiet() {
2828        // Both used to read `Tone::Neutral.token()`, which answered
2829        // `content-muted`, so the headline number sat at the colour of its own
2830        // caption. Neutral answers `content` now; each site states its own
2831        // claim rather than borrowing one from the status axis.
2832        let css = component_rules(&Emit::default());
2833
2834        assert!(
2835            css.contains(".figure > .figure-value {\n    color: var(--content);"),
2836            "{css}"
2837        );
2838        assert!(
2839            css.contains(".figure > .figure-caption {\n    color: var(--content-muted);"),
2840            "{css}"
2841        );
2842        assert!(
2843            css.contains(".badge {\n    color: var(--content-muted);"),
2844            "{css}"
2845        );
2846    }
2847
2848    #[test]
2849    fn a_described_list_is_not_a_bulleted_list() {
2850        let css = component_rules(&Emit::default());
2851        assert!(css.contains(".list {\n    list-style: none;"), "{css}");
2852    }
2853
2854    #[test]
2855    fn a_table_lays_itself_out_without_being_told_its_columns() {
2856        // The whole point of the CSS table. A described table's columns are
2857        // known at render time, so anything the stylesheet has to be told about
2858        // them would have to travel with the markup.
2859        let css = component_rules(&Emit::default());
2860
2861        assert!(css.contains(".table {\n    display: table;"), "{css}");
2862        assert!(css.contains("display: table-row;"), "{css}");
2863        assert!(css.contains("display: table-cell;"), "{css}");
2864        assert!(!css.contains("grid-template-columns"), "{css}");
2865    }
2866
2867    #[test]
2868    fn a_control_in_a_cell_is_not_painted_as_text() {
2869        // The point of makeover-layout 0.14.0's CellPart, and the table-side
2870        // twin of `the_three_text_parts_take_their_intents_and_actions_inherits`
2871        // above. One `.cell` and one content colour meant a button in a cell
2872        // inherited it.
2873        let css = table_rules(&Emit::default());
2874
2875        assert!(
2876            css.contains(".cell-value {\n    color: var(--content);"),
2877            "{css}"
2878        );
2879        assert!(!css.contains(".cell-actions {\n    color:"), "{css}");
2880        assert!(!css.contains(".cell-tokens {\n    color:"), "{css}");
2881        assert!(!css.contains(".cell-link {\n    color:"), "{css}");
2882
2883        // The colour is on the part that is text, never on the container. On
2884        // `.cell` it would cascade into the three parts that are not text,
2885        // which is the bug written as one rule.
2886        assert!(!css.contains(".cell {\n    color:"), "{css}");
2887    }
2888
2889    #[test]
2890    fn an_unknown_cell_part_renders_plainly_rather_than_failing_to_build() {
2891        // `part_class`'s obligation, taken on for the table side too. CellPart
2892        // is `#[non_exhaustive]`, so a member added upstream must land as a
2893        // bare class rather than as a build that stops.
2894        assert_eq!(cell_part_class(CellPart::Value), "cell-value");
2895        assert_eq!(cell_part_class(CellPart::Actions), "cell-actions");
2896    }
2897
2898    #[test]
2899    fn a_column_drops_by_its_priority_and_never_by_its_position() {
2900        let css = component_rules(&Emit::default());
2901
2902        // Optional goes at the narrowest class and secondary goes with it,
2903        // which is `kept_at`'s cutoff walk said as two queries.
2904        let compact = css
2905            .find(&format!("@media {}", SizeClass::Compact.media_condition()))
2906            .expect("a compact query");
2907        let medium = css
2908            .find(&format!("@media {}", SizeClass::Medium.media_condition()))
2909            .expect("a medium query");
2910        assert!(css[compact..].contains(".cell-drops-next"), "{css}");
2911        assert!(!css[medium..].contains(".cell-drops-next"), "{css}");
2912
2913        // Essential columns are never mentioned, because not being mentioned is
2914        // already what never dropping means.
2915        assert!(!css.contains(".cell-keeps"), "{css}");
2916
2917        // And nothing counts. `nth-child` is the bug the priority vocabulary
2918        // exists to end.
2919        assert!(!css.contains("nth-child"), "{css}");
2920    }
2921
2922    #[test]
2923    fn a_track_places_by_custom_property_and_never_by_a_size() {
2924        let css = track_rules(&Emit::default());
2925
2926        // Placement arrives from the caller, computed once by Track::fraction.
2927        // If either of these becomes a literal, three renderers have started
2928        // disagreeing about where 09:30 is.
2929        assert!(css.contains("top: var(--track-at"), "{css}");
2930        assert!(css.contains("height: var(--track-for"), "{css}");
2931
2932        // Overlap lanes default so an entry naming neither is full width.
2933        assert!(css.contains("--track-lane, 0"), "{css}");
2934        assert!(css.contains("--track-lanes, 1"), "{css}");
2935
2936        // The refusal that matters. A slot height here would be this crate
2937        // deciding how tall a quarter of an hour is, which is the thing
2938        // makeover-geometry owns and the reason `.track` gets no height at all.
2939        assert!(!css.contains("height: var(--track-slot"), "{css}");
2940        for size in ["px", "rem", "em", "vh"] {
2941            let bare = css
2942                .lines()
2943                .filter(|l| !l.contains("var(--"))
2944                .any(|l| l.contains(size));
2945            assert!(!bare, "track_rules named a {size} outside a var(): {css}");
2946        }
2947    }
2948
2949    #[test]
2950    fn a_relaxed_part_clamps_and_a_tight_one_says_nothing() {
2951        let css = stylesheet(&Emit::default());
2952        assert!(css.contains(".row-relaxed {"));
2953        assert!(css.contains("-webkit-line-clamp: 2;"));
2954        // The count is `Flow`'s, not this crate's. If the tier ever means three
2955        // lines, this fails here rather than in an app.
2956        assert!(css.contains(&format!("line-clamp: {};", Flow::Relaxed.lines())));
2957        // Tight gets no rule at all: one line is what a run already does, and a
2958        // class per part saying so is a declaration that changes nothing.
2959        assert!(!css.contains("row-tight"));
2960        assert_eq!(crate::list::flow_class(Flow::Relaxed), Some("row-relaxed"));
2961        assert_eq!(crate::list::flow_class(Flow::Tight), None);
2962        // Emitted, therefore checkable: an app's dead-vocabulary seal and the
2963        // overlap check both read `vocabulary::names`, so a class the renderer
2964        // can write and that list does not carry is invisible to both.
2965        assert!(crate::vocabulary::names(&Emit::default()).contains("row-relaxed"));
2966    }
2967
2968    #[test]
2969    fn the_two_kinds_of_wait_stop_rendering_identically() {
2970        // The state `d43ea1c5` fixes: the emitter had been writing
2971        // `data-awaiting` for months and nothing styled either value, so a
2972        // measured wait and an unmeasured one drew the same nothing.
2973        let css = stylesheet(&Emit::default());
2974        assert!(css.contains("[data-awaiting]::after"), "{css}");
2975        assert!(
2976            css.contains("[data-awaiting][aria-busy=\"true\"]::after"),
2977            "the mark is drawn only while something is actually waiting"
2978        );
2979        assert!(
2980            css.contains("[data-awaiting=\"determinate\"][aria-busy=\"true\"]::after"),
2981            "the measured half is its own drawing"
2982        );
2983    }
2984
2985    #[test]
2986    fn the_blink_takes_its_cadence_and_never_names_one() {
2987        // Three renderers draw this mark. A number written here would be a
2988        // second heartbeat for one wait.
2989        let css = stylesheet(&Emit::default());
2990        assert!(
2991            css.contains("calc(var(--cadence-activity) * 2)"),
2992            "a half-period doubled, not a literal"
2993        );
2994        assert!(!css.contains("500ms"), "{css}");
2995    }
2996
2997    #[test]
2998    fn a_bar_nobody_is_counting_is_empty_rather_than_full() {
2999        // Rule 1. The share defaults to zero, so a determinate control with no
3000        // binder watching bytes draws a trough. A default of 1 would be the
3001        // confidently-wrong drawing the rule exists to forbid.
3002        let css = stylesheet(&Emit::default());
3003        assert!(css.contains("var(--awaiting-share, 0)"), "{css}");
3004    }
3005
3006    #[test]
3007    fn motion_off_leaves_the_mark_lit_because_the_keyframes_do_the_dimming() {
3008        // `makeover_timing::reduced_motion_css` sets `--cadence-activity: 0ms`,
3009        // and a zero-length animation leaves the element in its base style
3010        // rather than at its last keyframe. So the base has to be the lit one.
3011        // The inverted spelling would blank the mark for a reader who asked for
3012        // less motion, which answers a request nobody made.
3013        let css = stylesheet(&Emit::default());
3014        let busy = css
3015            .split("[data-awaiting][aria-busy=\"true\"]::after {")
3016            .nth(1)
3017            .expect("the busy rule");
3018        let busy = busy.split('}').next().expect("its body");
3019        assert!(
3020            busy.contains("background: var(--action);"),
3021            "the base state is lit: {busy}"
3022        );
3023    }
3024
3025    #[test]
3026    fn the_whole_sheet_still_names_every_colour() {
3027        // The crate's founding property, asserted over the component layer and
3028        // not only the primitives.
3029        let css = stylesheet(&Emit::default());
3030        assert!(!css.contains('#'));
3031        assert!(!css.contains("rgb"));
3032        for line in css.lines() {
3033            // Declarations only: a selector or an at-rule can carry a colon of
3034            // its own (`:root`, `:hover`, `@media (hover: hover)`) and declares
3035            // nothing. Keyed on the trailing semicolon rather than on leading
3036            // indentation, which only ever worked as a proxy for nesting depth
3037            // and stopped when the sheet gained a cascade layer around it.
3038            let trimmed = line.trim();
3039            if !trimmed.ends_with(';') {
3040                continue;
3041            }
3042            let Some((_, value)) = trimmed.split_once(": ") else {
3043                continue;
3044            };
3045            if value.contains("var(--") {
3046                continue;
3047            }
3048            // Everything left has to be a keyword, a number or a
3049            // caller-supplied length, never a colour.
3050            //
3051            // The length arm is what the comment above always claimed and the
3052            // list never covered: `border_width` arrives from `Emit` and lands
3053            // bare in the focus ring's offset, where the bevel had only ever
3054            // used it inside an `inset` shadow.
3055            let opts = Emit::default();
3056            assert!(
3057                value.contains("inset")
3058                    || value.contains(opts.border_width)
3059                    || value.contains(opts.focus_width)
3060                    // 0.12.0's two: a sortable header is a control and says so
3061                    // with the pointer, and the caret is this renderer's own
3062                    // expression of `aria-sort`. Neither is a colour, which is
3063                    // what this test is actually about, and neither is a size,
3064                    // which is the other thing this crate must not name. The
3065                    // leading space inside the glyph is the same thing the other
3066                    // two renderers write into theirs, so it is part of the
3067                    // caret rather than spacing this crate decided on.
3068                    || value
3069                        .trim_start_matches('"')
3070                        .trim_start()
3071                        .starts_with("\\2")
3072                    || matches!(
3073                        value.trim_end_matches(';'),
3074                        "0" | "1"
3075                            // The wait's mark and bar, 0.60.0. An empty
3076                            // `content` is what brings a pseudo-element into
3077                            // existence with nothing in it, and `step-end`
3078                            // is an easing: the blink is two states, not a
3079                            // slide between them. Neither is a colour, and
3080                            // neither is a magnitude.
3081                            | "\"\""
3082                            // Where the mark sits on the line it joins.
3083                            // Alignment is structure, the same way `display`
3084                            // is, and `baseline` is the initial value said out
3085                            // loud so a host stylesheet cannot leave it
3086                            // wherever an earlier rule put it.
3087                            | "baseline"
3088                            | "inline-block"
3089                            | "none"
3090                            | "auto"
3091                            | "not-allowed"
3092                            | "pointer"
3093                            // The caret's reserved box. Visibility is presence,
3094                            // not magnitude and not colour.
3095                            | "hidden"
3096                            | "visible"
3097                            // The link's two signals. `underline` is a line and
3098                            // `inherit` defers to whatever the app set, so
3099                            // neither names a colour or a magnitude.
3100                            | "underline"
3101                            | "inherit"
3102                            // The table frame. `display` is structure and not a
3103                            // size; `nowrap` is what makes a content column
3104                            // content. The two widths are the awkward pair and
3105                            // they are still not sizes: `100%` is "all of
3106                            // whatever you were given" and `1%` is the CSS
3107                            // table idiom for "shrink to fit", which is a
3108                            // behaviour spelled as a number because CSS has no
3109                            // keyword for it. Neither names a magnitude, which
3110                            // is the thing this crate leaves to
3111                            // makeover-geometry.
3112                            | "table"
3113                            | "table-row"
3114                            | "table-cell"
3115                            | "nowrap"
3116                            | "100%"
3117                            | "1%"
3118                            // The time axis, 0.42.0. `position` is the one
3119                            // property whose whole job is where a thing sits,
3120                            // which is exactly what this crate spent its life
3121                            // refusing to say -- so it is worth being exact
3122                            // about why these two are not that refusal
3123                            // breaking.
3124                            //
3125                            // Neither names a magnitude. `relative` says the
3126                            // track is what its entries resolve against, and
3127                            // `absolute` says an entry is placed rather than
3128                            // flowed. *Where* each entry lands is
3129                            // `--track-at` and `--track-for`, custom
3130                            // properties the caller sets from
3131                            // `Track::fraction`, and they are skipped by the
3132                            // `var(--` arm above like every other value this
3133                            // crate refuses to decide.
3134                            //
3135                            // The rule that would break the refusal is a slot
3136                            // height, and there is none: the track's height is
3137                            // the app's, so the percentages have something to
3138                            // resolve against and this crate still never says
3139                            // how tall a day is.
3140                            | "relative"
3141                            | "absolute"
3142                            // A run's five, 0.49.0. `flex`, `wrap` and
3143                            // `center` are structure and alignment, the same
3144                            // reading `table` gets: which way members are laid
3145                            // out and how they line up, never how much of
3146                            // anything.
3147                            //
3148                            // The two intrinsic keywords are the interesting
3149                            // pair and they are the opposite of a size. A
3150                            // magnitude is a number somebody chose;
3151                            // `min-content` and `max-content` are the browser
3152                            // being asked what the members themselves come to,
3153                            // which is the derived minimum the room ruling
3154                            // requires and the reason no breakpoint appears
3155                            // anywhere in these rules. `1 1 max-content` is
3156                            // grow, shrink and that basis, so its two digits
3157                            // are ratios rather than lengths.
3158                            | "flex"
3159                            | "wrap"
3160                            | "center"
3161                            | "min-content"
3162                            | "1 1 max-content"
3163                            // A menu run's overflow control, 0.64.0.
3164                            // `column` and `stretch` are the same reading
3165                            // `flex` and `center` get one line up: which way
3166                            // the shed members stack inside the panel and how
3167                            // they line up across it. Neither is a magnitude.
3168                            //
3169                            // `100%` is already above and reused here as the
3170                            // panel's `inset-block-start`, which is "the whole
3171                            // of the control it hangs from" rather than a
3172                            // distance anybody picked.
3173                            | "column"
3174                            | "stretch"
3175                            // A picture's three, 0.36.0. `block` is structure
3176                            // for the reason `table` is: an inline image sits
3177                            // on the baseline and carries a descender's worth
3178                            // of space under it, which is a fact about
3179                            // replaced elements rather than a size this crate
3180                            // chose. `cover` and `contain` are `Fit`'s two
3181                            // named members reaching CSS unchanged, which is
3182                            // an intent arriving rather than a value being
3183                            // picked.
3184                            | "block"
3185                            | "cover"
3186                            | "contain"
3187                            // A relaxed part's three, and the third is the
3188                            // awkward one. `-webkit-box` and `vertical` are
3189                            // structure: they say the part is a box of lines
3190                            // stacked downward, which is the only way CSS lets
3191                            // anyone ask for a clamp at all.
3192                            //
3193                            // `2` is a count of lines, not a length. The
3194                            // distinction this crate holds is between naming a
3195                            // magnitude -- a padding, a height, a font size,
3196                            // all of which belong to makeover-geometry -- and
3197                            // naming how many of something there are. A line's
3198                            // height is still the app's, so two lines is
3199                            // whatever two of the app's lines come to, and
3200                            // nothing here decides how tall that is. It is also
3201                            // not a value picked here: it is `Flow::Relaxed`'s
3202                            // own answer arriving unchanged, the same way
3203                            // `cover` and `contain` are `Fit`'s.
3204                            | "-webkit-box"
3205                            | "vertical"
3206                            | "2"
3207                    ),
3208                "unrecognised literal value: {line}"
3209            );
3210        }
3211    }
3212
3213    #[test]
3214    fn flat_emits_nothing_at_all() {
3215        assert_eq!(depth_class(Depth::Flat, &Emit::default()), None);
3216        assert!(!depth_rules(&Emit::default()).contains("flat"));
3217    }
3218
3219    #[test]
3220    fn a_prefix_namespaces_every_class() {
3221        let opts = Emit {
3222            class_prefix: "mo-",
3223            ..Emit::default()
3224        };
3225        let css = depth_rules(&opts);
3226        assert!(css.contains(".mo-raised {"));
3227        assert!(css.contains(".mo-well {"));
3228        assert!(!css.contains(".raised {"));
3229    }
3230
3231    #[test]
3232    fn the_border_width_is_the_callers() {
3233        let opts = Emit {
3234            border_width: "2px",
3235            ..Emit::default()
3236        };
3237        assert!(bevel_shadow(Bevel::Raised, &opts).contains("inset 2px 2px 0"));
3238    }
3239
3240    #[test]
3241    fn edges_agree_with_the_description() {
3242        // Not a tautology: it is the guard that a CSS-shaped convenience never
3243        // quietly reverses which side is lit.
3244        let (tl, br) = Bevel::Raised.edges();
3245        assert_eq!(tl.token(), Edge::Light.token());
3246        assert_eq!(br.token(), Edge::Dark.token());
3247    }
3248}