uzor-text 1.5.0

Pretext-pattern text layout engine for uzor — greedy word-wrap, rich multi-run spans, InlineBox, and a mixed-run baseline pass over the cosmic-text shaper (Arc 2 Phase 1+2; kinetics/ascii/hyphenation land in later phases).
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197

//! Typography track T1 (2026-07-25): tabular-figures (`tnum`) measurement.
//!
//! **Test-only, zero production surface** — this module exists purely to
//! answer, empirically, a question this crate's own `CLAUDE.md`
//! (Typography-gap WAVE 2 §5) had already logged but never actually
//! checked: *"cosmic-text 0.12.1's public API has no seam to inject `tnum`
//! — but do the fonts this workspace actually SHIPS even need it?"* Per
//! this task's own brief ("measure before building... if already equal,
//! say so plainly and do not build anything; if they differ, report by how
//! much and stop there"), the answer below is a measurement, not new
//! machinery.
//!
//! Measures the raw `hmtx` advance width of every ASCII digit glyph
//! (`'0'..='9'`) directly from each embedded font FILE via [`ttf_parser`] —
//! deliberately bypassing `uzor::shaper`/cosmic-text's `FontFamily`
//! selection entirely, since `uzor::shaper::parse_css_font`'s own
//! `resolve_family` only recognizes 3 logical families (Roboto/PT Root
//! UI/JetBrains Mono) and has no seam to select a specific STYLE FILE
//! (Roboto Bold vs Italic vs BoldItalic) or DejaVu Sans at all (see
//! `uzor/src/ui/assets/fonts/fonts.rs`'s own `resolve_family` — "dejavu" is
//! not a recognized keyword, so a CSS family string naming it would
//! silently resolve to Roboto instead). Reading `hmtx` directly per FILE
//! sidesteps that gap and is also the more DIRECT answer to "does this
//! font's own default figures have equal advances" — the digit advance
//! recorded in `hmtx` is a per-glyph fact of the font file itself; a plain
//! per-codepoint shape (no ligature, since digits never ligate with each
//! other) never differs from it unless a `tnum`/`pnum`-class GSUB
//! substitution is explicitly requested (the exact feature cosmic-text
//! 0.12.1 has no seam to request), so reading it directly (once, per file)
//! answers the *default* (`tnum`-OFF) question exactly, at every render
//! size (an advance is a font-unit value scaled linearly by `size_px /
//! units_per_em` — the equal/unequal verdict is scale-invariant; measuring
//! it at 16px AND 32px below is a direct demonstration of that, not a
//! second independent fact).
//!
//! ## Verdict (measured, NOT assumed — the numbers are the point)
//!
//! **Mixed, and genuinely mixed — not a clean "non-issue across the
//! board."** Roboto (all 4 style files), DejaVu Sans, and JetBrains Mono
//! (the monospace control — tabular by definition, every glyph shares one
//! advance) all already have EQUAL-width default digits — `tnum` is a
//! real NON-ISSUE for any document set in one of those. **PT Root UI (the
//! variable font, default instance) does NOT** — its digit `'1'` measures
//! `350` font units against `550`-`610` for the others (upm 1000), a
//! ~30-43% narrower advance, a real ~4px gap at 16px / ~8px at 32px (see
//! `digit_width_audit_table`'s printed numbers for the exact per-digit
//! set). This is NOT a "user-supplied proportional-figure font" case —
//! PT Root UI is one of only 3 [`uzor::fonts::FontFamily`] variants
//! `FontSpec`/`uzor::shaper` can select at all, so any first-party document
//! that chooses PT Root UI for numeric/tabular content (financial tables,
//! KPI figures) inherits this gap TODAY. Per this task's own instruction
//! ("if they differ, report by how much and stop there — the fix is a
//! separate decision, not this task"), the number is reported and closed
//! here; no `tnum` plumbing is built by this pass.

#[cfg(test)]
mod tests {
    use ttf_parser::Face;

    /// Every embedded font FILE this task named — `(label, bytes)`. Exactly
    /// the 8 files the task brief enumerates (DejaVu Sans; Roboto Regular/
    /// Bold/Italic/BoldItalic; JetBrains Mono Regular/Bold; PT Root UI VF).
    fn embedded_fonts() -> [(&'static str, &'static [u8]); 8] {
        [
            ("DejaVuSans", uzor_fonts::DEJAVU_SANS),
            ("Roboto-Regular", uzor_fonts::ROBOTO_REGULAR),
            ("Roboto-Bold", uzor_fonts::ROBOTO_BOLD),
            ("Roboto-Italic", uzor_fonts::ROBOTO_ITALIC),
            ("Roboto-BoldItalic", uzor_fonts::ROBOTO_BOLD_ITALIC),
            ("JetBrainsMono-Regular", uzor_fonts::JETBRAINS_MONO_REGULAR),
            ("JetBrainsMono-Bold", uzor_fonts::JETBRAINS_MONO_BOLD),
            ("PTRootUI-VF", uzor_fonts::PT_ROOT_UI_VF),
        ]
    }

    /// `(raw hmtx advance per digit '0'..='9', units_per_em)` for one font
    /// file — default instance (no variation-axis coordinates applied),
    /// no shaping features involved (a plain `cmap` codepoint lookup +
    /// `hmtx` read).
    fn digit_advances(bytes: &[u8]) -> (Vec<u16>, u16) {
        let face = Face::parse(bytes, 0).expect("embedded font must parse");
        let upm = face.units_per_em();
        let advances: Vec<u16> = ('0'..='9')
            .map(|c| {
                let gid = face.glyph_index(c).unwrap_or_else(|| panic!("font must contain digit '{c}'"));
                face.glyph_hor_advance(gid).unwrap_or_else(|| panic!("digit '{c}' must have a horizontal advance"))
            })
            .collect();
        (advances, upm)
    }

    fn all_equal(advances: &[u16]) -> bool {
        advances.windows(2).all(|w| w[0] == w[1])
    }

    /// Scaled px advance at `size_px` — linear from the raw font-unit
    /// value, matching how every shaper in this workspace scales `hmtx`
    /// (no hinting-driven per-size rounding at this measurement layer).
    fn to_px(units: u16, upm: u16, size_px: f64) -> f64 {
        units as f64 * size_px / upm as f64
    }

    /// T1 measurement table: every embedded font's raw digit advances
    /// (font units) + scaled px advances at 16px and 32px, one line per
    /// font. Run with `--nocapture` to reproduce the table this task's own
    /// report/handoff quotes verbatim.
    #[test]
    fn digit_width_audit_table() {
        for (label, bytes) in embedded_fonts() {
            let (advances, upm) = digit_advances(bytes);
            let tabular = all_equal(&advances);
            let px16: Vec<f64> = advances.iter().map(|&u| (to_px(u, upm, 16.0) * 100.0).round() / 100.0).collect();
            let px32: Vec<f64> = advances.iter().map(|&u| (to_px(u, upm, 32.0) * 100.0).round() / 100.0).collect();
            println!("{label:22} upm={upm:5} tabular={tabular:5} units(0-9)={advances:?} px16={px16:?} px32={px32:?}");
        }
    }

    /// Control: JetBrains Mono is monospace, so EVERY glyph (not just
    /// digits) shares one advance — tabular by definition, both bundled
    /// weights. The ground truth this module's own methodology is checked
    /// against.
    #[test]
    fn jetbrains_mono_digits_are_tabular_by_definition_both_weights() {
        let (regular, _) = digit_advances(uzor_fonts::JETBRAINS_MONO_REGULAR);
        let (bold, _) = digit_advances(uzor_fonts::JETBRAINS_MONO_BOLD);
        assert!(all_equal(&regular), "JetBrains Mono Regular digits must be tabular (monospace)");
        assert!(all_equal(&bold), "JetBrains Mono Bold digits must be tabular (monospace)");
    }

    /// T1 verdict (non-issue half), locked in as a regression guard: Roboto
    /// (all 4 style files) and DejaVu Sans already have equal-width default
    /// digits — measured directly, never assumed. A future font swap that
    /// silently makes this false for any of them fails this test, forcing
    /// an explicit acknowledgment rather than a silent tabular-alignment
    /// regression in a first-party document that relies on it.
    #[test]
    fn roboto_and_dejavu_sans_already_have_tabular_default_digits() {
        let already_tabular = [
            ("DejaVuSans", uzor_fonts::DEJAVU_SANS),
            ("Roboto-Regular", uzor_fonts::ROBOTO_REGULAR),
            ("Roboto-Bold", uzor_fonts::ROBOTO_BOLD),
            ("Roboto-Italic", uzor_fonts::ROBOTO_ITALIC),
            ("Roboto-BoldItalic", uzor_fonts::ROBOTO_BOLD_ITALIC),
        ];
        for (label, bytes) in already_tabular {
            let (advances, _) = digit_advances(bytes);
            assert!(all_equal(&advances), "{label}: expected already-tabular default digits (T1 measurement), got {advances:?}");
        }
    }

    /// T1 verdict (real-gap half), locked in as a regression guard the
    /// OTHER direction: PT Root UI's variable-font default instance does
    /// NOT have equal-width digits — `'1'` (index 1) is measurably the
    /// narrowest, a genuine ~30%+ advance spread against the widest digit,
    /// not a rounding artifact. PT Root UI is one of only 3
    /// [`uzor::fonts::FontFamily`] variants this workspace's own shaper can
    /// select — this is a real, currently-live gap for first-party content
    /// choosing that family for numeric text, reported here per this
    /// task's own "report by how much and stop there" instruction (no
    /// `tnum` plumbing is built by this pass).
    #[test]
    fn pt_root_ui_vf_default_digits_are_not_tabular_a_real_measured_gap() {
        let (advances, upm) = digit_advances(uzor_fonts::PT_ROOT_UI_VF);
        assert!(!all_equal(&advances), "PT Root UI VF: expected a NON-tabular default (T1 measurement), got uniformly {advances:?}");

        let widest = *advances.iter().max().expect("10 digits");
        let narrowest = *advances.iter().min().expect("10 digits");
        assert_eq!(advances[1], narrowest, "digit '1' must be the narrowest — the classic proportional-figures tell, got advances={advances:?}");
        let spread_fraction = (widest - narrowest) as f64 / widest as f64;
        assert!(
            spread_fraction > 0.25,
            "the widest/narrowest digit spread must be a REAL gap (>25% of the widest digit's own advance), got {:.1}% (upm={upm})",
            spread_fraction * 100.0
        );
    }

    /// The size-invariance claim this module's own doc comment makes: the
    /// tabular/non-tabular VERDICT for a font must be identical whether
    /// read in raw font units or scaled to any render size (16px/32px) —
    /// proves measuring once in font units and reasoning about it at every
    /// size (rather than re-measuring per size) is sound, not an assumption.
    /// Exercised against BOTH the tabular fonts and the one non-tabular
    /// font (PT Root UI) — the invariance must hold either way.
    #[test]
    fn tabular_verdict_is_identical_across_render_sizes() {
        for (label, bytes) in embedded_fonts() {
            let (advances, upm) = digit_advances(bytes);
            let units_verdict = all_equal(&advances);
            let px16: Vec<f64> = advances.iter().map(|&u| to_px(u, upm, 16.0)).collect();
            let px32: Vec<f64> = advances.iter().map(|&u| to_px(u, upm, 32.0)).collect();
            let px16_verdict = px16.windows(2).all(|w| (w[0] - w[1]).abs() < 1e-9);
            let px32_verdict = px32.windows(2).all(|w| (w[0] - w[1]).abs() < 1e-9);
            assert_eq!(units_verdict, px16_verdict, "{label}: 16px verdict must match the font-unit verdict");
            assert_eq!(units_verdict, px32_verdict, "{label}: 32px verdict must match the font-unit verdict");
        }
    }
}