Files
vectorcraft/crates/pdf/src/import_text.rs
T
Taha-Mahmoodi b6b1ec7300 PDF import: keep a line's glyphs on course when the file's widths are a little off the font's advances
Cause: since #3742 place_each measures a line's spacing from the file's
own widths, so tracking is only the file's Tc/Tw/TJ. Where the widths
are systematically off the installed font's advances (up to the 1% that
still counts as the same font), those differences add up along a line
and nothing made up for them. With no kerning and every tracking value
0, place_each returned the untracked layout early, without the 0.02 em
placement check (CodeRabbit's review of #3742): a line of embedded
Source Sans 3 with /Widths 0.8% wide opened 3.7 pt off. With the early
return removed the check rejects the line and the line-wide fit takes
over: one whole-thousandth tracking value with the font's kerning,
0.30 pt off (as before #3715). Before #3742 main placed it exactly but
with a run per character (49 and 62 runs).

Change: the early return is gone; every line goes through the check.
Before it, on_course() keeps the file's spacing where the glyphs stay
within 0.01 em of the file's; otherwise each stretch of the line takes
one share after every character (the longest stretch one share keeps
within 0.01 em), so ordinary text stays one run, and a line whose
differences change course (30 m then 30 i) takes a run per stretch (2).

Verification: new test
a_long_line_of_widths_a_little_off_the_fonts_advances_lands_in_one_run
(embedded Source Sans 3, widths x1.008 and x0.992, a kerned and an
unkerned line): one run each, glyphs under 0.24 pt (0.02 em) from where
the file draws them. Without on_course it fails after #3742, with and
without the early return ("glyphs 0.300 pt"); before #3742 it fails on
the run count. It replaces an Inter test drafted for the review: Inter
isn't embedded there, so the file's outlines are the interpreter's
stand-in stretched to the widths, not Inter, and the contours differ by
0.6 pt even with exact widths. tests_text_placement and
text_becomes_point_type pass; cargo xtask ci: all 10 steps passed.

Thanks to the CodeRabbit review on #3742 for the early-return finding.
2026-10-23 10:42:45 +07:00

1001 lines
47 KiB
Rust

//! Text as editable point type: glyphs drawn one after another on the same baseline make one
//! point text object of their Unicode text, a run per font, size and paint (a gap wider than a
//! fifth of the smaller size beside it reads as a space; a glyph the file raised or lowered off the
//! baseline, text rise, is baseline shift). Each glyph stays where the file drew it: the kerning
//! and tracking the type takes put it there (#722). Fonts are named from the file's base font name (its
//! subset prefix dropped, the family matched against the fonts available); a font that isn't
//! available keeps its name, and the text shows in the fallback font until it is.
//!
//! Glyphs each turned a little further along a curve (type set on a path: apps write each glyph
//! with its own placement) make one type-on-a-path object, its path through the glyphs' baseline.
use std::collections::{HashMap, HashSet};
use kurbo::{Affine, BezPath, Point, Rect, Vec2};
use vectorcraft_color::Paint;
use vectorcraft_doc::{CharStyle, Node, ParaDirection, TextKind, TextObject, TextRun};
use vectorcraft_text::TextLayout;
/// One glyph's placement: its baseline origin, advance direction, size and horizontal scale.
#[derive(Clone, Copy, Debug)]
pub(crate) struct Placement {
pub origin: Point,
/// Unit vector along the baseline.
pub dir: Vec2,
/// Font size in points (the em's height).
pub size: f64,
/// Width of the em ÷ its height, in percent.
pub h_scale: f64,
/// How far the em leans along the baseline per unit of its height (a text matrix with a
/// shear: faux italic, or type slanted as a whole); 0 when it stands square on the baseline.
pub slant: f64,
}
impl Placement {
/// Where a glyph drawn with `m` (glyph space, 1000 units per em, y up → document) sits;
/// `None` when it is mirrored, degenerate or not finite (those keep their outlines).
pub fn of(m: Affine) -> Option<Self> {
let origin = m * Point::ORIGIN;
let ex = m * Point::new(1000.0, 0.0) - origin;
let ey = m * Point::new(0.0, 1000.0) - origin;
let w = ex.hypot();
// Upright in y-down document space: x to the right of up.
let upright = ex.cross(ey) < 0.0;
if !(w.is_finite() && w > 0.01) {
return None;
}
let dir = ex / w;
// The em's height is across the baseline; what's left of `ey` along it is a slant.
let up = Vec2::new(dir.y, -dir.x);
let size = ey.dot(up);
let slant = ey.dot(dir) / size;
(origin.is_finite() && size.is_finite() && slant.is_finite() && size > 0.01 && size < 1e5 && upright).then(|| Self {
origin,
dir,
size,
h_scale: w / size * 100.0,
slant,
})
}
fn up(&self) -> Vec2 {
Vec2::new(self.dir.y, -self.dir.x)
}
}
/// A glyph set vertically (WMode 1: no horizontal advance) and the top centre of its em box: a
/// column of them is vertical type.
#[derive(Clone, Copy, Debug)]
pub(crate) struct Upright {
pub top: Point,
}
/// What a run of type is drawn with: its font (cache key, family and style), size, horizontal
/// scale, fill and stroke (paint and width).
#[derive(Clone, Debug, PartialEq)]
pub(crate) struct Look {
pub font: u128,
pub family: String,
pub style: String,
/// The installed version of the font when it isn't the one family and style resolve to.
pub version: Option<String>,
pub size: f64,
pub h_scale: f64,
pub fill: Option<Paint>,
pub stroke: Option<(Paint, f64)>,
/// The fill, the stroke overprints.
pub overprint_fill: bool,
pub overprint_stroke: bool,
}
impl Look {
/// Can a glyph drawn with `other` join a run drawn with this? A fill joins a run whose
/// glyphs were also stroked over (fill and stroke rendering draws each glyph twice); invisible
/// text (neither fill nor stroke) joins only invisible text.
fn takes(&self, other: &Look) -> bool {
self.font == other.font
&& (self.size - other.size).abs() <= self.size * 0.01
&& (self.h_scale - other.h_scale).abs() < 0.5
&& self.fill == other.fill
&& self.overprint_fill == other.overprint_fill
&& ((other.stroke.is_none() && other.fill.is_some()) || (self.stroke == other.stroke && self.overprint_stroke == other.overprint_stroke))
}
fn style(&self) -> CharStyle {
let (stroke, stroke_width) = self.stroke.clone().unwrap_or((Paint::None, 0.0));
CharStyle {
font_family: self.family.clone(),
font_style: self.style.clone(),
font_version: self.version.clone(),
size: round(self.size),
h_scale: round(self.h_scale),
fill: self.fill.clone().unwrap_or(Paint::None),
stroke,
stroke_width,
overprint_fill: self.overprint_fill,
overprint_stroke: self.overprint_stroke,
..CharStyle::default()
}
}
}
/// A gap between adjacent glyphs on a line.
#[derive(Clone, Copy, Debug)]
struct Gap {
/// Byte offset in the line's text of the glyph after the gap.
at: usize,
/// From the end of the glyph before to the start of the next (points).
width: f64,
/// The smaller of the two glyphs' sizes (points): a word gap beside a much larger glyph is
/// narrow for the large size (#3623).
em: f64,
/// Along the line, not turned along a curve.
on_line: bool,
}
/// The spacing a line's letters share (letter-spaced type), in ems: the gap a third of the way up
/// from the narrowest, of three or more along the line. Never below zero, and none at all past
/// half an em: glyphs set wider apart than that are words or columns, not tracked letters.
fn common_gap(gaps: &[Gap]) -> f64 {
let mut widths: Vec<f64> = gaps.iter().filter(|g| g.on_line).map(|g| g.width / g.em).filter(|w| w.is_finite()).collect();
if widths.len() < 3 {
return 0.0;
}
let pivot = widths.len() / 3;
let (_, common, _) = widths.select_nth_unstable_by(pivot, f64::total_cmp);
if *common > 0.5 { 0.0 } else { common.max(0.0) }
}
/// A line of type being gathered: runs of glyphs on one baseline.
pub(crate) struct TextLine {
/// The first glyph's placement.
at: Placement,
opacity: f32,
/// Where the next glyph would start without spacing.
next: Point,
/// The last glyph's origin.
last: Point,
/// Where the last glyph that isn't a space ends.
visible: Point,
/// The last glyph's baseline direction.
last_dir: Vec2,
/// The last glyph's size (points).
last_size: f64,
/// Each glyph's baseline origin, then where the last one ends.
baseline: Vec<Point>,
/// Runs of glyphs drawn alike: their look, how far the file raised them off the line's
/// baseline (points, up; text rise, `Ts`), their text.
runs: Vec<(Look, f64, String)>,
/// Where the file put each character of the runs' text, in order: the first character of each
/// glyph has its glyph's [`Slot`]; a space read from a gap and the other characters of a
/// glyph of several have none.
slots: Vec<Option<Slot>>,
/// The first glyph was set vertically ([`Upright`]).
upright: Option<Upright>,
/// The glyphs make a column (vertical type), each below the last.
column: bool,
/// Each glyph's ink box as drawn (document space), in drawing order; glyphs without ink are
/// left out.
pub ink: Vec<Rect>,
/// The gaps between adjacent glyphs that may read as spaces ([`finish`](Self::finish)
/// decides).
gaps: Vec<Gap>,
/// The spaces the file draws between two glyphs that aren't spaces: their byte offset in the
/// text before the spaces read from gaps go in, and the gaps on both sides of it plus its
/// advance (points).
drawn_spaces: Vec<(usize, f64)>,
/// A space the file drew after a glyph that isn't one, until the next glyph comes: its byte
/// offset, and the gap before it plus its advance (points).
drawn_space: Option<(usize, f64)>,
/// The strokes of its glyphs' outlines drawn between them, one per glyph with ink so far
/// (written right after the line).
pub held: Vec<Node>,
}
/// Where the file put a glyph: the glyph's index in the line, and its advance in the file's font
/// (points) when it shows one character (`None` for a ligature's several).
#[derive(Clone, Copy, Debug)]
struct Slot {
glyph: usize,
advance: Option<f64>,
}
/// What a finished line of type tells beyond its text object (see [`crate::import_lines`]).
#[derive(Clone, Debug, Default)]
pub(crate) struct LineFacts {
/// [`TextLine::ink`].
pub ink: Vec<Rect>,
/// Horizontal point type: it ended in a space (as a line an editor wrapped in a frame does).
pub wraps: bool,
/// Horizontal point type: how long the file set it, from its first glyph's origin to the end
/// of its last glyph that isn't a space (points, along the baseline).
pub length: f64,
}
/// The most a glyph on a curve turns from the one before it (about 20°).
const CURVE_TURN_COS: f64 = 0.94;
impl TextLine {
pub fn new(at: Placement, opacity: f32) -> Self {
Self {
at,
opacity,
next: at.origin,
last: at.origin,
visible: at.origin,
last_dir: at.dir,
last_size: at.size,
baseline: vec![],
runs: vec![],
slots: vec![],
upright: None,
column: false,
ink: vec![],
gaps: vec![],
drawn_spaces: vec![],
drawn_space: None,
held: vec![],
}
}
/// The paint the last glyph is filled with.
pub fn last_fill(&self) -> Option<&Paint> {
self.runs.last()?.0.fill.as_ref()
}
/// Add glyph `text` set vertically (`upright`) if it continues this column (or makes this
/// line's single glyph a column): below the last one, in the same column. `false`: it starts
/// another line.
pub fn push_upright(&mut self, look: &Look, at: Placement, opacity: f32, upright: Upright, text: &str) -> bool {
let down = -self.at.up();
let size = self.at.size.max(at.size);
let Some((last, _, run)) = self.runs.last_mut() else {
self.runs.push((look.clone(), 0.0, text.to_string()));
(self.upright, self.next, self.last) = (Some(upright), at.origin + down * at.size, at.origin);
return true;
};
let across = (at.origin - self.at.origin).dot(self.at.dir);
let step = (at.origin - self.next).dot(down);
// Only glyphs set vertically (WMode 1) come here, so a second one below the first, one em
// on or letter-spaced up to another em, makes a column.
let first_step = (at.origin - self.last).dot(down);
let starts_column = !self.column && self.upright.is_some() && first_step > size * 0.95 && first_step < size * 2.0;
let in_column = (self.column || starts_column)
&& opacity == self.opacity
&& at.dir.dot(self.at.dir) > 0.9995
&& across.abs() < size * 0.6
&& step > -size * 0.3
&& step < size * 1.2;
if !in_column {
return false;
}
self.column = true;
// Letter-spaced Japanese is common: a gap reads as a space only when a whole em is left
// out (the tracking set in `finish` keeps narrower gaps).
if step > size * 0.9 && !run.ends_with(' ') && !text.starts_with(' ') {
run.push(' ');
}
if last.takes(look) {
run.push_str(text);
} else {
self.runs.push((look.clone(), 0.0, text.to_string()));
}
self.next = at.origin + down * at.size;
self.last = at.origin;
true
}
/// Add glyph `text` drawn with `look` at `at` (advancing `advance` points) at `opacity` if
/// it continues this line; `false`: it starts another.
pub fn push(&mut self, look: &Look, at: Placement, opacity: f32, advance: f64, text: &str) -> bool {
// A vertical line (even a single upright glyph) takes no horizontal glyphs.
if self.column || self.upright.is_some() {
return false;
}
let bytes: usize = self.runs.iter().map(|(_, _, t)| t.len()).sum();
let glyph = self.baseline.len();
let curved = self.curved();
let mut shift = 0.0;
if let Some((last, last_shift, run)) = self.runs.last_mut() {
let size = self.at.size.max(at.size);
let gap = (at.origin - self.next).dot(self.last_dir);
// How far the glyph sits above the line's baseline. The file raises a superscript or
// lowers a subscript (text rise, `Ts`) with the glyphs turned exactly as the rest of
// the line (`level`): it stays on the line, shifted off its baseline. Up to half an em
// when it follows on from the glyph before (no wider gap than a space reads from):
// further apart, it is type of its own (the next cell of a table's row, its baseline a
// little higher).
let above = (at.origin - self.at.origin).dot(self.at.up());
let level = at.dir.dot(self.at.dir) > 1.0 - 1e-9 && !curved;
let on_line = opacity == self.opacity
&& at.dir.dot(self.at.dir) > 0.9995
&& (at.slant - self.at.slant).abs() < 1e-3
&& (above.abs() < size * 0.15 || (level && above.abs() < size * 0.5 && gap <= size * 0.2))
// Proportional widths close CJK marks up by up to half an em.
&& gap > -size * 0.6
&& gap < size * 3.0;
// Within a hair of the baseline (rounding in the file) is on it; a glyph turned a
// little (type on a path) sits on the curve, not shifted off it.
if on_line && level && above.abs() > size * 0.01 {
shift = above;
}
// Or on a curve: turned a little from the glyph before it, and starting about where
// that one ends.
let on_curve = !on_line
&& opacity == self.opacity
&& (at.slant - self.at.slant).abs() < 1e-3
&& at.dir.dot(self.last_dir) > CURVE_TURN_COS
&& at.dir.dot(self.last_dir) < 0.99999
&& (at.origin - self.next).hypot() < size * 0.6;
if !on_line && !on_curve {
return false;
}
if !run.ends_with(' ') && !text.starts_with(' ') {
self.gaps.push(Gap { at: bytes, width: gap, em: self.last_size.min(at.size), on_line });
}
// A space the file draws between two glyphs spans the gaps on both sides of it.
let drawn = self.drawn_space.take();
if on_line {
if text == " " && !run.ends_with(' ') {
self.drawn_space = Some((bytes, gap + advance));
} else if let Some((b, before)) = drawn.filter(|_| !text.starts_with(' ')) {
self.drawn_spaces.push((b, before + gap));
}
}
if last.takes(look) && (*last_shift - shift).abs() < 1e-3 {
run.push_str(text);
} else {
self.runs.push((look.clone(), shift, text.to_string()));
}
} else {
self.runs.push((look.clone(), 0.0, text.to_string()));
}
let one = text.chars().count() == 1;
self.slots.push(Some(Slot { glyph, advance: one.then_some(advance) }));
self.slots.extend(text.chars().skip(1).map(|_| None));
self.next = at.origin + at.dir * advance;
if !text.trim().is_empty() {
self.visible = self.next;
}
self.last = at.origin;
self.last_dir = at.dir;
self.last_size = at.size;
self.baseline.push(at.origin);
true
}
/// Has the baseline turned (more than about 3° from the first glyph to the last)?
fn curved(&self) -> bool {
self.last_dir.dot(self.at.dir) < 0.9986
}
/// A smooth path through the glyphs' baseline origins and the end of the last glyph, an em
/// further (Catmull-Rom through the points, as cubic Béziers).
fn baseline_path(&self) -> Option<BezPath> {
let mut pts = self.baseline.clone();
// On past the end of the last glyph by an em, so rounding in the spacing doesn't push it
// off the end of the path (where type on a path stops).
pts.push(self.next);
pts.push(self.next + self.last_dir * self.at.size);
let first = *pts.first()?;
// Finite placements far out can still overflow once an em is added: no path then.
if pts.len() < 3 || !pts.iter().all(|p| p.is_finite()) {
return None;
}
let mut bp = BezPath::new();
bp.move_to(first);
for (i, seg) in pts.windows(2).enumerate() {
let &[p1, p2] = seg else { continue };
let p0 = *pts.get(i.wrapping_sub(1)).unwrap_or(&p1);
let p3 = *pts.get(i + 2).unwrap_or(&p2);
bp.curve_to(p1 + (p2 - p0) / 6.0, p2 - (p3 - p1) / 6.0, p2);
}
Some(bp)
}
/// A stroke (`stroke`: paint and width) over the glyph just drawn at `at` in `font` (fill
/// and stroke rendering), which `overprint`s or not: the run is stroked. `false`: it isn't
/// that glyph.
pub fn stroke_last(&mut self, font: u128, at: Placement, stroke: (Paint, f64), overprint: bool) -> bool {
let Some((look, ..)) = self.runs.last_mut().filter(|(l, ..)| l.font == font && (at.origin - self.last).hypot() < self.at.size * 1e-3) else {
return false;
};
(look.stroke, look.overprint_stroke) = (Some(stroke), overprint);
true
}
/// The point type object, its opacity and what else the line tells.
pub fn finish(mut self) -> Option<(TextObject, f32, LineFacts)> {
let mut facts = LineFacts { ink: std::mem::take(&mut self.ink), ..LineFacts::default() };
// Turned along a curve: type on a path through the glyphs.
let path = if self.curved() { self.baseline_path() } else { None };
facts.wraps = path.is_none() && self.upright.is_none() && self.runs.last().is_some_and(|(.., t)| t.ends_with(' '));
if let Some((.., last)) = self.runs.last_mut() {
last.truncate(last.trim_end().len());
}
self.runs.retain(|(.., t)| !t.is_empty());
if self.runs.iter().all(|(.., t)| t.trim().is_empty()) {
return None;
}
// The spaces trimmed off the end have no place either.
let chars: usize = self.runs.iter().map(|(.., t)| t.chars().count()).sum();
self.slots.truncate(chars);
// The baseline most of the characters sit on is the line's: the others are shifted off it
// (a line starting with a raised footnote mark keeps the baseline of its text).
let mut on: Vec<(f64, usize)> = vec![];
for (_, shift, text) in &self.runs {
let n = text.chars().count();
match on.iter_mut().find(|(s, _)| (s - shift).abs() < 1e-3) {
Some((_, count)) => *count += n,
None => on.push((*shift, n)),
}
}
// A tie keeps the first glyph's ("m²" sits on the m's baseline).
let base = on.iter().fold(None::<(f64, usize)>, |best, &(s, n)| if best.is_none_or(|(_, b)| n > b) { Some((s, n)) } else { best });
let base = base.map_or(0.0, |(s, _)| s);
if path.is_some() {
// Type on a path sits on its curve: no glyph is shifted off it.
for (_, shift, _) in &mut self.runs {
*shift = 0.0;
}
} else if base != 0.0 {
// Up the slant the type leans by (a sheared text matrix), as the file raises it.
self.at.origin += (self.at.up() + self.at.dir * self.at.slant) * base;
for (_, shift, _) in &mut self.runs {
*shift -= base;
}
}
let mut runs = self.runs.into_iter().map(|(look, shift, text)| {
let style = CharStyle { baseline_shift: round(shift), ..look.style() };
TextRun { text, style, inline: None }
});
let first = runs.next()?;
let mut t = TextObject::point(Point::ORIGIN, &first.text, first.style);
t.runs.extend(runs);
// A gap wider than a fifth of an em, past the spacing the line's letters share, reads as a
// space (one set at exactly a fifth still counts, whatever the rounding). Letter spacing is
// read only in text left in the file's order.
let space = |gap: &Gap, common: f64| gap.width / gap.em - common > 0.199;
let directed = vectorcraft_text::logical_order(&t.plain_text()).is_some();
let mut common = if directed { 0.0 } else { common_gap(&self.gaps) };
// Evenly spaced glyphs alone could be letters set apart or words of one letter each: the
// shared spacing is letter spacing only on a line that also shows where its words break,
// with spaces it draws or with gaps wider than the rest.
if self.drawn_spaces.is_empty() && !self.gaps.iter().any(|gap| space(gap, common)) {
common = 0.0;
}
let mut spaces = vec![];
for gap in &self.gaps {
if !space(gap, common) {
continue;
}
// After the earlier spaces, at the end of the run before it (the run it follows).
let at = gap.at + spaces.len();
let mut offset = 0;
let Some(i) = t.runs.iter().position(|run| {
offset += run.text.len();
at <= offset
}) else {
break;
};
// The characters before it, in the runs before and in its own.
let earlier: usize = t.runs.iter().take(i).map(|r| r.text.chars().count()).sum();
let Some(run) = t.runs.get_mut(i) else { break };
let at = at - (offset - run.text.len());
let Some(head) = run.text.get(..at) else { continue };
// A space read from a gap has no slot of its own.
let char_at = earlier + head.chars().count();
if char_at <= self.slots.len() {
self.slots.insert(char_at, None);
}
run.text.insert(at, ' ');
spaces.push((gap.at, gap.width, gap.on_line));
}
// Each by its offset in the text with the spaces in.
let mut gaps: Vec<(usize, f64)> = spaces.iter().enumerate().filter(|(_, s)| s.2).map(|(i, s)| (s.0 + i, s.1)).collect();
gaps.extend(self.drawn_spaces.iter().map(|&(b, width)| (b + spaces.iter().filter(|s| s.0 < b).count(), width)));
gaps.sort_by_key(|&(b, _)| b);
if self.upright.is_none() {
to_logical(&mut t);
}
let angle = self.at.dir.atan2();
// The glyphs' tops lean along the baseline (+x), up being -y in the type's own space.
t.xf = Affine::translate(self.at.origin.to_vec2()) * Affine::rotate(angle) * Affine::skew(-self.at.slant, 0.0);
let db = vectorcraft_text::FontDb::global();
// Each glyph where the file put it (#722): the kerning (the font's, or none) and tracking
// (one value for the line, or each character its own) that place them there.
let placed = (self.upright.is_none() && t.para.direction.is_none())
.then(|| place_each(&t, &self.slots, &self.baseline, path.as_ref(), self.at, self.at.size))
.flatten()
.map(|(runs, laid)| {
t.runs = runs;
laid.bounds
});
if let Some(path) = path {
// Evenly spaced type keeps its length along the path: the path runs through the
// glyphs' origins, one curve each, to the end of the last glyph, and tracking makes up
// the difference from the advances.
if placed.is_none() {
let length: f64 = path.segments().take(self.baseline.len()).map(|seg| kurbo::ParamCurveArclen::arclen(&seg, 1e-3)).sum();
let laid = vectorcraft_text::layout(db, &t);
let natural: f64 = laid.glyphs.iter().map(|g| g.advance).sum();
let chars: usize = t.runs.iter().map(|r| r.text.chars().count()).sum();
let size = self.at.size;
if chars > 1 && length.is_finite() && (length - natural).abs() > size * 0.01 {
let (tracking, spaces) = spacing(&laid, &[], length - natural, chars, size);
t.runs = tracked(std::mem::take(&mut t.runs), tracking, &spaces);
}
}
t.kind = TextKind::OnPath { path: vectorcraft_geom::PathData::from_bezpath(&path), start: 0.0, end: None };
t.xf = Affine::IDENTITY;
t.cached_bounds = Some(vectorcraft_text::layout(db, &t).bounds);
return Some((t, self.opacity, facts));
}
let (start, along) = match self.upright {
// Vertical point type (a column, or a single upright glyph) is anchored at the top
// centre of its first em box.
Some(u) => {
t.vertical = true;
t.xf = Affine::translate(u.top.to_vec2()) * Affine::rotate(angle);
(self.at.origin, -self.at.up())
}
_ => (self.at.origin, self.at.dir),
};
// Tracking that makes the line as long as in the file: the PDF placed each glyph, the
// layout sets them by their advances. Spread over the gaps between characters (tracking
// follows each character, the last one's past the end of the line). A horizontal line is
// measured to its last glyph that isn't a space: the spaces after it were dropped.
let length = (if t.vertical { self.next } else { self.visible } - start).dot(along);
if !t.vertical {
facts.length = length;
}
// Placed glyph by glyph: laid out so already.
if let Some(bounds) = placed {
t.cached_bounds = Some(bounds);
return Some((t, self.opacity, facts));
}
let laid = vectorcraft_text::layout(db, &t);
let natural: f64 = laid.glyphs.iter().map(|g| g.advance).sum();
let chars: usize = t.runs.iter().map(|r| r.text.chars().count()).sum();
let size = self.at.size;
let mut bounds = laid.bounds;
if chars > 1 && length.is_finite() && (length - natural).abs() > size * 0.01 {
// The spaces read from gaps or drawn between glyphs keep their own place only in text
// left in the file's order.
let total: usize = t.runs.iter().map(|r| r.text.len()).sum();
let gaps: Vec<(usize, f64)> = if t.para.direction.is_none() { gaps.into_iter().filter(|(b, _)| *b < total).collect() } else { vec![] };
let (tracking, spaces) = spacing(&laid, &gaps, length - natural, chars, size);
t.runs = tracked(std::mem::take(&mut t.runs), tracking, &spaces);
// Tracked: laid out again for its bounds.
bounds = vectorcraft_text::layout(db, &t).bounds;
}
t.cached_bounds = Some(bounds);
Some((t, self.opacity, facts))
}
}
/// Where along the line (points) layout `laid` puts each character at a byte offset of `chars`
/// (in order), and its advance: `None` for one inside another's cluster. Text in the file's order
/// (left to right).
fn pens(laid: &TextLayout, chars: &[usize]) -> Vec<Option<(f64, f64)>> {
let mut at = vec![None; chars.len()];
let mut pen = 0.0;
for g in &laid.glyphs {
if let Some(slot) = chars.binary_search(&g.byte).ok().and_then(|i| at.get_mut(i)) {
slot.get_or_insert((pen, g.advance));
}
pen += g.advance;
}
at
}
/// `t`'s runs with kerning `kern` and each character's tracking from `own` (points, one per
/// character, in thousandths of each run's em), characters alike kept in one run.
fn restyled(t: &TextObject, kern: Option<f64>, own: &[f64]) -> Vec<TextRun> {
let mut out: Vec<TextRun> = Vec::with_capacity(t.runs.len());
// The run of `t` the last character came from.
let mut from = None;
let mut k = 0;
for (i, r) in t.runs.iter().enumerate() {
let size = r.style.size.max(1e-3);
for c in r.text.chars() {
let tracking = own.get(k).map_or(0.0, |e| (e / size * 1000.0).clamp(-1000.0, 10000.0));
k += 1;
// Hundredths of a thousandth of an em: no drift along a line, no noise in the panels.
let tracking = (tracking * 100.0).round() / 100.0;
match out.last_mut() {
Some(last) if from == Some(i) && last.style.tracking == tracking => last.text.push(c),
last => {
let style = CharStyle { tracking, kerning: kern, ..r.style.clone() };
match last {
Some(last) if last.style == style && last.inline == r.inline => last.text.push(c),
_ => out.push(TextRun { text: c.to_string(), style, inline: r.inline.clone() }),
}
}
}
from = Some(i);
}
}
out
}
/// The runs of `t` that put its glyphs where the file did (#722), and their layout: with the
/// font's kerning or none, whichever the file's spacing follows, and tracking that is the same
/// after every character when the file spaced them evenly, else each character's own (word
/// spacing, `TJ` adjustments, character tightening, spaces read from gaps), measured from the
/// file's widths (not the font's advances, which may be a hair off them) and kept on course where
/// the differences add up ([`on_course`]).
///
/// The file put the glyph of each character with a [`Slot`] at `baseline` (one origin per glyph),
/// measured along `path` (type on a path: one curve per glyph) or along `line`'s baseline from its
/// origin, where a character shifted off it by its baseline shift draws that shift times `line`'s
/// slant further along (the type's skew leans it so, as the file's text rise did). `None` when the
/// font laid out isn't the file's (its advances differ from the file's): spacing measured against
/// a stand-in's widths would keep the stand-in's differences once the font is installed; or when
/// the glyphs can't be placed so (the layout joins characters the file drew apart).
fn place_each(
t: &TextObject,
slots: &[Option<Slot>],
baseline: &[Point],
path: Option<&BezPath>,
line: Placement,
size: f64,
) -> Option<(Vec<TextRun>, TextLayout)> {
let chars: Vec<usize> = t
.runs
.iter()
.scan(0, |off, r| {
let base = *off;
*off += r.text.len();
Some(r.text.char_indices().map(move |(i, _)| base + i).collect::<Vec<_>>())
})
.flatten()
.collect();
if chars.len() < 2 || chars.len() != slots.len() || t.runs.iter().any(|r| r.inline.is_some()) {
return None;
}
// How far along the line the file put each glyph.
let along: Vec<f64> = match path {
Some(bp) => std::iter::once(0.0)
.chain(bp.segments().scan(0.0, |d, seg| {
*d += kurbo::ParamCurveArclen::arclen(&seg, 1e-3);
Some(*d)
}))
.take(baseline.len())
.collect(),
None => baseline.iter().map(|p| (*p - line.origin).dot(line.dir)).collect(),
};
// Each character's baseline shift, which the type's slant draws `slant` times as far along.
let shifts: Vec<f64> = t.runs.iter().flat_map(|r| r.text.chars().map(move |_| r.style.baseline_shift)).collect();
let slant = if path.is_some() { 0.0 } else { line.slant };
// Each character's target: where the file put it, if it put it anywhere.
let target: Vec<Option<(f64, Option<f64>)>> =
slots.iter().zip(&shifts).map(|(s, shift)| s.and_then(|s| Some((*along.get(s.glyph)? - slant * shift, s.advance)))).collect();
if target.first().is_none_or(Option::is_none) || target.iter().flatten().any(|(d, _)| !d.is_finite()) {
return None;
}
let db = vectorcraft_text::FontDb::global();
// The spacing each character takes after it, laid out with `pens`: whatever puts the next one
// the file placed where it did. Characters between two the file placed (a space read from a
// gap, the rest of a ligature) take the line's spacing, the last of them the rest.
let spacing = |pens: &[Option<(f64, f64)>]| -> Option<Vec<f64>> {
// How far each character the file placed is from where the layout puts it.
let placed: Vec<(usize, f64)> =
target.iter().zip(pens).enumerate().filter_map(|(i, (t, p))| Some((i, t.as_ref()?.0 - p.as_ref()?.0))).collect();
let steps: Vec<f64> = placed.windows(2).filter(|w| w[1].0 == w[0].0 + 1).map(|w| w[1].1 - w[0].1).collect();
let line = median(steps).unwrap_or(0.0);
let mut own = vec![line; chars.len()];
for w in placed.windows(2) {
let ((a, da), (b, db)) = (w[0], w[1]);
*own.get_mut(b.checked_sub(1)?)? = db - da - line * (b - a - 1) as f64;
}
// After the last character the file placed, spacing changes nothing: the line's.
let last = placed.last()?.0;
let tail = if last > 0 { own.get(last - 1).copied().unwrap_or(line) } else { line };
for e in own.get_mut(last..).into_iter().flatten() {
*e = tail;
}
Some(own)
};
// Evenly spaced: one tracking value for the line. (A file rounds its widths to thousandths of
// an em: a hair either way is even.)
let even = |own: &[f64]| {
let line = median(own.to_vec())?;
own.iter().all(|e| (e - line).abs() <= size * 0.002).then(|| vec![line; own.len()])
};
let changes = |own: &[f64]| own.windows(2).filter(|w| (w[1] - w[0]).abs() > size * 1e-4).count();
// The type as it is: untracked, with the font's kerning.
let auto = vectorcraft_text::layout(db, t);
// Each glyph's own advance (the font's, unkerned) tells whether the layout's font is the
// file's, whether the font kerned any of the line's pairs, and where the glyphs go unkerned
// (the same glyphs, each on its own advance: no layout needed to tell).
let mut faces: Vec<std::sync::Arc<vectorcraft_text::FontFace>> = vec![];
let mut unkerned_pens = vec![None; chars.len()];
let mut pen = 0.0;
let mut kerned_any = false;
for g in &auto.glyphs {
let face = match faces.iter().find(|f| f.id() == g.font_id) {
Some(f) => f.clone(),
None => {
let f = db.face_by_id(g.font_id)?;
faces.push(f.clone());
f
}
};
let style = &t.runs.get(g.run)?.style;
let own = face.advance(g.gid) * style.size / face.units_per_em().max(1e-9) * style.h_scale / 100.0;
kerned_any |= (own - g.advance).abs() > size * 1e-6;
if let Some(slot) = chars.binary_search(&g.byte).ok().and_then(|i| unkerned_pens.get_mut(i)) {
slot.get_or_insert((pen, own));
}
pen += own;
}
let same_font = unkerned_pens.iter().zip(&target).all(|(n, t)| match t {
Some((_, Some(adv))) => n.is_some_and(|(_, a): (f64, f64)| (a - adv).abs() <= size * 0.01),
_ => true,
});
if !same_font {
return None;
}
// Where the glyphs go on the file's own widths: each character's pen moved by how much wider
// the file says the characters before it are than the font's advances (a file gives widths to
// the thousandth of an em, so a font of 2048 units an em is a hair off), where the file gives
// a character's width and the layout sets it as a glyph of its own. Spacing measured from them
// is the file's (its `Tc`, `Tw`, `TJ`): measured from the installed font's advances, each
// character would also take its own hair of tracking (a run each), and keep it once a font
// rounded otherwise is installed.
let end = pen;
let mut wider = 0.0;
let mut drift = Vec::with_capacity(chars.len());
for (i, t) in target.iter().enumerate() {
drift.push(wider);
let (Some(Some((start, _))), Some((_, Some(width)))) = (unkerned_pens.get(i), t) else { continue };
let next = match unkerned_pens.get(i + 1) {
Some(n) => n.map(|(p, _)| p),
None => Some(end),
};
if let Some(next) = next {
wider += width - (next - start);
}
}
let by_file =
|pens: &[Option<(f64, f64)>]| -> Vec<Option<(f64, f64)>> { pens.iter().zip(&drift).map(|(p, d)| p.map(|(x, a)| (x + d, a))).collect() };
// The font's kerning when the file followed it (or there is none to follow), else none.
let kerned_pens = pens(&auto, &chars);
let kerned = spacing(&by_file(&kerned_pens))?;
let (kern, own) = match even(&kerned) {
Some(own) => (None, own),
// Without kerning the line lays out the same.
None if !kerned_any => (None, kerned),
None => {
let unkerned = spacing(&by_file(&unkerned_pens))?;
match even(&unkerned) {
Some(own) => (Some(0.0), own),
None if changes(&unkerned) < changes(&kerned) => (Some(0.0), unkerned),
None => (None, kerned),
}
}
};
let own = on_course(own, if kern.is_some() { &unkerned_pens } else { &kerned_pens }, &target, size);
let runs = restyled(t, kern, &own);
// Laid out so, are they where the file put them? (A ligature the layout forms where the file
// drew two glyphs moves those after it.)
let mut probe = t.clone();
probe.runs = runs;
let check = vectorcraft_text::layout(db, &probe);
let got = pens(&check, &chars);
let ok = got.iter().zip(&target).all(|(g, t)| match t {
Some((d, _)) => g.is_some_and(|(p, _)| (p - d).abs() <= size * 0.02),
None => true,
});
ok.then_some((probe.runs, check))
}
/// `own` (each character's spacing, points, measured from the file's widths) kept on course: laid
/// out from `pens` (where each character goes untracked), the glyphs land within a hundredth of an
/// em of `target` (where the file put them). The font's advances may each be a little off the
/// file's widths (by up to the 1% that still counts as the same font), and a line spaced evenly to
/// a hair takes the same value after every character: small differences that add up along a line.
/// Off course, each stretch of the line takes a share of the difference after every character
/// (one run), as long a stretch as one share keeps on course.
fn on_course(mut own: Vec<f64>, pens: &[Option<(f64, f64)>], target: &[Option<(f64, Option<f64>)>], size: f64) -> Vec<f64> {
let tolerance = size * 0.01;
// How much further along than `own` lays it out each glyph the file placed must go.
let mut at = 0.0;
let mut behind = vec![];
for (i, ((p, t), e)) in pens.iter().zip(target).zip(&own).enumerate() {
if let (Some((p, _)), Some((t, _))) = (p, t) {
behind.push((i, t - p - at));
}
at += e;
}
if behind.iter().all(|(_, d)| d.abs() <= tolerance) {
return own;
}
// The character the stretch starts at, how far the shares before it moved it, the last glyph
// the stretch keeps on course, and the shares (lowest, highest) that keep each since on it.
let (mut from, mut moved, mut last) = (0, 0.0, 0);
let (mut lo, mut hi) = (f64::NEG_INFINITY, f64::INFINITY);
for &(i, d) in &behind {
if i <= from {
continue;
}
let course = |from: usize, moved: f64| {
let n = (i - from) as f64;
((d - moved - tolerance) / n, (d - moved + tolerance) / n)
};
let (l, h) = course(from, moved);
if l > hi || h < lo {
// No share keeps this glyph on course too: the stretch ends at the last one it does.
let share = (lo + hi) / 2.0;
for e in own.get_mut(from..last).into_iter().flatten() {
*e += share;
}
moved += share * (last - from) as f64;
from = last;
(lo, hi) = course(from, moved);
} else {
(lo, hi) = (lo.max(l), hi.min(h));
}
last = i;
}
// The last stretch takes its share after the characters past the last glyph the file placed
// too (it changes nothing there).
if lo.is_finite() && hi.is_finite() {
let share = (lo + hi) / 2.0;
for e in own.get_mut(from..).into_iter().flatten() {
*e += share;
}
}
own
}
/// The middle value of `v` (the upper one of an even count).
pub(crate) fn median(mut v: Vec<f64>) -> Option<f64> {
v.sort_by(f64::total_cmp);
v.get(v.len() / 2).copied()
}
/// The tracking (thousandths of an em) that sets a line of `chars` characters, laid out untracked
/// as `laid`, `extra` points longer, as the file set it: the same after every character, except
/// the spaces read from gaps or drawn between glyphs (`gaps`: byte offset, gap in points) where
/// that would put the glyphs after them more than a tenth of an em off (a barcode's digits, set in
/// groups apart). Those take their own tracking, given with their byte offsets.
fn spacing(laid: &TextLayout, gaps: &[(usize, f64)], extra: f64, chars: usize, size: f64) -> (f64, Vec<(usize, f64)>) {
let pairs = chars.saturating_sub(1).max(1) as f64;
// The advance of the first glyph laid out at each byte offset.
let space: HashMap<usize, f64> = laid.glyphs.iter().rev().map(|g| (g.byte, g.advance)).collect();
// The space and the character before it take `need` (the gap less the space) between them;
// tracking gives them twice its share.
let needs: Vec<(usize, f64)> = gaps.iter().filter_map(|&(b, gap)| Some((b, gap - space.get(&b)?))).collect();
let mut own: Vec<(usize, f64)> = vec![];
let mut taken: HashSet<usize> = HashSet::new();
let mut tracking = extra / pairs;
// Each space taken apart changes the others' share: until no more are.
loop {
let more: Vec<(usize, f64)> =
needs.iter().filter(|(b, need)| !taken.contains(b) && (need - 2.0 * tracking).abs() > size * 0.1).copied().collect();
if more.is_empty() {
break;
}
taken.extend(more.iter().map(|(b, _)| *b));
own.extend(more);
let rest = pairs - 2.0 * own.len() as f64;
tracking = if rest >= 1.0 { (extra - own.iter().map(|(_, need)| need).sum::<f64>()) / rest } else { 0.0 };
}
own.sort_by_key(|(b, _)| *b);
let em = |pts: f64| (pts / size * 1000.0).clamp(-1000.0, 10000.0).round();
(em(tracking), own.into_iter().map(|(b, need)| (b, em(need - tracking))).collect())
}
/// `runs` with tracking `tracking`, and each space at a byte offset of `spaces` (in order) a run
/// of its own with the tracking given with it.
fn tracked(runs: Vec<TextRun>, tracking: f64, spaces: &[(usize, f64)]) -> Vec<TextRun> {
let mut out = Vec::with_capacity(runs.len() + 2 * spaces.len());
let mut off = 0;
for r in runs {
let style = CharStyle { tracking, ..r.style };
let end = off + r.text.len();
let mut from = 0;
for &(b, own) in spaces.iter().filter(|(b, _)| (off..end).contains(b)) {
let at = b - off;
let (Some(before), Some(space)) = (r.text.get(from..at), r.text.get(at..at + 1)) else { continue };
if !before.is_empty() {
out.push(TextRun { text: before.to_string(), style: style.clone(), inline: None });
}
out.push(TextRun { text: space.to_string(), style: CharStyle { tracking: own, ..style.clone() }, inline: None });
from = at + 1;
}
if let Some(rest) = r.text.get(from..).filter(|t| !t.is_empty()) {
out.push(TextRun { text: rest.to_string(), style, inline: None });
}
off = end;
}
out
}
fn round(v: f64) -> f64 {
(v * 1000.0).round() / 1000.0
}
/// Lower-case letters and digits only (`"Source Sans 3"` → `"sourcesans3"`).
fn norm(s: &str) -> String {
s.chars().filter(char::is_ascii_alphanumeric).map(|c| c.to_ascii_lowercase()).collect()
}
/// `"TimesNewRoman"` → `"Times New Roman"`, `"SourceSans3"` → `"Source Sans 3"`.
fn spaced(s: &str) -> String {
let mut out = String::with_capacity(s.len() + 4);
let mut prev: Option<char> = None;
for c in s.chars() {
if let Some(p) = prev
&& ((p.is_ascii_lowercase() && c.is_ascii_uppercase()) || (p.is_ascii_alphabetic() && c.is_ascii_digit()))
{
out.push(' ');
}
out.push(c);
prev = Some(c);
}
out
}
/// `name` without a PostScript name's trailing `PSMT`, `MT` or `PS`.
fn strip_ps(name: &str) -> &str {
["PSMT", "MT", "PS"].iter().find_map(|s| name.strip_suffix(s).filter(|r| !r.is_empty())).unwrap_or(name)
}
/// The families available, by [`norm`]ed name.
pub(crate) struct Families(HashMap<String, String>);
impl Families {
pub fn available() -> Self {
Self(vectorcraft_text::FontDb::global().families().into_iter().map(|f| (norm(&f), f)).collect())
}
/// The family and style of a font with base (PostScript) name `name`, or of weight `weight`
/// and slant `italic` when the name has no style; whether the family is available.
pub fn resolve(&self, name: &str, weight: Option<u32>, italic: bool) -> (String, String, bool) {
// A subset's six-letter tag.
let name = match name.split_once('+') {
Some((tag, rest)) if tag.len() == 6 && tag.chars().all(|c| c.is_ascii_uppercase()) => rest,
_ => name,
};
// An installed face of that exact PostScript name: its own family and style.
if let Some((family, style)) = vectorcraft_text::FontDb::global().by_postscript_name(name) {
return (family, style, true);
}
let (fam, style) = name.split_once(['-', ',']).unwrap_or((name, ""));
let found = [fam, strip_ps(fam)].iter().find_map(|f| self.0.get(&norm(f)).cloned());
let style = match spaced(strip_ps(style)).as_str() {
"" | "Roman" | "Book" | "Normal" | "Plain" => match (weight.is_some_and(|w| w >= 600) || style.contains("Bold"), italic) {
(true, true) => "Bold Italic".to_string(),
(true, false) => "Bold".to_string(),
(false, true) => "Italic".to_string(),
(false, false) => "Regular".to_string(),
},
s => s.to_string(),
};
match found {
Some(f) => (f, style, true),
None => (spaced(strip_ps(fam)).trim().to_string(), style, false),
}
}
}
/// Hebrew or Arabic comes from a PDF in visual order (each glyph where it is drawn): put `t`'s text
/// back in logical order, with the paragraph direction that shows it as drawn.
fn to_logical(t: &mut TextObject) {
let Some((order, rtl)) = vectorcraft_text::logical_order(&t.plain_text()) else { return };
let chars: Vec<(char, usize)> = t.runs.iter().enumerate().flat_map(|(i, r)| r.text.chars().map(move |c| (c, i))).collect();
let mut runs: Vec<(usize, String)> = Vec::with_capacity(t.runs.len());
for &(c, i) in order.iter().filter_map(|&k| chars.get(k)) {
match runs.last_mut() {
Some((run, text)) if *run == i => text.push(c),
_ => runs.push((i, c.to_string())),
}
}
let runs = runs.into_iter().filter_map(|(i, text)| Some(TextRun { text, style: t.runs.get(i)?.style.clone(), inline: None })).collect();
t.runs = runs;
t.para.direction = Some(if rtl { ParaDirection::RightToLeft } else { ParaDirection::LeftToRight });
}