Compare commits

..
5 changed files with 105 additions and 49 deletions
+1 -12
View File
@@ -390,7 +390,6 @@ pub fn pdf_options(p: &Value, count: usize) -> Result<deckcraft_pdf::PdfOptions>
pub fn parse_slide_range(text: &str, count: usize) -> Result<Vec<usize>> {
let invalid = || bad("file.export", format!("“{text}” isn't a valid slide range: use numbers and ranges such as 1-3, 5"));
let mut out: Vec<usize> = vec![];
let mut seen: std::collections::HashSet<usize> = std::collections::HashSet::new();
for part in text.split([',', ';']).map(str::trim).filter(|p| !p.is_empty()) {
let (a, b) = match part.split_once('-') {
Some((a, b)) => (a.trim(), b.trim()),
@@ -406,7 +405,7 @@ pub fn parse_slide_range(text: &str, count: usize) -> Result<Vec<usize>> {
return Err(bad("file.export", format!("the presentation has {count} slides; “{part}” is outside them")));
}
for i in lo - 1..hi {
if seen.insert(i) {
if !out.contains(&i) {
out.push(i);
}
}
@@ -770,16 +769,6 @@ mod tests {
use super::{parse_slide_range, write_file};
#[test]
fn broad_and_repeated_slide_ranges_expand_once() {
// 200 copies of a 100,000-slide range: each index once, in order, without a scan per index.
let text = vec!["1-100000"; 200].join(",");
let v = parse_slide_range(&text, 100_000).unwrap();
assert_eq!(v.len(), 100_000);
assert!(v.windows(2).all(|w| w[0] < w[1]));
assert_eq!(parse_slide_range("3, 1-3, 2", 3).unwrap(), vec![2, 0, 1]);
}
#[test]
fn slide_range_rejects_plus_sign() {
assert!(parse_slide_range("+2", 3).is_err());
+1 -1
View File
@@ -367,7 +367,7 @@ fn spelling(s: &mut Session, p: &Value) -> Result<Value> {
}
out.push(v);
}
let rep: std::collections::HashSet<usize> = crate::spell::repeated(para, &default).iter().map(|w| w.start).collect();
let rep: Vec<usize> = crate::spell::repeated(para, &default).iter().map(|w| w.start).collect();
for w in crate::spell::paragraph_words(para, &default) {
if ignore.contains(&w.text.to_lowercase()) || rep.contains(&w.start) {
continue;
+1 -2
View File
@@ -405,9 +405,8 @@ fn select_slides(s: &mut Session, p: &Value) -> Result<Value> {
};
s.select(|doc, sel| {
if add {
let mut have: std::collections::HashSet<_> = sel.slides.iter().copied().collect();
for i in &ids {
if have.insert(*i) {
if !sel.slides.contains(i) {
sel.slides.push(*i);
}
}
+89 -34
View File
@@ -758,17 +758,20 @@ impl Chart {
}
fn stacked_value_range(&self, n: usize) -> (f64, f64) {
let mut lo: f64 = 0.0;
let mut hi: f64 = 0.0;
for i in 0..n {
let (mut p, mut m) = (0.0, 0.0);
for s in &self.series {
let v = s.values.get(i).copied().flatten().unwrap_or(0.0);
if v >= 0.0 { p += v } else { m += v }
// Sums per category, from the values the series hold (an absent one adds nothing).
let (mut pos, mut neg) = (vec![0.0f64; n], vec![0.0f64; n]);
for s in &self.series {
for (i, v) in s.values.iter().enumerate().take(n) {
let v = v.unwrap_or(0.0);
if v >= 0.0 {
pos[i] += v;
} else {
neg[i] += v;
}
}
hi = hi.max(p);
lo = lo.min(m);
}
let hi = pos.iter().fold(0.0f64, |a, &p| a.max(p));
let lo = neg.iter().fold(0.0f64, |a, &m| a.min(m));
(lo, hi)
}
@@ -1348,40 +1351,50 @@ impl Chart {
/// The box chart's groups (distinct category labels in order of appearance; one unnamed
/// group without categories) and, per series, each group's summary (None when it has no
/// points).
///
/// Each value is put in its group's bucket once (groups are found by name in a map), and the
/// series x groups result is held to [`MAX_BOX_CELLS`]: groups past it are left out.
pub fn box_groups(&self) -> (Vec<String>, Vec<Vec<Option<BoxStats>>>) {
let max_groups = (MAX_BOX_CELLS / self.series.len().max(1)).max(1);
let mut groups: Vec<String> = Vec::new();
let mut index: std::collections::HashMap<&str, usize> = std::collections::HashMap::new();
for c in &self.categories {
if !groups.contains(c) {
if !index.contains_key(c.as_str()) && groups.len() < max_groups {
index.insert(c, groups.len());
groups.push(c.clone());
}
}
if groups.is_empty() {
groups.push(String::new());
index.insert("", 0);
}
// A value past the categories belongs to the unnamed group, when there is one.
let past_end = index.get("").copied();
let stats = self
.series
.iter()
.map(|s| {
groups
.iter()
.map(|g| {
let pts: Vec<f64> = s
.values
.iter()
.enumerate()
.filter(|(i, _)| self.categories.get(*i).map_or(g.is_empty() || self.categories.is_empty(), |c| c == g))
.filter_map(|(_, v)| *v)
.filter(|v| v.is_finite())
.collect();
box_stats_by(pts, self.inclusive_quartiles)
})
.collect()
let mut buckets: Vec<Vec<f64>> = vec![Vec::new(); groups.len()];
for (i, v) in s.values.iter().enumerate() {
let Some(v) = v.filter(|v| v.is_finite()) else { continue };
let g = match self.categories.get(i) {
Some(c) => index.get(c.as_str()).copied(),
None => past_end,
};
if let Some(b) = g.and_then(|g| buckets.get_mut(g)) {
b.push(v);
}
}
buckets.into_iter().map(|pts| box_stats_by(pts, self.inclusive_quartiles)).collect()
})
.collect();
(groups, stats)
}
}
/// Most series x group summaries [`Chart::box_groups`] makes.
pub const MAX_BOX_CELLS: usize = 1_000_000;
/// Widens the (min, max) extent `ext` to include `v`.
fn widen(ext: &mut (f64, f64), v: f64) {
ext.0 = ext.0.min(v);
@@ -1454,33 +1467,41 @@ impl Chart {
let n = self.categories.len().max(s.values.len());
let levels: Vec<&Vec<String>> = self.category_levels.iter().take(MAX_LEVELS).collect();
// Each leaf's path from the top level down, blanks filled from the row above.
let mut last: Vec<String> = vec![String::new(); levels.len()];
let paths: Vec<Vec<String>> = (0..n)
// Labels are shared, not copied: an outer label given once stands for every leaf under it.
use std::rc::Rc;
let empty: Rc<str> = Rc::from("");
let mut last: Vec<Rc<str>> = vec![empty.clone(); levels.len()];
let paths: Vec<Vec<Rc<str>>> = (0..n)
.map(|i| {
let mut path = Vec::with_capacity(levels.len() + 1);
for (d, lv) in levels.iter().enumerate().rev() {
let l = lv.get(i).map(|x| x.trim()).unwrap_or("");
if !l.is_empty() {
last[d] = l.to_string();
last[d] = Rc::from(l);
// A new outer label starts fresh inner groups.
for inner in last.iter_mut().take(d) {
inner.clear();
*inner = empty.clone();
}
}
path.push(last[d].clone());
}
path.push(self.categories.get(i).cloned().unwrap_or_default());
path.push(self.categories.get(i).map_or_else(|| empty.clone(), |c| Rc::from(c.as_str())));
path
})
.collect();
let value = |i: usize| s.values.get(i).copied().flatten().filter(|v| v.is_finite()).unwrap_or(0.0).max(0.0);
fn build(rows: &[(usize, &[String])], value: &dyn Fn(usize) -> f64) -> Vec<TreeNode> {
fn build(rows: &[(usize, &[std::rc::Rc<str>])], value: &dyn Fn(usize) -> f64) -> Vec<TreeNode> {
let mut out: Vec<TreeNode> = Vec::new();
let mut i = 0;
while i < rows.len() {
let (idx, path) = rows[i];
if path.len() <= 1 {
out.push(TreeNode { name: path.first().cloned().unwrap_or_default(), value: value(idx), leaf: Some(idx), children: vec![] });
out.push(TreeNode {
name: path.first().map(|p| p.to_string()).unwrap_or_default(),
value: value(idx),
leaf: Some(idx),
children: vec![],
});
i += 1;
continue;
}
@@ -1489,14 +1510,14 @@ impl Chart {
while j < rows.len() && rows[j].1.len() > 1 && &rows[j].1[0] == head {
j += 1;
}
let sub: Vec<(usize, &[String])> = rows[i..j].iter().map(|(k, p)| (*k, &p[1..])).collect();
let sub: Vec<(usize, &[std::rc::Rc<str>])> = rows[i..j].iter().map(|(k, p)| (*k, &p[1..])).collect();
let children = build(&sub, value);
out.push(TreeNode { name: head.clone(), value: children.iter().map(|c| c.value).sum(), leaf: None, children });
out.push(TreeNode { name: head.to_string(), value: children.iter().map(|c| c.value).sum(), leaf: None, children });
i = j;
}
out
}
let rows: Vec<(usize, &[String])> = paths.iter().enumerate().map(|(i, p)| (i, p.as_slice())).collect();
let rows: Vec<(usize, &[Rc<str>])> = paths.iter().enumerate().map(|(i, p)| (i, p.as_slice())).collect();
build(&rows, &value)
}
}
@@ -1607,6 +1628,40 @@ mod chartex_tests {
assert!(c.category_tree().iter().all(|n| n.leaf.is_some()));
}
#[test]
fn large_box_hierarchy_and_stacked_charts_take_linear_work() {
// 50,000 distinct categories in one series: grouped through a map, one pass.
let n = 50_000;
let mut c = Chart::sample(ChartType::BoxWhisker);
c.series.truncate(1);
c.categories = (0..n).map(|i| format!("g{i}")).collect();
c.series[0].values = (0..n).map(|i| Some(i as f64)).collect();
let (groups, stats) = c.box_groups();
assert_eq!((groups.len(), stats[0].len()), (n, n));
assert!(stats[0].iter().all(Option::is_some));
// 2,000 series cap the groups so the result stays within MAX_BOX_CELLS.
c.series = vec![c.series[0].clone(); 2000];
let (groups, stats) = c.box_groups();
assert_eq!(groups.len(), MAX_BOX_CELLS / 2000);
assert!(stats.iter().all(|s| s.len() == groups.len()));
// One long outer label over 50,000 leaves is shared, not copied per leaf.
let mut t = Chart::sample(ChartType::Treemap);
t.series.truncate(1);
t.categories = (0..n).map(|i| format!("c{i}")).collect();
t.series[0].values = (0..n).map(|_| Some(1.0)).collect();
let mut outer = vec![String::new(); n];
outer[0] = "o".repeat(1 << 16);
t.category_levels = vec![outer];
let tree = t.category_tree();
assert_eq!(tree.len(), 1);
assert_eq!(tree[0].children.len(), n);
// Stacked sums over many sparse series.
let mut st = Chart::sample(ChartType::StackedBar);
st.categories = (0..n).map(|i| i.to_string()).collect();
st.series = (0..2000).map(|k| Series { values: vec![Some(1.0); if k == 0 { n } else { 1 }], ..Default::default() }).collect();
assert!(st.value_range().1 >= 2000.0, "{:?}", st.value_range());
}
#[test]
fn box_groups_follow_category_labels() {
let c = Chart::sample(ChartType::BoxWhisker);
+13
View File
@@ -143,9 +143,14 @@ fn group(digits: &str, sep: char) -> String {
out
}
/// Longest number format applied, as in Excel (255 characters); a longer one formats as
/// General, so a chart cannot make every label expand a huge format.
pub const MAX_CODE_LEN: usize = 255;
/// `v` formatted with the Excel number format `code` in language `lang` (axis labels: the text
/// section of a format doesn't apply).
pub fn format(code: &str, v: f64, lang: Option<&str>) -> String {
let code = if code.len() > MAX_CODE_LEN * 4 || code.chars().count() > MAX_CODE_LEN { "General" } else { code };
let (sep, dec) = marks(lang);
if !v.is_finite() {
return String::new();
@@ -299,6 +304,14 @@ mod tests {
const RUB: &str = r#"_-* #\ ##0\ "₽"_-;\-* #\ ##0\ "₽"_-;_-* "-"??\ "₽"_-;_-@_-"#;
#[test]
fn an_overlong_format_is_general() {
let code = "0".repeat(100_000);
assert_eq!(format(&code, 12.5, None), format("General", 12.5, None));
let ok = "0".repeat(MAX_CODE_LEN);
assert_eq!(format(&ok, 7.0, None).len(), MAX_CODE_LEN);
}
#[test]
fn scientific_subnormal() {
assert_eq!(format("0.00E+00", 5e-324, None), "4.94E-324");