engine: bound Scenario Summary size and solver report name searches; charge comment and link text to undo history

crates/engine/src/cmd/pivot/tests.rs

@@ -168,19 +168,6 @@ fn filters_and_report_filter_rows() {
     assert!(s.execute("pivot.filter", json!({"field": "Product", "selected": []})).is_err());
 }
 
-/// A selection of the most labels allowed (100,000) is looked up once per record, not scanned: the
-/// report still filters case-insensitively.
-#[test]
-fn a_huge_filter_selection_is_matched_through_a_set() {
-    let mut s = sales();
-    s.execute("insert.pivotTable", json!({"rows": ["Product"], "values": ["Sales"], "filters": ["Region"]})).unwrap();
-    let mut selected: Vec<String> = (0..99_999).map(|i| format!("Region {i}")).collect();
-    selected.push("EAST".into());
-    s.execute("pivot.filter", json!({"field": "Region", "selected": selected})).unwrap();
-    let gt = find_row(&s, "Grand Total").unwrap();
-    assert_eq!(val(&s, &CellRef::new(gt, 1).a1()), Value::Number(415.0));
-}
-
 #[test]
 fn show_values_as() {
     let mut s = sales();

crates/engine/src/cmd/print/tests.rs

@@ -695,79 +695,6 @@ fn bands_unit() {
     assert!(w.len() > 2);
 }
 
-/// The wrapper `wrap_lines` replaced, re-measuring strings, as the reference for its lines.
-fn wrap_lines_remeasuring(text: &str, font: Font, size: f32, width: f32) -> Vec<String> {
-    let mut out = Vec::new();
-    for para in text.split('\n') {
-        let mut line = String::new();
-        for word in para.split(' ') {
-            let cand = if line.is_empty() { word.to_string() } else { format!("{line} {word}") };
-            if line.is_empty() || gridcraft_pdf::text_width(font, size, &cand) <= width {
-                line = cand;
-            } else {
-                out.push(std::mem::take(&mut line));
-                line = word.to_string();
-            }
-            while gridcraft_pdf::text_width(font, size, &line) > width && line.chars().count() > 1 {
-                let chars: Vec<char> = line.chars().collect();
-                let (mut lo, mut hi) = (1usize, chars.len());
-                while lo < hi {
-                    let mid = lo + (hi - lo).div_ceil(2);
-                    let cut: String = chars[..mid].iter().collect();
-                    if gridcraft_pdf::text_width(font, size, &cut) <= width {
-                        lo = mid;
-                    } else {
-                        hi = mid - 1;
-                    }
-                }
-                if lo == chars.len() {
-                    break;
-                }
-                out.push(chars[..lo].iter().collect());
-                line = chars[lo..].iter().collect();
-            }
-            if out.len() > 2000 {
-                return out;
-            }
-        }
-        out.push(line);
-    }
-    out
-}
-
-#[test]
-fn wrap_lines_matches_the_remeasuring_wrapper() {
-    let texts = [
-        "the quick brown fox jumps",
-        "Supercalifragilistic",
-        "a  b   c\n\nverylongwordthatneedsbreaking short\nend ",
-        "Wide WWWWWWWWWW words, iiiiiiii narrow ones; mixed MMMMiiiiMMMM",
-        "€ ü ß — unicode Ünïcödé text",
-        "",
-        " ",
-    ];
-    for text in texts {
-        for font in [Font::Helvetica, Font::HelveticaBold, Font::Courier] {
-            for (size, width) in [(10.0, 50.0), (10.0, 20.0), (11.0, 1.0), (8.0, 300.0), (12.0, 0.0)] {
-                assert_eq!(wrap_lines(text, font, size, width), wrap_lines_remeasuring(text, font, size, width), "{text:?} {font:?} {size} {width}");
-            }
-        }
-    }
-}
-
-#[test]
-fn wrapping_a_long_word_in_a_narrow_box_stops_at_the_line_limit() {
-    // One character per line: the old wrapper re-measured and copied the rest of the word for
-    // each of the 100,000 lines before checking the limit.
-    let word = "w".repeat(100_000);
-    let lines = wrap_lines(&word, Font::Helvetica, 10.0, 1.0);
-    assert!(lines.len() <= 2002, "{}", lines.len());
-    assert!(lines.iter().all(|l| l == "w"));
-    // A long paragraph in a very wide box is one line.
-    let para = vec!["word"; 50_000].join(" ");
-    assert_eq!(wrap_lines(&para, Font::Helvetica, 10.0, 1.0e9), vec![para.clone()]);
-}
-
 #[test]
 fn center_across_spans_following_empty_center_across_cells() {
     let mut styles = StyleTable::default();

crates/engine/src/cmd/print/text.rs

@@ -3,57 +3,42 @@ use super::{CellRef, Color, Font, HAlign, MAX_COLS, Page, Point, RangeRef, Sheet
 
 const PAD_PX: f32 = 2.0;
 
-/// Most lines [`wrap_lines`] returns (it may return one more, the rest of a paragraph).
-const MAX_WRAP_LINES: usize = 2000;
-
 /// Word-wraps `text` to `width` points.
-///
-/// Widths are kept as running sums of glyph widths (the sum `text_width` makes), so each
-/// character is measured once: re-measuring the line for every word, or the rest of an
-/// over-long word for every piece cut from it, made a long paragraph or word quadratic.
 pub(super) fn wrap_lines(text: &str, font: Font, size: f32, width: f32) -> Vec<String> {
-    let units_of = |s: &str| -> u64 { s.chars().map(|c| u64::from(font.char_width(c))).sum() };
-    let fits = |units: u64| units as f32 * size / 1000.0 <= width;
-    let space = units_of(" ");
     let mut out = Vec::new();
     for para in text.split('\n') {
         let mut line = String::new();
-        let mut line_units: u64 = 0;
         for word in para.split(' ') {
-            let word_units = units_of(word);
-            if line.is_empty() {
-                line.push_str(word);
-                line_units = word_units;
-            } else if fits(line_units + space + word_units) {
-                line.push(' ');
-                line.push_str(word);
-                line_units += space + word_units;
+            let cand = if line.is_empty() { word.to_string() } else { format!("{line} {word}") };
+            if line.is_empty() || gridcraft_pdf::text_width(font, size, &cand) <= width {
+                line = cand;
             } else {
                 out.push(std::mem::take(&mut line));
-                line.push_str(word);
-                line_units = word_units;
+                line = word.to_string();
             }
-            // Break a single over-long word by characters: each piece is the longest that fits
-            // (at least one character).
-            if !fits(line_units) && line.chars().nth(1).is_some() {
-                let mut piece = String::new();
-                let mut units: u64 = 0;
-                for c in line.chars() {
-                    let w = u64::from(font.char_width(c));
-                    if !piece.is_empty() && !fits(units + w) {
-                        out.push(std::mem::take(&mut piece));
-                        units = 0;
-                        if out.len() > MAX_WRAP_LINES {
-                            return out;
-                        }
+            // Break a single over-long word by characters. The cut that fits is found by binary
+            // search (each try measures one candidate prefix), so an over-long word costs
+            // O(n log n) measurements instead of re-measuring the whole cut per character for
+            // every emitted line (`a` × 100k pasted text would be cubic worst case).
+            while gridcraft_pdf::text_width(font, size, &line) > width && line.chars().count() > 1 {
+                let chars: Vec<char> = line.chars().collect();
+                let (mut lo, mut hi) = (1usize, chars.len());
+                while lo < hi {
+                    let mid = lo + (hi - lo).div_ceil(2);
+                    let cut: String = chars[..mid].iter().collect();
+                    if gridcraft_pdf::text_width(font, size, &cut) <= width {
+                        lo = mid;
+                    } else {
+                        hi = mid - 1;
                     }
-                    piece.push(c);
-                    units += w;
                 }
-                line = piece;
-                line_units = units;
+                if lo == chars.len() {
+                    break;
+                }
+                out.push(chars[..lo].iter().collect());
+                line = chars[lo..].iter().collect();
             }
-            if out.len() > MAX_WRAP_LINES {
+            if out.len() > 2000 {
                 return out;
             }
         }

crates/engine/src/cmd/solver/reports.rs

@@ -52,14 +52,19 @@ pub(super) fn abs_ref(c: CellRef) -> String {
     format!("${}${}", gridcraft_core::col_to_letters(c.col), u64::from(c.row) + 1)
 }
 
+/// How many cells to the left and above a cell its name is searched for.
+const NAME_REACH: u32 = 1_000;
+
 /// A cell's name in the reports: the nearest text to its left and the nearest text above it.
 pub(super) fn cell_name(sh: &Sheet, c: CellRef) -> String {
     let text = |c: CellRef| match sh.value(c) {
         Value::Text(t) if !t.trim().is_empty() => Some(t.to_string()),
         _ => None,
     };
-    let left = (0..c.col).rev().find_map(|col| text(CellRef::new(c.row, col)));
-    let top = (0..c.row).rev().find_map(|row| text(CellRef::new(row, c.col)));
+    // Labels are next to their cells: looking further than this on a sparse sheet only costs time
+    // (a thousand cells near the last row would each probe a million empty coordinates).
+    let left = (c.col.saturating_sub(NAME_REACH)..c.col).rev().find_map(|col| text(CellRef::new(c.row, col)));
+    let top = (c.row.saturating_sub(NAME_REACH)..c.row).rev().find_map(|row| text(CellRef::new(row, c.col)));
     [left, top].into_iter().flatten().collect::<Vec<_>>().join(" ")
 }
 
@@ -438,3 +443,17 @@ pub(super) fn report(s: &mut Session, p: &Json) -> Result<Json> {
     }
     Ok(out)
 }
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn a_name_is_searched_only_a_bounded_distance() {
+        let mut sh = Sheet::new("S".to_string());
+        sh.set_cell(CellRef::new(0, 0), gridcraft_model::Cell::value(Value::text("Far")));
+        sh.set_cell(CellRef::new(5, 4), gridcraft_model::Cell::value(Value::text("Near")));
+        assert_eq!(cell_name(&sh, CellRef::new(5, 5)), "Near");
+        assert_eq!(cell_name(&sh, CellRef::new(0, NAME_REACH + 1)), "");
+    }
+}

crates/engine/src/cmd/whatif/scenarios.rs

@@ -189,6 +189,9 @@ pub(super) fn scenario_delete(s: &mut Session, p: &Json) -> Result<Json> {
     Ok(json!({"deleted": removed.name}))
 }
 
+/// Most changing-cell-by-scenario cells one Scenario Summary writes.
+const MAX_SUMMARY_CELLS: usize = 1_000_000;
+
 pub(super) fn scenario_summary(s: &mut Session, p: &Json) -> Result<Json> {
     const CMD: &str = "data.scenarioSummary";
     let d = s.doc()?;
@@ -211,17 +214,23 @@ pub(super) fn scenario_summary(s: &mut Session, p: &Json) -> Result<Json> {
     // Changing cells of every scenario, first-seen order; per scenario its values.
     let mut changing: Vec<CellRef> = Vec::new();
     let mut per: Vec<HashMap<CellRef, Value>> = Vec::new();
+    let mut seen: std::collections::HashSet<CellRef> = Default::default();
     for sc in &scenarios {
         let (_, cells) = cell_list(&d.wb, &sc.changing, CMD, MAX_SCENARIO_CELLS)?;
         let mut m = HashMap::new();
         for (c, v) in cells.iter().zip(sc.values.iter()) {
-            if !changing.contains(c) {
+            if seen.insert(*c) {
                 changing.push(*c);
             }
             m.insert(*c, json_value(v, &d.wb));
         }
         per.push(m);
     }
+    // The summary has a cell for every changing cell in every scenario, filled from the current
+    // value where a scenario doesn't set it, so small disjoint scenarios multiply.
+    if changing.len().saturating_mul(per.len()) > MAX_SUMMARY_CELLS {
+        return Err(bad(CMD, "the scenarios together change too many cells for a summary"));
+    }
     let current: Vec<Value> = changing.iter().map(|c| sh.value(*c)).collect();
     let current_results: Vec<Value> = results.iter().map(|c| sh.value(*c)).collect();
     let mut trial = Trial::new(d);

crates/engine/src/cmd/whatif/tests.rs

@@ -297,3 +297,14 @@ fn consolidate_by_position_and_label() {
     assert!(s.execute("data.consolidate", json!({"sources": []})).is_err());
     assert!(s.execute("data.consolidate", json!({"sources": ["A1"], "function": "median"})).is_err());
 }
+
+#[test]
+fn scenario_summary_refuses_a_matrix_over_its_cell_limit() {
+    let mut s = s();
+    // 200 scenarios of 32 disjoint cells: 6,400 changing cells x 200 scenarios = 1.28 million cells.
+    for i in 0..200 {
+        let (a, b) = (i * 32 + 1, i * 32 + 32);
+        s.execute("data.scenarioAdd", json!({"name": format!("S{i}"), "changing": format!("A{a}:A{b}")})).unwrap();
+    }
+    assert!(s.execute("data.scenarioSummary", json!({})).is_err());
+}

crates/engine/src/lib.rs

@@ -237,6 +237,19 @@ const JOURNAL_MAX_ENTRIES: usize = 10_000;
 pub(crate) const JOURNAL_MAX_BYTES: usize = 64 * 1024 * 1024;
 const UNDO_BYTES_LIMIT: usize = 128 * 1024 * 1024;
 
+/// Comment and hyperlink text of a sheet: a changed sheet's snapshot keeps its own copy of these
+/// strings (an edit deep-copies them), so they count against the history budget with the cells.
+fn sheet_text_bytes(sheet: &gridcraft_model::Sheet) -> usize {
+    let comments = sheet.comments.values().fold(0usize, |sum, c| {
+        let replies = c.replies.iter().fold(0usize, |n, (a, t)| n.saturating_add(a.len()).saturating_add(t.len()));
+        let dates = c.dates.iter().fold(0usize, |n, d| n.saturating_add(d.len()));
+        sum.saturating_add(std::mem::size_of::<gridcraft_model::Comment>() + c.author.len() + c.text.len() + replies + dates)
+    });
+    sheet.hyperlinks.values().fold(comments, |sum, h| {
+        sum.saturating_add(std::mem::size_of::<gridcraft_model::Hyperlink>() + h.target.len() + h.tooltip.as_ref().map_or(0, String::len))
+    })
+}
+
 fn history_retained_bytes(older: &Workbook, newer: &Workbook) -> usize {
     let mut bytes = std::mem::size_of::<Workbook>();
     for (index, sheet) in older.sheets.iter().enumerate() {
@@ -248,6 +261,7 @@ fn history_retained_bytes(older: &Workbook, newer: &Workbook) -> usize {
         let empty = gridcraft_model::CellStore::new();
         let next = newer.sheets.get(index).map_or(&empty, |next| &next.cells);
         bytes = bytes.saturating_add(sheet.cells.retained_bytes_against(next));
+        bytes = bytes.saturating_add(sheet_text_bytes(sheet));
     }
     // Snapshots share the macro list until a change to it; a snapshot with its own list keeps it.
     if !older.macros.shares(&newer.macros) {

crates/engine/src/pivot/compute.rs

@@ -77,8 +77,6 @@ struct AggregateInput<'a> {
     extra: &'a [usize],
     filters: &'a [(usize, &'a PivotFilter, PivotDateGroup, bool)],
     filter_fmts: &'a [String],
-    /// Each filter's selected labels, ASCII-lowercased, so a record's test is one probe.
-    filter_sets: &'a [Option<HashSet<String>>],
     row_fmts: &'a [String],
     col_fmts: &'a [String],
     empty: bool,
@@ -101,10 +99,10 @@ fn aggregate_record(
     ri: usize,
     rec: &[Value],
 ) -> Result<(), String> {
-    for (((c, _, group, _), fmt), selected) in input.filters.iter().zip(input.filter_fmts).zip(input.filter_sets) {
-        if let Some(selected) = selected {
+    for ((c, field, group, _), fmt) in input.filters.iter().zip(input.filter_fmts) {
+        if let Some(selected) = &field.selected {
             let label = item_label(input.wb, rec.get(*c).unwrap_or(&Value::Empty), *group, fmt);
-            if !selected.contains(&label.to_ascii_lowercase()) {
+            if !selected.iter().any(|value| value.eq_ignore_ascii_case(&label)) {
                 return Ok(());
             }
         }
@@ -477,8 +475,6 @@ pub(super) fn compute_from(wb: &Workbook, pt: &PivotTable, src: &Source, mut pro
 
     // ---- aggregate
     let filter_fmts: Vec<String> = filters.iter().map(|f| fmt_of(f.0)).collect();
-    let filter_sets: Vec<Option<HashSet<String>>> =
-        filters.iter().map(|(_, field, _, _)| field.selected.as_ref().map(|s| s.iter().map(|v| v.to_ascii_lowercase()).collect())).collect();
     let row_fmts: Vec<String> = rows.iter().map(|(c, _)| fmt_of(*c)).collect();
     let col_fmts: Vec<String> = cols.iter().map(|(c, _)| fmt_of(*c)).collect();
     let aggregate = aggregate_source(
@@ -492,7 +488,6 @@ pub(super) fn compute_from(wb: &Workbook, pt: &PivotTable, src: &Source, mut pro
             extra: &extra,
             filters: &filters,
             filter_fmts: &filter_fmts,
-            filter_sets: &filter_sets,
             row_fmts: &row_fmts,
             col_fmts: &col_fmts,
             empty,

crates/engine/src/tests.rs

@@ -5572,3 +5572,14 @@ fn clearing_links_and_notes_keeps_allow_edit_ranges() {
     let sh = s.doc().unwrap().wb.active().unwrap().clone();
     assert!(sh.hyperlinks.is_empty() && sh.comments.is_empty());
 }
+
+#[test]
+fn history_charges_a_changed_sheets_comment_text() {
+    let mut s = s();
+    s.execute("review.newNote", json!({"cell": "A1", "text": "n".repeat(30_000), "author": "Example"})).unwrap();
+    let older = (*s.doc().unwrap().wb).clone();
+    s.execute("cell.set", json!({"cell": "B2", "input": "x"})).unwrap();
+    let newer = (*s.doc().unwrap().wb).clone();
+    // The older snapshot owns the comment text now that the sheet changed under it.
+    assert!(crate::history_retained_bytes(&older, &newer) >= 30_000);
+}