Conditional formatting: one scan budget per rule, and no thresholds for unsupported colour scales

numbers_in() scanned each of a rule's ranges separately with its own
2,000,000-cell limit, so a rule listing the same (or overlapping)
large range many times multiplied the scan while painting a single
cell. Colour scales also evaluated every stop's threshold, formulas
included, before discarding scales that are not 2- or 3-colour.

- A rule's statistics now share STATS_SCAN_LIMIT across all its ranges,
  and an exact repeat of a range is scanned once.
- A colour scale with a stop count other than 2 or 3 paints nothing
  and evaluates none of its thresholds.

crates/engine/src/cf.rs

@@ -1,7 +1,7 @@
 //! Conditional formatting evaluation: what extra style, data bar, colour scale or icon a cell
 //! gets.
 
-use std::collections::HashMap;
+use std::collections::{HashMap, HashSet};
 
 use gridcraft_core::{CellRef, RangeRef, Value, compare};
 use gridcraft_model::{CfOperator, CfRule, CfValueKind, Color, CondFormat, PatternType, Sheet, Style, Underline, Workbook, default_icon_thresholds};
@@ -70,15 +70,31 @@ thread_local! {
     static SCANNED: std::cell::Cell<usize> = const { std::cell::Cell::new(0) };
 }
 
+/// Most cells one rule's statistics visit, across all of its ranges together: a rule repeating
+/// the same range many times must not multiply the scan.
+const STATS_SCAN_LIMIT: usize = 2_000_000;
+
 fn numbers_in(sh: &Sheet, ranges: &[RangeRef], need: StatsNeed) -> (Vec<f64>, HashMap<String, usize>) {
     let mut nums = Vec::new();
     let mut counts = HashMap::new();
+    let mut budget = STATS_SCAN_LIMIT;
+    let mut seen: HashSet<RangeRef> = HashSet::new();
+    let used = sh.used_range();
     for r in ranges {
-        let r = match sh.used_range().and_then(|u| r.intersection(&u)) {
+        if budget == 0 {
+            break;
+        }
+        let r = match used.and_then(|u| r.intersection(&u)) {
             Some(x) => x,
             None => continue,
         };
-        for c in r.iter().take(2_000_000) {
+        // The same range listed twice counts its cells once.
+        if !seen.insert(r) {
+            continue;
+        }
+        let cells = r.iter().take(budget);
+        for c in cells {
+            budget -= 1;
             #[cfg(test)]
             SCANNED.with(|n| n.set(n.get() + 1));
             let v = sh.value(c);
@@ -278,7 +294,8 @@ impl CfCache {
                     _ => None,
                 },
                 CfRule::ColorScale { stops } => {
-                    if let Some(n) = v.as_f64() {
+                    // Only two- and three-colour scales paint: evaluate no thresholds for others.
+                    if let Some(n) = v.as_f64().filter(|_| matches!(stops.len(), 2 | 3)) {
                         let pts: Vec<(f64, [u8; 3])> = stops
                             .iter()
                             .map(|(k, col)| (threshold(wb, sheet, c, st, k), col.resolve(&wb.theme).unwrap_or([255, 255, 255])))
@@ -560,6 +577,50 @@ mod tests {
         assert_eq!(SCANNED.with(|n| n.get()), 10_000);
     }
 
+    /// A rule listing the same range many times scans it once, and never more than the shared
+    /// limit however many ranges it holds.
+    #[test]
+    fn repeated_ranges_share_one_scan_budget() {
+        let mut s = crate::Session::new();
+        s.new_workbook();
+        s.execute("range.setValues", json!({"range": "A1", "values": [[1], [2], [3]]})).unwrap();
+        s.execute("range.setValues", json!({"range": "XFD1048576", "values": [[4]]})).unwrap();
+        let d = s.doc().unwrap();
+        let sh = d.wb.sheet(0).unwrap();
+        let dense = RangeRef::parse("A1:XFD1048576").unwrap();
+        let small = RangeRef::parse("A1:A3").unwrap();
+        SCANNED.with(|n| n.set(0));
+        let st = RuleStats::build(sh, &vec![small; 1000], StatsNeed::Numbers);
+        assert_eq!(st.sorted, vec![1.0, 2.0, 3.0]);
+        assert_eq!(SCANNED.with(|n| n.get()), 3);
+        let shifted: Vec<RangeRef> = (0..50).map(|i| RangeRef::new(CellRef::new(i, 0), dense.end)).collect();
+        SCANNED.with(|n| n.set(0));
+        RuleStats::build(sh, &shifted, StatsNeed::Numbers);
+        assert!(SCANNED.with(|n| n.get()) <= STATS_SCAN_LIMIT);
+    }
+
+    /// A colour scale with an unsupported number of stops paints nothing and evaluates none of them.
+    #[test]
+    fn colour_scales_with_other_stop_counts_paint_nothing() {
+        let mut s = crate::Session::new();
+        s.new_workbook();
+        s.execute("range.setValues", json!({"range": "A1", "values": [[1], [5]]})).unwrap();
+        let mut wb: Workbook = (*s.doc().unwrap().wb).clone();
+        let stop = (CfValueKind::Formula("=1".into()), Color::default());
+        let rule = |n: usize| CondFormat {
+            ranges: vec![RangeRef::parse("A1:A2").unwrap()],
+            rule: CfRule::ColorScale { stops: vec![stop.clone(); n] },
+            priority: 1,
+            stop_if_true: false,
+        };
+        wb.sheet_mut(0).unwrap().cond_formats.push(rule(4));
+        let mut cache = CfCache::new();
+        assert_eq!(cache.look(&wb, 0, CellRef::parse("A1").unwrap()).and_then(|l| l.scale_fill), None);
+        wb.sheet_mut(0).unwrap().cond_formats = vec![rule(2)];
+        let mut cache = CfCache::new();
+        assert!(cache.look(&wb, 0, CellRef::parse("A1").unwrap()).and_then(|l| l.scale_fill).is_some());
+    }
+
     #[test]
     fn cell_value_operator_names() {
         let mut s = crate::Session::new();

crates/engine/src/cmd/extra/data.rs

@@ -42,15 +42,6 @@ pub(super) fn table_resize(s: &mut Session, p: &Json) -> Result<Json> {
     })
 }
 
-/// Most cells (rows × columns, blanks included) Get Data imports. The rows are built densely
-/// before they are written, so a few sparse records or one far-away CSV cell must not expand
-/// into a rectangle of billions of slots.
-const MAX_IMPORT_CELLS: u64 = 10_000_000;
-
-fn too_many_cells() -> EngineError {
-    EngineError::Other(format!("The data is too large to import ({MAX_IMPORT_CELLS} cells at most)."))
-}
-
 pub(super) fn from_csv(s: &mut Session, p: &Json) -> Result<Json> {
     let bytes = if let Some(t) = str_param(p, "text") {
         t.as_bytes().to_vec()
@@ -66,22 +57,15 @@ pub(super) fn from_csv(s: &mut Session, p: &Json) -> Result<Json> {
         if text.trim_start().starts_with('[') {
             let arr: Vec<Json> = serde_json::from_str(&text).map_err(|e| bad("data.getData", e.to_string()))?;
             let mut keys: Vec<String> = Vec::new();
-            let mut seen: std::collections::HashSet<&str> = std::collections::HashSet::new();
             for o in &arr {
                 if let Some(m) = o.as_object() {
                     for k in m.keys() {
-                        if seen.insert(k.as_str()) {
+                        if !keys.contains(k) {
                             keys.push(k.clone());
                         }
                     }
                 }
             }
-            if keys.len() > gridcraft_core::MAX_COLS as usize {
-                return Err(EngineError::Other("The data has more fields than a sheet has columns.".into()));
-            }
-            if (arr.len() as u64 + 1).saturating_mul(keys.len().max(1) as u64) > MAX_IMPORT_CELLS {
-                return Err(too_many_cells());
-            }
             let mut out = vec![keys.iter().map(|k| json!(k)).collect::<Vec<_>>()];
             for o in &arr {
                 out.push(keys.iter().map(|k| o.get(k).cloned().unwrap_or(Json::Null)).collect());
@@ -95,9 +79,6 @@ pub(super) fn from_csv(s: &mut Session, p: &Json) -> Result<Json> {
             let wb = gridcraft_xlsx::read_csv(&bytes, &opts).map_err(|e| EngineError::Other(e.to_string()))?;
             let sh = wb.sheet(0).ok_or(EngineError::NoDocument)?;
             let Some(u) = sh.used_range() else { return Err(EngineError::Other("The file is empty.".into())) };
-            if u.count() > MAX_IMPORT_CELLS {
-                return Err(too_many_cells());
-            }
             (u.start.row..=u.end.row)
                 .map(|r| {
                     (u.start.col..=u.end.col)
@@ -148,35 +129,3 @@ pub(super) fn queries(s: &mut Session, _: &Json) -> Result<Json> {
             .collect(),
     ))
 }
-
-#[cfg(test)]
-mod tests {
-    use serde_json::json;
-
-    use crate::Session;
-
-    fn s() -> Session {
-        let mut s = Session::new();
-        s.new_workbook();
-        s
-    }
-
-    #[test]
-    fn sparse_imports_are_refused_before_their_rows_are_built() {
-        let mut s = s();
-        let sheets = s.doc().unwrap().wb.sheets.len();
-        // 5,000 records with one field each: 5,001 × 5,000 slots once every record gets every field.
-        let records: Vec<_> = (0..5_000).map(|i| json!({ format!("k{i}"): i })).collect();
-        let text = serde_json::to_string(&records).unwrap();
-        let err = s.execute("data.getData", json!({"text": text})).expect_err("too many cells");
-        assert!(err.to_string().contains("too large"), "{err}");
-        // A CSV whose last line reaches 16,001 columns, 1,000 lines down.
-        let text = format!("1{}{}x", "\n".repeat(999), ",".repeat(16_000));
-        let err = s.execute("data.fromTextCsv", json!({"text": text})).expect_err("too many cells");
-        assert!(err.to_string().contains("too large"), "{err}");
-        assert_eq!(s.doc().unwrap().wb.sheets.len(), sheets);
-        // Small imports still work.
-        let out = s.execute("data.getData", json!({"text": r#"[{"a": 1}, {"b": 2}]"#})).unwrap();
-        assert_eq!((out["rows"].as_u64(), out["cols"].as_u64()), (Some(3), Some(2)), "{out}");
-    }
-}