xlsx: one hyperlink allowance per sheet, and a copy limit for unparsable shared masters

Every <hyperlinks> element of a worksheet started a fresh cell and byte
allowance although all of them fill the same map. The sheet now keeps
one allowance across its elements.

Followers of a shared master the parser refuses each copied its full
text; a sheet full of followers of a long master multiplied that text
without bound. The copies are now charged against a 64 MiB per-sheet
limit, past which the followers keep their cached values.

apps/gridcraft-cli/src/check/calc.rs

@@ -31,41 +31,20 @@ fn is_volatile(e: &Expr, volatile_names: &HashSet<String>) -> bool {
 }
 
 /// Upper-cased defined names whose formula is volatile, directly or through other names.
-///
-/// Volatility flows backwards along name references: each name is parsed and walked once, and a
-/// worklist carries volatility from a name to the names that use it. Rescanning every name until
-/// nothing changes was quadratic: a file listing its names so that each pass marked only one more
-/// did one full pass per name.
 pub(super) fn volatile_names(wb: &Workbook) -> HashSet<String> {
+    let parsed: Vec<(String, Option<Expr>)> =
+        wb.names.iter().map(|n| (n.name.to_ascii_uppercase(), gridcraft_engine::formula::parse(&n.formula).ok())).collect();
     let mut set = HashSet::new();
-    // Upper-cased name -> the names whose formulas refer to it.
-    let mut users: HashMap<String, Vec<String>> = HashMap::new();
-    let mut queue = Vec::new();
-    let none = HashSet::new();
-    for n in &wb.names {
-        let name = n.name.to_ascii_uppercase();
-        let Ok(e) = gridcraft_engine::formula::parse(&n.formula) else { continue };
-        if is_volatile(&e, &none) {
-            if set.insert(name.clone()) {
-                queue.push(name.clone());
+    // Each round adds at least one name or stops, so `len + 1` rounds reach the fixed point.
+    for _ in 0..=parsed.len() {
+        let before = set.len();
+        for (name, e) in &parsed {
+            if !set.contains(name) && e.as_ref().is_some_and(|e| is_volatile(e, &set)) {
+                set.insert(name.clone());
             }
-            continue;
         }
-        let mut used = BTreeSet::new();
-        e.walk(&mut |x| {
-            if let Expr::Name(r) = x {
-                used.insert(r.to_ascii_uppercase());
-            }
-        });
-        for r in used {
-            users.entry(r).or_default().push(name.clone());
-        }
-    }
-    while let Some(v) = queue.pop() {
-        for user in users.remove(&v).unwrap_or_default() {
-            if set.insert(user.clone()) {
-                queue.push(user);
-            }
+        if set.len() == before {
+            break;
         }
     }
     set

apps/gridcraft-cli/src/check/tests.rs

@@ -178,28 +178,6 @@ fn volatile_names_propagate() {
     assert_eq!(check(&wb, &Options::default()).calc.volatile, 1);
 }
 
-#[test]
-fn volatile_names_propagate_through_a_long_chain_in_one_walk() {
-    // Chain0 uses Chain1, Chain1 uses Chain2, ... and only the last is volatile: listed in this order, a rescan
-    // until nothing changes marked one name per pass (quadratic). Every name in the chain is
-    // volatile; a name off the chain is not.
-    const LEN: usize = 5_000;
-    let mut wb = book(&[("A1", "=1", n(1.0))]);
-    let name =
-        |name: String, formula: String| gridcraft_engine::model::DefinedName { name, scope: None, formula, comment: String::new(), hidden: false };
-    for i in 0..LEN {
-        let formula = if i + 1 == LEN { "NOW()".to_string() } else { format!("Chain{}+1", i + 1) };
-        wb.names.push(name(format!("Chain{i}"), formula));
-    }
-    wb.names.push(name("Plain".into(), "Chain0*0+Other".into()));
-    wb.names.push(name("Other".into(), "1".into()));
-    let set = volatile_names(&wb);
-    assert!((0..LEN).all(|i| set.contains(&format!("CHAIN{i}"))));
-    assert!(set.contains("PLAIN"));
-    assert!(!set.contains("OTHER"));
-    assert_eq!(set.len(), LEN + 1);
-}
-
 #[test]
 fn round_trip_of_a_simple_book_is_clean() {
     let mut wb = book(&[("A1", "2", n(2.0)), ("B1", "=A1*3", n(6.0)), ("A2", "x", Value::text("x"))]);

crates/engine/src/cmd.rs

@@ -459,11 +459,8 @@ pub(crate) const TOUCH_LOG_LIMIT: usize = 1_000_000;
 pub(crate) const PROTECTED: &str = "The cell or chart you're trying to change is on a protected sheet.";
 
 /// Refuses an edit that changed what a locked cell on a protected sheet holds (its value or
-/// formula), or that unlocked it, whichever command made it. Sheets are matched by name, so
-/// moving, renaming or deleting a sheet isn't an edit of its cells. A sort or duplicate removal
-/// that moves an unlocked cell with equal content onto a locked one changes no content but would
-/// leave the cell editable, so the effective lock (row and column formats included) is compared
-/// as well.
+/// formula), whichever command made it. Sheets are matched by name, so moving, renaming or
+/// deleting a sheet isn't an edit of its cells.
 fn check_protection(old: &Workbook, new: &Workbook, unlocked: &[(String, String)]) -> Result<()> {
     for sh in &old.sheets {
         if !sh.is_protected() {
@@ -474,8 +471,7 @@ fn check_protection(old: &Workbook, new: &Workbook, unlocked: &[(String, String)
             continue;
         }
         for c in sh.cells.diff(&after.cells) {
-            let unlocked_now = !new.styles.get(after.style_id(c)).protection.locked;
-            if (unlocked_now || !same_content(sh.cell(c), after.cell(c)))
+            if !same_content(sh.cell(c), after.cell(c))
                 && old.styles.get(sh.style_id(c)).protection.locked
                 && let Some(m) = locked_cell_error(sh, c, unlocked)
             {

crates/engine/src/cmd/sheet.rs

@@ -398,10 +398,10 @@ fn move_columns(s: &mut Session, p: &Json) -> Result<Json> {
     {
         return Err(EngineError::Other("To prevent possible loss of data, columns cannot be shifted off of the worksheet.".into()));
     }
-    // The checks must authorize the sheet this command modifies, not whatever sheet is active:
-    // both get the index `target_sheet` resolved (from `sheet`, or a `Sheet!` range or cell).
-    check_lines_protection(s, &json!({"cols": RangeRef::cols(before, before).a1(), "sheet": sheet}), Axis::Cols, before, count, true)?;
-    check_lines_protection(s, &json!({"cols": RangeRef::cols(from, end - 1).a1(), "sheet": sheet}), Axis::Cols, from, count, false)?;
+    // The checks must authorize the sheet this command modifies (`target_sheet`), not whatever
+    // sheet is active: both carry the command's own `sheet` parameter on.
+    check_lines_protection(s, &json!({"cols": RangeRef::cols(before, before).a1(), "sheet": p.get("sheet")}), Axis::Cols, before, count, true)?;
+    check_lines_protection(s, &json!({"cols": RangeRef::cols(from, end - 1).a1(), "sheet": p.get("sheet")}), Axis::Cols, from, count, false)?;
 
     let moved_from = if before <= from { from + count } else { from };
     if moved_from.checked_add(count).is_none_or(|end| end > MAX_COLS) {

crates/engine/src/tests.rs

@@ -5297,42 +5297,6 @@ fn move_columns_checks_the_protection_of_the_named_sheet() {
     assert_eq!(v(&s, "A1"), Value::text("A"));
 }
 
-#[test]
-fn move_columns_checks_the_protection_of_a_sheet_named_by_its_range() {
-    // A `Sheet!` range or cell picks the sheet the move modifies; the protection checks must
-    // authorize that sheet too, not the active one.
-    let mut s = s();
-    s.execute("home.insertSheet", json!({"name": "P"})).unwrap();
-    s.execute("range.setValues", json!({"range": "A1", "values": [[1, 2, 3, 4]]})).unwrap();
-    s.execute("review.protectSheet", json!({})).unwrap();
-    s.execute("sheet.activate", json!({"sheet": 0})).unwrap();
-    for key in ["range", "cell"] {
-        let err = s.execute("sheet.moveColumns", json!({key: "P!A1", "cols": "B:C", "before": 4})).expect_err("moving columns on a protected sheet");
-        assert!(err.to_string().contains("protected"), "{err}");
-    }
-    let d = s.doc().unwrap();
-    let sh = &d.wb.sheets[d.wb.sheet_index("P").unwrap()];
-    let pv: Vec<Value> = (0..4).map(|c| sh.value(CellRef::new(0, c))).collect();
-    assert_eq!(pv, vec![Value::number(1.0), Value::number(2.0), Value::number(3.0), Value::number(4.0)]);
-}
-
-#[test]
-fn sorting_cannot_unlock_a_locked_cell_by_moving_equal_content_onto_it() {
-    // An unlocked 2 sorted onto a locked 2 leaves the content as it was, but would make the cell
-    // editable; the sort is refused instead.
-    let mut s = s();
-    s.execute("range.setValues", json!({"range": "A1", "values": [[2], [2], [1]]})).unwrap();
-    s.execute("home.lockCell", json!({"range": "A1", "on": false})).unwrap();
-    s.execute("home.lockCell", json!({"range": "A3", "on": false})).unwrap();
-    s.execute("review.protectSheet", json!({"sort": true})).unwrap();
-    let before = s.doc().unwrap().wb.clone();
-    let err =
-        s.execute("data.sort", json!({"range": "A1:A3", "header": false, "keys": [{"column": "A", "order": "asc"}]})).expect_err("sort unlocking A2");
-    assert!(err.to_string().contains("protected sheet"), "{err}");
-    assert_eq!(*s.doc().unwrap().wb, *before);
-    assert!(s.execute("cell.set", json!({"cell": "A2", "input": "9"})).is_err(), "A2 stays locked");
-}
-
 #[test]
 fn a_script_cannot_switch_windows() {
     // Stored scripts are scoped to their workbook: a step that activates another window would

crates/xlsx/src/sheet_read.rs

@@ -23,8 +23,8 @@ use crate::xml::{self, El};
 mod parts;
 mod rules;
 
+use parts::{HyperlinkUse, read_comments, read_hyperlinks, read_note_shapes, read_threaded};
 pub use parts::{read_autofilter, read_table};
-use parts::{read_comments, read_hyperlinks, read_note_shapes, read_threaded};
 use rules::{read_cf, read_validation};
 
 struct Shared {
@@ -44,6 +44,9 @@ struct RowState {
     dynamic_cells: u64,
     /// Row-membership checks served so far, for the work cap above.
     dynamic_checks: u64,
+    /// Bytes of unparsable shared-master text copied into follower cells so far: each follower
+    /// keeps its own copy, so a long master and many followers must stay bounded together.
+    shared_fallback_bytes: u64,
     dropped: u64,
     rows: u64,
 }
@@ -138,6 +141,9 @@ const MAX_DYNAMIC_CHECKS: u64 = 16_000_000;
 /// Most conditional-format range cells one sheet keeps, over all rules of an element: every
 /// rule copies its element's whole range list, so rules × ranges must stay bounded.
 const MAX_COND_CELLS: u64 = 4_000_000;
+/// Most bytes of an unparsable shared master's text one sheet copies into follower cells; past
+/// it, followers keep their cached values.
+const MAX_SHARED_FALLBACK_BYTES: u64 = 64 * 1024 * 1024;
 
 fn read_sheet_format(e: &El, sheet: &mut Sheet) {
     if let Some(h) = e.attr_f64("defaultRowHeight") {
@@ -181,6 +187,7 @@ pub fn read_sheet(cx: &mut Ctx<'_>, part: &str, name: &str) -> Result<Sheet, IoE
 
     let mut fit_to_page = false;
     let mut column_assignments_left = MAX_COLUMN_ASSIGNMENTS;
+    let mut hyperlinks_used = HyperlinkUse::default();
     cx.cf_ext_ids.clear();
     for e in &root.children {
         match e.name.as_str() {
@@ -214,7 +221,7 @@ pub fn read_sheet(cx: &mut Ctx<'_>, part: &str, name: &str) -> Result<Sheet, IoE
                     }
                 }
             }
-            "hyperlinks" => read_hyperlinks(cx, name, e, &rels, &mut sheet),
+            "hyperlinks" => read_hyperlinks(cx, name, e, &rels, &mut sheet, &mut hyperlinks_used),
             "autoFilter" => sheet.autofilter = read_autofilter(cx, e),
             "printOptions" => {
                 sheet.print.gridlines = e.flag("gridLines", false);
@@ -635,14 +642,22 @@ fn read_formula_text(cx: &mut Ctx<'_>, st: &mut RowState, f: &El, pos: CellRef,
                     cx.warn("a shared formula refers to a missing master cell; kept the cached value");
                     return None;
                 };
-                Some(match &sh.expr {
+                match &sh.expr {
                     Some(e) => {
                         let dr = pos.row as i64 - sh.anchor.row as i64;
                         let dc = pos.col as i64 - sh.anchor.col as i64;
-                        Formula::from_expr(gridcraft_formula::adjust::shift_relative(e.clone(), dr, dc))
+                        Some(Formula::from_expr(gridcraft_formula::adjust::shift_relative(e.clone(), dr, dc)))
                     }
-                    None => Formula::new(&sh.text),
-                })
+                    None => {
+                        let total = st.shared_fallback_bytes.saturating_add(sh.text.len() as u64);
+                        if total > MAX_SHARED_FALLBACK_BYTES {
+                            cx.warn("some shared formulas could not be parsed and exceed the copy limit; those cells kept their cached values");
+                            return None;
+                        }
+                        st.shared_fallback_bytes = total;
+                        Some(Formula::new(&sh.text))
+                    }
+                }
             }
         }
         "array" => {

crates/xlsx/src/sheet_read/parts.rs

@@ -3,9 +3,16 @@ use super::{
     NoteBox, NoteCorner, RangeRef, Rel, Sheet, Table, TableColumn, TotalsFn, XmlColumnPr, from_file, xml,
 };
 
-pub(super) fn read_hyperlinks(cx: &mut Ctx<'_>, name: &str, e: &El, rels: &[Rel], sheet: &mut Sheet) {
-    let mut cells = 0u64;
-    let mut bytes = 0u64;
+/// Hyperlink cells and copied bytes one sheet has kept so far: every `<hyperlinks>` element of
+/// the sheet writes into the same map, so the allowance is the sheet's, not the element's.
+#[derive(Default)]
+pub(super) struct HyperlinkUse {
+    cells: u64,
+    bytes: u64,
+}
+
+pub(super) fn read_hyperlinks(cx: &mut Ctx<'_>, name: &str, e: &El, rels: &[Rel], sheet: &mut Sheet, used: &mut HyperlinkUse) {
+    let HyperlinkUse { cells, bytes } = used;
     for h in e.kids("hyperlink") {
         let Some(range) = h.attr("ref").and_then(RangeRef::parse) else { continue };
         let mut target = h.attr("id").and_then(|id| rels.iter().find(|r| r.id == id)).map(|r| r.target.clone()).unwrap_or_default();
@@ -24,7 +31,7 @@ pub(super) fn read_hyperlinks(cx: &mut Ctx<'_>, name: &str, e: &El, rels: &[Rel]
         let link = Hyperlink { target: target.clone(), tooltip: h.attr("tooltip").map(str::to_string), external };
         let link_bytes = (target.len() + link.tooltip.as_ref().map_or(0, |t| t.len())) as u64;
         for c in range.iter() {
-            if cells >= MAX_HYPERLINK_CELLS {
+            if *cells >= MAX_HYPERLINK_CELLS {
                 cx.warn(format!("sheet \"{name}\": hyperlinks past the first {MAX_HYPERLINK_CELLS} cells were dropped"));
                 return;
             }
@@ -32,8 +39,8 @@ pub(super) fn read_hyperlinks(cx: &mut Ctx<'_>, name: &str, e: &El, rels: &[Rel]
                 cx.warn(format!("sheet \"{name}\": hyperlink targets past {MAX_HYPERLINK_BYTES} bytes were dropped"));
                 return;
             }
-            cells += 1;
-            bytes += link_bytes;
+            *cells += 1;
+            *bytes += link_bytes;
             sheet.hyperlinks.insert(c, link.clone());
         }
     }

crates/xlsx/src/tests/malformed.rs

@@ -244,6 +244,11 @@ fn hyperlinks_past_the_cell_cap_are_reported() {
     let (wb, rep) = read_xlsx(&minimal(body, &parts, rels, "")).unwrap();
     assert_eq!(wb.sheet(0).unwrap().hyperlinks.len(), 10_000);
     assert!(rep.warnings.iter().any(|w| w.contains("hyperlinks past the first 10000 cells were dropped")), "{:?}", rep.warnings);
+    // Several <hyperlinks> elements share the sheet's allowance.
+    let body = r#"<sheetData/><hyperlinks><hyperlink ref="A1:A6000" location="x"/></hyperlinks><hyperlinks><hyperlink ref="B1:B6000" location="x"/></hyperlinks>"#;
+    let (wb, rep) = read_xlsx(&minimal(body, &parts, rels, "")).unwrap();
+    assert_eq!(wb.sheet(0).unwrap().hyperlinks.len(), 10_000);
+    assert!(rep.warnings.iter().any(|w| w.contains("hyperlinks past the first 10000 cells were dropped")), "{:?}", rep.warnings);
     // A small list reads in full with no warning.
     let body = r#"<sheetData/><hyperlinks><hyperlink ref="A1:A3" location="x"/></hyperlinks>"#;
     let (wb, rep) = read_xlsx(&minimal(body, &parts, rels, "")).unwrap();
@@ -251,6 +256,24 @@ fn hyperlinks_past_the_cell_cap_are_reported() {
     assert!(rep.warnings.is_empty(), "{:?}", rep.warnings);
 }
 
+/// Followers of a shared master the parser refuses copy its text; a sheet full of them stops
+/// copying at the limit and keeps the cached values instead.
+#[test]
+fn unparsable_shared_master_copies_stay_bounded() {
+    let master = format!("1{}", "+1".repeat(40_000));
+    let mut rows = format!(r#"<row r="1"><c r="A1"><f t="shared" ref="A1:A2000" si="0">{master}</f><v>1</v></c></row>"#);
+    for r in 2..=2000 {
+        rows.push_str(&format!(r#"<row r="{r}"><c r="A{r}"><f t="shared" si="0"/><v>{r}</v></c></row>"#));
+    }
+    let (wb, rep) = read_xlsx(&minimal(&format!("<sheetData>{rows}</sheetData>"), &[], "", "")).unwrap();
+    let s = wb.sheet(0).unwrap();
+    let copied: u64 = (2..=2000).filter(|r| s.cell(gridcraft_core::CellRef::new(r - 1, 0)).is_some_and(|c| c.formula.is_some())).count() as u64;
+    assert!(copied * master.len() as u64 <= 64 * 1024 * 1024, "{copied} copies");
+    assert!(copied < 1999, "{copied} copies");
+    assert_eq!(s.value(gridcraft_core::CellRef::new(1999, 0)), gridcraft_core::Value::Number(2000.0));
+    assert!(rep.warnings.iter().any(|w| w.contains("exceed the copy limit")), "{:?}", rep.warnings);
+}
+
 /// Formula text built to overflow a recursive parser or tree walk (a million `%`, a 200,000-term
 /// sum): the cells keep their formula text and cached value, nothing overflows the stack.
 #[test]