address review: keep the original code in the public number_format for overlong formats

The display cache hashed the full format code on every lookup, so one
huge code shared by many cells cost its length for every cell drawn.
A code with more than 4 × 255 bytes can't be within Excel's 255
characters and is now formatted as General straight away.

crates/engine/src/display.rs

@@ -8,8 +8,13 @@ use gridcraft_model::{Sheet, Workbook};
 use gridcraft_numfmt::{Formatted, NumberFormat, format_value};
 
 /// Parsed number formats are cached by code.
-fn parsed(code: &str) -> NumberFormat {
+pub(crate) fn parsed(code: &str) -> NumberFormat {
     static CACHE: std::sync::LazyLock<Mutex<HashMap<String, NumberFormat>>> = std::sync::LazyLock::new(|| Mutex::new(HashMap::new()));
+    // A code of more bytes than 4 × the character limit can't be within it: it shows as General
+    // without being hashed or scanned again for every cell that shares it.
+    if code.len() > 4 * gridcraft_numfmt::MAX_FORMAT_CODE_CHARS {
+        return NumberFormat::parse("General");
+    }
     let lock = || CACHE.lock().unwrap_or_else(std::sync::PoisonError::into_inner);
     if let Some(f) = lock().get(code) {
         return f.clone();
@@ -24,7 +29,12 @@ fn parsed(code: &str) -> NumberFormat {
     f
 }
 
+/// The parsed format of `code`; one longer than the character limit is a General format that keeps
+/// the code it was made from. The engine itself formats through `parsed`, which doesn't copy it.
 pub fn number_format(code: &str) -> NumberFormat {
+    if code.len() > 4 * gridcraft_numfmt::MAX_FORMAT_CODE_CHARS {
+        return NumberFormat::parse(code);
+    }
     parsed(code)
 }
 
@@ -226,3 +236,31 @@ pub fn stats(sheet: &Sheet, ranges: &[RangeRef]) -> Stats {
     }
     s
 }
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    /// A format code far past Excel's 255 characters shows its values as General, for every cell
+    /// that shares it, without being hashed into the cache each time.
+    #[test]
+    fn an_overlong_code_formats_as_general() {
+        let wb = Workbook::new();
+        let code = format!("0.00{}", "\"x\"".repeat(400_000));
+        for _ in 0..1_000 {
+            assert_eq!(format(&Value::Number(1.5), &code, &wb).text, "1.5");
+        }
+        assert_eq!(format(&Value::Number(1.5), "0.00", &wb).text, "1.50");
+    }
+
+    /// The public `number_format` keeps the code of an overlong format; it is General, but not the
+    /// same format as a real `General`.
+    #[test]
+    fn number_format_keeps_an_overlong_code() {
+        let code = format!("0.00{}", "\"x\"".repeat(2_000));
+        let f = number_format(&code);
+        assert!(f.is_general());
+        assert_eq!(f.code(), code);
+        assert_ne!(f, number_format("General"));
+    }
+}

crates/engine/src/io.rs

@@ -443,7 +443,7 @@ pub fn csv_values_bytes(wb: &Workbook) -> Result<Vec<u8>> {
     let Some(sh) = wb.sheet(wb.active_sheet()) else { return Ok(Vec::new()) };
     gridcraft_xlsx::write_csv(sh, wb.csv_delimiter.unwrap_or(b','), |c| match sh.value(c) {
         gridcraft_core::Value::Number(n) if n.is_finite() => {
-            let shape = crate::display::number_format(wb.styles.get(sh.style_id(c)).num_fmt.as_str()).date_shape(n);
+            let shape = crate::display::parsed(wb.styles.get(sh.style_id(c)).num_fmt.as_str()).date_shape(n);
             match shape.zip(gridcraft_core::date::datetime_from_serial(wb.date_system, n)) {
                 Some(((date, time), d)) => iso(&d, date, time),
                 None => n.to_string(),

crates/engine/src/macros/recorder.rs

@@ -440,7 +440,7 @@ pub fn infer_input(wb: &Workbook, sheet: usize, r: RangeRef) -> NamedInput {
         Some(match v {
             Value::Empty => return None,
             Value::Text(t) if t.is_empty() => return None,
-            Value::Number(_) if crate::display::number_format(wb.styles.get(sh.style_id(c)).num_fmt.as_str()).is_date() => ColType::Date,
+            Value::Number(_) if crate::display::parsed(wb.styles.get(sh.style_id(c)).num_fmt.as_str()).is_date() => ColType::Date,
             Value::Number(_) => ColType::Number,
             Value::Text(_) => ColType::Text,
             Value::Bool(_) => ColType::Bool,

crates/engine/src/macros/run.rs

@@ -335,7 +335,7 @@ fn check_input(wb: &Workbook, input: &gridcraft_model::NamedInput, place: Place)
                 gridcraft_model::ColType::Number => v.is_number(),
                 gridcraft_model::ColType::Text => v.is_text(),
                 gridcraft_model::ColType::Bool => matches!(v, Value::Bool(_)),
-                gridcraft_model::ColType::Date => v.is_number() && crate::display::number_format(&date()).is_date(),
+                gridcraft_model::ColType::Date => v.is_number() && crate::display::parsed(&date()).is_date(),
             };
             if !ok {
                 wrong.push((c, holds(&v, ty)));

crates/engine/src/pivot.rs

@@ -182,7 +182,7 @@ pub fn read_source(wb: &Workbook, text: &str, default_sheet: usize) -> Result<So
     for (n, a) in numeric.iter_mut().zip(any.iter()) {
         *n = *n && *a;
     }
-    let dates = numeric.iter().zip(formats.iter()).map(|(n, f)| *n && crate::display::number_format(f).is_date()).collect();
+    let dates = numeric.iter().zip(formats.iter()).map(|(n, f)| *n && crate::display::parsed(f).is_date()).collect();
     Ok(Source { sheet: si, range: r, headers, formats, numeric, dates, rows })
 }