address review: stop collecting JSON keys at the column limit and reject oversized row counts before building rows

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

@@ -70,14 +70,19 @@ pub(super) fn from_csv(s: &mut Session, p: &Json) -> Result<Json> {
             for o in &arr {
                 if let Some(m) = o.as_object() {
                     for k in m.keys() {
-                        if seen.insert(k.as_str()) {
+                        if !seen.contains(k.as_str()) {
+                            // Stop collecting at the column limit.
+                            if keys.len() >= gridcraft_core::MAX_COLS as usize {
+                                return Err(EngineError::Other("The data has more fields than a sheet has columns.".into()));
+                            }
+                            seen.insert(k.as_str());
                             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 > u64::from(gridcraft_core::MAX_ROWS) {
+                return Err(EngineError::Other("The data has more rows than a sheet has.".into()));
             }
             if (arr.len() as u64 + 1).saturating_mul(keys.len().max(1) as u64) > MAX_IMPORT_CELLS {
                 return Err(too_many_cells());

crates/engine/src/tests.rs

@@ -1312,6 +1312,18 @@ fn scripts_leave_file_access_working_after_they_run() {
     let _ = std::fs::remove_dir_all(&dir);
 }
 
+#[test]
+fn get_data_refuses_json_with_too_many_fields_or_rows_before_building_rows() {
+    let mut s = s();
+    let wide = format!("[{}]", (0..=gridcraft_core::MAX_COLS).map(|i| format!("{{\"k{i}\":1}}")).collect::<Vec<_>>().join(","));
+    let err = s.execute("data.fromTextCsv", json!({"text": wide})).unwrap_err().to_string();
+    assert!(err.contains("more fields"), "{err}");
+    // MAX_ROWS objects plus the header row are one too many.
+    let tall = format!("[{}]", vec!["{}"; gridcraft_core::MAX_ROWS as usize].join(","));
+    let err = s.execute("data.fromTextCsv", json!({"text": tall})).unwrap_err().to_string();
+    assert!(err.contains("more rows"), "{err}");
+}
+
 #[test]
 fn scripts_cannot_reach_outside_the_workbook_even_when_trusted() {
     let dir = std::env::temp_dir().join(format!("gridcraft-script-deny-{}", std::process::id()));