xls: fix issue with incorrect SST record

Changelog.md

@@ -13,6 +13,9 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Fixed
 
+- Fixed issue #548 for xls files where the SST record had an incorrect number of
+  unique strings.
+
 
 ## [0.30.0] - 2025-08-07
 

src/xls.rs

@@ -892,6 +892,9 @@ fn parse_dimensions(r: &[u8]) -> Result<Dimensions, XlsError> {
     }
 }
 
+// Parse the Excel xls Shared String Table (SST). See [MS-XLS] 2.4.265.
+//
+// https://learn.microsoft.com/en-us/openspecs/office_file_formats/ms-xls/b6231b92-d32e-4626-badd-c3310a672bab
 fn parse_sst(r: &mut Record<'_>, encoding: &XlsEncoding) -> Result<Vec<String>, XlsError> {
     if r.data.len() < 8 {
         return Err(XlsError::Len {
@@ -900,13 +903,15 @@ fn parse_sst(r: &mut Record<'_>, encoding: &XlsEncoding) -> Result<Vec<String>,
             found: r.data.len(),
         });
     }
-    let len: usize = read_i32(&r.data[4..8]).try_into().unwrap();
-    let mut sst = Vec::with_capacity(len);
+    let mut sst = vec![];
+
+    // Skip cstTotal and cstUnique headers in SST record.
     r.data = &r.data[8..];
 
-    for _ in 0..len {
+    while !r.data.is_empty() || r.continue_record() {
         sst.push(read_rich_extended_string(r, encoding)?);
     }
+
     Ok(sst)
 }
 
@@ -958,7 +963,7 @@ fn read_rich_extended_string(
     r: &mut Record<'_>,
     encoding: &XlsEncoding,
 ) -> Result<String, XlsError> {
-    if r.data.is_empty() && !r.continue_record() || r.data.len() < 3 {
+    if r.data.len() < 3 {
         return Err(XlsError::Len {
             typ: "rich extended string",
             expected: 3,

tests/test.rs

@@ -2154,6 +2154,37 @@ fn test_oom_allocation() {
     assert_eq!(ws[0].0, "Colsale (Aug".to_string());
 }
 
+// Test for issue #548. The SST table in the test file has an incorrect unique
+// string count.
+#[test]
+fn test_incorrect_sst_unique_count() {
+    // Check for the string that appears last in the SST table: "11th May 2023".
+    // This appears in cell C10 of each worksheet in the workbook.
+    let mut xls: Xls<_> = wb("gh548_incorrect_sst_unique_count.xls");
+    let range = xls.worksheet_range("System Level Data").unwrap();
+
+    assert_eq!(
+        range.get_value((9, 2)).unwrap(),
+        &String("11th May 2023".into())
+    );
+}
+
+// Test the parsing of an SST table that finishes with a complete, untruncated,
+// string at the end of the block, and where the next string is in a CONTINUE
+// block. This is related to the previous test for gh548 to ensure that the edge
+// condition is met.
+#[test]
+fn test_sst_continue() {
+    let mut xls: Xls<_> = wb("sst_continue.xls");
+    let range = xls.worksheet_range("Sheet1").unwrap();
+
+    // Check for the string that appears last in the SST table.
+    assert_eq!(
+        range.get_value((135, 0)).unwrap(),
+        &String("New CONTINUE block".into())
+    );
+}
+
 // Test for issue #419 where the part name is sentence case instead of camel
 // case. The test file contains a sub-file called "xl/SharedStrings.xml" (note
 // the uppercase S in Shared). This is allowed by "Office Open XML File Formats