fix(parser): grow truncated rows to host inherited merge values (#21044)

internal/parser/parser/office_table_render.go

@@ -594,11 +594,14 @@ func inheritMergedHeader(records [][]string, headerRowIdx int, mm map[[2]int][2]
 
 // inheritMergedDataCells fills empty cells of non-header rows with the value
 // of their merge master, so a vertically merged group label (e.g. A2:A4) is
-// visible on every data row instead of only the first. Unlike the header-row
-// path above, data rows are never widened: only cells already present in the
-// row slice are filled, so a far merge cannot allocate a sea of empty <td>
-// cells (see padRowToWidth for the blow-up this avoids). Call with absolute
-// sheet row numbers, before records are reordered around the header row.
+// visible on every data row instead of only the first. excelize truncates
+// each row at its last valued cell, so a slave cell past that point is absent
+// from the slice: the row is then extended exactly up to the slave column to
+// host the real inherited value. The extension is bounded by the master row's
+// own width (the master cell must exist and be non-empty), so unlike header
+// padding it cannot chase a far merge across the sheet (see padRowToWidth).
+// Call with absolute sheet row numbers, before records are reordered around
+// the header row.
 func inheritMergedDataCells(rows [][]string, headerRow int, ranges []mergeRange) {
 	for _, r := range ranges {
 		// Vertical merges only: a horizontally merged data cell is already
@@ -629,8 +632,10 @@ func inheritMergedDataCells(rows [][]string, headerRow int, ranges []mergeRange)
 			if row == headerRow {
 				continue
 			}
-			if r.sc > len(rows[row-1]) {
-				continue
+			// Grow a truncated row exactly up to the slave column; the
+			// appended cells stay empty except the inherited value below.
+			for len(rows[row-1]) < r.sc {
+				rows[row-1] = append(rows[row-1], "")
 			}
 			if strings.TrimSpace(rows[row-1][r.sc-1]) == "" {
 				rows[row-1][r.sc-1] = master

internal/parser/parser/office_table_render_test.go

@@ -566,6 +566,38 @@ func TestXLSXParser_MergedDataCellHorizontalUntouched(t *testing.T) {
 	}
 }
 
+// TestXLSXParser_MergedDataCellBeyondTruncation asserts a merged slave cell
+// past its row's truncation point (excelize drops trailing empties) is still
+// filled: the row grows exactly up to the slave column to host the value.
+func TestXLSXParser_MergedDataCellBeyondTruncation(t *testing.T) {
+	data := newTestXLSX(t, func(f *excelize.File) {
+		mustSetCell(t, f, "Sheet1", "A1", "name")
+		mustSetCell(t, f, "Sheet1", "B1", "age")
+		mustSetCell(t, f, "Sheet1", "C1", "department")
+		mustSetCell(t, f, "Sheet1", "D1", "notes")
+		mustSetCell(t, f, "Sheet1", "A2", "Mike")
+		mustSetCell(t, f, "Sheet1", "B2", 28)
+		mustSetCell(t, f, "Sheet1", "C2", "Sales")
+		mustSetCell(t, f, "Sheet1", "D2", "Important customer")
+		mustMergeCell(t, f, "Sheet1", "A2", "A3")
+		mustMergeCell(t, f, "Sheet1", "B2", "B3")
+		mustMergeCell(t, f, "Sheet1", "D2", "D3")
+		// Row 3's own value sits before the merged columns, so GetRows
+		// truncates the row short of the D3 slave.
+		mustSetCell(t, f, "Sheet1", "C3", "IT")
+	})
+	p, _ := NewXLSXParser("")
+	res := p.ParseWithResult(t.Context(), "t.xlsx", data)
+	if res.Err != nil {
+		t.Fatalf("ParseWithResult: %v", res.Err)
+	}
+	text := spreadsheetText(res)
+	want := "<tr><td>Mike</td><td>28</td><td>IT</td><td>Important customer</td></tr>"
+	if !strings.Contains(text, want) {
+		t.Fatalf("truncated slave cells not filled, got:\n%s", text)
+	}
+}
+
 // TestXLSXParser_MergedHeaderDownIntoDataUntouched asserts a vertical merge
 // mastered on the header row does not leak the header label into data rows.
 func TestXLSXParser_MergedHeaderDownIntoDataUntouched(t *testing.T) {