Merge pull request apache#610 from rootvector2/quote-comment-marker-first-char

Quote value starting with comment marker in minimal quote mode

Author: garydgregory

src/main/java/org/apache/commons/csv/CSVFormat.java

@@ -2454,10 +2454,11 @@ private void printWithQuotes(final Object object, final CharSequence charSeq, fi
                 }
             } else {
                 char c = charSeq.charAt(pos);
-                if (c <= Constants.COMMENT) {
+                if (c <= Constants.COMMENT || isCommentMarkerSet() && c == commentMarker.charValue()) {
                     // Some other chars at the start of a value caused the parser to fail, so for now
                     // encapsulate if we start in anything less than '#'. We are being conservative
-                    // by including the default comment char too.
+                    // by including the default comment char and any configured comment marker too,
+                    // which the parser would otherwise read back as a comment line.
                     quote = true;
                 } else {
                     while (pos < len) {

src/test/java/org/apache/commons/csv/CSVPrinterTest.java

@@ -1829,6 +1829,33 @@ void testQuoteCommaFirstChar() throws IOException {
         }
     }
 
+    @Test
+    void testQuoteCommentMarkerFirstChar() throws IOException {
+        final CSVFormat format = CSVFormat.DEFAULT.builder().setCommentMarker(';').get();
+        final StringWriter sw = new StringWriter();
+        final String col1 = ";comment-like";
+        try (CSVPrinter printer = new CSVPrinter(sw, format)) {
+            // A real comment is written with the marker, unquoted.
+            printer.printComment("a real comment");
+            // A value starting with the marker is quoted, so it does not read back as a comment.
+            printer.printRecord(col1, "b");
+            // The marker past the first character does not start a comment, so only the leading-marker value is quoted.
+            printer.printRecord("a;b", ";c");
+        }
+        assertEquals("; a real comment" + RECORD_SEPARATOR +
+                "\";comment-like\",b" + RECORD_SEPARATOR +
+                "a;b,\";c\"" + RECORD_SEPARATOR, sw.toString());
+        // The comment is dropped on read; both data records survive intact.
+        try (CSVParser parser = CSVParser.parse(sw.toString(), format)) {
+            final List<CSVRecord> records = parser.getRecords();
+            assertEquals(2, records.size());
+            assertEquals(col1, records.get(0).get(0));
+            assertEquals("b", records.get(0).get(1));
+            assertEquals("a;b", records.get(1).get(0));
+            assertEquals(";c", records.get(1).get(1));
+        }
+    }
+
     @Test
     void testQuoteNonNumeric() throws IOException {
         final StringWriter sw = new StringWriter();