Fix multiple parsing bugs and add RFC 4180 quote support

- Change nread to size_t to prevent overflow.

- Fix state machine to handle consecutive newlines appropriately.

- Add csv_parser_finish to emit pending trailing fields at EOF.

- Fix nread boundary count on callback errors.

- Add support for quoted fields, including escaped quotes.

- Update and expand test suite to cover new cases and quote parsing.

1 files changed, 9 insertions, 1 deletions

diff --git a/src/csv_parser.h b/src/csv_parser.h
index 8c78e8b..267c768 100644
--- a/src/csv_parser.h
+++ b/src/csv_parser.h
@@ -35,6 +35,8 @@ typedef enum csv_parser_state_t {
     csvps_line_start = 0,
     csvps_field_start,
     csvps_field_value,
+    csvps_field_quoted_value,
+    csvps_field_quoted_quote,
     csvps_field_end,
     csvps_line_end_begin,
     csvps_line_end,
@@ -48,7 +50,7 @@ typedef struct csv_parser_t {
     int row;
     /* csv column */
     int col;
-    uint32_t nread;
+    size_t nread;
     /* user data */
     void *data;
 } csv_parser_t;
@@ -88,6 +90,12 @@ size_t csv_parser_execute(csv_parser_t *parser,
                           const char *data,
                           size_t data_len);
 
+/**
+ * Flush any pending state. This is useful when the file ends
+ * without a trailing newline, to flush the final empty field.
+ * Returns 0 on success, anything else on error.
+ */
+int csv_parser_finish(csv_parser_t *parser, const csv_parser_settings_t *settings);
 
 #ifdef __cplusplus
 }