fix(folderparser): use csv.reader to handle RFC 4180 quoted fields in annotation CSVs

roboflow/util/folderparser.py

@@ -1,3 +1,4 @@
+import csv
 import json
 import os
 import re
@@ -324,25 +325,37 @@ def _read_jsonl(path):
 
 
 def _parseAnnotationCSV(filename):
-    # TODO: use a proper CSV library?
+    # Open in text mode so line endings are normalized to \n (matches legacy behaviour).
+    # Annotation filenames never contain embedded newlines, so the csv multi-line-field
+    # caveat for newline="" does not apply here.
     with open(filename) as f:
-        lines = f.readlines()
-    headers = [h.strip() for h in lines[0].split(",")]
+        raw_lines = f.readlines()
+
+    rows = list(csv.reader(raw_lines))
+    if not rows:
+        return {"headers": "", "lines": []}
+
+    headers = [h.strip() for h in rows[0]]
+
     # Multi-label classification csv typically named _classes.csv
     if os.path.basename(filename) == "_classes.csv":
         parsed_lines = []
-        for line in lines[1:]:
-            parts = [p.strip() for p in line.split(",")]
+        for parts in rows[1:]:
+            parts = [p.strip() for p in parts]
+            if not parts:
+                continue
             file_name = parts[0]
             labels = [headers[i] for i, v in enumerate(parts[1:], start=1) if v == "1"]
             parsed_lines.append({"file_name": file_name, "labels": labels})
         return {"type": "multilabel_csv", "rows": parsed_lines, "headers": headers}
-    header_line = lines[0]
-    lines = [{"file_name": ld.split(",")[0].strip(), "line": ld} for ld in lines[1:]]
-    return {
-        "headers": header_line,
-        "lines": lines,
-    }
+
+    # For regular CSV, preserve raw lines so callers can reconstruct verbatim CSV text
+    # for upload, but use csv.reader to correctly extract file_name (handles quoted commas).
+    header_line = raw_lines[0] if raw_lines else ""
+    lines = [
+        {"file_name": row[0].strip() if row else "", "line": raw_line} for raw_line, row in zip(raw_lines[1:], rows[1:])
+    ]
+    return {"headers": header_line, "lines": lines}
 
 
 def _guessAnnotationFileFormat(parsed, extension):

tests/util/test_folderparser.py

@@ -205,6 +205,35 @@ def test_coco_root_annotation_matches_images_in_subdirs(self):
             self.assertEqual(len(ann_data["annotations"]), 1, "Should have one annotation")
             self.assertEqual(ann_data["annotations"][0]["bbox"], [10, 20, 100, 200])
 
+    def test_parse_csv_quoted_filename(self):
+        """_parseAnnotationCSV must handle filenames containing commas (RFC 4180 quoting)."""
+        with tempfile.NamedTemporaryFile(mode="w", suffix=".csv", delete=False, newline="") as f:
+            f.write("img_fName,class_label\n")
+            f.write('"image,with,commas.jpg",cat\n')
+            f.write("normal.jpg,dog\n")
+            tmppath = f.name
+        try:
+            parsed = folderparser._parseAnnotationCSV(tmppath)
+            names = [ld["file_name"] for ld in parsed["lines"]]
+            self.assertEqual(names[0], "image,with,commas.jpg")
+            self.assertEqual(names[1], "normal.jpg")
+        finally:
+            os.unlink(tmppath)
+
+    def test_parse_multilabel_csv_quoted_filename(self):
+        """_parseAnnotationCSV must handle quoted filenames in _classes.csv format."""
+        with tempfile.TemporaryDirectory() as tmpdir:
+            classes_csv = os.path.join(tmpdir, "_classes.csv")
+            with open(classes_csv, "w") as f:
+                f.write("filename,cat,dog\n")
+                f.write('"image,with,commas.jpg",1,0\n')
+                f.write("normal.jpg,0,1\n")
+            parsed = folderparser._parseAnnotationCSV(classes_csv)
+            self.assertEqual(parsed["type"], "multilabel_csv")
+            rows = {r["file_name"]: r["labels"] for r in parsed["rows"]}
+            self.assertEqual(rows["image,with,commas.jpg"], ["cat"])
+            self.assertEqual(rows["normal.jpg"], ["dog"])
+
 
 def _assertJsonMatchesFile(actual, filename):
     with open(filename) as file: