fix(csv): skip leading blank lines so they don't wipe the whole table

packages/markitdown/src/markitdown/converters/_csv_converter.py

@@ -51,6 +51,13 @@ def convert(
         reader = csv.reader(io.StringIO(content))
         rows = list(reader)
 
+        # A leading blank line (or any empty rows before the header) parses as an
+        # empty list. If that becomes the header its column count is 0, and every
+        # data row then gets truncated to nothing -- silent total data loss.
+        # Skip leading empty rows so the first real row becomes the header.
+        while rows and len(rows[0]) == 0:
+            rows.pop(0)
+
         if not rows:
             return DocumentConverterResult(markdown="")
 

packages/markitdown/tests/_test_vectors.py

@@ -152,6 +152,20 @@ class FileTestVector(object):
         ],
         must_not_include=[],
     ),
+    FileTestVector(
+        filename="test_blank_first_line.csv",
+        mimetype="text/csv",
+        charset="utf-8",
+        url=None,
+        must_include=[
+            "| name | age |",
+            "| bob | 3 |",
+            "| alice | 7 |",
+        ],
+        must_not_include=[
+            "|  |",
+        ],
+    ),
     FileTestVector(
         filename="test.json",
         mimetype="application/json",

packages/markitdown/tests/test_files/test_blank_first_line.csv

@@ -0,0 +1,4 @@
+
+name,age
+bob,3
+alice,7