Update data_extraction.py

data_extraction.py

@@ -89,6 +89,27 @@ def _context_snippet(raw, pos, window=120):
     return f"...{before!r} <-- ERROR HERE --> {after!r}..."
 
 
+def _preprocess_mongo_shell_syntax(raw):
+    """
+    Converts MongoDB *shell* export syntax into plain JSON.
+    mongoexport/mongodump produce valid JSON (e.g. {"$oid": "..."}), but a
+    raw copy-paste from the mongo shell or Compass often contains
+    constructor-style tokens that are NOT valid JSON, e.g.:
+        ObjectId("665f1...")   -> "665f1..."
+        ISODate("2024-01-01T00:00:00Z") -> "2024-01-01T00:00:00Z"
+        NumberLong(123)        -> 123
+        NumberInt(123)         -> 123
+        NumberDecimal("1.5")   -> "1.5"
+        Timestamp(123, 1)      -> "Timestamp(123, 1)"  (rare; kept as string)
+    This rewrites those into equivalent plain JSON values.
+    """
+    # ObjectId("...") / ISODate("...") / NumberDecimal("...") -> just the quoted string
+    raw = re.sub(r'\b(?:ObjectId|ISODate|NumberDecimal|UUID)\(\s*"([^"]*)"\s*\)', r'"\1"', raw)
+    # NumberLong(123) / NumberInt(123) -> bare number (handles quoted numbers too)
+    raw = re.sub(r'\b(?:NumberLong|NumberInt)\(\s*"?(-?\d+)"?\s*\)', r'\1', raw)
+    return raw
+
+
 def load_records(filepaths):
     """
     Handles several shapes robustly:
@@ -108,6 +129,8 @@ def load_records(filepaths):
             print(f"  [skip] {fp} is empty")
             continue
 
+        raw = _preprocess_mongo_shell_syntax(raw)
+
         whole_file_error = None
         try:
             obj = json.loads(raw)