From 60532312d016561eeb8879a491e9e734a92b18ef Mon Sep 17 00:00:00 2001
From: Ludwig Lierhammer <ludwig.lierhammer@dwd.de>
Date: Fri, 26 Jun 2026 13:20:25 +0100
Subject: [PATCH] restore string literals in columns for ParquetSttreamReader
 items

---
 src/cdm_reader_mapper/common/iterators.py | 55 +++++++++++++++++++++++
 1 file changed, 55 insertions(+)

diff --git a/src/cdm_reader_mapper/common/iterators.py b/src/cdm_reader_mapper/common/iterators.py
index 94e97bb2..7726463d 100755
--- a/src/cdm_reader_mapper/common/iterators.py
+++ b/src/cdm_reader_mapper/common/iterators.py
@@ -768,6 +768,59 @@ def _make_parquet_reader(d: TemporaryDirectory[str], t: type, s: str | None) ->
     return tuple(final_iterators + output_non_data)
 
 
+def _restore_columns(item: Any) -> Any:
+    """
+    Restore columns from string literals if `item` is a pandas DataFrame or Series.
+
+    Parameters
+    ----------
+    item : Any
+        Object to restore.
+
+    Returns
+    -------
+    Any
+        Restored object.
+    """
+
+    def _literal_eval(column: Any) -> Any:
+        """
+        Evaluate a string literal if possible.
+
+        Parameters
+        ----------
+        column : Any
+            Column that is possibly a string literal.
+
+        Returns
+        -------
+        Any
+            Evaluated column.
+        """
+        if not isinstance(column, str):
+            return column
+        try:
+            from ast import literal_eval
+
+            return literal_eval(column)
+        except (ValueError, SyntaxError):
+            return column
+
+    if isinstance(item, pd.DataFrame):
+        columns = item.columns
+        new_columns = []
+        for column in columns:
+            column = _literal_eval(column)
+            new_columns.append(column)
+
+        item.columns = new_columns
+
+    if isinstance(item, pd.Series):
+        item.name = _literal_eval(item.name)
+
+    return item
+
+
 def _process_chunks(
     readers: list[ParquetStreamReader],
     func: Callable[..., Any],
@@ -835,6 +888,8 @@ def _process_chunks(
     for items in zip(*readers, strict=True):
         _validate_chunk(items, requested_types)
 
+        items = tuple([_restore_columns(item) for item in items])
+
         result = func(*items, *static_args, **static_kwargs)
         data, meta = _process_result(result, requested_types, non_data_output, chunk_counter)