Revert transform_string perf penalty in _flatten (introduced in 3.2.0b2)

Author: ptmcg

CHANGES

@@ -96,6 +96,9 @@ Version 3.3.0 - December, 2025
     (3.1.1 through 3.3.0), aggregating results into `perf_pyparsing.csv` at the repo root.
   - See `tests/README.md` for usage instructions.
 
+- Used performance benchmarking to identify and revert an inefficient utility method used in
+  `transform_string` (introduced in pyparsing 3.2.0b2).
+
 
 Version 3.2.5 - September, 2025
 -------------------------------

pyparsing/util.py

@@ -312,11 +312,14 @@ def no_escape_re_range_char(c: str) -> str:
 
 def _flatten(ll: Iterable) -> list:
     ret = []
-    to_visit = [*ll]
-    while to_visit:
-        i = to_visit.pop(0)
-        if isinstance(i, Iterable) and not isinstance(i, str):
-            to_visit[:0] = i
+    for i in ll:
+        # Developer notes:
+        # - do not collapse this section of code, isinstance checks are done
+        # in optimal order
+        if isinstance(i, str):
+            ret.append(i)
+        elif isinstance(i, Iterable):
+            ret.extend(_flatten(i))
         else:
             ret.append(i)
     return ret