Fix bug in using eval with templates not at top level

galsim/config/process.py

@@ -202,7 +202,7 @@ def ProcessAllTemplates(config, logger=None, base=None):
     """
     if base is None: base = config
     ProcessTemplate(config, base, logger)
-    for (key, field) in config.items():
+    for (key, field) in list(config.items()):
         if isinstance(field, dict):
             ProcessAllTemplates(field, logger, base)
         elif isinstance(field, list):

tests/test_config_image.py

@@ -2547,6 +2547,21 @@ def test_template():
     config8['psf']['items'][2] = config4['image']['noise']
     assert config9 == config8
 
+    # Make sure evals work correltly for template string when not at top level.
+    # (This used to give an error about the config dict changing size during iteration.)
+    config10 = {
+        "eval_variables" : {
+            "iabc": 123,
+            "ddict": {
+                "template": "$os.path.join('config_input','dict.yaml')"
+            }
+        }
+    }
+    galsim.config.ProcessAllTemplates(config10)
+    assert config10['eval_variables']['ddict']['b'] == False
+    assert config10['eval_variables']['ddict']['s'] == "Brian"
+    assert config10['eval_variables']['ddict']['noise']['models'][0]['variance'] == 0.12
+
 
 @timer
 def test_variable_cat_size():