Fix infinite recursion when checking module of expression (#2159)

## Changes
Don't drop existing statements for globals when appending trees

### Linked issues
None

### Functionality
None

### Tests
- [x] added unit tests

Was discovered when running make solacc. Fixes 1 of the 2 issue. Other
issue is fixed by #2157

Co-authored-by: Eric Vergnaud <[email protected]>

Author: ericvergnaud

src/databricks/labs/ucx/source_code/linters/python_ast.py

@@ -12,6 +12,7 @@
     Attribute,
     Call,
     Const,
+    Expr,
     Import,
     ImportFrom,
     Module,
@@ -220,12 +221,16 @@ def append_statements(self, tree: Tree) -> Tree:
             stmt.parent = self_module
             self_module.body.append(stmt)
         for name, value in tree_module.globals.items():
-            self_module.globals[name] = value
+            statements: list[Expr] = self_module.globals.get(name, None)
+            if statements is None:
+                self_module.globals[name] = list(value)  # clone the source list to avoid side-effects
+                continue
+            statements.extend(value)
         # the following may seem strange but it's actually ok to use the original module as tree root
         return tree
 
     def is_from_module(self, module_name: str):
-        # if his is the call's root node, check it against the required module
+        # if this is the call's root node, check it against the required module
         if isinstance(self._node, Name):
             if self._node.name == module_name:
                 return True

tests/unit/source_code/linters/test_python_ast.py

@@ -159,3 +159,21 @@ def test_is_from_module():
         Call, [("saveAsTable", Attribute), ("format", Attribute), ("write", Attribute), ("df", Name)]
     )[0]
     assert Tree(save_call).is_from_module("spark")
+
+
+def test_supports_recursive_refs_when_checking_module():
+    source_1 = """
+    df = spark.read.csv("hi")
+    """
+    source_2 = """
+    df = df.withColumn(stuff)
+    """
+    source_3 = """
+    df = df.withColumn(stuff2)
+    """
+    main_tree = Tree.normalize_and_parse(source_1)
+    main_tree.append_statements(Tree.normalize_and_parse(source_2))
+    tree = Tree.normalize_and_parse(source_3)
+    main_tree.append_statements(tree)
+    assign = tree.locate(Assign, [])[0]
+    assert Tree(assign.value).is_from_module("spark")