Fix get_codebase_session local filesystem corruption + Fix codebase verify_output (#558)

**This PR is broken down into three parts:**
- Fix rare bug where `get_codebase_session` would modify files in the
`codegen-sdk` repo instead of the test
- Fix & re-enable `verify_output` logic
- Fix any broken tests caused by `verify_output`

Author: EdwardJXLi

src/codegen/sdk/codebase/factory/get_session.py

@@ -9,6 +9,7 @@
 from codegen.sdk.codebase.config import CodebaseConfig, ProjectConfig, SessionOptions, TestFlags
 from codegen.sdk.codebase.factory.codebase_factory import CodebaseFactory
 from codegen.sdk.core.codebase import Codebase, PyCodebaseType, TSCodebaseType
+from codegen.sdk.core.file import SourceFile
 from codegen.sdk.tree_sitter_parser import print_errors
 from codegen.shared.configs.models.feature_flags import CodebaseFeatureFlags
 from codegen.shared.configs.models.secrets import SecretsConfig
@@ -83,9 +84,15 @@ def get_codebase_session(
     ):
         if verify_input:
             for file in codebase.files:
-                if os.path.exists(file.filepath):
-                    print_errors(file.filepath, file.content)
-                    assert not file.ts_node.has_error, "Invalid syntax in test case"
+                # NOTE: We only check SourceFiles for syntax errors
+                abs_filepath = os.path.join(tmpdir, file.filepath)
+                if os.path.exists(abs_filepath):
+                    if isinstance(file, SourceFile):
+                        # Check for syntax errors
+                        print_errors(abs_filepath, file.content)
+                        if file.ts_node.has_error:
+                            msg = "Invalid syntax in test case"
+                            raise SyntaxError(msg)
         yield codebase
 
     if verify_output:

tests/unit/codegen/sdk/codebase/codebase_graph/test_codebase_reset.py

@@ -200,18 +200,18 @@ def __init__(self):
 @pytest.mark.parametrize(
     "original, expected",
     [
-        ({"src/a.py": "original content"}, {"src/a.py": "modified content", "src/b.py": "new file content"}),
+        ({"src/a.py": "# original content"}, {"src/a.py": "# modified content", "src/b.py": "# new file content"}),
     ],
     indirect=["original", "expected"],
 )
 def test_codebase_reset_preserves_external_changes(codebase: Codebase, assert_expected, tmp_path):
     # Make external changes to existing file
     src_dir = tmp_path / "src"
     src_dir.mkdir(exist_ok=True)
-    (src_dir / "a.py").write_text("modified content")
+    (src_dir / "a.py").write_text("# modified content")
 
     # Add new file externally
-    (src_dir / "b.py").write_text("new file content")
+    (src_dir / "b.py").write_text("# new file content")
 
     # Reset should detect and preserve these changes
     codebase.commit()

tests/unit/codegen/sdk/codebase/session/target_python_file.py

@@ -0,0 +1 @@
+print("Hello World!")

tests/unit/codegen/sdk/codebase/session/test_codebase_session_corruption.py

@@ -0,0 +1,36 @@
+import os
+
+from codegen.sdk.codebase.factory.get_session import get_codebase_session
+from codegen.shared.enums.programming_language import ProgrammingLanguage
+
+
+def test_get_codebase_session(tmpdir) -> None:
+    # READ THIS CONTEXT BEFORE EDITING THIS TEST!
+    # Essentially, there was this bug where the file in the base codegen-sdk repo was being edited instead of the file in the tmpdir
+    # This test is to ensure that this bug does not happen again.
+    # target_file must be a file that shares the same relative path in the codegen-sdk repo
+
+    # Init Test
+    target_file = "tests/unit/codegen/sdk/codebase/session/target_python_file.py"
+    assert os.path.exists(target_file), f"Target file {target_file} does not exist! Please change this to a file that exists in codegen-sdk"
+    target_orig_content = open(target_file).read()
+
+    # Setup dummy test
+    try:
+        with get_codebase_session(
+            tmpdir=tmpdir,
+            # NOTE: For this edge case in particular, the final file SHOULD be an invalid python file!
+            files={target_file: "This should not be here\n" + target_orig_content},
+            programming_language=ProgrammingLanguage.PYTHON,
+            verify_output=True,
+        ) as codebase:
+            file = codebase.get_file(target_file)
+            assert file.content is not None
+            assert file.content != ""
+            assert file.content != target_orig_content
+            assert "This should not be here" in file.content
+    except SyntaxError:
+        pass  # This is expected!
+
+    # Verify the file on codegen-sdk was not edited
+    assert open(target_file).read() == target_orig_content

tests/unit/codegen/sdk/codebase/session/test_verify_output.py

@@ -0,0 +1,19 @@
+from codegen.sdk.codebase.factory.get_session import get_codebase_session
+from codegen.shared.enums.programming_language import ProgrammingLanguage
+
+
+def test_verify_output_python(tmpdir) -> None:
+    try:
+        with get_codebase_session(
+            tmpdir=tmpdir,
+            files={"invalid.py": "print('Hello, world!'", "valid.py": "print('Hello, world!')"},
+            programming_language=ProgrammingLanguage.PYTHON,
+            verify_output=True,
+        ) as codebase:
+            pass
+
+        # If we reach here, the verify_output failed
+        msg = "Verify output failed"
+        raise Exception(msg)
+    except SyntaxError:
+        pass