The folders here contains golden files and test file that are used for testing parser.

There are two kinds of files. .scope which contains print of the scope tree, which is generated during first phase of parsing and .tast files, which contains truffle AST. For one test can be generated one .scope file and one .tast file

Folder goldenFiles contains golden files for tests, where the parsed source for the test is a string. The structure is that the folder is class name, where the test is and the name of file is to the of the test. So for example:

public class AssignmentTests extends ParserTestBase {

    @Test
    public void assignment01() throws Exception {
        checkTreeResult("a = 1");
    }

}

creates file assignment01.tast (if not exists yet) in folder AssignmentTests in goldenFiles folder.

The testFiles folder has the same structure, but also can contains tested .py file. The name of the test is the name of file with .py extension that is tested.

For example:

public class RuntimeFileTests extends ParserTestBase {

    @Test
    public void _collections_abc() throws Exception {
        checkScopeAndTree();
    }

}

This test is looking for file testFiles/RuntimeFileTests/_collections_abc.py and after first run are created testFiles/RuntimeFileTests/_collections_abc.scope and testFiles/RuntimeFileTests/_collections_abc.tast golden files.

As is mentioned the first run of these test generate the golden file for testing source. The golden file is generated with the old parser and compares the result from new parser. If the old parser generate wrong tree, then it should be manually corrected or reused the output from new parser.

If there is a change in a tree, that causes failing many tests, then it can be done by deleting all .tast files, then run parser tests and check, whether the diffs are correct.

These commands should do it:

cd graalpython/com.oracle.graal.python.test/testData
find . -name "*.tast" -delete
mx punittest com.oracle.graal.python.test.parser

After this all .tast files are regenerated. Currently after the regeneration there are two test failing DictAndSetTests.dict09 and DictAndSetTests.dict10. It's because the old parser generated the tree in wrong way. These are small tests and should be enough just the revert the modifications and provide needed changes manually if there are necessary.

The same aproach can be used for .scope files.