Preprocessing unit tests

[sfluegel05]

We already have some tests for data preprocessing. However, those are more integration tests that capture the behaviour of the tool as a whole than unit tests for specific functions. In order to efficiently test the different preprocessing functionalities, we need to add some smaller-scale unit tests. Those should not include real data, but sample input values that can be generated from scratch.

Here are the classes / functions that should be covered (from the implementation in the protein_prediction branch reader.py:

• DataReader: to_data()
• ChemDataReader: _read_data()
• DeepChemDataReader: _read_data()
• SelfiesReader: _read_data()
• ProteinDataReader: _read_data() collate.py:
• DefaultCollator: __call__()
• RaggedCollator: __call__(), process_label_rows() datasets/base.py
• XYBaseDataModule: _filter_labels()
• DynamicDataset: get_test_split(), get_train_val_splits_given_test() datasets/chebi.py
• _ChEBIDataExtractor: _extract_class_hierarchy(), _graph_to_raw_dataset(), _load_dict(), _setup_pruned_test_set()
• ChEBIOverX: select_classes()
• ChEBIOverXPartial: extract_class_hierarchy()
• term_callback() datasets/go_uniprot.py:
• _GOUniprotDataExtractor: _extract_class_hierarchy(), term_callback(), _graph_to_raw_dataset(), _get_swiss_to_go_mapping(), _load_dict()
• _GoUniProtOverX: select_classes() datasets/tox21.py:
• ~Tox21MolNet: setup_processed(), _load_data_from_file()~
• Tox21Challenge: setup_processed(), _load_data_from_file(), _load_dict()

For some functions, it is necessary to read from / write to files. Instead of real files, I would suggest to use mock objects (see e.g. this comment)

[aditya0by0]

How to Run All Tests in a Given Folder

To run all unit tests in a specific folder using Python's unittest framework, you can use the unittest discover command.

Command Example:

python -m unittest discover -s <test_folder> -t <top-level-directory> -p "<pattern>" 

Explanation of Parameters:

• -s <test_folder>: This specifies the folder to search for test files. In our case, <test_folder> would be the path to the directory containing the tests (e.g., tests/unit).

• -p <pattern>: (Optional) This allows you to define a file name pattern for matching the test files. For example:

  • -p "*_test.py" will match files like example_test.py.
  • -p "test_*.py" will match files like test_example.py.
  • *If omitted, the default is `test.py(files starting withtest`).**

• -t <top_level_directory>: Specifies the top-level directory of the project (root). This is necessary when the tests are located in a subdirectory but need to be executed within the context of the top-level project directory. This is where Python will consider the root of the project for resolving module imports during testing.

Running All Tests in tests/unit Folder Example:

python -m unittest discover -s G:\github-aditya0by0\python-chebai\tests\unit -t G:\github-aditya0by0\python-chebai

[aditya0by0]

• Tox21MolNet: setup_processed(), _load_data_from_file()

The test case for above class is not added/covered in this issue, due to issue #53. It will added after completion of the issue.

[sfluegel05]

The toxicity-related tests have not been merged yet. They will be included in #56 (see also issue #53)