[TEST] Test descriptor_calculator for various input types

[akalikadien]

---

We need to implement test(s) for the following scenario:

Description

Create csv files with descriptors given a set of xyz files, CREST folders or DFT log files

Code Reference

• case for xyz files: https://github.com/EPiCs-group/obelix/blob/d95e1ee3d7f345a0490abe7dc8992a7bf3a4854f/obelix/descriptor_calculator.py#L651
• case for CREST folders: https://github.com/EPiCs-group/obelix/blob/d95e1ee3d7f345a0490abe7dc8992a7bf3a4854f/obelix/descriptor_calculator.py#L685
• case for DFT log files: https://github.com/EPiCs-group/obelix/blob/d95e1ee3d7f345a0490abe7dc8992a7bf3a4854f/obelix/descriptor_calculator.py#L778
• examples can be found here: https://github.com/EPiCs-group/obelix/blob/d95e1ee3d7f345a0490abe7dc8992a7bf3a4854f/obelix/descriptor_calculator.py#L864

  Test Cases

  Case:

• Take a set of input files and calculate descriptors
• Preconditions: pre-generated xyz files, CREST folders and DFT log files of complexes with a bidentate ligand that needs to be featurized
• Test Steps:
• Run calculate_morfeus_descriptors on xyz files and CREST folders (separately)
• Run calculate_dft_descriptors_from_log on folder with DFT log files
• Test Data: [ if required, data to execute the test cases ]
• Expected Result:
• A csv file with a list of descriptors
• Post-condition:
• Check whether csv with descriptors is generated
• Compare csv with a predefined reference

---

[akalikadien]

The examples for testing the descriptors calculator can be found here: descriptor_calculator_test_examples.zip. As discussed, for maximum code coverage we can have 4 test:

• Use an xyz file generated with mace, metal_adduct = pristine
• Use L1_example.xyz, metal_adduct = nbd
• Use L1_example.log to test the DFT descriptors, metal_adduct = nbd
• Use L40_example/ to test CREST input, metal_adduct = pristine

[niketagrawal]

Implementation of this test is being done in the branch 25-test-descriptor-calculator

Done

• Testcases parametrized for the following metal_adduct and output_type combination: [nbd, xyz files], [pristine, xyz files], [nbd, dft log file]
• testcases: assert descriptor values, index values, filename valus, and element values in the csv file produced

Issues to fix

• [x] The descriptor values generated by the dft log file (L1_example.log) contains non numeric data. It contains a ‘True’. numpy array comparison fails, so assert descriptor values fails All descriptor values are supposed to be a number, however, there are also ‘metadata’ columns in the dataset like the ‘element_x’ or ‘index_x’ column. This also goes for the ‘optimization_success_dft’ column, it is practically an indication of whether the DFT optimization actually succeeded or not, if it failed then the descriptors are not calculated. Fix: Convert the boolean to binary in the numpy array

Besides the boolean, there were other non-numeric types in the array. They were converted to numeric types using expected_descriptor_values = pd.to_numeric( expected_descriptor_values.flatten(), errors="coerce" ).reshape(expected_descriptor_values.shape)

• [x] The L40_example/ generates an empty descriptors.csv file.