[SYSTEMDS-3971] Improve Scuro test coverage

[shieru1214]

This change adds 13 tests for persistence and results retrieval in unimodal_optimizer.py.

Persistence

• store_results: It saves the optimizer results and execution statistics in two separate files. The tests check the saved contents, the automatically generated file name, and the fallback when the file name does not end with .pkl.
• load_results: It loads previously saved results. The test stores one result, clears it from memory, and then checks that it can be loaded again.
• _count_results_by_modality: It counts the results from the first task of each modality for checkpoint progress. The test checks the returned count for both a modality with results and an empty modality.
• resume_from_checkpoint: It restores results from a checkpoint. The tests cover both restoring saved results and keeping the current results when no checkpoint is found.

Results Retrieval

• get_k_best_results: It returns the top k results according to their scores. The tests check the score order, using existing cached data, executing the result DAGs when the cache is empty, and returning empty lists when there are no results.
• get_dag_by_id: It searches a list of DAGs by dag_id. The tests cover both returning the matching DAG and returning None when the ID is not found.

[shieru1214]

This change improves the test coverage of hyperparameter_tuner.py. The added tests focus on three parts: search space construction, trial parameter, and tuning result handling.

1. Search space construction

This part converts the parameter ranges declared by operators into search spaces that Optuna can use. _build_param_specs is the entry function and it was already fully covered, so I mainly added tests for three helper functions. For _param_values_to_spec, I tested list, tuple, scalar, other iterable, and invalid inputs.
For _window_input_stats, I tested whether it can create input length information from window_size. This information is later used to remove candidate values that do not fit the input.
For _expand_aggregation_param_specs, I added five cases for flattening nested aggregation parameters, narrowing the domain based on the input length, and inputs that cannot be expanded.

2. Trial parameter

The first part flattens the parameters, while this part puts the values selected by Optuna back into the DAG nodes. The main function is _apply_trial_params_to_node. It first takes the parameters belonging to the current node, and then decides where to place them based on the form of the node. I added tests for two branches: a node with a pushed-down aggregation and a node that is itself an AggregatedRepresentation.
For _apply_pushdown_trial_params, I tested whether aggregation parameters are correctly stored in _pushdown_aggregation, whether other parameters stay at the top level, and whether the original parameters remain unchanged. I also added tests for _materialize_node_params and the two operation checks, including non-class inputs, import failures, leaf nodes, and empty parameters.

3. Tuning result handling

This part tests how tuning results are stored and read. For add_result, I added two cases: storing multimodal results in mm_results, and skipping a tuning result when it is None. setup_mm uses optimize_unimodal to decide whether the result structure should be replaced with an mm_results entry. This flag indicates whether the unimodal representations at the leaf nodes are also optimized. The tests for get_k_best_dags check that best_params are applied to the rebuilt DAG and that the original DAG is not changed. The test for get_k_best_results checks that the rebuilt DAG is executed and the last value is taken from its output dictionary.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 75.66%. Comparing base (8c7d556) to head (b3134b4).
⚠️ Report is 19 commits behind head on main.

Additional details and impacted files

@@              Coverage Diff              @@
##               main    #2615       +/-   ##
=============================================
+ Coverage     71.81%   75.66%    +3.84%     
=============================================
  Files          2075      436     -1639     
  Lines        222176    25465   -196711     
  Branches      38264        0    -38264     
=============================================
- Hits         159559    19268   -140291     
+ Misses        51561     6197    -45364     
+ Partials      11056        0    -11056     

Flag	Coverage Δ
java	?
python	75.66% <100.00%> (+0.70%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[shieru1214]

This change adds tests for multimodal_optimizer.py, which cover three parts: fusion DAG generation, constructor setup, and the optimize loop.

Fusion DAG Generation

Before this change, the three methods that generate the fusion search space had no active tests.

• _generate_modality_combinations: test_modality_combinations_respect_min_max_modalities uses three modalities and checks the exact subsets for max_modalities values of 2, 3, and 5. This also checks the case where the maximum is larger than the number of available modalities.
• _generate_representation_combinations: test_representation_combinations_pick_one_per_modality uses one modality with two representations and another with three. It checks all six possible combinations.
• _generate_fusion_dags: test_fusion_dags_contain_every_selected_representation checks every generated DAG. It checks that the selected modality and representation leaves are present, the number of fusion nodes is correct, each fusion node has two inputs, and the root is a fusion node.
• test_fusion_dags_use_every_fusion_operator checks that both configured operators, Concatenation and Average, are used. This is tested separately because a DAG can have the correct structure but still miss one operator.

Constructor Setup

The generator tests already called __init__, but it did not directly check the values created during construction. I added three tests for this part.

• test_min_modalities_clamped_to_two checks that min_modalities=1 is changed to 2. This prevents single-modality DAGs from being treated as fusion DAGs.
• test_max_modalities_defaults_to_number_of_modalities checks that the default maximum is the number of available modalities.
• test_k_best_representations_hold_the_cached_data covers _extract_k_best_representations. It checks that every modality is stored by its ID and that the optimizer keeps the cached representation data returned by the unimodal results, instead of the score records.

Optimize Loop

Before this change, none of the tests called optimize. I added three tests for the main logic of this loop.

• test_stops_at_max_combinations uses three modalities and sets max_combinations=5. It checks that _evaluate_dag is called exactly five times and that five results are stored.
• test_drops_failed_evaluations makes every second evaluation return None. It checks that six evaluations are attempted but only three results are stored. Failed evaluations still count towards the limit, and the loop continues after them.
• test_keeps_results_and_budget_per_task uses two tasks with a limit of two. It checks that each task receives two evaluations and that every result is stored under the correct task name.