Candidate Duplicate Test Cases For Cleanup

[!TIP]
For best results, initiate chat on the files or code changes.

#!/bin/bash
python3 - <<'EOF'
import ast
import os
import sys
from pathlib import Path

def find_duplicates_in_list(elements):
    seen = []
    dupes = []
    for i, el in enumerate(elements):
        key = ast.dump(el)
        for j, (prev_key, prev_idx) in enumerate(seen):
            if key == prev_key:
                dupes.append((prev_idx, i))
                break
        else:
            seen.append((key, i))
    return dupes

results = []

for path in Path("tests").rglob("*.py"):
    try:
        src = path.read_text(errors="replace")
        tree = ast.parse(src, filename=str(path))
    except SyntaxError:
        continue

    for node in ast.walk(tree):
        if not isinstance(node, ast.Assign):
            continue
        for target in node.targets:
            if not isinstance(target, ast.Name):
                continue
            name = target.id
            if "TEST" not in name.upper() and "CASES" not in name.upper() and "PARAMS" not in name.upper():
                continue
            val = node.value
            if not isinstance(val, ast.List):
                continue
            elems = val.elts
            if len(elems) < 2:
                continue
            dupes = find_duplicates_in_list(elems)
            if dupes:
                results.append((str(path), name, dupes))

for fpath, varname, dupes in sorted(results):
    for (i, j) in dupes:
        print(f"{fpath}  {varname}  indices {i} and {j} are identical")

if not results:
    print("No duplicates found.")
EOF

@ericspod Here's the full list of files with byte-identical duplicate entries in their test case lists:

apps/

• tests/apps/deepgrow/test_deepgrow_dataset.py — TEST_CASE_4, TEST_CASE_5
• tests/apps/detection/utils/test_detector_utils.py — TEST_CASE_3
• tests/apps/pathology/transforms/post/test_generate_succinct_contour.py — TEST_CASE_1
• tests/apps/pathology/transforms/post/test_generate_succinct_contourd.py — TEST_CASE_1

bundle/

• tests/bundle/test_bundle_ckpt_export.py — TEST_CASE_1
• tests/bundle/test_bundle_trt_export.py — TEST_CASE_1, TEST_CASE_2

data/

• tests/data/test_arraydataset.py — TEST_CASE_1/2/3, test_labels
• tests/data/test_lmdbdataset.py — TEST_CASE_3
• tests/data/test_nrrd_reader.py — TEST_CASE_1 through TEST_CASE_8 (indices 0 and 2 identical in each)
• tests/data/test_persistentdataset.py — TEST_CASE_4/5/6/7
• tests/data/test_pil_reader.py — TEST_CASE_1 through TEST_CASE_6
• tests/data/utils/test_decollate.py — test_case

handlers/

• tests/handlers/test_handler_checkpoint_saver.py — TEST_CASE_1 through TEST_CASE_8 (many cross-index duplicates — likely worth a separate look; the list entries may be intentional combinations of primitives)
• tests/handlers/test_handler_hausdorff_distance.py — TEST_SAMPLE_2, TEST_SAMPLE_4
• tests/handlers/test_handler_surface_distance.py — TEST_SAMPLE_2, TEST_SAMPLE_4

inferers/

• tests/inferers/test_patch_inferer.py — 9 different TEST_CASE_* variables (indices 0 and 3 identical in each)

metrics/

• tests/metrics/test_calibration_metric.py — TEST_VALUE_1B1C

networks/

• tests/networks/blocks/test_mlp.py, test_se_blocks.py, test_selfattention.py — test_case
• tests/networks/layers/test_savitzky_golay_filter.py — TEST_CASE_1D_REP, TEST_CASE_SINGLE_VALUE_REP
• tests/networks/nets/test_ahnet.py — 5 cases
• tests/networks/nets/test_autoencoder.py — TEST_CASE_0/1/2
• tests/networks/nets/test_autoencoderkl.py — CASES_ATTENTION
• tests/networks/nets/test_cell_sam_wrapper.py, test_fullyconnectednet.py, test_spade_autoencoderkl.py, test_varautoencoder.py, test_vnet.py, test_voxelmorph.py
• tests/networks/test_save_state.py — TEST_CASE_6, TEST_CASE_10

transforms/

• tests/transforms/test_border_pad.py, test_border_padd.py
• tests/transforms/test_center_spatial_crop.py
• tests/transforms/test_cucim_dict_transform.py, test_cucim_transform.py (and rand variants)
• tests/transforms/test_fill_holes.py, test_fill_holesd.py — 8 cases each
• tests/transforms/test_load_image.py — 5 cases
• tests/transforms/test_rand_torchio.py, test_rand_torchiod.py, test_torchio.py, test_torchiod.py
• tests/transforms/test_savitzky_golay_smooth.py, test_savitzky_golay_smoothd.py
• tests/transforms/test_spatial_padd.py

The test_handler_checkpoint_saver.py cases are likely combining shared primitive values intentionally (not pure parametrize duplicates), so worth manually verifying those. The rest look like straightforward cleanup candidates similar to this PR.

Originally posted by @coderabbitai[bot] in https://github.com/Project-MONAI/MONAI/issues/8896#issuecomment-4773223542

Start by reviewing the listed duplicate test-case definitions under tests/, especially tests/handlers/test_handler_checkpoint_saver.py where combinations may be intentional. Compare each reported pair in its surrounding test file, then run the affected tests after removing only confirmed duplicates. Done means redundant entries are cleaned up without losing distinct coverage.