Testora/src/testora/evaluation/ClassificationResultsSummarizer.py at main · michaelpradel/Testora · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
from collections import Counter
import json

# Select the combination of prompt and model to evaluate:

# single-question, GPT-4o
# result_files = ["data/classification_results_03_2025/keras/single-question_GPT-4o.json",
#                 "data/classification_results_03_2025/marshmallow/single-question_GPT-4o.json",
#                 "data/classification_results_03_2025/pandas/single-question_GPT-4o.json",
#                 "data/classification_results_03_2025/scipy/single-question_GPT-4o.json"]

# single-question, GPT-4o-mini
# result_files = ["data/classification_results_03_2025/keras/single-question_GPT-4o-mini.json",
#                 "data/classification_results_03_2025/marshmallow/single-question_GPT-4o-mini.json",
#                 "data/classification_results_03_2025/pandas/single-question_GPT-4o-mini.json",
#                 "data/classification_results_03_2025/scipy/single-question_GPT-4o-mini.json"]

# single-question, DeepSeek-R1
# result_files = ["data/classification_results_03_2025/keras/single-question_DeepSeek-R1.json",
#                 "data/classification_results_03_2025/marshmallow/single-question_DeepSeek-R1.json",
#                 "data/classification_results_03_2025/pandas/single-question_DeepSeek-R1.json",
#                 "data/classification_results_03_2025/scipy/single-question_DeepSeek-R1.json"]

# multi-question, GPT-4o
# result_files = ["data/classification_results_03_2025/keras/multi-question_GPT-4o.json",
#                 "data/classification_results_03_2025/marshmallow/multi-question_GPT-4o.json",
#                 "data/classification_results_03_2025/pandas/multi-question_GPT-4o.json",
#                 "data/classification_results_03_2025/scipy/multi-question_GPT-4o.json"]

# multi-question, GPT-4o-mini
result_files = ["data/classification_results_03_2025/keras/multi-question_GPT-4o-mini.json",
                "data/classification_results_03_2025/marshmallow/multi-question_GPT-4o-mini.json",
                "data/classification_results_03_2025/pandas/multi-question_GPT-4o-mini.json",
                "data/classification_results_03_2025/scipy/multi-question_GPT-4o-mini.json"]

# multi-question, DeepSeek-R1
# result_files = ["data/classification_results_03_2025/keras/multi-question_DeepSeek-R1.json",
#                 "data/classification_results_03_2025/marshmallow/multi-question_DeepSeek-R1.json",
#                 "data/classification_results_03_2025/pandas/multi-question_DeepSeek-R1.json",
#                 "data/classification_results_03_2025/scipy/multi-question_DeepSeek-R1.json"]

# extract results
print("Project, PR, Prediction, Label, Result")
nb_fps = 0
nb_tps = 0
nb_fns = 0
nb_tns = 0
variance_ctr = Counter()
config_used = None
for result_file in result_files:
    if "keras" in result_file:
        project = "keras"
    elif "marshmallow" in result_file:
        project = "marshmallow"
    elif "pandas" in result_file:
        project = "pandas"
    elif "scipy" in result_file:
        project = "scipy"
    else:
        raise ValueError(f"Couldn't determine project from file name {result_file}")

    with open(result_file, "r") as f:
        result_json = json.load(f)
        config_used_here = result_json[0]["message"]
        if config_used is None:
            config_used = config_used_here
        else:
            assert config_used == config_used_here, f"Config mismatch:\nUsed before:\n{config_used}\nvs used now in {project}:\n {config_used_here}"
        for entry in result_json:
            if entry["message"] == "Classification result":
                # compare label and predictions
                results = []
                if entry["label"] in ["unintended", "coincidental fix"]:
                    for prediction in entry["predictions"].split("#"):
                        if prediction == "unintended":
                            results.append("TP")
                            nb_tps += 1
                        elif prediction == "intended":
                            results.append("FN")
                            nb_fns += 1
                        else:
                            raise ValueError(
                                f"Invalid prediction: {entry['prediction']}")
                elif entry["label"] in ["intended"]:
                    for prediction in entry["predictions"].split("#"):
                        if prediction == "intended":
                            results.append("TN")
                            nb_tns += 1
                        elif prediction == "unintended":
                            results.append("FP")
                            nb_fps += 1
                        else:
                            raise ValueError(
                                f"Invalid prediction: {entry['prediction']}")
                else:
                    raise ValueError(
                        f"Invalid label: {entry['label']}, {entry['pr_nb']}")

                # check variance of predictions
                results_counter = Counter(results)
                variance_str = str(
                    sorted(list(results_counter.values()), reverse=True))
                variance_ctr[variance_str] += 1

                # print into CSV
                print(f"{project}, "
                      f"{entry['pr_nb']}, "
                      f"{entry['predictions']}, "
                      f"{entry['label']}, "
                      f"{', '.join(results)}"
                      )

print(config_used)
print()
print(f"TP: {nb_tps}, FP: {nb_fps}, FN: {nb_fns}, TN: {nb_tns}")
precision = 0 if (nb_tps + nb_fps) == 0 else nb_tps / (nb_tps + nb_fps)
print(f"Precision: {precision}")
recall = 0 if (nb_tps + nb_fns) == 0 else nb_tps / (nb_tps + nb_fns)
print(f"Recall: {recall}")
f1 = 0 if (precision + recall) == 0 else 2 * \
    precision * recall / (precision + recall)
print(f"F1: {f1}")
print()
print(f"Variance of predictions: {variance_ctr}")
print(f"Total data points: {nb_tps+nb_fps+nb_fns+nb_tns}")