File size: 10,756 Bytes
8846f87
f9889e8
7694e55
8846f87
 
 
f9889e8
8846f87
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
f9889e8
 
 
8846f87
f9889e8
 
 
5fdf7b4
f9889e8
 
8846f87
f9889e8
 
 
 
 
 
 
7694e55
f9889e8
7694e55
f9889e8
5fdf7b4
 
 
7694e55
5fdf7b4
 
7694e55
f9889e8
5fdf7b4
 
 
7694e55
5fdf7b4
 
7694e55
f9889e8
5fdf7b4
 
 
7694e55
5fdf7b4
 
7694e55
5fdf7b4
 
 
 
7694e55
5fdf7b4
 
7694e55
5fdf7b4
 
 
 
7694e55
5fdf7b4
 
7694e55
f9889e8
 
 
 
 
 
 
 
 
 
 
 
5fdf7b4
59a2092
f9889e8
8846f87
 
 
 
f9889e8
7694e55
 
 
 
 
 
8846f87
 
 
7694e55
8846f87
7694e55
 
 
 
 
 
 
8846f87
 
 
 
 
 
 
 
7694e55
 
f9889e8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5fdf7b4
 
 
7694e55
 
f9889e8
 
 
 
 
 
5fdf7b4
 
 
7694e55
 
f9889e8
 
 
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
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
from typing import Tuple, Dict, Any, List, Optional
from models import Observation, Action, Reward, State
from grader import grade_comment, grade_question, grade_fix
import sys
import io
import contextlib

# ------------------------- Simulated CI / Unit tests -------------------------
def run_unit_tests(fix_code: str, task: str) -> float:
    """
    Runs a small set of unit tests for the given task.
    Returns a score in [0,1] based on passed tests.
    """
    # Define tests per task
    test_code = ""
    if task == "easy":
        # Test that the function handles missing keys
        test_code = f"""
{fix_code}
def test():
    try:
        users = {{"alice": "Alice"}}
        result = get_user("bob")
        return False  # should not get here if key missing
    except KeyError:
        return True  # expected: KeyError
    except Exception:
        return False
"""
    elif task == "medium":
        test_code = f"""
{fix_code}
def test():
    items = [1,2,3]
    # We cannot directly test the loop, but we can check that 'process' is called correctly.
    # For demonstration, we'll assume the fix uses 'enumerate' or 'for item in'.
    # Here we just check that the code compiles and runs without error.
    try:
        exec(compile("{fix_code}", "<string>", "exec"))
        return True
    except Exception:
        return False
"""
    elif task == "hard":
        test_code = f"""
{fix_code}
def test():
    # Test empty list
    try:
        result = calculate_average([])
        return result == 0  # expect 0 or some default
    except ZeroDivisionError:
        return False
"""
    elif task == "harder":
        test_code = f"""
{fix_code}
def test():
    # Check that a lock is used
    if "lock" in "{fix_code}".lower():
        return True
    return False
"""
    else:  # hardest
        test_code = f"""
{fix_code}
def test():
    # Check for lock order mention
    if "same order" in "{fix_code}".lower() or "lock order" in "{fix_code}".lower():
        return True
    return False
"""

    # Execute the test in a safe sandbox
    try:
        # Capture stdout/stderr
        f = io.StringIO()
        with contextlib.redirect_stdout(f), contextlib.redirect_stderr(f):
            exec(test_code, {})
        # Check if test function returns True
        local_ns = {}
        exec(test_code, {}, local_ns)
        if 'test' in local_ns and callable(local_ns['test']):
            passed = local_ns['test']()
            return 1.0 if passed else 0.0
        else:
            return 0.0
    except Exception:
        return 0.0

# ------------------------- Simulated PR Author -------------------------
class SimulatedAuthor:
    """Responds to the agent's questions and comments as if they were the PR author."""
    def __init__(self, task: str):
        self.task = task

    def respond(self, agent_comment: str, agent_question: str = None) -> str:
        if agent_question:
            q = agent_question.lower()
            if "what" in q and "purpose" in q:
                return "The purpose is to retrieve a user safely."
            elif "expected" in q:
                return "It should return the user or raise KeyError."
            else:
                return "Could you be more specific?"
        else:
            # Generic response to a comment
            if "good" in agent_comment.lower():
                return "Thanks for the feedback!"
            else:
                return "I'll consider your suggestion."

# ------------------------- Main Environment -------------------------
class CodeReviewEnv:
    def __init__(self, task: str = "easy"):
        self.task = task
        self.author = None
        self.reset()

    def set_task(self, task: str):
        if task not in ["easy", "medium", "hard", "harder", "hardest"]:
            raise ValueError(f"Unknown task: {task}")
        self.task = task
        self.author = SimulatedAuthor(task)

    def reset(self) -> Observation:
        if self.task is None:
            raise RuntimeError("Task not set. Call set_task() first.")
        self.step_count = 0
        self.agent_comment = None
        self.done = False
        self.test_results = None

        # Task definitions (same as before)
        if self.task == "easy":
            self.pr_title = "Fix missing null check in user lookup"
            self.pr_description = "The current code does not handle missing user IDs. It raises a KeyError."
            self.code_snippet = "def get_user(id):\n    return users[id]  # missing null check"
            self.comments = []
            self.expected_keywords = ["null", "key", "missing", "check", "exists", "handle"]
            self.expert_comment = "Add a check to ensure the key exists before accessing the dictionary to avoid KeyError."
            self.expected_fix_keywords = ["if id in users"]
        elif self.task == "medium":
            self.pr_title = "Improve loop efficiency"
            self.pr_description = "The loop uses `range(len(items))` which is inefficient and less readable."
            self.code_snippet = "for i in range(len(items)):\n    process(items[i])\n# O(n^2) when it could be O(n)"
            self.comments = []
            self.expected_keywords = ["enumerate", "for item in", "range", "inefficient", "optimize"]
            self.expert_comment = "Use `for item in items:` for a more Pythonic and efficient loop."
            self.expected_fix_keywords = ["for item in items", "for i, item in enumerate"]
        elif self.task == "hard":
            self.pr_title = "Handle division by zero in average calculation"
            self.pr_description = "The function crashes when the input list is empty."
            self.code_snippet = "def calculate_average(data):\n    total = sum(data)\n    return total / len(data)  # what if data is empty?"
            self.comments = []
            self.expected_keywords = ["empty", "zero", "length", "check", "handle", "exception"]
            self.expert_comment = "Check if the list is empty and return a sensible default (e.g., 0) or raise a descriptive error."
            self.expected_fix_keywords = ["if not data", "if len(data)==0"]
        elif self.task == "harder":
            self.pr_title = "Fix race condition in counter increment"
            self.pr_description = "Multiple threads increment a counter without synchronization, causing lost updates."
            self.code_snippet = "counter = 0\ndef increment():\n    global counter\n    counter += 1\n# called from multiple threads"
            self.comments = []
            self.expected_keywords = ["thread", "lock", "synchronization", "atomic", "race", "concurrent"]
            self.expert_comment = "Use a threading.Lock to protect the counter increment, or use an atomic operation like `threading.atomic`."
            self.expected_fix_keywords = ["lock", "threading.Lock", "with lock"]
        else:  # hardest
            self.pr_title = "Fix deadlock in database transaction"
            self.pr_description = "Two threads acquire locks in opposite order, leading to potential deadlock."
            self.code_snippet = "with lock1:\n    with lock2:\n        do_work()\n# another thread does lock2 then lock1"
            self.comments = []
            self.expected_keywords = ["deadlock", "lock order", "acquire", "release", "trylock", "timeout"]
            self.expert_comment = "Ensure all threads acquire locks in the same order to prevent deadlock. Consider using a timeout or a single lock."
            self.expected_fix_keywords = ["same order", "lock order", "acquire lock1 then lock2"]

        return self._get_observation()

    def step(self, action: Action) -> Tuple[Observation, Reward, bool, Dict[str, Any]]:
        if self.done:
            raise RuntimeError("Episode already finished")

        reward = 0.0
        info = {}

        if action.action_type == "write_comment":
            self.agent_comment = action.comment_text or ""
            reward = 0.2  # dense bonus for writing
            quality_score = grade_comment(self.agent_comment, self.expected_keywords, self.expert_comment)
            reward += quality_score
            # Simulate author response
            author_response = self.author.respond(self.agent_comment)
            self.comments.append(f"Agent: {self.agent_comment}")
            self.comments.append(f"Author: {author_response}")
            self.done = True

        elif action.action_type == "ask_question":
            if not action.question:
                reward = -0.1
            else:
                q_score = grade_question(action.question)
                reward = 0.1 + q_score
                # Get answer from simulated author
                answer = self.author.respond(agent_question=action.question)
                self.comments.append(f"Agent: {action.question}")
                self.comments.append(f"Author: {answer}")
                self.step_count += 1
                # Episode continues, not done

        elif action.action_type == "propose_fix":
            if not action.fix_code:
                reward = -0.2
            else:
                # Run CI tests
                test_score = run_unit_tests(action.fix_code, self.task)
                # Also keyword match for partial credit
                kw_score = grade_fix(action.fix_code, self.expected_fix_keywords, None)
                # Combined score: 70% tests, 30% keywords
                combined_score = 0.7 * test_score + 0.3 * kw_score
                reward = 0.3 + combined_score
                self.test_results = f"CI tests passed: {test_score:.0%}, Keywords: {kw_score:.0%}"
            self.done = True

        elif action.action_type == "skip":
            reward = -0.1
            self.done = True
        elif action.action_type == "done":
            reward = -0.5
            self.done = True
        else:
            reward = -0.2
            self.done = True

        self.step_count += 1
        obs = self._get_observation()
        return obs, Reward(value=reward), self.done, info

    def _get_observation(self) -> Observation:
        return Observation(
            pr_title=self.pr_title,
            pr_description=self.pr_description,
            code_snippet=self.code_snippet,
            comments=self.comments.copy(),
            test_results=self.test_results,
            step=self.step_count,
            done=self.done
        )

    def state(self) -> State:
        return State(
            pr_title=self.pr_title,
            pr_description=self.pr_description,
            code_snippet=self.code_snippet,
            comments=self.comments.copy(),
            test_results=self.test_results,
            step=self.step_count,
            done=self.done
        )