File size: 12,542 Bytes
ff10d5b
 
 
9f883be
ff10d5b
 
 
9f883be
ff10d5b
 
 
 
9f883be
 
 
 
 
 
 
 
ff10d5b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9f883be
 
 
 
ff10d5b
 
 
 
 
 
 
 
 
 
 
 
 
9f883be
ff10d5b
 
 
9f883be
ff10d5b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9f883be
ff10d5b
 
 
 
 
9f883be
ff10d5b
 
 
 
9f883be
ff10d5b
 
 
9f883be
 
 
ff10d5b
 
 
 
9f883be
ff10d5b
 
9f883be
ff10d5b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9f883be
 
 
 
ff10d5b
 
9f883be
 
 
 
 
ff10d5b
9f883be
 
 
 
 
 
ff10d5b
 
9f883be
 
ff10d5b
9f883be
ff10d5b
 
 
 
9f883be
 
 
ff10d5b
 
9f883be
 
 
ff10d5b
 
 
9f883be
ff10d5b
 
 
 
9f883be
ff10d5b
9f883be
 
ff10d5b
9f883be
 
 
 
 
ff10d5b
 
 
 
 
 
 
9f883be
ff10d5b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9f883be
ff10d5b
 
 
 
9f883be
 
ff10d5b
 
 
 
 
 
9f883be
d366e7a
9f883be
 
 
 
 
 
 
 
 
ff10d5b
 
 
9f883be
 
ff10d5b
 
 
 
 
 
 
9f883be
ff10d5b
 
 
9f883be
 
ff10d5b
 
9f883be
ff10d5b
9f883be
ff10d5b
 
9f883be
 
 
 
 
 
 
 
ff10d5b
 
9f883be
 
 
 
ff10d5b
 
 
9f883be
 
ff10d5b
 
9f883be
 
ff10d5b
 
9f883be
 
ff10d5b
 
9f883be
 
 
ff10d5b
9f883be
 
ff10d5b
 
9f883be
 
ff10d5b
 
9f883be
d366e7a
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
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
"""
env/db_simulator.py β€” SQL Database Engineer Agent
Simulates a production database responding to optimization actions.
Core mechanism: index coverage reduces query execution time by up to 85-90%.
"""

import math
import random
from typing import Optional


class DatabaseSimulator:
    """
    Simulates a production database that degrades over time.
    The agent applies optimization actions and sees performance scores change.

    Performance score: 0-100 (100 = all queries running at target speed).
    The agent's goal: get performance_score >= target_score.
    """

    def __init__(self, scenario: dict):
        self.scenario     = scenario
        self.tables       = {t["name"]: dict(t) for t in scenario["tables"]}
        self.queries      = [dict(q) for q in scenario["slow_queries"]]
        self.indexes      = {
            name: list(t.get("indexes", ["PRIMARY"]))
            for name, t in self.tables.items()
        }
        self.stats_fresh  = {name: False for name in self.tables}
        self.partitioned  = {name: False for name in self.tables}
        self.baseline     = self._compute_score()
        self.history      = [self.baseline]
        self.best_score   = self.baseline
        self.target_score = scenario.get("target_score", 85.0)

    # ─────────────────────────────────────────────
    #  PUBLIC ACTIONS
    # ─────────────────────────────────────────────

    def apply_action(self, action_type: str, payload: dict) -> dict:
        """
        Apply an optimization action to the database.
        Returns delta showing performance change.
        """
        old_score = self._compute_score()
        affected  = []

        if action_type == "create_index":
            table = payload.get("table", "")
            cols  = payload.get("columns", [])
            if isinstance(cols, str):
                cols = [c.strip() for c in cols.split(",")]
            idx_name = "idx_" + "_".join(cols)
            if table in self.indexes and idx_name not in self.indexes[table]:
                self.indexes[table].append(idx_name)
                affected = self._queries_benefiting_from_index(table, cols)
            else:
                # Duplicate index β€” no benefit
                return {
                    "old_score": old_score, "new_score": old_score,
                    "delta": 0.0, "affected_queries": [],
                    "improved": False, "message": "Index already exists or table not found."
                }

        elif action_type == "rewrite_query":
            qid     = payload.get("query_id", "")
            new_sql = payload.get("new_sql", "")
            for q in self.queries:
                if q["id"] == qid:
                    improvement = self._estimate_rewrite(new_sql, q)
                    q["avg_ms"]  = max(10, int(q["avg_ms"] * (1 - improvement)))
                    affected     = [qid]
                    break

        elif action_type == "partition_table":
            table = payload.get("table", "")
            if table in self.tables and not self.partitioned.get(table):
                self.partitioned[table] = True
                affected = [q["id"] for q in self.queries if table in q.get("sql", "")]

        elif action_type == "analyze_statistics":
            table = payload.get("table", "")
            if table in self.tables:
                self.stats_fresh[table] = True
                affected = [q["id"] for q in self.queries if table in q.get("sql", "")]

        elif action_type == "drop_index":
            table    = payload.get("table", "")
            idx_name = payload.get("index_name", "")
            if idx_name in self.indexes.get(table, []) and idx_name != "PRIMARY":
                self.indexes[table].remove(idx_name)

        elif action_type == "add_column":
            table   = payload.get("table", "")
            col     = payload.get("column", "")
            purpose = payload.get("purpose", "")
            if table in self.tables:
                if "extra_columns" not in self.tables[table]:
                    self.tables[table]["extra_columns"] = []
                self.tables[table]["extra_columns"].append(col)
                # Denormalization can help JOINy queries
                affected = [
                    q["id"] for q in self.queries
                    if "join" in q.get("sql", "").lower() and table in q.get("sql", "")
                ]

        new_score = self._compute_score()
        self.history.append(new_score)
        if new_score > self.best_score:
            self.best_score = new_score

        return {
            "old_score":        round(old_score, 2),
            "new_score":        round(new_score, 2),
            "delta":            round(new_score - old_score, 2),
            "affected_queries": affected,
            "improved":         new_score > old_score,
        }

    def inspect_query(self, query_id: str) -> dict:
        """
        EXPLAIN a slow query β€” reveals scan type, rows examined, cost.
        This is the agent's primary investigation tool.
        """
        for q in self.queries:
            if q["id"] == query_id:
                has_index    = self._check_query_index_coverage(q) > 0.1
                is_partition = self.partitioned.get(q.get("main_table", ""), False)
                rows_examined = 50 if has_index else q.get("rows_examined",
                    self.tables.get(q.get("main_table", ""), {}).get("rows", 50000))

                return {
                    "query_id":         query_id,
                    "sql":              q["sql"],
                    "avg_ms":           q["avg_ms"],
                    "scan_type":        "INDEX RANGE SCAN" if has_index else "FULL TABLE SCAN",
                    "rows_examined":    rows_examined,
                    "partitioned":      is_partition,
                    "optimization_hint": (
                        "Query is using index efficiently."
                        if has_index
                        else "No index covering WHERE columns. Consider adding composite index."
                    ),
                    "main_table":       q.get("main_table", "unknown"),
                }
        return {"error": f"Query '{query_id}' not found"}

    def analyze_indexes(self, table: str) -> dict:
        """
        Show all indexes on a table + usage stats + missing index hints.
        """
        if table not in self.tables:
            return {"error": f"Table '{table}' not found"}

        existing   = self.indexes.get(table, [])
        hints      = [
            h for h in self.scenario.get("missing_index_hints", [])
            if h.get("table") == table
        ]
        used_by    = []
        for q in self.queries:
            cov = self._check_query_index_coverage(q)
            if table in q.get("sql", "") and cov > 0.1:
                used_by.append(q["id"])

        return {
            "table":           table,
            "row_count":       self.tables[table].get("rows", 0),
            "existing_indexes": existing,
            "indexes_used_by": used_by,
            "missing_hints":   hints,
            "stats_fresh":     self.stats_fresh.get(table, False),
            "partitioned":     self.partitioned.get(table, False),
            "size_mb":         self.tables[table].get("size_mb", 0),
        }

    # ─────────────────────────────────────────────
    #  STATE
    # ─────────────────────────────────────────────

    def get_current_state(self) -> dict:
        """Returns the full current DB state for the Observation."""
        return {
            "performance_score": round(self._compute_score(), 2),
            "baseline_score":    round(self.baseline, 2),
            "target_score":      self.target_score,
            "tables":            list(self.tables.values()),
            "slow_queries":      self.queries,
            "indexes":           self.indexes,
            "history":           self.history,
            "best_score":        round(self.best_score, 2),
        }

    def get_performance_score(self) -> float:
        return round(self._compute_score(), 2)

    def is_target_reached(self) -> bool:
        return self._compute_score() >= self.target_score

    # ─────────────────────────────────────────────
    #  INTERNAL SCORING ENGINE
    # ─────────────────────────────────────────────

    def _compute_score(self) -> float:
        """
        Core scoring: calculates performance score 0-100.
        Higher = better. Based on how fast queries run given current indexes.
        """
        if not self.queries:
            return 0.0

        scores = []
        for q in self.queries:
            table       = q.get("main_table", "")
            coverage    = self._check_query_index_coverage(q)
            part_bonus  = 0.30 if self.partitioned.get(table, False) else 0.0
            stats_bonus = 0.05 if self.stats_fresh.get(table, False) else 0.0
            total_reduction = min(coverage * 0.85 + part_bonus + stats_bonus, 0.97)
            effective_ms    = q["avg_ms"] * (1 - total_reduction)
            # Score formula: 100ms = score 99, 1000ms = score 90, 8500ms = ~14
            score = max(0.0, 100.0 - (effective_ms / 100.0))
            scores.append(score)

        return round(sum(scores) / len(scores), 2)

    def _check_query_index_coverage(self, query: dict) -> float:
        """
        Returns 0.0-1.0 representing how well indexes cover this query's WHERE clause.
        0.0 = full table scan, 1.0 = perfect index coverage.
        """
        sql = query.get("sql", "").lower()
        for table, indexes in self.indexes.items():
            if table not in sql:
                continue
            for idx in indexes:
                if idx == "PRIMARY":
                    # Primary key only helps if query filters by primary key
                    if "where id=" in sql or "where id =" in sql:
                        return 0.95
                    continue
                # Extract columns from index name (idx_col1_col2)
                cols = idx.replace("idx_", "").split("_")
                matches = sum(1 for c in cols if c in sql)
                if matches >= 2:
                    return 0.90  # Composite index β€” excellent coverage
                if matches == 1:
                    return 0.60  # Single column β€” partial coverage
        return 0.0

    def _queries_benefiting_from_index(self, table: str, cols: list) -> list:
        """Returns query IDs that would benefit from an index on given table/columns."""
        benefiting = []
        for q in self.queries:
            sql = q.get("sql", "").lower()
            if table in sql and any(c.lower() in sql for c in cols):
                benefiting.append(q["id"])
        return benefiting

    def _estimate_rewrite(self, new_sql: str, query: dict) -> float:
        """
        Estimates improvement factor from a query rewrite (0.0 to 0.70).
        Checks for common optimization patterns.
        """
        new_lower = new_sql.lower()
        old_lower = query.get("sql", "").lower()
        improvement = 0.0

        # Remove SELECT * β†’ specific columns
        if "select *" not in new_lower and "select *" in old_lower:
            improvement += 0.20

        # Add LIMIT clause
        if "limit " in new_lower and "limit " not in old_lower:
            improvement += 0.15

        # Use EXISTS instead of IN subquery
        if "exists" in new_lower and "in (select" in old_lower:
            improvement += 0.25

        # Use INNER JOIN instead of implicit cross join
        if "inner join" in new_lower and "," in old_lower and "join" not in old_lower:
            improvement += 0.30

        # Add WHERE clause that was missing
        if "where" in new_lower and "where" not in old_lower:
            improvement += 0.35

        # Use COALESCE / ISNULL
        if "coalesce" in new_lower:
            improvement += 0.05

        return min(improvement, 0.70)