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 | class SnowflakeSnowparkBudgetRuntime(BaseBudgetRuntime[SnowflakeSnowparkBudgetExecutor]):
"""Snowflake Snowpark budget runtime using EXPLAIN for estimation."""
DEFAULT_GUARD = BudgetGuard(
env_var="FF_SF_MAX_BYTES",
estimator_attr="runtime_budget_estimate_query_bytes",
engine_label="Snowflake",
what="query",
)
def estimate_query_bytes(self, sql: str) -> int | None:
"""
Best-effort Snowflake bytes estimation using EXPLAIN USING JSON.
Mirrors the previous executor-side logic.
"""
try:
body = self.executor._selectable_body(sql)
except Exception:
body = sql
try:
rows = self.executor.session.sql(f"EXPLAIN USING JSON {body}").collect()
if not rows:
return None
parts: list[str] = []
for r in rows:
try:
parts.append(str(r[0]))
except Exception:
as_dict: dict[str, Any] = getattr(r, "asDict", lambda: {})()
if as_dict:
parts.extend(str(v) for v in as_dict.values())
plan_text = "\n".join(parts).strip()
if not plan_text:
return None
try:
plan_data = json.loads(plan_text)
except Exception:
return None
bytes_val = self._extract_bytes_from_plan(plan_data)
if bytes_val is None or bytes_val <= 0:
return None
return bytes_val
except Exception:
# Any parsing / EXPLAIN issues → no estimate, guard skipped
return None
def dataframe_bytes(self, df: Any) -> int | None:
"""
Best-effort bytes estimate for a Snowpark DataFrame.
"""
try:
sql_text = self._snowpark_df_sql(df)
if not isinstance(sql_text, str) or not sql_text.strip():
return None
return self.estimate_query_bytes(sql_text)
except Exception:
return None
def _extract_bytes_from_plan(self, plan_data: Any) -> int | None:
def _to_int(value: Any) -> int | None:
if value is None:
return None
try:
return int(value)
except Exception:
return None
if isinstance(plan_data, dict):
global_stats = plan_data.get("GlobalStats") or plan_data.get("globalStats")
if isinstance(global_stats, dict):
candidate = _to_int(
global_stats.get("bytesAssigned") or global_stats.get("bytes_assigned")
)
if candidate:
return candidate
for val in plan_data.values():
bytes_val = self._extract_bytes_from_plan(val)
if bytes_val:
return bytes_val
elif isinstance(plan_data, list):
for item in plan_data:
bytes_val = self._extract_bytes_from_plan(item)
if bytes_val:
return bytes_val
return None
def _snowpark_df_sql(self, df: Any) -> str | None:
"""
Extract the main SQL statement for a Snowpark DataFrame.
Uses the documented public APIs:
- DataFrame.queries -> {"queries": [sql1, sql2, ...], "post_actions": [...]}
- Optionally falls back to df._plan.sql() if needed.
"""
queries_dict = getattr(df, "queries", None)
if isinstance(queries_dict, dict):
queries = queries_dict.get("queries")
if isinstance(queries, list) and queries:
candidates = [q.strip() for q in queries if isinstance(q, str) and q.strip()]
if candidates:
return max(candidates, key=len)
plan = getattr(df, "_plan", None)
if plan is not None:
with suppress(Exception):
simplify = getattr(plan, "simplify", None)
if callable(simplify):
simplified = simplify()
to_sql = getattr(simplified, "sql", None)
if callable(to_sql):
sql = to_sql()
if isinstance(sql, str) and sql.strip():
return sql.strip()
with suppress(Exception):
to_sql = getattr(plan, "sql", None)
if callable(to_sql):
sql = to_sql()
if isinstance(sql, str) and sql.strip():
return sql.strip()
return None
|