178 lines
8.5 KiB
Python
178 lines
8.5 KiB
Python
#!/usr/bin/env python3
|
||
# -*- coding: utf-8 -*-
|
||
# M11b-1a 定点补丁脚本（一次性工具，跑完即删）：
|
||
# Patch1 maintenance_sql(): 百分号格式化渲染 → token + str.replace，避开 % 格式化崩溃（QC #1）
|
||
# Patch2 verify_db(): 删空转查询；落分区/分区裁剪改用 information_schema 分区清单
|
||
# + SELECT ... PARTITION(p) 反查 + EXPLAIN FORMAT=JSON 解析 partitions（QC #6）
|
||
# Patch3 新增连库辅助函数 _explain_partitions/_row_in_partition/_locate_row_partition
|
||
# DDL 列定义/索引/分区计划文本一律不动。
|
||
import os
|
||
|
||
HERE = os.path.dirname(os.path.abspath(__file__))
|
||
P = os.path.join(os.path.dirname(HERE), "scripts", "pbl_runtime_event_ddl.py")
|
||
src = open(P, encoding="utf-8").read()
|
||
orig = src
|
||
|
||
# ================================================================ Patch 1
|
||
i0 = src.index("def maintenance_sql(")
|
||
i1 = src.index("def migrate_from_legacy_sql()")
|
||
region = src[i0:i1]
|
||
assert "%(table)s" in region and "%(pmax)s" in region and "%(months)s" in region, "占位符缺失"
|
||
|
||
new_region = (region
|
||
.replace("%(table)s", "__TABLE__")
|
||
.replace("%(pmax)s", "__PMAX__")
|
||
.replace("%(months)s", "__MONTHS__"))
|
||
|
||
old_doc = ' """分区维护：过程（幂等补建月分区 / 到期 DROP 旧分区）+ 每月 EVENT。"""'
|
||
new_doc = (' """分区维护：过程（幂等补建月分区 / 到期 DROP 旧分区）+ 每月 EVENT。\n'
|
||
'\n'
|
||
" QC #1 修复：正文含大量字面百分号（DATE_FORMAT(...,'%Y-%m-01')、'%Y%m'），\n"
|
||
' 原先用「三引号 % 字典」渲染会抛 ValueError: unsupported format character。\n'
|
||
' 现改用 __TABLE__/__PMAX__/__MONTHS__ token + str.replace，\n'
|
||
' 字面百分号原样输出、无需再写双写转义。\n'
|
||
' """')
|
||
assert old_doc in new_region, "docstring 未命中"
|
||
new_region = new_region.replace(old_doc, new_doc)
|
||
|
||
old_tail = ('""" % {"table": TABLE, "pmax": PARTITION_MAXVALUE, "months": months,\n'
|
||
' "today": (today or date.today()).isoformat()}')
|
||
new_tail = ('""".replace("__TABLE__", TABLE)\n'
|
||
' .replace("__PMAX__", PARTITION_MAXVALUE)\n'
|
||
' .replace("__MONTHS__", str(months))')
|
||
assert old_tail in new_region, "maintenance_sql 渲染尾段未命中"
|
||
new_region = new_region.replace(old_tail, new_tail)
|
||
|
||
src = src[:i0] + new_region + src[i1:]
|
||
|
||
# ================================================================ Patch 2
|
||
s2 = src.index(' cur.execute("SELECT PARTITION_NAME FROM information_schema.PARTITIONS "')
|
||
e2_marker = (' results.append(("分区裁剪", pruned.count(",") == 0 and "p" in pruned,'
|
||
' "EXPLAIN partitions=%s" % pruned))')
|
||
e2 = src.index(e2_marker) + len(e2_marker)
|
||
old_block = src[s2:e2]
|
||
assert "EXPLAIN PARTITIONS SELECT id FROM `%s` WHERE event_uid" in old_block, "旧判定块未命中"
|
||
assert "for k in" in old_block, "空转 for 循环不在预期块内"
|
||
|
||
new_block = ''' # 分区清单：以 information_schema.PARTITIONS 为权威来源（按 ordinal 排序），
|
||
# 后面「落哪个分区」「分区裁剪」两项判定都基于这份清单（QC #6）
|
||
cur.execute("SELECT PARTITION_NAME, PARTITION_DESCRIPTION FROM information_schema.PARTITIONS "
|
||
"WHERE TABLE_SCHEMA=DATABASE() AND TABLE_NAME=%s "
|
||
"AND PARTITION_NAME IS NOT NULL ORDER BY PARTITION_ORDINAL_POSITION", (TABLE,))
|
||
part_rows = [(r[0], r[1]) for r in cur.fetchall()]
|
||
part_names = [n for n, _ in part_rows if n]
|
||
results.append(("MAXVALUE 兜底", PARTITION_MAXVALUE in part_names, ",".join(part_names)))
|
||
|
||
# 插入两条不同月份的探针行 → 应落在两个不同的物理分区
|
||
probe = "m11bprobe%s" % date.today().strftime("%H%M%S")
|
||
probes = (("%s_a" % probe, "2026-01-05 00:00:00"),
|
||
("%s_b" % probe, "2027-03-05 00:00:00"))
|
||
try:
|
||
for i, (uid, ts) in enumerate(probes):
|
||
cur.execute("INSERT INTO `%s` (tenant_id,event_uid,event_type,scene_id,seq,created_at)"
|
||
" VALUES ('t_probe',%%s,'probe',0,%%s,%%s)" % TABLE, (uid, i + 1, ts))
|
||
except Exception as exc: # noqa: BLE001
|
||
results.append(("跨月插入", False, str(exc)[:160]))
|
||
else:
|
||
results.append(("跨月插入", True, "%d 条探针行（不同月份）" % len(probes)))
|
||
located = [_locate_row_partition(cur, uid, ts, part_names) for uid, ts in probes]
|
||
distinct = len(set(located)) == 2
|
||
in_list = all(p in part_names for p in located)
|
||
results.append(("跨月落不同分区", distinct and in_list,
|
||
" | ".join("%s -> %s" % (uid, p)
|
||
for (uid, _), p in zip(probes, located))))
|
||
|
||
# 分区裁剪：created_at 区间只应命中该区间所属的少数分区，而非全部分区。
|
||
# 旧写法 EXPLAIN PARTITIONS 取 row[3]：列序跨版本不稳定，且该语法新版 MySQL 已废弃。
|
||
pruned = _explain_partitions(cur,
|
||
"SELECT id FROM `%s` WHERE created_at>='2026-01-01'"
|
||
" AND created_at<'2026-02-01'" % TABLE)
|
||
results.append(("分区裁剪",
|
||
bool(pruned) and all(p in part_names for p in pruned)
|
||
and len(pruned) < max(1, len(part_names)),
|
||
"命中 %d/%d 分区: %s" % (len(pruned), len(part_names),
|
||
",".join(pruned) or "N/A")))'''
|
||
|
||
src = src[:s2] + new_block + src[e2:]
|
||
|
||
# ================================================================ Patch 3
|
||
helpers = '''def _explain_partitions(cur, sql, params=None):
|
||
"""EXPLAIN FORMAT=JSON 取执行计划实际扫描的分区清单（QC #6）。
|
||
|
||
不再用 EXPLAIN PARTITIONS：其分区列固定在结果 row[3]，列序随 MariaDB/MySQL
|
||
版本变化，且该扩展语法在新版 MySQL 已废弃。JSON 计划里的 partitions 字段两版一致。
|
||
取不到（引擎不支持/解析失败）返回 []，由调用方判失败，不猜。
|
||
"""
|
||
try:
|
||
cur.execute("EXPLAIN FORMAT=JSON " + sql, tuple(params or ()))
|
||
row = cur.fetchone()
|
||
except Exception: # noqa: BLE001
|
||
return []
|
||
raw = row[0] if row else None
|
||
if isinstance(raw, (bytes, bytearray)):
|
||
raw = raw.decode("utf-8", "replace")
|
||
if not raw:
|
||
return []
|
||
try:
|
||
plan = json.loads(raw)
|
||
except (ValueError, TypeError):
|
||
return []
|
||
found = []
|
||
|
||
def walk(node):
|
||
if isinstance(node, dict):
|
||
for key, val in node.items():
|
||
if key == "partitions" and isinstance(val, list):
|
||
for p in val:
|
||
if isinstance(p, str) and p and p not in found:
|
||
found.append(p)
|
||
else:
|
||
walk(val)
|
||
elif isinstance(node, list):
|
||
for val in node:
|
||
walk(val)
|
||
|
||
walk(plan)
|
||
return found
|
||
|
||
|
||
def _row_in_partition(cur, part, uid):
|
||
"""SELECT ... FROM t PARTITION(p) 反查：探针行是否物理落在该分区。"""
|
||
try:
|
||
cur.execute("SELECT id FROM `%s` PARTITION (`%s`)"
|
||
" WHERE tenant_id='t_probe' AND event_uid=%%s" % (TABLE, part), (uid,))
|
||
except Exception: # noqa: BLE001
|
||
return False
|
||
return bool(cur.fetchone())
|
||
|
||
|
||
def _locate_row_partition(cur, uid, ts, part_names):
|
||
"""定位某探针行实际所在分区（QC #6）。
|
||
|
||
1) 首选按 information_schema 分区清单逐个 PARTITION(p) 反查物理位置——与 EXPLAIN
|
||
输出列序无关，最可靠；
|
||
2) 反查不中（引擎不支持 PARTITION 子句等）时退回 EXPLAIN FORMAT=JSON 解析
|
||
partitions（带 tenant_id+event_uid+created_at 等值条件，裁剪后应只剩 1 个分区）。
|
||
"""
|
||
for part in part_names:
|
||
if _row_in_partition(cur, part, uid):
|
||
return part
|
||
planned = _explain_partitions(cur,
|
||
"SELECT id FROM `%s` WHERE tenant_id='t_probe'"
|
||
" AND event_uid=%%s AND created_at=%%s" % TABLE,
|
||
(uid, ts))
|
||
if len(planned) == 1:
|
||
return planned[0]
|
||
return ",".join(planned) or "?"
|
||
|
||
|
||
'''
|
||
marker = "def verify_db():"
|
||
assert src.count(marker) == 1, "verify_db 定位异常"
|
||
src = src.replace(marker, helpers + marker)
|
||
|
||
open(P, "w", encoding="utf-8").write(src)
|
||
print("PATCHED %s: %d -> %d bytes" % (P, len(orig), len(src)))
|
||
print("residual 'EXPLAIN PARTITIONS ' count:", src.count("EXPLAIN PARTITIONS "))
|
||
print("residual 'for k in' count:", src.count("for k in"))
|