26.0.x: higher client CPU per call for `fetchone()` / `execute()` compared to 4.0.2

未关闭
  • 哪里不对？
#621 3 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

还没有人认领这个 Issue。

评估

难度
4/5
预计耗时
3-5 天
新手友好度
32/100
Issue 类型
缺陷
描述清晰度
基本清楚
活跃度
活跃
技术栈
python
领域
databases, performance

调研方向

Start by running the included repro_overhead.py against 4.0.2 and 26.0.1 in both thin and thick modes to confirm the fetchone() and execute() CPU gap while fetchmany() stays flat. Then compare how the two versions handle the per-row fetch path, since that is where the regression appears. Done means the per-call overhead is traced to a specific change and the numbers come back within a few percent of 4.0.2 in both modes.

由索引模型根据 Issue 内容生成。

描述

bug
  1. What versions are you using?
  • python-oracledb: 26.0.1, compared with 4.0.2 (stock PyPI wheels)
  • Oracle Database: 12.1.0.2.0
  • Oracle Client (Thick mode): 19.18.0.0.0
  • Python: 3.11.13
  • platform.platform: Linux-6.18.54.1-microsoft-standard-WSL2-x86_64-with-glibc2.35
  • sys.maxsize > 2**32: True
  1. Is it an error or a hang or a crash?

Neither: performance regression.

  1. What error(s) or behavior you are seeing?

fetchone() and execute() use more client CPU in 26.0.1 than in 4.0.2, in both Thin and Thick modes. fetchmany() is unchanged.

Client CPU time measured with the script below (seconds, average of two runs, each run being the median of 5):

Operation Mode 4.0.2 26.0.1 Change
fetchone() loop, 50k rows Thin 0.182 0.237 +31 %
fetchone() loop, 50k rows Thick 0.199 0.303 +52 %
2000 × (execute() + fetchone()) Thin 0.383 0.445 +16 %
2000 × (execute() + fetchone()) Thick 0.397 0.477 +20 %
fetchmany() scan, 50k rows Thin 0.169 0.179 +6 %
fetchmany() scan, 50k rows Thick 0.190 0.190 0 %
  1. Does your application call init_oracle_client()?

Both modes were measured (Thin and Thick).

  1. Include a runnable Python script that shows the problem.

No schema needed. Run once with each oracledb version:

ORA_USER=... ORA_PASSWORD=... ORA_DSN=host:port/service python repro_overhead.py thin
ORA_USER=... ORA_PASSWORD=... ORA_DSN=host:port/service python repro_overhead.py thick

import os
import statistics
import sys
import time

import oracledb

MODE = sys.argv[1] if len(sys.argv) > 1 else "thin"
ROWS = 50_000
N_EXEC = 2_000
REPEAT = 5

if MODE == "thick":
    oracledb.init_oracle_client()
oracledb.defaults.arraysize = 100
oracledb.defaults.prefetchrows = 101

conn = oracledb.connect(user=os.environ["ORA_USER"], password=os.environ["ORA_PASSWORD"],
                        dsn=os.environ["ORA_DSN"])

SCAN = f"""select level, level / 7, date '2026-01-01' + mod(level, 365), 'row_' || level
           from dual connect by level <= {ROWS}"""
ONE = "select :k, 'x' from dual"


def s1_fetchmany():
    with conn.cursor() as cur:
        cur.execute(SCAN)
        while cur.fetchmany():
            pass


def s2_execute_fetchone():
    with conn.cursor() as cur:
        for k in range(N_EXEC):
            cur.execute(ONE, k=k)
            cur.fetchone()


def s3_fetchone():
    with conn.cursor() as cur:
        cur.execute(SCAN)
        while cur.fetchone() is not None:
            pass


def cpu(fn):
    fn()                                    # warm-up
    samples = []
    for _ in range(REPEAT):
        c0 = time.process_time()
        fn()
        samples.append(time.process_time() - c0)
    return statistics.median(samples)


print("oracledb", oracledb.__version__, MODE)
print(f"fetchmany() scan of {ROWS} rows      : CPU {cpu(s1_fetchmany):.3f} s")
print(f"{N_EXEC} x (execute() + fetchone())  : CPU {cpu(s2_execute_fetchone):.3f} s")
print(f"fetchone() loop over {ROWS} rows     : CPU {cpu(s3_fetchone):.3f} s")
主要语言
Python
星标
456
派生
119
30 天内没有已合并 PR

环境准备

从这里开始

  1. 先读完整个 Issue，再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库，在一个分支上完成修改。
  4. 提交 Pull Request，并在描述里引用这个 Issue 编号。

oracle/python-oracledb 的其他 Issue

查看 oracle/python-oracledb 的全部 Issue

相似的 Issue

更多 Python Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。