测试框架实践--多线程

前面几次的分享，我从一个数据驱动的实现展开去，先后讨论了什么是数据驱动，如何实现数据驱动，数据驱动在自动化框架里如何应用。
测试框架–教你用Python实现数据驱动1
测试框架–教你用Python实现数据驱动2
测试框架–数据驱动动态增加测试用例
测试框架实践–动态挑选待运行测试用例

为什么要讲这些呢？

因为这些是一个测试框架必不可少的部分，看看前面4次的分享，虽然可以成功的运行，但我们采用了顺序运行的方式，跟实际使用相差很远，现实中我们一般实现“并发”运行我们的测试用例。

说起并发，Python 有多线程（Multiple Threading）和多进程（Multiple Process）之分， 由于GIL即Global Interpreter Lock(全局解释器锁)的存在，python多线程并不能实现真正的并发（无论你的CPU是否多核），相反，多进程因为有独占的内存空间可以真正的实现并发。但在我们的自动化测试框架里，我们还是希望利用线程的公用内存空间特性，特别是同一个测试类我们希望有统一的setup， teardown特性。而这个多进程就不太适用，加上我们测试用例也不是CPU计算密集型，多线程方案的“并发”看起来是最佳选择。

但是是不是就一定要用threading.Thread呢？

我们先看看”传统“的多线程并发。 一个通用的多线程模板是这样的：

import threading
import queue
import time
exitFlag = 0
class MyThread(threading.Thread):
    def __init__(self, q):
        threading.Thread.__init__(self)
        self.q = q
    def run(self):
        do_something(self.q)
def do_something(q):
    while not exitFlag:
        queue_lock.acquire()
        if not work_queue.empty():
            data = q.get()
            queue_lock.release()
            print("Now thread %s is processing %s" % (threading.currentThread(), data))
            time.sleep(1)
        else:
            queue_lock.release()
if __name__ == "__main__":
    num_worker_threads = 3
    target_case = ['case1', 'case2', 'case3', 'case4', 'case5']
    queue_lock = threading.Lock()
    work_queue = queue.Queue()
    threads = []
    start_time = time.time()
    # Create new thread
    for case in range(num_worker_threads):
        thread = MyThread(work_queue)
        thread.start()
        threads.append(thread)
    # Fill queue
    queue_lock.acquire()
    for case in target_case:
        work_queue.put(case)
    queue_lock.release()
    # Wait queue empty
    while not work_queue.empty():
        pass
    # Notify thread quite
    exitFlag = 1
    # Wait all threads done
    for t in threads:
        t.join()
    end_time = time.time()
    print("All cases finished with running time --%s" % (end_time - start_time))

把我们前面实现的顺序运行改成并发， 只需要改成如下就可以实现多线程：

import threading
import queue
import time
from common.test_case_finder import DiscoverTestCases, unpack_test_cases_from_functions
exitFlag = 0
def f(case):
    name, func, value = case
    try:
        if value:
            func.__call__(name, *value)
        else:
            func.__call__(name)
    except:
        # traceback.print_exc()
        cases_run_fail.append(name)
    else:
        cases_run_success.append(name)
    return cases_run_fail, cases_run_success
class MyThread(threading.Thread):
    def __init__(self, q):
        threading.Thread.__init__(self)
        self.q = q
    def run(self):
        while not exitFlag:
            queue_lock.acquire()
            if not work_queue.empty():
                data = self.q.get()
                f(data)
                time.sleep(5)
                queue_lock.release()
                print("Now thread %s is processing %s" % (threading.currentThread(), data))
            else:
                queue_lock.release()
if __name__ == "__main__":
    num_worker_threads = 5
    mypath = r"D:\ktest\tests\test_page1"
    cases_to_run = []
    cases_run_success = []
    cases_run_fail = []
    discover_cases = DiscoverTestCases(mypath)
    mds = discover_cases.get_modules_spec()
    raw_test_cases = discover_cases.find_classes_in_module(mds)
    cases_to_run = unpack_test_cases_from_functions(raw_test_cases)
    queue_lock = threading.Lock()
    work_queue = queue.Queue()
    threads = []
    start_time = time.time()
    # Create new thread
    for case in range(num_worker_threads):
        thread = MyThread(work_queue)
        thread.start()
        threads.append(thread)
    # Fill queue
    queue_lock.acquire()
    for case in cases_to_run:
        work_queue.put(case)
    queue_lock.release()
    # Wait queue empty
    while not work_queue.empty():
        pass
    # Notify thread quite
    exitFlag = 1
    # Wait all threads done
    for t in threads:
        t.join()
    end_time = time.time()
    print("All cases finished with running time --{:.2f} seconds" .format(end_time - start_time))
    print('Below cases are passed:\n %s' %cases_run_success)
    print('Below cases are failed:\n %s' % cases_run_fail)

可以看到，你自己要做很多事情来保证多线程的正常运行，你要维护queue，要注意资源锁，你要使用join来等待子线程都完成。 多线程难道都这么麻烦吗？

“人生苦短，我用python”不知道你们听过没 ：）

multiprocessing.dummy 来助你一臂之力！

dummy是multiprocessing的一个克隆体，唯一的区别在于，dummy使用线程而multiprocessing使用进程。

from multiprocessing.dummy import Pool as ThreadPool
import time
from common.test_case_finder import DiscoverTestCases, unpack_test_cases_from_functions
def f(case):
    name, func, value = case
    try:
        if value:
            func.__call__(name, *value)
        else:
            func.__call__(name)
    except:
        # traceback.print_exc()
        cases_run_fail.append(name)
    else:
        cases_run_success.append(name)
    return cases_run_fail, cases_run_success
if __name__ == "__main__":
    number_of_threads = 5
    mypath = r"D:\ktest\tests\test_page1"
    cases_to_run = []
    cases_run_success = []
    cases_run_fail = []
    discover_cases = DiscoverTestCases(mypath)
    mds = discover_cases.get_modules_spec()
    raw_test_cases = discover_cases.find_classes_in_module(mds)
    cases_to_run = unpack_test_cases_from_functions(raw_test_cases)
    start_time = time.time()
    with ThreadPool(number_of_threads) as p:
        p.map(f, cases_to_run)
    p.close()
    p.join()
    end_time = time.time()
    print('Below cases are passed:\n %s' %cases_run_success)
    print('Below cases are failed:\n %s' % cases_run_fail)
    print("All cases finished with running time --{:.2f} seconds" .format(end_time - start_time))

看下图，左边是传统的threading.Thread的多线程用法， 右边是multipleprocess.dummy的用法。

除开定义Thread类外，threading.Thread还用了这么多篇幅才能令多线程工作，但multiprocessing.dummy只用一个Pool就全搞定了。

好了，今天我们讲到这里，并发也实现了，动态挑选也实现了，一个测试框架就搭起来来， 后面就是增强了， 下一期来实现test fixture， 敬请期待！