仅仅执行一堆'or'可能不是那么糟糕.使用python构建xpath，以免获得编写者的抽筋，然后对其进行预编译.实际的xpath代码在libxml中，应该很快.

It might not be that bad just to do a bunch of 'or's. Build the xpath with python so that you don't get writer's cramp and then precompile it. The actual xpath code is in libxml and should be fast.

sites=['aaa', 'bbb']
contains = ' or '.join('contains(@href,(%s))' % site for site in sites)
anchor_xpath = etree.XPath('//a[%s][descendant::img]' % contains)