用字典理解过滤字典

我在使用dict理解时过滤dict的字典时遇到问题。

I have problem with filtering dict of dict while using dict comprehension.

我有dict：

clients = {
        'Shop': {'url' : 'url_v', 'customer' : 'cumoster_v', 
                 'some_other_key1' : 'some_value'},
        'Gym': {'url' : 'url_v1', 'customer_1' : 'customer_v1', 'customer_2': 'customer_v2',
                 'some_other_key2' : 'some_value'},
        'Bank': {'url' : 'url_v2', 'customer_3' : 'customer_v3',    
                  'some_other_key3' : 'some_value'}
}

我想再做一个仅具有 customer。*键的字典。
因此，新字典应该像这样：

I would like to make another dict which will have only 'customer.*' keys. So, new dict should looks like:

dict_only_cust = {
               'Shop': {'customer' : 'cumoster_v'},
               'Gym': {'customer_1' : 'customer_v1', 'customer_2': 'customer_v2'},
               'Bank': {'customer_3' : 'customer_v3'}
}

由于我是列表和字典理解的忠实拥护者，我想知道

As I'm big fan of lists and dicts comprehension, I'm wondering if it is possible to do it with this.

到目前为止，我已经写过：

So far, I've written:

dict_only_cust = {v.pop(t) for k, v in clients.items()
                for t, vv in v.items()
                if not re.match('.*customer.*', t)}

代码失败，并带有' RuntimeError：词典在迭代过程中更改了大小'

第二次尝试：

dict_only_cust = {k:{t: vv} for k, v in clients.items()
                for t, vv in v.items()
                if re.match('.*customer.*', t)}

这几乎可以，但是正在返回

It is almost OK, but it is returning

dict_only_cust = {
              'Shop' : {'customer' : 'cumoster_v'},
              'Gym' : {'customer_1' : 'customer_v1'},
              'Bank' : {'customer_3' : 'customer_v3'}
}

如何解决使用dict理解会出现此问题吗？
我正在使用python 3.4。

How to solve this problem using dict comprehension? I'm using python 3.4.

谢谢！