Docs demonstrate initializing a metric's labels with a list, but labelnames is internally represented as a tuple. If labelnames is truthy, it gets cast to a tuple (and a reasonable error is thrown if it can't be cast), but _MetricWrapper simply forwards labelnames as-is if it is falsy.

This is normally not a problem, but when initializing a histogram with an empty labels list:

from prometheus_client import Histogram
Histogram('h', 'help', [])

this error occurs:

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/local/lib/python2.7/dist-packages/prometheus_client/core.py", line 539, in init
    collector = cls(name, labelnames, (), **kwargs)
  File "/usr/local/lib/python2.7/dist-packages/prometheus_client/core.py", line 835, in __init__
    bucket_labelnames = labelnames + ('le',)
TypeError: can only concatenate list (not "tuple") to list

I see no reason why any iterable that can be cast to a tuple shouldn't reasonably be expected to work here (use case for the empty list: building up a list of labels conditionally; it's inconvenient/inconsistent to check if empty, or to cast to tuple when lists are used for all other cases).

I think a reasonable fix is to set labelnames to an empty tuple if it evaluates to falsy. (Possible alternative: casting even the falsy value to tuple, in case you would like to throw a clear error if someone calls the function with a falsy non-iterable?)

The added test might be a bit simplistic, but it passes with this change and fails without it.