Yuji 'Tomita' Tomita Yuji 'Tomita' Tomita - 5 months ago 32
Python Question

How to hash *args **kwargs for function cache?

I'm working with

which has a 4k limit on how many styles can be defined in an excel doc.

Normally, one creates styles like so:

style = xlwt.easyxf("font: bold 1")

Which I simply replaced with

def cached_easyxf(self, format):
return self._cache.setdefault(format, xlwt.easyxf(format))

Which works perfectly. Now, I've found out that I need to pass in keyword arguments sometimes which got me thinking: how should I hash the args/kwargs signature?

Should I create a cache key based on str(value)? Pickle? What's most robust?

For my situation it looks like I can just convert key/values to strings and add it to my keys... but I'm now curious about a generic way to handle this with say unhashable types like
arg=[1, 2, 3]

def cached_call(*args, **kwargs):
return cache.get(what_here)
cached_call([1, 2, 3], {'1': True})


Here is the technique used in functools.lru_cache():

kwd_mark = object()     # sentinel for separating args from kwargs

def cached_call(*args, **kwargs):
    key = args + (kwd_mark,) + tuple(sorted(kwargs.items()))
    return cache.get(key)

Note, the above code handles keyword arguments but makes no attempt to handle non-hashable values like lists. Your idea for using the str of a list is a reasonable start. For set objects, you would need to sort the entries first, str(sorted(someset)). Other objects may not have a useful __repr__ or __str__ (i.e. they may show only the object type and location in memory). In summary, handling arbitrary unhashable arguments requires careful consideration of each object type.