Python 'TypeError: unhashable type: 'list'' (or 'dict', 'set', 'numpy.ndarray') when using a value as a dict key, in a set, with lru_cache, or in pandas groupby: how do I fix it?

Cause: dictionary keys, set members and functools.lru_cache arguments must be hashable: their hash value must never change during their lifetime, and equal objects must hash the same. Mutable built-in containers such as list, dict and set are therefore unhashable, and a tuple or frozenset is hashable only if everything inside it is hashable (a tuple containing a list still raises the error). User-defined classes are hashable by default, but defining __eq__ without __hash__ makes instances unhashable. Numpy arrays are also mutable and unhashable. Fix (best first): 1. Find which object is the culprit: the traceback line shows the dict/set operation; print type(x) for the key or element. 2. Convert the value to an immutable equivalent before using it as a key: list to tuple(x) (convert nested lists recursively), set to frozenset(x), dict to tuple(sorted(d.items())) or a frozenset of items when all values are hashable, numpy array to arr.tobytes() or tuple(arr.tolist()). 3. For functools.lru_cache or cache, the docs require hashable positional and keyword arguments, so pass tuples instead of lists, or wrap the call with a conversion layer. 4. If you wanted to group or count lists, use a list of pairs, a dict keyed by a tuple form, or a dict of lists (collections.defaultdict(list)) instead of using the list itself as a key. 5. For your own classes with __eq__, add a __hash__ built from the same immutable fields, or use @dataclass(frozen=True), which generates a matching __hash__. Do not hash on fields you mutate while the object sits in a dict or set. Verify: the same dict/set/cached call runs without TypeError, and two equal inputs produce equal keys (key(a) == key(b) and hash(key(a)) == hash(key(b))). Sources: https://docs.python.org/3/glossary.html#term-hashable https://docs.python.org/3/library/functools.html#functools.lru_cache