TypeError: unhashable type: 'list'

What this error means

When you see it

Traceback (most recent call last):
  File "dedupe.py", line 5, in <module>
    seen.add(row)
TypeError: unhashable type: 'list'

Why it happens

1. A set of lists. Deduplicating rows with set(rows) or seen.add(row), when each row is a list, which is exactly what csv.reader gives you.
2. A list as a dict key. Counting or grouping by a pair of values stored as [city, year].
3. A list hiding inside a tuple. (1, [2, 3]) is a tuple, but it's only hashable if everything inside it is.
4. The dict version. set_of_records.add({"id": 7}) fails the same way, with unhashable type: 'dict'.

rows = [["Oslo", 2024], ["Lima", 2024], ["Oslo", 2024]]
unique = set(rows)       # TypeError: unhashable type: 'list'

How to fix it

Turn the list into a tuple. A tuple can't change, so it can be hashed. For a row of fixed values, a tuple is also the more accurate type:

rows = [["Oslo", 2024], ["Lima", 2024], ["Oslo", 2024]]
unique = {tuple(row) for row in rows}
print(unique)            # {('Oslo', 2024), ('Lima', 2024)}

Need to keep the original order while removing duplicates? A set won't keep the order, but a dictionary keyed by tuples will:

unique_rows = list(dict.fromkeys(tuple(row) for row in rows))
# [('Oslo', 2024), ('Lima', 2024)]  (the first appearance wins)

rows = ["alpha", "beta", "gamma"]

# Earlier blocks on this page that this one builds on, re-run quietly first.
_pm_earlier = [
    'rows = [["Oslo", 2024], ["Lima", 2024], ["Oslo", 2024]]\nunique = set(rows)       # TypeError: unhashable type: \'list\'',
    'rows = [["Oslo", 2024], ["Lima", 2024], ["Oslo", 2024]]\nunique = {tuple(row) for row in rows}\nprint(unique)            # {(\'Oslo\', 2024), (\'Lima\', 2024)}',
]
def _pm_run_earlier(_srcs, _g):
    import sys as _s, io as _io, warnings as _w
    for _src in _srcs:
        _o, _e = _s.stdout, _s.stderr
        _s.stdout = _s.stderr = _io.StringIO()
        try:
            with _w.catch_warnings():
                _w.simplefilter('ignore')
                exec(compile(_src, '<earlier block>', 'exec'), _g)
        except (Exception, SystemExit):
            pass
        finally:
            _s.stdout, _s.stderr = _o, _e
    if 'matplotlib.pyplot' in _s.modules:
        _s.modules['matplotlib.pyplot'].close('all')   # their plots belong to their own blocks
_pm_run_earlier(_pm_earlier, globals())
del _pm_earlier, _pm_run_earlier

Counting by a combination of values? Key by a tuple.

from collections import Counter

visits = [["Oslo", 2024], ["Lima", 2024], ["Oslo", 2024]]
by_city_year = Counter(tuple(v) for v in visits)
print(by_city_year[("Oslo", 2024)])   # 2

Order of the items doesn't matter? Use frozenset. Then {"python", "sql"} and {"sql", "python"} become the same key:

skills_seen = {frozenset(["python", "sql"])}
print(frozenset(["sql", "python"]) in skills_seen)   # True

A dict as the key? Use its identifying field instead (records_by_id[record["id"]]), and don't try to hash the whole dict.

When you'd see this in real code

• Removing duplicate rows from a CSV with set(csv.reader(f)).
• A cache decorator, functools.lru_cache, wrapped around a function you call with a list argument. The cache uses the arguments as dictionary keys.
• pandas df.groupby(...) or df.drop_duplicates() on a column whose cells hold lists.
• Building a graph as edges = {[a, b]: weight} instead of {(a, b): weight}.

Related errors

• TypeError: list indices must be integers or slices, not str: using a list where a dict was expected, the other way round.
• KeyError: 'X': the key was hashable, but it isn't in the dictionary.
• RuntimeError: dictionary changed size during iteration: another rule that exists so a dict can keep track of its keys.

See Also

• All Python errors: the full index, by type and by when it happens.
• Tuples: the immutable sequence, and why that makes it a good key.
• Sets: what can go in one, and why.
• Dictionaries: keys, hashing and lookups.