list(bucket()) hangs with 100% CPU

[alexchandel]

From the docs:

b = bucket(['a1', 'b1', 'c1', 'a2', 'b2', 'c2', 'b3'], lambda x: x[0])

The following uses 100% CPU and never returns:

list(b)

Additionally, dict(b) fails with dictionary update sequence element #0 has length 0; 2 is required, leaving no way to inspect the opaque result without prior knowledge of all keys.

This occurs in all versions of more-itertools.

[MSeifert04]

bucket itself is an infinite iterable. That's why it hangs.

However maybe bucket is only iterable by accident (it implements __getitem__ so it's iterable) - I'm not sure.

But yeah, bucket (as it currently is) isn't meant to be iterated over without knowledge of the keys.

[bbayles]

Correct - this was discussed at the time of implementation.

[bbayles]

Because __getitem__ is implemented, when calling list() or dict() on bucket(something) causes Python to look up 0, 1,, 2... on to infinity.

#371 prevents this by explicitly defining __iter__. Now calling list() on bucket(something) will extract the keys from the iterable and return them.