TopolearnThese cover the remaining loop shapes, so most stream logic can be expressed without manual state.
Filtering by condition:
takewhile(pred, it)yields items whilepredis true, then stops for good (it consumes the first failing item).dropwhile(pred, it)skips items whilepredis true, then yields everything after, without testing again.filterfalse(pred, it)is the complement of the built-infilter.compress(data, selectors)keeps items whose matching selector is truthy, handy with a precomputed mask.
Transforming:
starmap(func, it)callsfunc(*args)for each tuple:starmap(pow, [(2, 5), (3, 2)])gives32, 9. It'smapfor pre-zipped arguments.zip_longest(a, b, fillvalue=None)pairs items likezipbut continues to the longest input, filling gaps.
These differ from comprehensions in two ways: they're lazy, and takewhile and dropwhile express "until" logic that a comprehension can't, since a comprehension always scans everything.
A frequent gotcha: takewhile silently swallows the first element that fails the predicate. If you need that element, use a manual loop or the before_and_after recipe from the docs.
Also note the built-in zip(..., strict=True) (3.10+) raises if lengths differ, the opposite of zip_longest.
Example
from itertools import compress, dropwhile, filterfalse, starmap, takewhile, zip_longest
readings = [3, 5, 9, 12, 4, 2]
print(list(takewhile(lambda x: x < 10, readings))) # [3, 5, 9]
print(list(dropwhile(lambda x: x < 10, readings))) # [12, 4, 2]
print(list(filterfalse(lambda x: x % 2, readings))) # [12, 4, 2]
log = ["# header", "# more", "data1", "# note", "data2"]
print(list(dropwhile(lambda l: l.startswith("#"), log)))
# ['data1', '# note', 'data2']
print(list(compress("abcdef", [1, 0, 1, 0, 0, 1]))) # ['a', 'c', 'f']
print(list(starmap(pow, [(2, 5), (3, 2)]))) # [32, 9]
print(list(zip_longest("ab", [1, 2, 3], fillvalue="-")))
# [('a', 1), ('b', 2), ('-', 3)]
it = iter([1, 2, 30, 4])
print(list(takewhile(lambda x: x < 10, it)), list(it)) # [1, 2] [4]Remember
takewhile/dropwhileexpress 'until' logic lazilycompressfilters by a selector mask;starmapapplies functions to argument tuplestakewhileconsumes and discards the first failing item
Watch out for
- Losing the first non-matching item from a shared iterator after
takewhile.