fix: _cut_iter yields wrong-length parts when step > n
partition()/chunks() on an iterator (_cut_iter) only correctly skipped step items between parts when step <= n. For step > n, slicing an n-length pool by pool[step:] silently clips to an empty list, then islice(it, step) pulled step (not n) fresh items into the pool instead of first discarding the step - n items between parts. This produced wrong-length, wrong-content chunks that diverged from the Sequence code path (_cut_seq) for the same input, e.g. list(range(10)) vs iter(range(10)) with n=2, step=3. Branch on step < n to keep the existing sliding-window logic, and for step >= n explicitly discard the gap before taking the next pool.
This commit is contained in:
1 parent
f15f2b80fb
commit
a12abd3a5e
2 files changed
+18
-2
No files matched your search
+6
-2
@@ -363,8 +363,12 @@ def _cut_iter(drop_tail, n, step, seq):
|
||||
if len(pool) < n:
|
||||
break
|
||||
yield pool
|
||||
pool = pool[step:]
|
||||
pool.extend(islice(it, step))
|
||||
if step < n:
|
||||
pool = pool[step:]
|
||||
pool.extend(islice(it, step))
|
||||
else:
|
||||
take(step - n, it) # discard the gap between parts
|
||||
pool = take(n, it)
|
||||
if not drop_tail:
|
||||
for item in _cut_seq(drop_tail, n, step, pool):
|
||||
yield item
|
||||
|
||||
Reference in new issue
Block a user