[RudraDudhat2509]

Follow-up to #6279, splitting out the non-blocking nit from the AI code review since @rsareddy0329 suggested a separate PR for it.

remainder was re-read via self.buffer.seek(self.read_pos) + self.buffer.read(), but line (already read a few lines above in the same iteration) holds the exact same bytes, since nothing touches the buffer or read_pos between that read and the StopIteration handler. Just returns line directly now, no behavior change.

All 21 tests in test_iterators.py pass, including test_line_iterator_no_trailing_newline_at_end which covers this exact path.

[github-actions]

Claude finished @rsareddy0329's task in 38s —— View job

---

PR Review: Avoid redundant seek/read in LineIterator.__next__

• Read the PR diff (sagemaker-core/src/sagemaker/core/iterators.py)
• Verify correctness of the equivalence
• Check for edge cases and test coverage

---

Verdict: ✅ Looks correct — safe to merge

The change replaces a redundant seek + read with a direct return of the already-read line:

# before
self.buffer.seek(self.read_pos)
remainder = self.buffer.read()
self.read_pos += len(remainder)
return remainder

# after
self.read_pos += len(line)
return line

I traced the equivalence and it holds:

• line comes from self.buffer.readline() at self.read_pos (line 170–171).
• This branch is only reached when line does not end with \n (the if line and line[-1] == ord("\n") check failed above). That means there is no newline between read_pos and EOF, so readline() consumed everything to the end of the buffer — identical to what buffer.read() from read_pos would return.
• Nothing mutates the buffer or read_pos between the readline() and this handler: the only thing that happens in between is next(self.byte_iterator) raising StopIteration (caught here), and buffer writes only occur later on the success path (lines 205–206). So line is still the exact remainder.
• Therefore len(line) == len(remainder), read_pos advances identically, and the returned bytes are the same. No behavior change.

The guard self.read_pos < self.buffer.getbuffer().nbytes guarantees line is non-empty in this path, so returning it (rather than an empty bytes) matches the old behavior.

Minor observations (non-blocking)

• Correctly returns the full line here (not line[:-1]), matching the old remainder since there's no trailing newline to strip. 👍
• Test coverage: test_line_iterator_no_trailing_newline_at_end already exercises this exact path, so the existing suite is adequate for this change.

No correctness, security, or performance concerns. Nice small cleanup that removes one unnecessary buffer seek/read per trailing-partial-line case.
· fix/lineiterator-avoid-redundant-seek-read