Ruben Fiszel 13b521651b fix(python-client): return at most size bytes from S3BufferedReader.read (#10623) ...

* feat: add unit tests for S3BufferedReader.read and improve read method implementation

* feat: refactor S3BufferedReader.read method and add unit tests for its functionality

* feat: implement peek() on S3BufferedReader with buffered reads

* fix(python-client): keep the read(size) contract and trim the test surface

Drop the duplicated `TestS3BufferedReaderRead` class from
`python-client/tests/wmill_client_test.py`: CI runs `pytest tests/` from
`python-client/wmill`, so that legacy manual harness never executes, and the
same assertions already live in `python-client/wmill/tests/test_s3_reader.py`.

Narrow that file to the four behaviours a future change could break, and make
the `bytes_generator` guard actually call `bytes_generator`.

Align `peek()` with `io.BufferedReader.peek`, which does at most one read on
the underlying stream, rather than looping until `size` bytes are buffered.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

* fix(python-client): hold read1 to one underlying read

read1 forwarded to read, so read1(-1) drained the whole object — the same
unbounded buffering this branch removes from read. Now that a buffer exists,
read1 can honour its own contract: fill only when the buffer is empty, then
serve from it.

Also treat read(None) as read(-1), per the BufferedReader contract, and pin
that read(0) does not pull from the stream: that holds only because the
drain sentinel is a negative size, and widening it to any falsy size would
reintroduce whole-file buffering.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

* fix(python-client): return from read1(0) without touching the stream

A zero-length read has nothing to serve, so pulling a chunk to satisfy it
both wastes a round trip and advances the stream. Guard it ahead of the
fill, and pin it with a chunk source that counts pulls.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Tushar <tusharanshu18@gmail.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

2026-08-10 23:04:27 +02:00

..

test_s3_reader.py

fix(python-client): return at most size bytes from S3BufferedReader.read (#10623)

2026-08-10 23:04:27 +02:00

test_workflow.py

test(wac): pin the failure record with one corpus both SDKs read (#10385)

2026-07-28 16:54:40 +02:00