StreamingBody.__enter__ should return self

[dizel3d]

Describe the bug

f = s3_client.get_object(...)["Body"]  # f is StreamingBody object
f.read()  # At the end of the stream this will check content-length and raise an error if mismatch
f.close()

is not equal to

with s3_client.get_object(...)["Body"] as f:  # f is NOT StreamingBody object
  f.read()  # This will NOT check content-length

Regression Issue

• Select this option if this issue appears to be a regression.

Expected Behavior

The problem is in StreamingBody.enter. I believe it should return self instead of self._raw_stream.

Current Behavior

StreamingBody.__enter__ returns self._raw_stream instead of self.

Reproduction Steps

See description.

Possible Solution

No response

Additional Information/Context

No response

SDK version used

1.40.34

Environment details (OS name and version, etc.)

Python 3.10.12

[adev-code]

Hello @dizel3d, thank you for reaching out. Could you please give a full minimal reproducible code to help us repro and see the issue?

[dizel3d]

Hello @dizel3d, thank you for reaching out. Could you please give a full minimal reproducible code to help us repro and see the issue?

Run python shell:

python3 -m venv venv
source venv/bin/activate
pip install boto3
python

Insert this:

import boto3
from botocore import UNSIGNED
from botocore.client import Config

s3 = boto3.client(
    's3',
    endpoint_url='https://s3.eu-central-1.amazonaws.com',
    region_name='eu-central-1',
    config=Config(signature_version=UNSIGNED),
)
response = s3.get_object(
    Bucket='levdik',
    Key='key.txt',
)
stream = response['Body']

with stream as s:
    assert type(stream) == type(s), f"{type(stream)} == {type(s)}"

You will get an error:

AssertionError: <class 'botocore.response.StreamingBody'> == <class 'urllib3.response.HTTPResponse'>

[adev-code]

Thanks for the information and for the repro of the issue. Could you provide more context on why this would be a bug and context for the fix? We would like to know the use case and reasoning.

[dizel3d]

Thanks for the information and for the repro of the issue. Could you provide more context on why this would be a bug and context for the fix? We would like to know the use case and reasoning.

I use such a function to read a file from S3:

def get_object(*args, **kwargs):
  with s3_client.get_object(*args, **kwargs)["Body"] as f:
    return f.read()

In a small percent of cases, f.read() returns only part of file. (I think it happens when the server closes the connection.)
HTTPResponse does not guarantee that the entire file will be returned. But StreamingBody would raise IncompleteReadError and I would retry.

Current behavior of with statement is not obvious and undocumented. Nobody expects that f is not StreamingBody but its underlying stream. I don't see any reason for this behavior. Built-in Python streams like io.BytesIO don't work that way, they return self.