boto3

boto3 copied to clipboard

The Binary class has a problematic implementation of the __str__() method that assumes str and bytes are interchangeable. Example:

Python 3.5.2 (default, Sep 28 2016, 18:11:34)
>>> from boto3.dynamodb.types import Binary
>>> b = Binary(b'hello')

>>> b
Binary(b'hello')

>>> print(b)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: __str__ returned non-string (type bytes)

>>> bytes(b)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: 'Binary' object is not iterable

The above examples all succeed under Python 2.7.

Makes sense to have. We will take a look at your PR as well to make sure this gets implemented.

Bumping this issue as I just encountered it while using python3. Any idea if the fix presented by @miguelgrinberg is acceptable?

Why is this marked as feature-request? It's a bug that happens in Python3.

Also, why is the Binary wrapper class needed at all?

This is currently preventing us from properly utilizing the Binary field in DynamoDB in our application. Is there an ETA on a potential fix for this issue?

FYI: I submitted a PR that addresses this issue (more than two years ago!!!): https://github.com/boto/boto3/pull/848. So if your process allows it, you can install the PR branch to avoid the issue. Unfortunately the maintainers are slow or not interested in getting this fix out. :(

I encountered this issue also. Found I could work around it by accessing the value directly. E.g.

Python 3.5.2 (default, Nov 12 2018, 13:43:14)
>>> from boto3.dynamodb.types import Binary
>>> b = Binary(b'hello')
>>> b
Binary(b'hello')

>>> print(b.value)
b'hello'

>>> print(b.value.decode("utf-8"))
hello

Just wanted to check in here and provide some updates. This seems like Python 2 vs Python 3 issue and can potentially be fixed but we would need to do more investigation to find backwards-compatible solution. This is currently a low priority for us but please feel free to check in back here for updates.

For now, I would refer to the workaround provided by @mhball in previous comment. Thanks for sharing the workaround with others ❤️