Django-storages: s3boto3 en/decoding issue

[AGASS007]

I recently started testing the new s3boto3 backend, and stumbled upon some unicode/str mixups (I'm still rolling with python2.7).

I have a file field with a path containing non ASCII characters and I encountered an issue at file loading (here). After further investigation, it seems that boto3 expects each name argument (the one given to each resource) to be a python2 unicode/python3 str, as it goes though this method. But on django-storages side, every path/filename goes though the internal method _encode_name, which converts to bytes in python2 (with smart_str, an alias for smart_bytes in python2). Hence my UnicodeDecodeError.

I'll open a PR soon. Any feedback welcome.

[jschneier]

Thanks for bringing this up...a while ago :)

[jschneier]

Hi. Does anyone watching this issue have a traceback? I'm eager to merge in this fix but cannot seem to cause the UnicodeDecodeError to happen.

[AGASS007]

Here's a case where I'm getting it. The file name contains non ascii characters.

----> 1 my_model.my_image_field.file.read()

/home/vagrant/code/.virtualenvs/lite/lib/python2.7/site-packages/django/db/models/fields/files.pyc in _get_file(self)
     49         self._require_file()
     50         if not hasattr(self, '_file') or self._file is None:
---> 51             self._file = self.storage.open(self.name, 'rb')
     52         return self._file
     53

/home/vagrant/code/.virtualenvs/lite/lib/python2.7/site-packages/django/core/files/storage.pyc in open(self, name, mode)
     35         Retrieves the specified file from storage.
     36         """
---> 37         return self._open(name, mode)
     38
     39     def save(self, name, content, max_length=None):

/home/vagrant/code/.virtualenvs/lite/lib/python2.7/site-packages/storages/backends/s3boto3.pyc in _open(self, name, mode)
    422         name = self._normalize_name(self._clean_name(name))
    423         try:
--> 424             f = self.file_class(name, mode, self)
    425         except self.connection_response_error as err:
    426             if err.response['ResponseMetadata']['HTTPStatusCode'] == 404:

/home/vagrant/code/.virtualenvs/lite/lib/python2.7/site-packages/storages/backends/s3boto3.pyc in __init__(self, name, mode, storage, buffer_size)
     96         if 'w' not in mode:
     97             # Force early RAII-style exception if object does not exist
---> 98             self.obj.load()
     99         self._is_dirty = False
    100         self._file = None

/home/vagrant/code/.virtualenvs/lite/lib/python2.7/site-packages/boto3/resources/factory.pyc in do_action(self, *args, **kwargs)
    503             # instance via ``self``.
    504             def do_action(self, *args, **kwargs):
--> 505                 response = action(self, *args, **kwargs)
    506                 self.meta.data = response
    507             # Create the docstring for the load/reload mehtods.

/home/vagrant/code/.virtualenvs/lite/lib/python2.7/site-packages/boto3/resources/action.pyc in __call__(self, parent, *args, **kwargs)
     81                     operation_name, params)
     82
---> 83         response = getattr(parent.meta.client, operation_name)(**params)
     84
     85         logger.debug('Response: %r', response)

/home/vagrant/code/.virtualenvs/lite/lib/python2.7/site-packages/botocore/client.pyc in _api_call(self, *args, **kwargs)
    249                     "%s() only accepts keyword arguments." % py_operation_name)
    250             # The "self" in this scope is referring to the BaseClient.
--> 251             return self._make_api_call(operation_name, kwargs)
    252
    253         _api_call.__name__ = str(py_operation_name)

/home/vagrant/code/.virtualenvs/lite/lib/python2.7/site-packages/botocore/client.pyc in _make_api_call(self, operation_name, api_params)
    511         }
    512         request_dict = self._convert_to_request_dict(
--> 513             api_params, operation_model, context=request_context)
    514
    515         handler, event_response = self.meta.events.emit_until_response(

/home/vagrant/code/.virtualenvs/lite/lib/python2.7/site-packages/botocore/client.pyc in _convert_to_request_dict(self, api_params, operation_model, context)
    564
    565         request_dict = self._serializer.serialize_to_request(
--> 566             api_params, operation_model)
    567         prepare_request_dict(request_dict, endpoint_url=self._endpoint.host,
    568                              user_agent=self._client_config.user_agent,

/home/vagrant/code/.virtualenvs/lite/lib/python2.7/site-packages/botocore/validate.pyc in serialize_to_request(self, parameters, operation_model)
    270                 raise ParamValidationError(report=report.generate_report())
    271         return self._serializer.serialize_to_request(parameters,
--> 272                                                      operation_model)

/home/vagrant/code/.virtualenvs/lite/lib/python2.7/site-packages/botocore/serialize.pyc in serialize_to_request(self, parameters, operation_model)
    406         serialized['url_path'] = self._render_uri_template(
    407             operation_model.http['requestUri'],
--> 408             partitioned['uri_path_kwargs'])
    409         # Note that we lean on the http implementation to handle the case
    410         # where the requestUri path already has query parameters.

/home/vagrant/code/.virtualenvs/lite/lib/python2.7/site-packages/botocore/serialize.pyc in _render_uri_template(self, uri_template, params)
    428             if template_param.endswith('+'):
    429                 encoded_params[template_param] = percent_encode(
--> 430                     params[template_param[:-1]], safe='/~')
    431             else:
    432                 encoded_params[template_param] = percent_encode(

/home/vagrant/code/.virtualenvs/lite/lib/python2.7/site-packages/botocore/utils.pyc in percent_encode(input_str, safe)
    308     if not isinstance(input_str, string_types):
    309         input_str = text_type(input_str)
--> 310     return quote(text_type(input_str).encode('utf-8'), safe=safe)
    311
    312

UnicodeDecodeError: 'ascii' codec can't decode byte 0xc3 in position 30: ordinal not in range(128)

traceback.txt

[AGASS007]

@jschneier ping!

[RailTracker]

Just experienced this issue while trying to upload an image with a Chinese filename to S3. I'm also on 2.7.

[jschneier]

Merged #217, will be in the next release this week.

[anuj9196]

I'm using version 1.7.1

When I try to upload a file from Django application, it closes the application server without any error or any message.

I tried uploading from Django's shell

file = File(open('/path/to/file'))
m = Media(file=file)
m.save()

But it gives an error

File "/home/user/.virtualenvs/qcg-TqOLHEIu/lib/python3.7/site-packages/s3transfer/upload.py", line 86, in read
    return self._fileobj.read(amount)
  File "/home/user/.virtualenvs/qcg-TqOLHEIu/lib/python3.7/codecs.py", line 322, in decode
    (result, consumed) = self._buffer_decode(data, self.errors, final)
UnicodeDecodeError: 'utf-8' codec can't decode byte 0x89 in position 0: invalid start byte