uploadGeneratetoken: Add support to optionally create new revision

[jcfr]

This commit updates the "uploadGeneratetoken" API introducing the new
parameter "create_additional_revision" allowing to optionally change
the default behavior and ensure a new revision is created when:

• a bitstream associated with the file already exists on the server
• and the checksum parameter is specified
• and the item updated has at least one revision

Since the http uploader doesn't specify the "checksum" parameter, the
problem is only reproducible using the web API.

For example, considering the following configuration:

 folder
   |--- item1 --- revision1
   |                 |--- bitstream1 (filename1)
   |
   |--- item2 --- revision1
   |                 |--- bitstream2 (filename2)

By default, trying to associate bitstream1 with item2 will result in
this configuration:

 folder
   |--- item1 --- revision1
   |                 |--- bitstream1 (filename1)
   |
   |--- item2 --- revision1
   |                 |--- bitstream2 (filename2)
   |                 |--- bitstream1 (filename1)

By setting the option "create_additional_revision" to True, the
following configuration will be obtained:

 folder
   |--- item1 --- revision1
   |                 |--- bitstream1 (filename1)
   |
   |--- item2 -|--- revision1
   |           |     |--- bitstream2 (filename2)
   |           |
   |           |--- revision2
   |           |     |--- bitstream1 (filename1)

[jcfr]

Cc: @jamiesnape @cpatrick

[jcfr]

The following snippet allow to reproduce the configuration:

%reload_ext autoreload
%autoreload 1
%aimport pydas
import hashlib
import os
import time

midas_url = 'http://slicer.kitware.com/midas3'
email = os.environ["MIDAS_PACKAGE_EMAIL"]
apikey = os.environ["MIDAS_PACKAGE_API_KEY"]
communicator = pydas.core.Communicator(midas_url)

folder_id = 1654 # Update this with a folder you have write access

def token(email, apikey):
    return communicator.login_with_api_key(email, apikey)

def md5(fname):
    hash = hashlib.md5()
    with open(fname) as f:
        for chunk in iter(lambda: f.read(4096), ""):
            hash.update(chunk)
    return hash.hexdigest()

def generate_file(filepath):
    with open(filepath, 'wb') as fout:
        fout.write(os.urandom(1024))
    filename = os.path.basename(filepath)
    checksum = md5(filepath)
    print("filepath: %s, md5: %s" % (filepath, checksum))
    return (filename, checksum)

prefix = str(int(time.time()))

# Create original file for item #1
filepath_item1 = './%s-bistream-1.bin' % prefix
(filename_item1, checksum_item1) = generate_file(filepath_item1)

# Create item #1
item = communicator.create_item(token(email, apikey), '%s-item1' % prefix, folder_id)
item_id = item['item_id']
print("item #1 created: %s" % (item_id))

# Upload initial revision into item #1
upload_token = communicator.generate_upload_token(token(email, apikey), item_id, filename_item1)
communicator.perform_upload(upload_token, filename_item1, item_id=item_id, filepath=filepath_item1)
print("item #1: initial revision created")

# Create original file for item #2
filepath_item2 = './%s-bistream-2.bin' % prefix
(filename_item2, checksum_item2) = generate_file(filepath_item2)

# Create item #2
item = communicator.create_item(token(email, apikey), '%s-item2' % prefix, folder_id)
item_id = item['item_id']
print("item #2 created: %s" % (item_id))

# Upload initial revision into item #2
upload_token = communicator.generate_upload_token(token(email, apikey), item_id, filename_item2)
communicator.perform_upload(upload_token, filename_item2, item_id=item_id, filepath=filepath_item2)
print("item #2: initial revision created")

# Upload an other revision into item #2 associating bitstream created for item #1
upload_token = communicator.generate_upload_token(token(email, apikey), item_id, filename_item1, checksum=checksum_item1)
print("item #2: other revision created")
if upload_token:
    raise Exception('item #2: upload_token is expected to be empty')

Item1:

Item2:

[jcfr]

And now by using an updated pydas adding support for create_additional_revision to generate_upload_token (See midasplatform/pydas#29) and updating the snippet posted in #146 (comment) changing:

# Upload an other revision into item #2 associating bitstream created for item #1
upload_token = communicator.generate_upload_token(token(email, apikey), item_id, filename_item1, checksum=checksum_item1)

with

# Upload an other revision into item #2 associating bitstream created for item #1
upload_token = communicator.generate_upload_token(token(email, apikey), item_id, filename_item1, checksum=checksum_item1, create_additional_revision=True)

If you scroll right, you can see the introduction of create_additional_revision=True

We obtain the expected results:

Item 1:

Item 2:

[jcfr]

I just realized the test are broken, I will fix the style.

[jamiesnape]

===

[jcfr]

Thanks for your review, comments have been addressed in a follow up commit.

[jcfr]

And now Style-CI is 😄