allow copy_object with content-type in metadata

[vlcinsky]

Trying to use copy_object to copy an object from one bucket to another.

To keep the target object well functional, I need to preserve Content-Type and Content-Encoding headers.

Content-Encoding is possible passing it's value via metadata.

However, I did not succeed with Content-Type.

The only solution was to edit sources of minio/helpers.py function is_supported_header adding value "content-type".

# returns true if a standard supported header
def is_supported_header(key):
    ## Supported headers for object.
    supported_headers = [
       "cache-control",
       "content-encoding",
       "content-type",  # added one
       "content-disposition",
       "content-language",
       "x-amz-website-redirect-location",
            ## Add more supported headers here.
        ]
    return key.lower() in supported_headers

This might be the solution, but I am not sure, it does not break anything else.

[poornas]

@vlcinsky , source metadata will be copied automatically to destination if you leave metadata as None in the copy_object call. You should pass metadata in the api call only if it is your intention to replace metadata in the source object.

copy_result = client.copy_object("testbucket", "test/image2.jpg",
                                     "testbucket/test/image1.jpg",
                                     copy_conditions)

[vlcinsky]

@poornas thanks, I confirm your suggestion to leave setting metadata copied source content-type and content-encoding to target object (I did not use copy_condition too, but this is probably irrelevant).

For my use case this works well now. Thanks a lot.

In other cases, when user copying object and trying to fix content-type e.g. from text/xml to application/xml it would still be an issue.

[poornas]

s3 CopyObject api allows you to entirely replace the metadata or retain the existing metadata on source side at destination, not both. If you want to keep some of the metadata at source and also add different metadata headers - then application needs to first do a stat of the object and get the metadata, then pass it in your copy_object with modifications you might need.

[vlcinsky]

@poornas the concept you describe is clear and I like it (either implicitly copy all metadata or explicitly define all).

Anyway, attempting to change content-type failed while with content-encoding it succeeded. So the current implementation does not completely follow the concept.

See the PR #739. It allows modifying content-type via metadata, but test case https://github.com/minio/minio-py/blob/master/tests/unit/header_value_test.py#L24 is rejecting such an option.

[poornas]

you are right @vlcinsky