How do I add or modify the Content-Disposition of an existing object in Amazon S3?

Question

We have hundreds of object in an AWS S3 bucket which don't have content disposition set.

I'm using Ruby aws-sdk gem.

How do you add or change content disposition to these objects, WITHOUT re-uploading the files again?

I have tried

obj.write(:content_disposition => 'attachment')
obj.copy_from(obj.key, :content_disposition => 'attachment')

and also copy_to(), move_to(), but none of these seem to work in adding the content disposition to the objects. In a few cases, the objects don't seem to have been modified at all (the modification time didn't change), in other cases, the object file is corrupted!

---

I'm aware of the alternative of using :response_content_disposition when requesting the s3 object via HTTP, which sets the Content-Disposition header,

obj.url_for(:read, :response_content_disposition => "attachment")

Thank you!

Answer 1

We found the solution by changing the aws-sdk source code.

In s3/s3_object.rb

add the following lines to copy_from() (similar to how :content_type was handled)

if options[:content_disposition]
    copy_opts[:content_disposition] = options[:content_disposition]
    copy_opts[:metadata_directive] = "REPLACE"
end

also, in s3/client.rb

add the line marked below

object_method(:copy_object, :put,
                    :header_options => {
                    :copy_source => 'x-amz-copy-source',
                    :cache_control => 'Cache-Control',
                    :metadata_directive => 'x-amz-metadata-directive',
                    :storage_class => 'x-amz-storage-class',
                    :server_side_encryption => 'x-amz-server-side-encryption',
                    :content_type => 'Content-Type',
                    :content_disposition => 'Content-Disposition', # add this line here
                }) do

Once you've done the above, you can do the following to add content disposition to your existing object:

obj.copy_from(obj.key, :content_disposition => 'attachment', :content_type => 'image/png', :acl => :public_read)

Answer 2

for future generations, here's a cli-based solution. if first lists the objects in /mybucket/brand_img/ios/ and then sets the content-disposition on each of the items to the filename. I needed this because my filenames had '@' in them and I did not want to force clients to url-encode the path.

aws s3 ls s3://mybucket/brand_img/ios/|awk {'print $4'} > objects.txt

while read line; do aws s3api copy-object --bucket mybucket  \
--copy-source /mybucket/brand_img/ios/$line --key brand_img/ios/$line \
--metadata-directive REPLACE --metadata Content-Disposition=$line --acl public-read; done < objects.txt