Support multipart range requests

[fatkodima]

Closes #559

[ioquatix]

Sorry, I only have a short time to review this, but I'm excited to see this PR.

Is AaB03x sufficient as a boundary?

[fatkodima]

Yes, why not? The same boundary is used for multipart forms

rack/lib/rack/multipart.rb

Line 14 in 93dfcdf

MULTIPART_BOUNDARY = "AaB03x"

[ioquatix]

Can we encode this as a string rather than a heredoc.

[fatkodima]

I tried to stick here with existing code styles, like

rack/lib/rack/multipart/generator.rb

Lines 71 to 89 in 93dfcdf

	def content_for_tempfile(io, file, name)
	<<-EOF
	--#{MULTIPART_BOUNDARY}\r
	Content-Disposition: form-data; name="#{name}"; filename="#{Utils.escape(file.original_filename)}"\r
	Content-Type: #{file.content_type}\r
	Content-Length: #{::File.stat(file.path).size}\r
	\r
	#{io.read}\r
	EOF
	end
	def content_for_other(file, name)
	<<-EOF
	--#{MULTIPART_BOUNDARY}\r
	Content-Disposition: form-data; name="#{name}"\r
	\r
	#{file}\r
	EOF
	end

and like examples in many test cases.
So should I change this to string?

[jeremyevans]

@fatkodima I think the implementation looks pretty good. I think you can keep using the existing static boundary. I also think you can keep using the heredoc, you don't need to change it to a string. Can you please rebase this against master? Please make sure that it works for multiple ranges the same as for single ranges (which now use BaseIterator, see 3936b90).

[fatkodima]

@jeremyevans Updated.

[jeremyevans]

Looks good. I tested locally with curl -r and got the expected results. I plan on merging in a few days unless other committer objects.