mirror of
https://github.com/fog/fog.git
synced 2022-11-09 13:51:43 -05:00
6716f37c5f
Done with `rubocop --auto-correct --only DeprecatedHashMethods`
163 lines
6.5 KiB
Ruby
163 lines
6.5 KiB
Ruby
module Fog
|
|
module Storage
|
|
class InternetArchive
|
|
class Real
|
|
# Get an object from S3
|
|
#
|
|
# @param bucket_name [String] Name of bucket to read from
|
|
# @param object_name [String] Name of object to read
|
|
# @param options [Hash]
|
|
# @option options If-Match [String] Returns object only if its etag matches this value, otherwise returns 412 (Precondition Failed).
|
|
# @option options If-Modified-Since [Time] Returns object only if it has been modified since this time, otherwise returns 304 (Not Modified).
|
|
# @option options If-None-Match [String] Returns object only if its etag differs from this value, otherwise returns 304 (Not Modified)
|
|
# @option options If-Unmodified-Since [Time] Returns object only if it has not been modified since this time, otherwise returns 412 (Precodition Failed).
|
|
# @option options Range [String] Range of object to download
|
|
#
|
|
# @return [Excon::Response] response:
|
|
# * body [String]- Contents of object
|
|
# * headers [Hash]:
|
|
# * Content-Length [String] - Size of object contents
|
|
# * Content-Type [String] - MIME type of object
|
|
# * ETag [String] - Etag of object
|
|
# * Last-Modified [String] - Last modified timestamp for object
|
|
#
|
|
# @see http://docs.amazonwebservices.com/AmazonS3/latest/API/RESTObjectGET.html
|
|
|
|
def get_object(bucket_name, object_name, options = {}, &block)
|
|
unless bucket_name
|
|
raise ArgumentError.new('bucket_name is required')
|
|
end
|
|
unless object_name
|
|
raise ArgumentError.new('object_name is required')
|
|
end
|
|
|
|
params = { :headers => {} }
|
|
params[:headers].merge!(options)
|
|
if options['If-Modified-Since']
|
|
params[:headers]['If-Modified-Since'] = Fog::Time.at(options['If-Modified-Since'].to_i).to_date_header
|
|
end
|
|
if options['If-Unmodified-Since']
|
|
params[:headers]['If-Unmodified-Since'] = Fog::Time.at(options['If-Unmodified-Since'].to_i).to_date_header
|
|
end
|
|
|
|
if block_given?
|
|
params[:response_block] = Proc.new
|
|
end
|
|
|
|
request(params.merge!({
|
|
:expects => [ 200, 206 ],
|
|
:host => "#{bucket_name}.#{@host}",
|
|
:idempotent => true,
|
|
:method => 'GET',
|
|
:path => CGI.escape(object_name),
|
|
}))
|
|
end
|
|
end
|
|
|
|
class Mock # :nodoc:all
|
|
def get_object(bucket_name, object_name, options = {}, &block)
|
|
unless bucket_name
|
|
raise ArgumentError.new('bucket_name is required')
|
|
end
|
|
|
|
unless object_name
|
|
raise ArgumentError.new('object_name is required')
|
|
end
|
|
|
|
response = Excon::Response.new
|
|
if (bucket = self.data[:buckets][bucket_name])
|
|
object = nil
|
|
if bucket[:objects].key?(object_name)
|
|
object = bucket[:objects][object_name].first
|
|
end
|
|
|
|
if (object && !object[:delete_marker])
|
|
if options['If-Match'] && options['If-Match'] != object['ETag']
|
|
response.status = 412
|
|
elsif options['If-Modified-Since'] && options['If-Modified-Since'] >= Time.parse(object['Last-Modified'])
|
|
response.status = 304
|
|
elsif options['If-None-Match'] && options['If-None-Match'] == object['ETag']
|
|
response.status = 304
|
|
elsif options['If-Unmodified-Since'] && options['If-Unmodified-Since'] < Time.parse(object['Last-Modified'])
|
|
response.status = 412
|
|
else
|
|
response.status = 200
|
|
for key, value in object
|
|
case key
|
|
when 'Cache-Control', 'Content-Disposition', 'Content-Encoding', 'Content-Length', 'Content-MD5', 'Content-Type', 'ETag', 'Expires', 'Last-Modified', /^x-amz-meta-/
|
|
response.headers[key] = value
|
|
end
|
|
end
|
|
|
|
body = object[:body]
|
|
if options['Range']
|
|
# since AWS S3 itself does not support multiple range headers, we will use only the first
|
|
ranges = byte_ranges(options['Range'], body.size)
|
|
unless ranges.nil? || ranges.empty?
|
|
response.status = 206
|
|
body = body[ranges.first]
|
|
end
|
|
end
|
|
|
|
unless block_given?
|
|
response.body = body
|
|
else
|
|
data = StringIO.new(body)
|
|
remaining = data.length
|
|
while remaining > 0
|
|
chunk = data.read([remaining, Excon::CHUNK_SIZE].min)
|
|
block.call(chunk)
|
|
remaining -= Excon::CHUNK_SIZE
|
|
end
|
|
end
|
|
end
|
|
else
|
|
response.status = 404
|
|
response.body = "...<Code>NoSuchKey<\/Code>..."
|
|
raise(Excon::Errors.status_error({:expects => 200}, response))
|
|
end
|
|
else
|
|
response.status = 404
|
|
response.body = "...<Code>NoSuchBucket</Code>..."
|
|
raise(Excon::Errors.status_error({:expects => 200}, response))
|
|
end
|
|
response
|
|
end
|
|
|
|
private
|
|
|
|
# === Borrowed from rack
|
|
# Parses the "Range:" header, if present, into an array of Range objects.
|
|
# Returns nil if the header is missing or syntactically invalid.
|
|
# Returns an empty array if none of the ranges are satisfiable.
|
|
def byte_ranges(http_range, size)
|
|
# See <http://www.w3.org/Protocols/rfc2616/rfc2616-sec14.html#sec14.35>
|
|
return nil unless http_range
|
|
ranges = []
|
|
http_range.split(/,\s*/).each do |range_spec|
|
|
matches = range_spec.match(/bytes=(\d*)-(\d*)/)
|
|
return nil unless matches
|
|
r0,r1 = matches[1], matches[2]
|
|
if r0.empty?
|
|
return nil if r1.empty?
|
|
# suffix-byte-range-spec, represents trailing suffix of file
|
|
r0 = [size - r1.to_i, 0].max
|
|
r1 = size - 1
|
|
else
|
|
r0 = r0.to_i
|
|
if r1.empty?
|
|
r1 = size - 1
|
|
else
|
|
r1 = r1.to_i
|
|
return nil if r1 < r0 # backwards range is syntactically invalid
|
|
r1 = size-1 if r1 >= size
|
|
end
|
|
end
|
|
ranges << (r0..r1) if r0 <= r1
|
|
end
|
|
ranges
|
|
end
|
|
end
|
|
end
|
|
end
|
|
end
|