3412: Make non-database fields selectable from API server.
[arvados.git] / services / api / app / models / collection.rb
index 64a6bb05304b84877bb4f6d55231c3e561bc8ee6..ee3b29c5bdc5244523d72aa9a5963e600ebc6ac2 100644 (file)
@@ -12,6 +12,12 @@ class Collection < ArvadosModel
     t.add :manifest_text
   end
 
+  def self.attributes_required_columns
+    super.merge({ "data_size" => ["manifest_text"],
+                  "files" => ["manifest_text"],
+                })
+  end
+
   def redundancy_status
     if redundancy_confirmed_as.nil?
       'unconfirmed'
@@ -29,26 +35,38 @@ class Collection < ArvadosModel
   end
 
   def assign_uuid
-    if self.manifest_text.nil? and self.uuid.nil?
-      super
-    elsif self.manifest_text and self.uuid
+    if not self.manifest_text
+      errors.add :manifest_text, 'not supplied'
+      return false
+    end
+    expect_uuid = Digest::MD5.hexdigest(self.manifest_text)
+    if self.uuid
       self.uuid.gsub! /\+.*/, ''
-      if self.uuid == Digest::MD5.hexdigest(self.manifest_text)
-        self.uuid.gsub! /$/, '+' + self.manifest_text.length.to_s
-        true
-      else
-        errors.add :uuid, 'does not match checksum of manifest_text'
-        false
+      if self.uuid != expect_uuid
+        errors.add :uuid, 'must match checksum of manifest_text'
+        return false
       end
-    elsif self.manifest_text
-      errors.add :uuid, 'not supplied (must match checksum of manifest_text)'
-      false
     else
-      errors.add :manifest_text, 'not supplied'
-      false
+      self.uuid = expect_uuid
     end
+    self.uuid.gsub! /$/, '+' + self.manifest_text.length.to_s
+    true
   end
 
+  # TODO (#3036/tom) replace above assign_uuid method with below assign_uuid and self.generate_uuid
+  # def assign_uuid
+  #   # Even admins cannot assign collection uuids.
+  #   self.uuid = self.class.generate_uuid
+  # end
+  # def self.generate_uuid
+  #   # The last 10 characters of a collection uuid are the last 10
+  #   # characters of the base-36 SHA256 digest of manifest_text.
+  #   [Server::Application.config.uuid_prefix,
+  #    self.uuid_prefix,
+  #    rand(2**256).to_s(36)[-5..-1] + Digest::SHA256.hexdigest(self.manifest_text).to_i(16).to_s(36)[-10..-1],
+  #   ].join '-'
+  # end
+
   def data_size
     inspect_manifest_text if @data_size.nil? or manifest_text_changed?
     @data_size
@@ -66,15 +84,6 @@ class Collection < ArvadosModel
       return
     end
 
-    #normalized_manifest = ""
-    #IO.popen(['arv-normalize'], 'w+b') do |io|
-    #  io.write manifest_text
-    #  io.close_write
-    #  while buf = io.read(2**20)
-    #    normalized_manifest += buf
-    #  end
-    #end
-
     @data_size = 0
     tmp = {}
 
@@ -148,7 +157,7 @@ class Collection < ArvadosModel
     [hash_part, size_part].compact.join '+'
   end
 
-  def self.for_latest_docker_image(search_term, search_tag=nil, readers=nil)
+  def self.uuids_for_docker_image(search_term, search_tag=nil, readers=nil)
     readers ||= [Thread.current[:user]]
     base_search = Link.
       readable_by(*readers).
@@ -156,12 +165,15 @@ class Collection < ArvadosModel
       joins("JOIN collections ON links.head_uuid = collections.uuid").
       order("links.created_at DESC")
 
-    # If the search term is a Collection locator with an associated
-    # Docker image hash link, return that Collection.
-    coll_matches = base_search.
-      where(link_class: "docker_image_hash", collections: {uuid: search_term})
-    if match = coll_matches.first
-      return find_by_uuid(match.head_uuid)
+    # If the search term is a Collection locator that contains one file
+    # that looks like a Docker image, return it.
+    if loc = Locator.parse(search_term)
+      loc.strip_hints!
+      coll_match = readable_by(*readers).where(uuid: loc.to_s).first
+      if coll_match and (coll_match.files.size == 1) and
+          (coll_match.files[0][1] =~ /^[0-9A-Fa-f]{64}\.tar$/)
+        return [loc.to_s]
+      end
     end
 
     # Find Collections with matching Docker image repository+tag pairs.
@@ -176,20 +188,24 @@ class Collection < ArvadosModel
               "docker_image_hash", "#{search_term}%")
     end
 
-    # Select the image that was created most recently.  Note that the
-    # SQL search order and fallback timestamp values are chosen so
-    # that if image timestamps are missing, we use the image with the
-    # newest link.
-    latest_image_link = nil
-    latest_image_timestamp = "1900-01-01T00:00:00Z"
+    # Generate an order key for each result.  We want to order the results
+    # so that anything with an image timestamp is considered more recent than
+    # anything without; then we use the link's created_at as a tiebreaker.
+    uuid_timestamps = {}
     matches.find_each do |link|
-      link_timestamp = link.properties.fetch("image_timestamp",
-                                             "1900-01-01T00:00:01Z")
-      if link_timestamp > latest_image_timestamp
-        latest_image_link = link
-        latest_image_timestamp = link_timestamp
-      end
+      uuid_timestamps[link.head_uuid] =
+        [(-link.properties["image_timestamp"].to_datetime.to_i rescue 0),
+         -link.created_at.to_i]
+    end
+    uuid_timestamps.keys.sort_by { |uuid| uuid_timestamps[uuid] }
+  end
+
+  def self.for_latest_docker_image(search_term, search_tag=nil, readers=nil)
+    image_uuid = uuids_for_docker_image(search_term, search_tag, readers).first
+    if image_uuid.nil?
+      nil
+    else
+      find_by_uuid(image_uuid)
     end
-    latest_image_link.nil? ? nil : find_by_uuid(latest_image_link.head_uuid)
   end
 end