Fixes API server to return default columns for collections
[arvados.git] / services / api / app / controllers / arvados / v1 / collections_controller.rb
index 517e84bc4909648bea12bb3ecc3cb8d524ae41dc..922cf7dac16b87741013c23e4073d4070a6fbe43 100644 (file)
+require "arvados/keep"
+
 class Arvados::V1::CollectionsController < ApplicationController
+  def self.limit_index_columns_read
+    ["manifest_text"]
+  end
+
   def create
-    # Collections are owned by system_user. Creating a collection has
-    # two effects: The collection is added if it doesn't already
-    # exist, and a "permission" Link is added (if one doesn't already
-    # exist) giving the current user (or specified owner_uuid)
-    # permission to read it.
-    owner_uuid = resource_attrs.delete(:owner_uuid) || current_user.uuid
-    owner_kind = if owner_uuid.match(/-(\w+)-/)[1] == User.uuid_prefix
-                   'arvados#user'
-                 else
-                   'arvados#group'
-                 end
-    unless current_user.can? write: owner_uuid
-      logger.warn "User #{current_user.andand.uuid} tried to set collection owner_uuid to #{owner_uuid}"
-      raise ArvadosModel::PermissionDeniedError
+    if resource_attrs[:uuid] and (loc = Keep::Locator.parse(resource_attrs[:uuid]))
+      resource_attrs[:portable_data_hash] = loc.to_s
+      resource_attrs.delete :uuid
     end
-    act_as_system_user do
-      @object = model_class.new resource_attrs.reject { |k,v| k == :owner_uuid }
-      begin
-        @object.save!
-      rescue ActiveRecord::RecordNotUnique
-        logger.debug resource_attrs.inspect
-        if resource_attrs[:manifest_text] and resource_attrs[:uuid]
-          @existing_object = model_class.
-            where('uuid=? and manifest_text=?',
-                  resource_attrs[:uuid],
-                  resource_attrs[:manifest_text]).
-            first
-          @object = @existing_object || @object
-        end
-      end
+    super
+  end
 
-      if @object
-        link_attrs = {
-          owner_uuid: owner_uuid,
-          link_class: 'permission',
-          name: 'can_read',
-          head_kind: 'arvados#collection',
-          head_uuid: @object.uuid,
-          tail_kind: owner_kind,
-          tail_uuid: owner_uuid
+  def find_object_by_uuid
+    if loc = Keep::Locator.parse(params[:id])
+      loc.strip_hints!
+      if c = Collection.readable_by(*@read_users).where({ portable_data_hash: loc.to_s }).limit(1).first
+        @object = {
+          uuid: c.portable_data_hash,
+          portable_data_hash: c.portable_data_hash,
+          manifest_text: c.signed_manifest_text,
         }
-        ActiveRecord::Base.transaction do
-          if Link.where(link_attrs).empty?
-            Link.create! link_attrs
-          end
-        end
       end
+    else
+      super
     end
-    show
+    true
   end
 
-  def collection_uuid(uuid)
-    m = /([a-f0-9]{32}(\+[0-9]+)?)(\+.*)?/.match(uuid)
-    if m
-      m[1]
+  def show
+    if @object.is_a? Collection
+      super
     else
-      nil
+      send_json @object
     end
   end
 
-  def script_param_edges(visited, sp)
-    if sp and not sp.empty?
-      case sp
-      when Hash
-        sp.each do |k, v|
-          script_param_edges(visited, v)
-        end
-      when Array
-        sp.each do |v|
-          script_param_edges(visited, v)
-        end
-      else
-        m = collection_uuid(sp)
-        if m
-          generate_provenance_edges(visited, m)
-        end
+  def find_collections(visited, sp, &b)
+    case sp
+    when ArvadosModel
+      sp.class.columns.each do |c|
+        find_collections(visited, sp[c.name.to_sym], &b) if c.name != "log"
+      end
+    when Hash
+      sp.each do |k, v|
+        find_collections(visited, v, &b)
+      end
+    when Array
+      sp.each do |v|
+        find_collections(visited, v, &b)
+      end
+    when String
+      if m = /[a-f0-9]{32}\+\d+/.match(sp)
+        yield m[0], nil
+      elsif m = Collection.uuid_regex.match(sp)
+        yield nil, m[0]
       end
     end
   end
 
-  def generate_provenance_edges(visited, uuid)
-    m = collection_uuid(uuid)
-    uuid = m if m
+  def search_edges(visited, uuid, direction)
+    if uuid.nil? or uuid.empty? or visited[uuid]
+      return
+    end
 
-    if not uuid or uuid.empty? or visited[uuid]
-      return ""
+    if loc = Keep::Locator.parse(uuid)
+      loc.strip_hints!
+      return if visited[loc.to_s]
     end
 
-    #puts "visiting #{uuid}"
+    logger.debug "visiting #{uuid}"
 
-    if m  
-      # uuid is a collection
-      Collection.where(uuid: uuid).each do |c|
-        visited[uuid] = c.as_api_response
-        visited[uuid][:files] = []
-        c.files.each do |f|
-          visited[uuid][:files] << f
+    if loc
+      # uuid is a portable_data_hash
+      collections = Collection.readable_by(*@read_users).where(portable_data_hash: loc.to_s)
+      c = collections.limit(2).all
+      if c.size == 1
+        visited[loc.to_s] = c[0]
+      elsif c.size > 1
+        name = collections.limit(1).where("name <> ''").first
+        if name
+          visited[loc.to_s] = {
+            portable_data_hash: c[0].portable_data_hash,
+            name: "#{name.name} + #{collections.count-1} more"
+          }
+        else
+          visited[loc.to_s] = {
+            portable_data_hash: c[0].portable_data_hash,
+            name: loc.to_s
+          }
         end
       end
 
-      Job.where(output: uuid).each do |job|
-        generate_provenance_edges(visited, job.uuid)
-      end
+      if direction == :search_up
+        # Search upstream for jobs where this locator is the output of some job
+        Job.readable_by(*@read_users).where(output: loc.to_s).each do |job|
+          search_edges(visited, job.uuid, :search_up)
+        end
 
-      Job.where(log: uuid).each do |job|
-        generate_provenance_edges(visited, job.uuid)
+        Job.readable_by(*@read_users).where(log: loc.to_s).each do |job|
+          search_edges(visited, job.uuid, :search_up)
+        end
+      elsif direction == :search_down
+        if loc.to_s == "d41d8cd98f00b204e9800998ecf8427e+0"
+          # Special case, don't follow the empty collection.
+          return
+        end
+
+        # Search downstream for jobs where this locator is in script_parameters
+        Job.readable_by(*@read_users).where(["jobs.script_parameters like ?", "%#{loc.to_s}%"]).each do |job|
+          search_edges(visited, job.uuid, :search_down)
+        end
+
+        Job.readable_by(*@read_users).where(["jobs.docker_image_locator = ?", "#{loc.to_s}"]).each do |job|
+          search_edges(visited, job.uuid, :search_down)
+        end
       end
-      
     else
-      # uuid is something else
+      # uuid is a regular Arvados UUID
       rsc = ArvadosModel::resource_class_for_uuid uuid
       if rsc == Job
-        Job.where(uuid: uuid).each do |job|
+        Job.readable_by(*@read_users).where(uuid: uuid).each do |job|
           visited[uuid] = job.as_api_response
-          script_param_edges(visited, job.script_parameters)
+          if direction == :search_up
+            # Follow upstream collections referenced in the script parameters
+            find_collections(visited, job) do |hash, uuid|
+              search_edges(visited, hash, :search_up) if hash
+              search_edges(visited, uuid, :search_up) if uuid
+            end
+          elsif direction == :search_down
+            # Follow downstream job output
+            search_edges(visited, job.output, direction)
+          end
+        end
+      elsif rsc == Collection
+        if c = Collection.readable_by(*@read_users).where(uuid: uuid).limit(1).first
+          search_edges(visited, c.portable_data_hash, direction)
+          visited[c.portable_data_hash] = c.as_api_response
         end
       elsif rsc != nil
         rsc.where(uuid: uuid).each do |r|
@@ -123,39 +146,46 @@ class Arvados::V1::CollectionsController < ApplicationController
       end
     end
 
-    Link.where(head_uuid: uuid, link_class: "provenance").each do |link|
-      visited[link.uuid] = link.as_api_response
-      generate_provenance_edges(visited, link.tail_uuid)
+    if direction == :search_up
+      # Search for provenance links pointing to the current uuid
+      Link.readable_by(*@read_users).
+        where(head_uuid: uuid, link_class: "provenance").
+        each do |link|
+        visited[link.uuid] = link.as_api_response
+        search_edges(visited, link.tail_uuid, direction)
+      end
+    elsif direction == :search_down
+      # Search for provenance links emanating from the current uuid
+      Link.readable_by(current_user).
+        where(tail_uuid: uuid, link_class: "provenance").
+        each do |link|
+        visited[link.uuid] = link.as_api_response
+        search_edges(visited, link.head_uuid, direction)
+      end
     end
-
-    #puts "finished #{uuid}"
   end
 
   def provenance
     visited = {}
-    generate_provenance_edges(visited, @object[:uuid])
-    render json: visited
+    search_edges(visited, @object[:portable_data_hash], :search_up)
+    search_edges(visited, @object[:uuid], :search_up)
+    send_json visited
   end
 
+  def used_by
+    visited = {}
+    search_edges(visited, @object[:uuid], :search_down)
+    search_edges(visited, @object[:portable_data_hash], :search_down)
+    send_json visited
+  end
 
   protected
-  def find_object_by_uuid
+
+  def load_limit_offset_order_params *args
     super
-    if !@object and !params[:uuid].match(/^[0-9a-f]+\+\d+$/)
-      # Normalize the given uuid and search again.
-      hash_part = params[:uuid].match(/^([0-9a-f]*)/)[1]
-      collection = Collection.where('uuid like ?', hash_part + '+%').first
-      if collection
-        # We know the collection exists, and what its real uuid is in
-        # the database. Now, throw out @objects and repeat the usual
-        # lookup procedure. (Returning the collection at this point
-        # would bypass permission checks.)
-        @objects = nil
-        @where = { uuid: collection.uuid }
-        find_objects_for_index
-        @object = @objects.first
-      end
+    if action_name == 'index'
+      # Omit manifest_text from index results unless expressly selected.
+      @select ||= model_class.selectable_attributes - ["manifest_text"]
     end
   end
-
 end