apps/workbench/app/controllers/collections_controller.rb

   1 require "arvados/keep"
   2
   3 class CollectionsController < ApplicationController
   4   include ActionController::Live
   5
   6   skip_around_filter(:require_thread_api_token,
   7                      only: [:show_file, :show_file_links])
   8   skip_before_filter(:find_object_by_uuid,
   9                      only: [:provenance, :show_file, :show_file_links])
  10   # We depend on show_file to display the user agreement:
  11   skip_before_filter :check_user_agreements, only: :show_file
  12   skip_before_filter :check_user_profile, only: :show_file
  13
  14   RELATION_LIMIT = 5
  15
  16   def show_pane_list
  17     %w(Files Provenance_graph Used_by Advanced)
  18   end
  19
  20   def set_persistent
  21     case params[:value]
  22     when 'persistent', 'cache'
  23       persist_links = Link.filter([['owner_uuid', '=', current_user.uuid],
  24                                    ['link_class', '=', 'resources'],
  25                                    ['name', '=', 'wants'],
  26                                    ['tail_uuid', '=', current_user.uuid],
  27                                    ['head_uuid', '=', @object.uuid]])
  28       logger.debug persist_links.inspect
  29     else
  30       return unprocessable "Invalid value #{value.inspect}"
  31     end
  32     if params[:value] == 'persistent'
  33       if not persist_links.any?
  34         Link.create(link_class: 'resources',
  35                     name: 'wants',
  36                     tail_uuid: current_user.uuid,
  37                     head_uuid: @object.uuid)
  38       end
  39     else
  40       persist_links.each do |link|
  41         link.destroy || raise
  42       end
  43     end
  44
  45     respond_to do |f|
  46       f.json { render json: @object }
  47     end
  48   end
  49
  50   def choose
  51     # Find collections using default find_objects logic, then search for name
  52     # links, and preload any other links connected to the collections that are
  53     # found.
  54     # Name links will be obsolete when issue #3036 is merged,
  55     # at which point this entire custom #choose function can probably be
  56     # eliminated.
  57
  58     params[:limit] ||= 40
  59
  60     find_objects_for_index
  61     @collections = @objects
  62
  63     @filters += [['link_class','=','name'],
  64                  ['head_uuid','is_a','arvados#collection']]
  65
  66     @objects = Link
  67     find_objects_for_index
  68
  69     @name_links = @objects
  70
  71     @objects = Collection.
  72       filter([['uuid','in',@name_links.collect(&:head_uuid)]])
  73
  74     preload_links_for_objects (@collections.to_a + @objects.to_a)
  75     super
  76   end
  77
  78   def index
  79     # API server index doesn't return manifest_text by default, but our
  80     # callers want it unless otherwise specified.
  81     @select ||= Collection.columns.map(&:name)
  82     base_search = Collection.select(@select)
  83     if params[:search].andand.length.andand > 0
  84       tags = Link.where(any: ['contains', params[:search]])
  85       @collections = (base_search.where(uuid: tags.collect(&:head_uuid)) |
  86                       base_search.where(any: ['contains', params[:search]])).
  87         uniq { |c| c.uuid }
  88     else
  89       if params[:limit]
  90         limit = params[:limit].to_i
  91       else
  92         limit = 100
  93       end
  94
  95       if params[:offset]
  96         offset = params[:offset].to_i
  97       else
  98         offset = 0
  99       end
 100
 101       @collections = base_search.limit(limit).offset(offset)
 102     end
 103     @links = Link.limit(1000).
 104       where(head_uuid: @collections.collect(&:uuid))
 105     @collection_info = {}
 106     @collections.each do |c|
 107       @collection_info[c.uuid] = {
 108         tag_links: [],
 109         wanted: false,
 110         wanted_by_me: false,
 111         provenance: [],
 112         links: []
 113       }
 114     end
 115     @links.each do |link|
 116       @collection_info[link.head_uuid] ||= {}
 117       info = @collection_info[link.head_uuid]
 118       case link.link_class
 119       when 'tag'
 120         info[:tag_links] << link
 121       when 'resources'
 122         info[:wanted] = true
 123         info[:wanted_by_me] ||= link.tail_uuid == current_user.uuid
 124       when 'provenance'
 125         info[:provenance] << link.name
 126       end
 127       info[:links] << link
 128     end
 129     @request_url = request.url
 130
 131     render_index
 132   end
 133
 134   def show_file_links
 135     Thread.current[:reader_tokens] = [params[:reader_token]]
 136     return if false.equal?(find_object_by_uuid)
 137     render layout: false
 138   end
 139
 140   def show_file
 141     # We pipe from arv-get to send the file to the user.  Before we start it,
 142     # we ask the API server if the file actually exists.  This serves two
 143     # purposes: it lets us return a useful status code for common errors, and
 144     # helps us figure out which token to provide to arv-get.
 145     coll = nil
 146     tokens = [Thread.current[:arvados_api_token], params[:reader_token]].compact
 147     usable_token = find_usable_token(tokens) do
 148       coll = Collection.find(params[:uuid])
 149     end
 150
 151     file_name = params[:file].andand.sub(/^(\.\/|\/|)/, './')
 152     if usable_token.nil?
 153       return  # Response already rendered.
 154     elsif file_name.nil? or not coll.manifest.has_file?(file_name)
 155       return render_not_found
 156     end
 157
 158     opts = params.merge(arvados_api_token: usable_token)
 159
 160     # Handle Range requests. Currently we support only 'bytes=0-....'
 161     if request.headers.include? 'HTTP_RANGE'
 162       if m = /^bytes=0-(\d+)/.match(request.headers['HTTP_RANGE'])
 163         opts[:maxbytes] = m[1]
 164         size = params[:size] || '*'
 165         self.response.status = 206
 166         self.response.headers['Content-Range'] = "bytes 0-#{m[1]}/#{size}"
 167       end
 168     end
 169
 170     ext = File.extname(params[:file])
 171     self.response.headers['Content-Type'] =
 172       Rack::Mime::MIME_TYPES[ext] || 'application/octet-stream'
 173     if params[:size]
 174       size = params[:size].to_i
 175       if opts[:maxbytes]
 176         size = [size, opts[:maxbytes].to_i].min
 177       end
 178       self.response.headers['Content-Length'] = size.to_s
 179     end
 180     self.response.headers['Content-Disposition'] = params[:disposition] if params[:disposition]
 181     begin
 182       file_enumerator(opts).each do |bytes|
 183         response.stream.write bytes
 184       end
 185     ensure
 186       response.stream.close
 187     end
 188   end
 189
 190   def sharing_scopes
 191     ["GET /arvados/v1/collections/#{@object.uuid}", "GET /arvados/v1/collections/#{@object.uuid}/", "GET /arvados/v1/keep_services/accessible"]
 192   end
 193
 194   def search_scopes
 195     begin
 196       ApiClientAuthorization.filter([['scopes', '=', sharing_scopes]]).results
 197     rescue ArvadosApiClient::AccessForbiddenException
 198       nil
 199     end
 200   end
 201
 202   def find_object_by_uuid
 203     if not Keep::Locator.parse params[:id]
 204       super
 205     end
 206   end
 207
 208   def show
 209     return super if !@object
 210     if current_user
 211       if Keep::Locator.parse params["uuid"]
 212         @same_pdh = Collection.filter([["portable_data_hash", "=", @object.portable_data_hash]]).limit(1000)
 213         if @same_pdh.results.size == 1
 214           redirect_to collection_path(@same_pdh[0]["uuid"])
 215           return
 216         end
 217         owners = @same_pdh.map(&:owner_uuid).to_a.uniq
 218         preload_objects_for_dataclass Group, owners
 219         preload_objects_for_dataclass User, owners
 220         render 'hash_matches'
 221         return
 222       else
 223         jobs_with = lambda do |conds|
 224           Job.limit(RELATION_LIMIT).where(conds)
 225             .results.sort_by { |j| j.finished_at || j.created_at }
 226         end
 227         @output_of = jobs_with.call(output: @object.portable_data_hash)
 228         @log_of = jobs_with.call(log: @object.portable_data_hash)
 229         @project_links = Link.limit(RELATION_LIMIT).order("modified_at DESC")
 230           .where(head_uuid: @object.uuid, link_class: 'name').results
 231         project_hash = Group.where(uuid: @project_links.map(&:tail_uuid)).to_hash
 232         @projects = project_hash.values
 233
 234         @permissions = Link.limit(RELATION_LIMIT).order("modified_at DESC")
 235           .where(head_uuid: @object.uuid, link_class: 'permission',
 236                  name: 'can_read').results
 237         @logs = Log.limit(RELATION_LIMIT).order("created_at DESC")
 238           .where(object_uuid: @object.uuid).results
 239         @is_persistent = Link.limit(1)
 240           .where(head_uuid: @object.uuid, tail_uuid: current_user.uuid,
 241                  link_class: 'resources', name: 'wants')
 242           .results.any?
 243         @search_sharing = search_scopes
 244
 245         if params["tab_pane"] == "Provenance_graph"
 246           @prov_svg = ProvenanceHelper::create_provenance_graph(@object.provenance, "provenance_svg",
 247                                                                 {:request => request,
 248                                                                   :direction => :bottom_up,
 249                                                                   :combine_jobs => :script_only}) rescue nil
 250         end
 251         if params["tab_pane"] == "Used_by"
 252           @used_by_svg = ProvenanceHelper::create_provenance_graph(@object.used_by, "used_by_svg",
 253                                                                    {:request => request,
 254                                                                      :direction => :top_down,
 255                                                                      :combine_jobs => :script_only,
 256                                                                      :pdata_only => true}) rescue nil
 257         end
 258       end
 259     end
 260     super
 261   end
 262
 263   def sharing_popup
 264     @search_sharing = search_scopes
 265     respond_to do |format|
 266       format.html
 267       format.js
 268     end
 269   end
 270
 271   helper_method :download_link
 272
 273   def download_link
 274     collections_url + "/download/#{@object.uuid}/#{@search_sharing.first.api_token}/"
 275   end
 276
 277   def share
 278     a = ApiClientAuthorization.create(scopes: sharing_scopes)
 279     @search_sharing = search_scopes
 280     render 'sharing_popup'
 281   end
 282
 283   def unshare
 284     @search_sharing = search_scopes
 285     @search_sharing.each do |s|
 286       s.destroy
 287     end
 288     @search_sharing = search_scopes
 289     render 'sharing_popup'
 290   end
 291
 292   protected
 293
 294   def find_usable_token(token_list)
 295     # Iterate over every given token to make it the current token and
 296     # yield the given block.
 297     # If the block succeeds, return the token it used.
 298     # Otherwise, render an error response based on the most specific
 299     # error we encounter, and return nil.
 300     most_specific_error = [401]
 301     token_list.each do |api_token|
 302       begin
 303         # We can't load the corresponding user, because the token may not
 304         # be scoped for that.
 305         using_specific_api_token(api_token, load_user: false) do
 306           yield
 307           return api_token
 308         end
 309       rescue ArvadosApiClient::ApiError => error
 310         if error.api_status >= most_specific_error.first
 311           most_specific_error = [error.api_status, error]
 312         end
 313       end
 314     end
 315     case most_specific_error.shift
 316     when 401, 403
 317       redirect_to_login
 318     when 404
 319       render_not_found(*most_specific_error)
 320     end
 321     return nil
 322   end
 323
 324   def file_enumerator(opts)
 325     FileStreamer.new opts
 326   end
 327
 328   class FileStreamer
 329     include ArvadosApiClientHelper
 330     def initialize(opts={})
 331       @opts = opts
 332     end
 333     def each
 334       return unless @opts[:uuid] && @opts[:file]
 335
 336       env = Hash[ENV].dup
 337
 338       require 'uri'
 339       u = URI.parse(arvados_api_client.arvados_v1_base)
 340       env['ARVADOS_API_HOST'] = "#{u.host}:#{u.port}"
 341       env['ARVADOS_API_TOKEN'] = @opts[:arvados_api_token]
 342       env['ARVADOS_API_HOST_INSECURE'] = "true" if Rails.configuration.arvados_insecure_https
 343
 344       bytesleft = @opts[:maxbytes].andand.to_i || 2**16
 345       io = IO.popen([env, 'arv-get', "#{@opts[:uuid]}/#{@opts[:file]}"], 'rb')
 346       while bytesleft > 0 && (buf = io.read([bytesleft, 2**16].min)) != nil
 347         # shrink the bytesleft count, if we were given a maximum byte
 348         # count to read
 349         if @opts.include? :maxbytes
 350           bytesleft = bytesleft - buf.length
 351         end
 352         yield buf
 353       end
 354       io.close
 355       # "If ios is opened by IO.popen, close sets $?."
 356       # http://www.ruby-doc.org/core-2.1.3/IO.html#method-i-close
 357       Rails.logger.warn("#{@opts[:uuid]}/#{@opts[:file]}: #{$?}") if $? != 0
 358     end
 359   end
 360 end