apps/workbench/app/controllers/collections_controller.rb

   1 require "arvados/keep"
   2 require "uri"
   3 require "cgi"
   4
   5 class CollectionsController < ApplicationController
   6   include ActionController::Live
   7
   8   skip_around_filter :require_thread_api_token, if: proc { |ctrl|
   9     Rails.configuration.anonymous_user_token and
  10     'show' == ctrl.action_name
  11   }
  12   skip_around_filter(:require_thread_api_token,
  13                      only: [:show_file, :show_file_links])
  14   skip_before_filter(:find_object_by_uuid,
  15                      only: [:provenance, :show_file, :show_file_links])
  16   # We depend on show_file to display the user agreement:
  17   skip_before_filter :check_user_agreements, only: :show_file
  18   skip_before_filter :check_user_profile, only: :show_file
  19
  20   RELATION_LIMIT = 5
  21
  22   def show_pane_list
  23     panes = %w(Files Upload Provenance_graph Used_by Advanced)
  24     panes = panes - %w(Upload) unless (@object.editable? rescue false)
  25     panes
  26   end
  27
  28   def set_persistent
  29     case params[:value]
  30     when 'persistent', 'cache'
  31       persist_links = Link.filter([['owner_uuid', '=', current_user.uuid],
  32                                    ['link_class', '=', 'resources'],
  33                                    ['name', '=', 'wants'],
  34                                    ['tail_uuid', '=', current_user.uuid],
  35                                    ['head_uuid', '=', @object.uuid]])
  36       logger.debug persist_links.inspect
  37     else
  38       return unprocessable "Invalid value #{value.inspect}"
  39     end
  40     if params[:value] == 'persistent'
  41       if not persist_links.any?
  42         Link.create(link_class: 'resources',
  43                     name: 'wants',
  44                     tail_uuid: current_user.uuid,
  45                     head_uuid: @object.uuid)
  46       end
  47     else
  48       persist_links.each do |link|
  49         link.destroy || raise
  50       end
  51     end
  52
  53     respond_to do |f|
  54       f.json { render json: @object }
  55     end
  56   end
  57
  58   def index
  59     # API server index doesn't return manifest_text by default, but our
  60     # callers want it unless otherwise specified.
  61     @select ||= Collection.columns.map(&:name)
  62     base_search = Collection.select(@select)
  63     if params[:search].andand.length.andand > 0
  64       tags = Link.where(any: ['contains', params[:search]])
  65       @objects = (base_search.where(uuid: tags.collect(&:head_uuid)) |
  66                       base_search.where(any: ['contains', params[:search]])).
  67         uniq { |c| c.uuid }
  68     else
  69       if params[:limit]
  70         limit = params[:limit].to_i
  71       else
  72         limit = 100
  73       end
  74
  75       if params[:offset]
  76         offset = params[:offset].to_i
  77       else
  78         offset = 0
  79       end
  80
  81       @objects = base_search.limit(limit).offset(offset)
  82     end
  83     @links = Link.where(head_uuid: @objects.collect(&:uuid))
  84     @collection_info = {}
  85     @objects.each do |c|
  86       @collection_info[c.uuid] = {
  87         tag_links: [],
  88         wanted: false,
  89         wanted_by_me: false,
  90         provenance: [],
  91         links: []
  92       }
  93     end
  94     @links.each do |link|
  95       @collection_info[link.head_uuid] ||= {}
  96       info = @collection_info[link.head_uuid]
  97       case link.link_class
  98       when 'tag'
  99         info[:tag_links] << link
 100       when 'resources'
 101         info[:wanted] = true
 102         info[:wanted_by_me] ||= link.tail_uuid == current_user.uuid
 103       when 'provenance'
 104         info[:provenance] << link.name
 105       end
 106       info[:links] << link
 107     end
 108     @request_url = request.url
 109
 110     render_index
 111   end
 112
 113   def show_file_links
 114     Thread.current[:reader_tokens] = [params[:reader_token]]
 115     return if false.equal?(find_object_by_uuid)
 116     render layout: false
 117   end
 118
 119   def show_file
 120     # We pipe from arv-get to send the file to the user.  Before we start it,
 121     # we ask the API server if the file actually exists.  This serves two
 122     # purposes: it lets us return a useful status code for common errors, and
 123     # helps us figure out which token to provide to arv-get.
 124     # The order of searched tokens is important: because the anonymous user
 125     # token is passed along with every API request, we have to check it first.
 126     # Otherwise, it's impossible to know whether any other request succeeded
 127     # because of the reader token.
 128     coll = nil
 129     tokens = [(Rails.configuration.anonymous_user_token || nil),
 130               params[:reader_token],
 131               Thread.current[:arvados_api_token]].compact
 132     usable_token = find_usable_token(tokens) do
 133       coll = Collection.find(params[:uuid])
 134     end
 135     if usable_token.nil?
 136       # Response already rendered.
 137       return
 138     end
 139
 140     if Rails.configuration.keep_web_url or
 141         Rails.configuration.keep_web_download_url
 142       opts = {}
 143       if usable_token == params[:reader_token]
 144         opts[:path_token] = usable_token
 145       elsif usable_token == Rails.configuration.anonymous_user_token
 146         # Don't pass a token at all
 147       else
 148         # We pass the current user's real token only if it's necessary
 149         # to read the collection.
 150         opts[:query_token] = usable_token
 151       end
 152       opts[:disposition] = params[:disposition] if params[:disposition]
 153       return redirect_to keep_web_url(params[:uuid], params[:file], opts)
 154     end
 155
 156     file_name = params[:file].andand.sub(/^(\.\/|\/|)/, './')
 157     if file_name.nil? or not coll.manifest.has_file?(file_name)
 158       return render_not_found
 159     end
 160
 161     opts = params.merge(arvados_api_token: usable_token)
 162
 163     # Handle Range requests. Currently we support only 'bytes=0-....'
 164     if request.headers.include? 'HTTP_RANGE'
 165       if m = /^bytes=0-(\d+)/.match(request.headers['HTTP_RANGE'])
 166         opts[:maxbytes] = m[1]
 167         size = params[:size] || '*'
 168         self.response.status = 206
 169         self.response.headers['Content-Range'] = "bytes 0-#{m[1]}/#{size}"
 170       end
 171     end
 172
 173     ext = File.extname(params[:file])
 174     self.response.headers['Content-Type'] =
 175       Rack::Mime::MIME_TYPES[ext] || 'application/octet-stream'
 176     if params[:size]
 177       size = params[:size].to_i
 178       if opts[:maxbytes]
 179         size = [size, opts[:maxbytes].to_i].min
 180       end
 181       self.response.headers['Content-Length'] = size.to_s
 182     end
 183     self.response.headers['Content-Disposition'] = params[:disposition] if params[:disposition]
 184     begin
 185       file_enumerator(opts).each do |bytes|
 186         response.stream.write bytes
 187       end
 188     ensure
 189       response.stream.close
 190     end
 191   end
 192
 193   def sharing_scopes
 194     ["GET /arvados/v1/collections/#{@object.uuid}", "GET /arvados/v1/collections/#{@object.uuid}/", "GET /arvados/v1/keep_services/accessible"]
 195   end
 196
 197   def search_scopes
 198     begin
 199       ApiClientAuthorization.filter([['scopes', '=', sharing_scopes]]).results
 200     rescue ArvadosApiClient::AccessForbiddenException
 201       nil
 202     end
 203   end
 204
 205   def find_object_by_uuid
 206     if not Keep::Locator.parse params[:id]
 207       super
 208     end
 209   end
 210
 211   def show
 212     return super if !@object
 213
 214     @logs = []
 215
 216     if params["tab_pane"] == "Provenance_graph"
 217       @prov_svg = ProvenanceHelper::create_provenance_graph(@object.provenance, "provenance_svg",
 218                                                             {:request => request,
 219                                                              :direction => :bottom_up,
 220                                                              :combine_jobs => :script_only}) rescue nil
 221     end
 222
 223     if current_user
 224       if Keep::Locator.parse params["uuid"]
 225         @same_pdh = Collection.filter([["portable_data_hash", "=", @object.portable_data_hash]]).limit(20)
 226         if @same_pdh.results.size == 1
 227           redirect_to collection_path(@same_pdh[0]["uuid"])
 228           return
 229         end
 230         owners = @same_pdh.map(&:owner_uuid).to_a.uniq
 231         preload_objects_for_dataclass Group, owners
 232         preload_objects_for_dataclass User, owners
 233         uuids = @same_pdh.map(&:uuid).to_a.uniq
 234         preload_links_for_objects uuids
 235         render 'hash_matches'
 236         return
 237       else
 238         jobs_with = lambda do |conds|
 239           Job.limit(RELATION_LIMIT).where(conds)
 240             .results.sort_by { |j| j.finished_at || j.created_at }
 241         end
 242         @output_of = jobs_with.call(output: @object.portable_data_hash)
 243         @log_of = jobs_with.call(log: @object.portable_data_hash)
 244         @project_links = Link.limit(RELATION_LIMIT).order("modified_at DESC")
 245           .where(head_uuid: @object.uuid, link_class: 'name').results
 246         project_hash = Group.where(uuid: @project_links.map(&:tail_uuid)).to_hash
 247         @projects = project_hash.values
 248
 249         @permissions = Link.limit(RELATION_LIMIT).order("modified_at DESC")
 250           .where(head_uuid: @object.uuid, link_class: 'permission',
 251                  name: 'can_read').results
 252         @logs = Log.limit(RELATION_LIMIT).order("created_at DESC")
 253           .select(%w(uuid event_type object_uuid event_at summary))
 254           .where(object_uuid: @object.uuid).results
 255         @is_persistent = Link.limit(1)
 256           .where(head_uuid: @object.uuid, tail_uuid: current_user.uuid,
 257                  link_class: 'resources', name: 'wants')
 258           .results.any?
 259         @search_sharing = search_scopes
 260
 261         if params["tab_pane"] == "Used_by"
 262           @used_by_svg = ProvenanceHelper::create_provenance_graph(@object.used_by, "used_by_svg",
 263                                                                    {:request => request,
 264                                                                      :direction => :top_down,
 265                                                                      :combine_jobs => :script_only,
 266                                                                      :pdata_only => true}) rescue nil
 267         end
 268       end
 269     end
 270     super
 271   end
 272
 273   def sharing_popup
 274     @search_sharing = search_scopes
 275     render("sharing_popup.js", content_type: "text/javascript")
 276   end
 277
 278   helper_method :download_link
 279
 280   def download_link
 281     collections_url + "/download/#{@object.uuid}/#{@search_sharing.first.api_token}/"
 282   end
 283
 284   def share
 285     ApiClientAuthorization.create(scopes: sharing_scopes)
 286     sharing_popup
 287   end
 288
 289   def unshare
 290     search_scopes.each do |s|
 291       s.destroy
 292     end
 293     sharing_popup
 294   end
 295
 296   protected
 297
 298   def find_usable_token(token_list)
 299     # Iterate over every given token to make it the current token and
 300     # yield the given block.
 301     # If the block succeeds, return the token it used.
 302     # Otherwise, render an error response based on the most specific
 303     # error we encounter, and return nil.
 304     most_specific_error = [401]
 305     token_list.each do |api_token|
 306       begin
 307         # We can't load the corresponding user, because the token may not
 308         # be scoped for that.
 309         using_specific_api_token(api_token, load_user: false) do
 310           yield
 311           return api_token
 312         end
 313       rescue ArvadosApiClient::ApiError => error
 314         if error.api_status >= most_specific_error.first
 315           most_specific_error = [error.api_status, error]
 316         end
 317       end
 318     end
 319     case most_specific_error.shift
 320     when 401, 403
 321       redirect_to_login
 322     when 404
 323       render_not_found(*most_specific_error)
 324     end
 325     return nil
 326   end
 327
 328   def keep_web_url(uuid_or_pdh, file, opts)
 329     munged_id = uuid_or_pdh.sub('+', '-')
 330     fmt = {uuid_or_pdh: munged_id}
 331
 332     tmpl = Rails.configuration.keep_web_url
 333     if Rails.configuration.keep_web_download_url and
 334         (!tmpl or opts[:disposition] == 'attachment')
 335       # Prefer the attachment-only-host when we want an attachment
 336       # (and when there is no preview link configured)
 337       tmpl = Rails.configuration.keep_web_download_url
 338     else
 339       test_uri = URI.parse(tmpl % fmt)
 340       if opts[:query_token] and
 341           not test_uri.host.start_with?(munged_id + "--") and
 342           not test_uri.host.start_with?(munged_id + ".")
 343         # We're about to pass a token in the query string, but
 344         # keep-web can't accept that safely at a single-origin URL
 345         # template (unless it's -attachment-only-host).
 346         tmpl = Rails.configuration.keep_web_download_url
 347         if not tmpl
 348           raise ArgumentError, "Download precluded by site configuration"
 349         end
 350         logger.warn("Using download link, even though inline content " \
 351                     "was requested: #{test_uri.to_s}")
 352       end
 353     end
 354
 355     if tmpl == Rails.configuration.keep_web_download_url
 356       # This takes us to keep-web's -attachment-only-host so there is
 357       # no need to add ?disposition=attachment.
 358       opts.delete :disposition
 359     end
 360
 361     uri = URI.parse(tmpl % fmt)
 362     uri.path += '/' unless uri.path.end_with? '/'
 363     if opts[:path_token]
 364       uri.path += 't=' + opts[:path_token] + '/'
 365     end
 366     uri.path += '_/'
 367     uri.path += CGI::escape(file)
 368
 369     query = CGI::parse(uri.query || '')
 370     { query_token: 'api_token',
 371       disposition: 'disposition' }.each do |opt, param|
 372       if opts.include? opt
 373         query[param] = opts[opt]
 374       end
 375     end
 376     unless query.empty?
 377       uri.query = query.to_query
 378     end
 379
 380     uri.to_s
 381   end
 382
 383   # Note: several controller and integration tests rely on stubbing
 384   # file_enumerator to return fake file content.
 385   def file_enumerator opts
 386     FileStreamer.new opts
 387   end
 388
 389   class FileStreamer
 390     include ArvadosApiClientHelper
 391     def initialize(opts={})
 392       @opts = opts
 393     end
 394     def each
 395       return unless @opts[:uuid] && @opts[:file]
 396
 397       env = Hash[ENV].dup
 398
 399       require 'uri'
 400       u = URI.parse(arvados_api_client.arvados_v1_base)
 401       env['ARVADOS_API_HOST'] = "#{u.host}:#{u.port}"
 402       env['ARVADOS_API_TOKEN'] = @opts[:arvados_api_token]
 403       env['ARVADOS_API_HOST_INSECURE'] = "true" if Rails.configuration.arvados_insecure_https
 404
 405       bytesleft = @opts[:maxbytes].andand.to_i || 2**16
 406       io = IO.popen([env, 'arv-get', "#{@opts[:uuid]}/#{@opts[:file]}"], 'rb')
 407       while bytesleft > 0 && (buf = io.read([bytesleft, 2**16].min)) != nil
 408         # shrink the bytesleft count, if we were given a maximum byte
 409         # count to read
 410         if @opts.include? :maxbytes
 411           bytesleft = bytesleft - buf.length
 412         end
 413         yield buf
 414       end
 415       io.close
 416       # "If ios is opened by IO.popen, close sets $?."
 417       # http://www.ruby-doc.org/core-2.1.3/IO.html#method-i-close
 418       Rails.logger.warn("#{@opts[:uuid]}/#{@opts[:file]}: #{$?}") if $? != 0
 419     end
 420   end
 421 end