apps/workbench/app/controllers/collections_controller.rb

   1 require "arvados/keep"
   2 require "arvados/collection"
   3 require "uri"
   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     if Rails.configuration.keep_web_url || Rails.configuration.keep_web_download_url
 115       # show_file will redirect to keep-web's directory listing
 116       return show_file
 117     end
 118     Thread.current[:reader_tokens] = [params[:reader_token]]
 119     return if false.equal?(find_object_by_uuid)
 120     render layout: false
 121   end
 122
 123   def show_file
 124     # We pipe from arv-get to send the file to the user.  Before we start it,
 125     # we ask the API server if the file actually exists.  This serves two
 126     # purposes: it lets us return a useful status code for common errors, and
 127     # helps us figure out which token to provide to arv-get.
 128     # The order of searched tokens is important: because the anonymous user
 129     # token is passed along with every API request, we have to check it first.
 130     # Otherwise, it's impossible to know whether any other request succeeded
 131     # because of the reader token.
 132     coll = nil
 133     tokens = [(Rails.configuration.anonymous_user_token || nil),
 134               params[:reader_token],
 135               Thread.current[:arvados_api_token]].compact
 136     usable_token = find_usable_token(tokens) do
 137       coll = Collection.find(params[:uuid])
 138     end
 139     if usable_token.nil?
 140       # Response already rendered.
 141       return
 142     end
 143
 144     # If we are configured to use a keep-web server, just redirect to
 145     # the appropriate URL.
 146     if Rails.configuration.keep_web_url or
 147         Rails.configuration.keep_web_download_url
 148       opts = {}
 149       if usable_token == params[:reader_token]
 150         opts[:path_token] = usable_token
 151       elsif usable_token == Rails.configuration.anonymous_user_token
 152         # Don't pass a token at all
 153       else
 154         # We pass the current user's real token only if it's necessary
 155         # to read the collection.
 156         opts[:query_token] = usable_token
 157       end
 158       opts[:disposition] = params[:disposition] if params[:disposition]
 159       return redirect_to keep_web_url(params[:uuid], params[:file], opts)
 160     end
 161
 162     # No keep-web server available. Get the file data with arv-get,
 163     # and serve it through Rails.
 164
 165     file_name = params[:file].andand.sub(/^(\.\/|\/|)/, './')
 166     if file_name.nil? or not coll.manifest.has_file?(file_name)
 167       return render_not_found
 168     end
 169
 170     opts = params.merge(arvados_api_token: usable_token)
 171
 172     # Handle Range requests. Currently we support only 'bytes=0-....'
 173     if request.headers.include? 'HTTP_RANGE'
 174       if m = /^bytes=0-(\d+)/.match(request.headers['HTTP_RANGE'])
 175         opts[:maxbytes] = m[1]
 176         size = params[:size] || '*'
 177         self.response.status = 206
 178         self.response.headers['Content-Range'] = "bytes 0-#{m[1]}/#{size}"
 179       end
 180     end
 181
 182     ext = File.extname(params[:file])
 183     self.response.headers['Content-Type'] =
 184       Rack::Mime::MIME_TYPES[ext] || 'application/octet-stream'
 185     if params[:size]
 186       size = params[:size].to_i
 187       if opts[:maxbytes]
 188         size = [size, opts[:maxbytes].to_i].min
 189       end
 190       self.response.headers['Content-Length'] = size.to_s
 191     end
 192     self.response.headers['Content-Disposition'] = params[:disposition] if params[:disposition]
 193     begin
 194       file_enumerator(opts).each do |bytes|
 195         response.stream.write bytes
 196       end
 197     ensure
 198       response.stream.close
 199     end
 200   end
 201
 202   def sharing_scopes
 203     ["GET /arvados/v1/collections/#{@object.uuid}", "GET /arvados/v1/collections/#{@object.uuid}/", "GET /arvados/v1/keep_services/accessible"]
 204   end
 205
 206   def search_scopes
 207     begin
 208       ApiClientAuthorization.filter([['scopes', '=', sharing_scopes]]).results
 209     rescue ArvadosApiClient::AccessForbiddenException
 210       nil
 211     end
 212   end
 213
 214   def find_object_by_uuid
 215     if not Keep::Locator.parse params[:id]
 216       super
 217     end
 218   end
 219
 220   def show
 221     return super if !@object
 222
 223     @logs = []
 224
 225     if params["tab_pane"] == "Provenance_graph"
 226       @prov_svg = ProvenanceHelper::create_provenance_graph(@object.provenance, "provenance_svg",
 227                                                             {:request => request,
 228                                                              :direction => :top_down,
 229                                                              :combine_jobs => :script_only}) rescue nil
 230     end
 231
 232     if current_user
 233       if Keep::Locator.parse params["uuid"]
 234         @same_pdh = Collection.filter([["portable_data_hash", "=", @object.portable_data_hash]]).limit(20)
 235         if @same_pdh.results.size == 1
 236           redirect_to collection_path(@same_pdh[0]["uuid"])
 237           return
 238         end
 239         owners = @same_pdh.map(&:owner_uuid).to_a.uniq
 240         preload_objects_for_dataclass Group, owners
 241         preload_objects_for_dataclass User, owners
 242         uuids = @same_pdh.map(&:uuid).to_a.uniq
 243         preload_links_for_objects uuids
 244         render 'hash_matches'
 245         return
 246       else
 247         if Job.api_exists?(:index)
 248           jobs_with = lambda do |conds|
 249             Job.limit(RELATION_LIMIT).where(conds)
 250               .results.sort_by { |j| j.finished_at || j.created_at }
 251           end
 252           @output_of = jobs_with.call(output: @object.portable_data_hash)
 253           @log_of = jobs_with.call(log: @object.portable_data_hash)
 254         end
 255
 256         @project_links = Link.limit(RELATION_LIMIT).order("modified_at DESC")
 257           .where(head_uuid: @object.uuid, link_class: 'name').results
 258         project_hash = Group.where(uuid: @project_links.map(&:tail_uuid)).to_hash
 259         @projects = project_hash.values
 260
 261         @permissions = Link.limit(RELATION_LIMIT).order("modified_at DESC")
 262           .where(head_uuid: @object.uuid, link_class: 'permission',
 263                  name: 'can_read').results
 264         @search_sharing = search_scopes
 265
 266         if params["tab_pane"] == "Used_by"
 267           @used_by_svg = ProvenanceHelper::create_provenance_graph(@object.used_by, "used_by_svg",
 268                                                                    {:request => request,
 269                                                                     :direction => :top_down,
 270                                                                     :combine_jobs => :script_only,
 271                                                                     :pdata_only => true}) rescue nil
 272         end
 273       end
 274     end
 275     super
 276   end
 277
 278   def sharing_popup
 279     @search_sharing = search_scopes
 280     render("sharing_popup.js", content_type: "text/javascript")
 281   end
 282
 283   helper_method :download_link
 284
 285   def download_link
 286     collections_url + "/download/#{@object.uuid}/#{@search_sharing.first.api_token}/"
 287   end
 288
 289   def share
 290     ApiClientAuthorization.create(scopes: sharing_scopes)
 291     sharing_popup
 292   end
 293
 294   def unshare
 295     search_scopes.each do |s|
 296       s.destroy
 297     end
 298     sharing_popup
 299   end
 300
 301   def remove_selected_files
 302     uuids, source_paths = selected_collection_files params
 303
 304     arv_coll = Arv::Collection.new(@object.manifest_text)
 305     source_paths[uuids[0]].each do |p|
 306       arv_coll.rm "."+p
 307     end
 308
 309     if @object.update_attributes manifest_text: arv_coll.manifest_text
 310       show
 311     else
 312       self.render_error status: 422
 313     end
 314   end
 315
 316   def update
 317     updated_attr = params[:collection].each.select {|a| a[0].andand.start_with? 'rename-file-path:'}
 318
 319     if updated_attr.size > 0
 320       # Is it file rename?
 321       file_path = updated_attr[0][0].split('rename-file-path:')[-1]
 322
 323       new_file_path = updated_attr[0][1]
 324       if new_file_path.start_with?('./')
 325         # looks good
 326       elsif new_file_path.start_with?('/')
 327         new_file_path = '.' + new_file_path
 328       else
 329         new_file_path = './' + new_file_path
 330       end
 331
 332       arv_coll = Arv::Collection.new(@object.manifest_text)
 333
 334       if arv_coll.exist?(new_file_path)
 335         @errors = 'Duplicate file path. Please use a different name.'
 336         self.render_error status: 422
 337       else
 338         arv_coll.rename "./"+file_path, new_file_path
 339
 340         if @object.update_attributes manifest_text: arv_coll.manifest_text
 341           show
 342         else
 343           self.render_error status: 422
 344         end
 345       end
 346     else
 347       # Not a file rename; use default
 348       super
 349     end
 350   end
 351
 352   protected
 353
 354   def find_usable_token(token_list)
 355     # Iterate over every given token to make it the current token and
 356     # yield the given block.
 357     # If the block succeeds, return the token it used.
 358     # Otherwise, render an error response based on the most specific
 359     # error we encounter, and return nil.
 360     most_specific_error = [401]
 361     token_list.each do |api_token|
 362       begin
 363         # We can't load the corresponding user, because the token may not
 364         # be scoped for that.
 365         using_specific_api_token(api_token, load_user: false) do
 366           yield
 367           return api_token
 368         end
 369       rescue ArvadosApiClient::ApiError => error
 370         if error.api_status >= most_specific_error.first
 371           most_specific_error = [error.api_status, error]
 372         end
 373       end
 374     end
 375     case most_specific_error.shift
 376     when 401, 403
 377       redirect_to_login
 378     when 404
 379       render_not_found(*most_specific_error)
 380     end
 381     return nil
 382   end
 383
 384   def keep_web_url(uuid_or_pdh, file, opts)
 385     munged_id = uuid_or_pdh.sub('+', '-')
 386     fmt = {uuid_or_pdh: munged_id}
 387
 388     tmpl = Rails.configuration.keep_web_url
 389     if Rails.configuration.keep_web_download_url and
 390         (!tmpl or opts[:disposition] == 'attachment')
 391       # Prefer the attachment-only-host when we want an attachment
 392       # (and when there is no preview link configured)
 393       tmpl = Rails.configuration.keep_web_download_url
 394     elsif not Rails.configuration.trust_all_content
 395       check_uri = URI.parse(tmpl % fmt)
 396       if opts[:query_token] and
 397           not check_uri.host.start_with?(munged_id + "--") and
 398           not check_uri.host.start_with?(munged_id + ".")
 399         # We're about to pass a token in the query string, but
 400         # keep-web can't accept that safely at a single-origin URL
 401         # template (unless it's -attachment-only-host).
 402         tmpl = Rails.configuration.keep_web_download_url
 403         if not tmpl
 404           raise ArgumentError, "Download precluded by site configuration"
 405         end
 406         logger.warn("Using download link, even though inline content " \
 407                     "was requested: #{check_uri.to_s}")
 408       end
 409     end
 410
 411     if tmpl == Rails.configuration.keep_web_download_url
 412       # This takes us to keep-web's -attachment-only-host so there is
 413       # no need to add ?disposition=attachment.
 414       opts.delete :disposition
 415     end
 416
 417     uri = URI.parse(tmpl % fmt)
 418     uri.path += '/' unless uri.path.end_with? '/'
 419     if opts[:path_token]
 420       uri.path += 't=' + opts[:path_token] + '/'
 421     end
 422     uri.path += '_/'
 423     uri.path += URI.escape(file) if file
 424
 425     query = Hash[URI.decode_www_form(uri.query || '')]
 426     { query_token: 'api_token',
 427       disposition: 'disposition' }.each do |opt, param|
 428       if opts.include? opt
 429         query[param] = opts[opt]
 430       end
 431     end
 432     unless query.empty?
 433       uri.query = URI.encode_www_form(query)
 434     end
 435
 436     uri.to_s
 437   end
 438
 439   # Note: several controller and integration tests rely on stubbing
 440   # file_enumerator to return fake file content.
 441   def file_enumerator opts
 442     FileStreamer.new opts
 443   end
 444
 445   class FileStreamer
 446     include ArvadosApiClientHelper
 447     def initialize(opts={})
 448       @opts = opts
 449     end
 450     def each
 451       return unless @opts[:uuid] && @opts[:file]
 452
 453       env = Hash[ENV].dup
 454
 455       require 'uri'
 456       u = URI.parse(arvados_api_client.arvados_v1_base)
 457       env['ARVADOS_API_HOST'] = "#{u.host}:#{u.port}"
 458       env['ARVADOS_API_TOKEN'] = @opts[:arvados_api_token]
 459       env['ARVADOS_API_HOST_INSECURE'] = "true" if Rails.configuration.arvados_insecure_https
 460
 461       bytesleft = @opts[:maxbytes].andand.to_i || 2**16
 462       io = IO.popen([env, 'arv-get', "#{@opts[:uuid]}/#{@opts[:file]}"], 'rb')
 463       while bytesleft > 0 && (buf = io.read([bytesleft, 2**16].min)) != nil
 464         # shrink the bytesleft count, if we were given a maximum byte
 465         # count to read
 466         if @opts.include? :maxbytes
 467           bytesleft = bytesleft - buf.length
 468         end
 469         yield buf
 470       end
 471       io.close
 472       # "If ios is opened by IO.popen, close sets $?."
 473       # http://www.ruby-doc.org/core-2.1.3/IO.html#method-i-close
 474       Rails.logger.warn("#{@opts[:uuid]}/#{@opts[:file]}: #{$?}") if $? != 0
 475     end
 476   end
 477 end