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
 141       opts = {}
 142       if usable_token == params[:reader_token]
 143         opts[:path_token] = usable_token
 144       elsif usable_token == Rails.configuration.anonymous_user_token
 145         # Don't pass a token at all
 146       else
 147         # We pass the current user's real token only if it's necessary
 148         # to read the collection.
 149         opts[:query_token] = usable_token
 150       end
 151       opts[:disposition] = params[:disposition] if params[:disposition]
 152       return redirect_to keep_web_url(params[:uuid], params[:file], opts)
 153     end
 154
 155     file_name = params[:file].andand.sub(/^(\.\/|\/|)/, './')
 156     if file_name.nil? or not coll.manifest.has_file?(file_name)
 157       return render_not_found
 158     end
 159
 160     opts = params.merge(arvados_api_token: usable_token)
 161
 162     # Handle Range requests. Currently we support only 'bytes=0-....'
 163     if request.headers.include? 'HTTP_RANGE'
 164       if m = /^bytes=0-(\d+)/.match(request.headers['HTTP_RANGE'])
 165         opts[:maxbytes] = m[1]
 166         size = params[:size] || '*'
 167         self.response.status = 206
 168         self.response.headers['Content-Range'] = "bytes 0-#{m[1]}/#{size}"
 169       end
 170     end
 171
 172     ext = File.extname(params[:file])
 173     self.response.headers['Content-Type'] =
 174       Rack::Mime::MIME_TYPES[ext] || 'application/octet-stream'
 175     if params[:size]
 176       size = params[:size].to_i
 177       if opts[:maxbytes]
 178         size = [size, opts[:maxbytes].to_i].min
 179       end
 180       self.response.headers['Content-Length'] = size.to_s
 181     end
 182     self.response.headers['Content-Disposition'] = params[:disposition] if params[:disposition]
 183     begin
 184       file_enumerator(opts).each do |bytes|
 185         response.stream.write bytes
 186       end
 187     ensure
 188       response.stream.close
 189     end
 190   end
 191
 192   def sharing_scopes
 193     ["GET /arvados/v1/collections/#{@object.uuid}", "GET /arvados/v1/collections/#{@object.uuid}/", "GET /arvados/v1/keep_services/accessible"]
 194   end
 195
 196   def search_scopes
 197     begin
 198       ApiClientAuthorization.filter([['scopes', '=', sharing_scopes]]).results
 199     rescue ArvadosApiClient::AccessForbiddenException
 200       nil
 201     end
 202   end
 203
 204   def find_object_by_uuid
 205     if not Keep::Locator.parse params[:id]
 206       super
 207     end
 208   end
 209
 210   def show
 211     return super if !@object
 212
 213     @logs = []
 214
 215     if params["tab_pane"] == "Provenance_graph"
 216       @prov_svg = ProvenanceHelper::create_provenance_graph(@object.provenance, "provenance_svg",
 217                                                             {:request => request,
 218                                                              :direction => :bottom_up,
 219                                                              :combine_jobs => :script_only}) rescue nil
 220     end
 221
 222     if current_user
 223       if Keep::Locator.parse params["uuid"]
 224         @same_pdh = Collection.filter([["portable_data_hash", "=", @object.portable_data_hash]]).limit(20)
 225         if @same_pdh.results.size == 1
 226           redirect_to collection_path(@same_pdh[0]["uuid"])
 227           return
 228         end
 229         owners = @same_pdh.map(&:owner_uuid).to_a.uniq
 230         preload_objects_for_dataclass Group, owners
 231         preload_objects_for_dataclass User, owners
 232         uuids = @same_pdh.map(&:uuid).to_a.uniq
 233         preload_links_for_objects uuids
 234         render 'hash_matches'
 235         return
 236       else
 237         jobs_with = lambda do |conds|
 238           Job.limit(RELATION_LIMIT).where(conds)
 239             .results.sort_by { |j| j.finished_at || j.created_at }
 240         end
 241         @output_of = jobs_with.call(output: @object.portable_data_hash)
 242         @log_of = jobs_with.call(log: @object.portable_data_hash)
 243         @project_links = Link.limit(RELATION_LIMIT).order("modified_at DESC")
 244           .where(head_uuid: @object.uuid, link_class: 'name').results
 245         project_hash = Group.where(uuid: @project_links.map(&:tail_uuid)).to_hash
 246         @projects = project_hash.values
 247
 248         @permissions = Link.limit(RELATION_LIMIT).order("modified_at DESC")
 249           .where(head_uuid: @object.uuid, link_class: 'permission',
 250                  name: 'can_read').results
 251         @logs = Log.limit(RELATION_LIMIT).order("created_at DESC")
 252           .select(%w(uuid event_type object_uuid event_at summary))
 253           .where(object_uuid: @object.uuid).results
 254         @is_persistent = Link.limit(1)
 255           .where(head_uuid: @object.uuid, tail_uuid: current_user.uuid,
 256                  link_class: 'resources', name: 'wants')
 257           .results.any?
 258         @search_sharing = search_scopes
 259
 260         if params["tab_pane"] == "Used_by"
 261           @used_by_svg = ProvenanceHelper::create_provenance_graph(@object.used_by, "used_by_svg",
 262                                                                    {:request => request,
 263                                                                      :direction => :top_down,
 264                                                                      :combine_jobs => :script_only,
 265                                                                      :pdata_only => true}) rescue nil
 266         end
 267       end
 268     end
 269     super
 270   end
 271
 272   def sharing_popup
 273     @search_sharing = search_scopes
 274     render("sharing_popup.js", content_type: "text/javascript")
 275   end
 276
 277   helper_method :download_link
 278
 279   def download_link
 280     collections_url + "/download/#{@object.uuid}/#{@search_sharing.first.api_token}/"
 281   end
 282
 283   def share
 284     ApiClientAuthorization.create(scopes: sharing_scopes)
 285     sharing_popup
 286   end
 287
 288   def unshare
 289     search_scopes.each do |s|
 290       s.destroy
 291     end
 292     sharing_popup
 293   end
 294
 295   protected
 296
 297   def find_usable_token(token_list)
 298     # Iterate over every given token to make it the current token and
 299     # yield the given block.
 300     # If the block succeeds, return the token it used.
 301     # Otherwise, render an error response based on the most specific
 302     # error we encounter, and return nil.
 303     most_specific_error = [401]
 304     token_list.each do |api_token|
 305       begin
 306         # We can't load the corresponding user, because the token may not
 307         # be scoped for that.
 308         using_specific_api_token(api_token, load_user: false) do
 309           yield
 310           return api_token
 311         end
 312       rescue ArvadosApiClient::ApiError => error
 313         if error.api_status >= most_specific_error.first
 314           most_specific_error = [error.api_status, error]
 315         end
 316       end
 317     end
 318     case most_specific_error.shift
 319     when 401, 403
 320       redirect_to_login
 321     when 404
 322       render_not_found(*most_specific_error)
 323     end
 324     return nil
 325   end
 326
 327   def keep_web_url(uuid_or_pdh, file, opts)
 328     munged_id = uuid_or_pdh.sub('+', '-')
 329     fmt = {uuid_or_pdh: munged_id}
 330     uri = URI.parse(Rails.configuration.keep_web_url % fmt)
 331     uri.path += '/' unless uri.path.end_with? '/'
 332     if opts[:path_token]
 333       uri.path += 't=' + opts[:path_token] + '/'
 334     end
 335     uri.path += '_/'
 336     uri.path += CGI::escape(file)
 337
 338     query = CGI::parse(uri.query || '')
 339     { query_token: 'api_token',
 340       disposition: 'disposition' }.each do |opt, param|
 341       if opts.include? opt
 342         query[param] = opts[opt]
 343       end
 344     end
 345     unless query.empty?
 346       uri.query = query.to_query
 347     end
 348
 349     if query.include? 'api_token' and
 350         query['disposition'] != 'attachment' and
 351         not uri.host.start_with?(munged_id + "--") and
 352         not uri.host.start_with?(munged_id + ".")
 353       # keep-web refuses query tokens ("?api_token=X") unless it sees
 354       # the collection ID in the hostname, or is running in
 355       # attachment-only mode.
 356       logger.warn("Single-origin keep_web_url can't serve inline content, " \
 357                   "but redirecting anyway: #{uri.to_s}")
 358     end
 359
 360     uri.to_s
 361   end
 362
 363   # Note: several controller and integration tests rely on stubbing
 364   # file_enumerator to return fake file content.
 365   def file_enumerator opts
 366     FileStreamer.new opts
 367   end
 368
 369   class FileStreamer
 370     include ArvadosApiClientHelper
 371     def initialize(opts={})
 372       @opts = opts
 373     end
 374     def each
 375       return unless @opts[:uuid] && @opts[:file]
 376
 377       env = Hash[ENV].dup
 378
 379       require 'uri'
 380       u = URI.parse(arvados_api_client.arvados_v1_base)
 381       env['ARVADOS_API_HOST'] = "#{u.host}:#{u.port}"
 382       env['ARVADOS_API_TOKEN'] = @opts[:arvados_api_token]
 383       env['ARVADOS_API_HOST_INSECURE'] = "true" if Rails.configuration.arvados_insecure_https
 384
 385       bytesleft = @opts[:maxbytes].andand.to_i || 2**16
 386       io = IO.popen([env, 'arv-get', "#{@opts[:uuid]}/#{@opts[:file]}"], 'rb')
 387       while bytesleft > 0 && (buf = io.read([bytesleft, 2**16].min)) != nil
 388         # shrink the bytesleft count, if we were given a maximum byte
 389         # count to read
 390         if @opts.include? :maxbytes
 391           bytesleft = bytesleft - buf.length
 392         end
 393         yield buf
 394       end
 395       io.close
 396       # "If ios is opened by IO.popen, close sets $?."
 397       # http://www.ruby-doc.org/core-2.1.3/IO.html#method-i-close
 398       Rails.logger.warn("#{@opts[:uuid]}/#{@opts[:file]}: #{$?}") if $? != 0
 399     end
 400   end
 401 end