Merge branch '4295-efficient-websockets' closes #4295
[arvados.git] / apps / workbench / app / controllers / collections_controller.rb
index 2508d3d49a7c28404800a31ddb8c2b9a5b76b648..4e0008d93cf63887926c7d3ef7907d4f1f377bd3 100644 (file)
 class CollectionsController < ApplicationController
-  skip_before_filter :find_object_by_uuid, :only => [:provenance]
-  skip_before_filter :check_user_agreements, :only => [:show_file]
+  include ActionController::Live
+
+  skip_around_filter(:require_thread_api_token,
+                     only: [:show_file, :show_file_links])
+  skip_before_filter(:find_object_by_uuid,
+                     only: [:provenance, :show_file, :show_file_links])
+  # We depend on show_file to display the user agreement:
+  skip_before_filter :check_user_agreements, only: :show_file
+  skip_before_filter :check_user_profile, only: :show_file
+
+  RELATION_LIMIT = 5
+
+  def show_pane_list
+    %w(Files Provenance_graph Used_by Advanced)
+  end
+
+  def set_persistent
+    case params[:value]
+    when 'persistent', 'cache'
+      persist_links = Link.filter([['owner_uuid', '=', current_user.uuid],
+                                   ['link_class', '=', 'resources'],
+                                   ['name', '=', 'wants'],
+                                   ['tail_uuid', '=', current_user.uuid],
+                                   ['head_uuid', '=', @object.uuid]])
+      logger.debug persist_links.inspect
+    else
+      return unprocessable "Invalid value #{value.inspect}"
+    end
+    if params[:value] == 'persistent'
+      if not persist_links.any?
+        Link.create(link_class: 'resources',
+                    name: 'wants',
+                    tail_uuid: current_user.uuid,
+                    head_uuid: @object.uuid)
+      end
+    else
+      persist_links.each do |link|
+        link.destroy || raise
+      end
+    end
+
+    respond_to do |f|
+      f.json { render json: @object }
+    end
+  end
+
+  def choose
+    # Find collections using default find_objects logic, then search for name
+    # links, and preload any other links connected to the collections that are
+    # found.
+    # Name links will be obsolete when issue #3036 is merged,
+    # at which point this entire custom #choose function can probably be
+    # eliminated.
+
+    params[:limit] ||= 40
+
+    find_objects_for_index
+    @collections = @objects
+
+    @filters += [['link_class','=','name'],
+                 ['head_uuid','is_a','arvados#collection']]
+
+    @objects = Link
+    find_objects_for_index
+
+    @name_links = @objects
+
+    @objects = Collection.
+      filter([['uuid','in',@name_links.collect(&:head_uuid)]])
+
+    preload_links_for_objects (@collections.to_a + @objects.to_a)
+    super
+  end
 
   def index
+    # API server index doesn't return manifest_text by default, but our
+    # callers want it unless otherwise specified.
+    @select ||= Collection.columns.map(&:name)
+    base_search = Collection.select(@select)
     if params[:search].andand.length.andand > 0
       tags = Link.where(any: ['contains', params[:search]])
-      @collections = (Collection.where(uuid: tags.collect(&:head_uuid)) |
-                      Collection.where(any: ['contains', params[:search]])).
+      @collections = (base_search.where(uuid: tags.collect(&:head_uuid)) |
+                      base_search.where(any: ['contains', params[:search]])).
         uniq { |c| c.uuid }
     else
-      @collections = Collection.limit(100)
+      if params[:limit]
+        limit = params[:limit].to_i
+      else
+        limit = 100
+      end
+
+      if params[:offset]
+        offset = params[:offset].to_i
+      else
+        offset = 0
+      end
+
+      @collections = base_search.limit(limit).offset(offset)
     end
     @links = Link.limit(1000).
       where(head_uuid: @collections.collect(&:uuid))
     @collection_info = {}
     @collections.each do |c|
       @collection_info[c.uuid] = {
-        tags: [],
+        tag_links: [],
         wanted: false,
         wanted_by_me: false,
         provenance: [],
@@ -28,7 +115,7 @@ class CollectionsController < ApplicationController
       info = @collection_info[link.head_uuid]
       case link.link_class
       when 'tag'
-        info[:tags] << link.name
+        info[:tag_links] << link
       when 'resources'
         info[:wanted] = true
         info[:wanted_by_me] ||= link.tail_uuid == current_user.uuid
@@ -38,228 +125,229 @@ class CollectionsController < ApplicationController
       info[:links] << link
     end
     @request_url = request.url
+
+    render_index
+  end
+
+  def show_file_links
+    Thread.current[:reader_tokens] = [params[:reader_token]]
+    return if false.equal?(find_object_by_uuid)
+    render layout: false
   end
 
   def show_file
-    opts = params.merge(arvados_api_token: Thread.current[:arvados_api_token])
-    if r = params[:file].match(/(\.\w+)/)
-      ext = r[1]
+    # We pipe from arv-get to send the file to the user.  Before we start it,
+    # we ask the API server if the file actually exists.  This serves two
+    # purposes: it lets us return a useful status code for common errors, and
+    # helps us figure out which token to provide to arv-get.
+    coll = nil
+    tokens = [Thread.current[:arvados_api_token], params[:reader_token]].compact
+    usable_token = find_usable_token(tokens) do
+      coll = Collection.find(params[:uuid])
+    end
+    if usable_token.nil?
+      return  # Response already rendered.
+    elsif params[:file].nil? or not coll.manifest.has_file?(params[:file])
+      return render_not_found
     end
+
+    opts = params.merge(arvados_api_token: usable_token)
+
+    # Handle Range requests. Currently we support only 'bytes=0-....'
+    if request.headers.include? 'HTTP_RANGE'
+      if m = /^bytes=0-(\d+)/.match(request.headers['HTTP_RANGE'])
+        opts[:maxbytes] = m[1]
+        size = params[:size] || '*'
+        self.response.status = 206
+        self.response.headers['Content-Range'] = "bytes 0-#{m[1]}/#{size}"
+      end
+    end
+
+    ext = File.extname(params[:file])
     self.response.headers['Content-Type'] =
       Rack::Mime::MIME_TYPES[ext] || 'application/octet-stream'
-    self.response.headers['Content-Length'] = params[:size] if params[:size]
+    if params[:size]
+      size = params[:size].to_i
+      if opts[:maxbytes]
+        size = [size, opts[:maxbytes].to_i].min
+      end
+      self.response.headers['Content-Length'] = size.to_s
+    end
     self.response.headers['Content-Disposition'] = params[:disposition] if params[:disposition]
-    self.response_body = FileStreamer.new opts
-  end
-
-  def describe_node(uuid)
-    rsc = ArvadosBase::resource_class_for_uuid uuid
-    if rsc
-      "\"#{uuid}\" [label=\"#{rsc}\\n#{uuid}\",href=\"#{url_for rsc}/#{uuid}\"];"
-    else
-      ""
+    begin
+      file_enumerator(opts).each do |bytes|
+        response.stream.write bytes
+      end
+    ensure
+      response.stream.close
     end
   end
 
-  def describe_script(job)
-    #"""\"#{job.script_version}\" [label=\"#{job.script}: #{job.script_version}\"];
-    #   \"#{job.uuid}\" -> \"#{job.script_version}\" [label=\"script\"];"""
-    "\"#{job.uuid}\" [label=\"#{job.script}\\n#{job.script_version}\"];"
+  def sharing_scopes
+    ["GET /arvados/v1/collections/#{@object.uuid}", "GET /arvados/v1/collections/#{@object.uuid}/", "GET /arvados/v1/keep_services/accessible"]
   end
 
-  def job_uuid(job)
-    "#{job.script}\\n#{job.script_version}"
+  def search_scopes
+    begin
+      ApiClientAuthorization.filter([['scopes', '=', sharing_scopes]]).results
+    rescue ArvadosApiClient::AccessForbiddenException
+      nil
+    end
   end
 
-  def collection_uuid(uuid)
-    m = /([a-f0-9]{32}(\+[0-9]+)?)(\+.*)?/.match(uuid)
-    if m
-      m[1]
-    else
-      nil
+  def find_object_by_uuid
+    if not Keep::Locator.parse params[:id]
+      super
     end
   end
 
-  def script_param_edges(visited, job, prefix, sp)
-    gr = ""
-    if sp and not sp.empty?
-      case sp
-      when Hash
-        sp.each do |k, v|
-          if prefix.size > 0
-            k = prefix + "::" + k.to_s
-          end
-          gr += script_param_edges(visited, job, k.to_s, v)
-        end
-      when Array
-        sp.each do |v|
-          gr += script_param_edges(visited, job, prefix, v)
+  def show
+    return super if !@object
+    if current_user
+      if Keep::Locator.parse params["uuid"]
+        @same_pdh = Collection.filter([["portable_data_hash", "=", @object.portable_data_hash]]).limit(1000)
+        if @same_pdh.results.size == 1
+          redirect_to collection_path(@same_pdh[0]["uuid"])
+          return
         end
+        owners = @same_pdh.map(&:owner_uuid).to_a.uniq
+        preload_objects_for_dataclass Group, owners
+        preload_objects_for_dataclass User, owners
+        render 'hash_matches'
+        return
       else
-        m = collection_uuid(sp)
-        if m
-          gr += "\"#{job_uuid(job)}\" -> \"#{m}\" [label=\" #{prefix}\"];"
-          gr += generate_provenance_edges(visited, m)
+        jobs_with = lambda do |conds|
+          Job.limit(RELATION_LIMIT).where(conds)
+            .results.sort_by { |j| j.finished_at || j.created_at }
         end
-      end
-    end
-    gr
-  end
-
-  def generate_provenance_edges(visited, uuid)
-    gr = ""
-    m = collection_uuid(uuid)
+        @output_of = jobs_with.call(output: @object.portable_data_hash)
+        @log_of = jobs_with.call(log: @object.portable_data_hash)
+        @project_links = Link.limit(RELATION_LIMIT).order("modified_at DESC")
+          .where(head_uuid: @object.uuid, link_class: 'name').results
+        project_hash = Group.where(uuid: @project_links.map(&:tail_uuid)).to_hash
+        @projects = project_hash.values
 
-    if not uuid or uuid.empty? or visited[uuid] or visited[m]
-      return ""
-    end
-
-    #puts "visiting #{uuid}"
-
-    if m  
-      # uuid is a collection
-      uuid = m
-      visited[uuid] = true
-
-      gr += describe_node(uuid)
-
-      Job.where(output: uuid).each do |job|
-        #gr += describe_node(job_uuid(job)) 
-        gr += "\"#{uuid}\" -> \"#{job_uuid(job)}\" [label=\" output\"];"
-        gr += generate_provenance_edges(visited, job.uuid)
-      end
+        @permissions = Link.limit(RELATION_LIMIT).order("modified_at DESC")
+          .where(head_uuid: @object.uuid, link_class: 'permission',
+                 name: 'can_read').results
+        @logs = Log.limit(RELATION_LIMIT).order("created_at DESC")
+          .where(object_uuid: @object.uuid).results
+        @is_persistent = Link.limit(1)
+          .where(head_uuid: @object.uuid, tail_uuid: current_user.uuid,
+                 link_class: 'resources', name: 'wants')
+          .results.any?
+        @search_sharing = search_scopes
 
-      Job.where(log: uuid).each do |job|
-        #gr += describe_node(job_uuid(job))
-        gr += "\"#{uuid}\" -> \"#{job_uuid(job)}\" [label=\" log\"];"
-        gr += generate_provenance_edges(visited, job.uuid)
-      end
-      
-    else
-      visited[uuid] = true
-
-      # uuid is something else
-      rsc = ArvadosBase::resource_class_for_uuid uuid
-
-      if rsc == Job
-        Job.where(uuid: uuid).each do |job|
-          gr += script_param_edges(visited, job, "", job.script_parameters)
-          #gr += describe_script(job)
+        if params["tab_pane"] == "Provenance_graph"
+          @prov_svg = ProvenanceHelper::create_provenance_graph(@object.provenance, "provenance_svg",
+                                                                {:request => request,
+                                                                  :direction => :bottom_up,
+                                                                  :combine_jobs => :script_only}) rescue nil
+        end
+        if params["tab_pane"] == "Used_by"
+          @used_by_svg = ProvenanceHelper::create_provenance_graph(@object.used_by, "used_by_svg",
+                                                                   {:request => request,
+                                                                     :direction => :top_down,
+                                                                     :combine_jobs => :script_only,
+                                                                     :pdata_only => true}) rescue nil
         end
-      else
-        gr += describe_node(uuid)
       end
     end
+    super
+  end
 
-    Link.where(head_uuid: uuid, link_class: "provenance").each do |link|
-      gr += describe_node(link.tail_uuid)
-      gr += "\"#{link.head_uuid}\" -> \"#{link.tail_uuid}\" [label=\" #{link.name}\", href=\"/links/#{link.uuid}\"];"
-      gr += generate_provenance_edges(visited, link.tail_uuid)
+  def sharing_popup
+    @search_sharing = search_scopes
+    respond_to do |format|
+      format.html
+      format.js
     end
+  end
 
-    #puts "finished #{uuid}"
+  helper_method :download_link
 
-    gr
+  def download_link
+    collections_url + "/download/#{@object.uuid}/#{@search_sharing.first.api_token}/"
   end
 
-  def create_provenance_graph(uuid)
-    require 'open3'
-    
-    gr = """strict digraph {
-//rankdir=LR;
-node [fontsize=8,shape=box];
-edge [dir=back,fontsize=8];"""
-
-    visited = {}
-    gr += generate_provenance_edges(visited, uuid)
-
-    gr += "}"
-    svg = ""
-
-    Open3.popen2("dot", "-Tsvg") do |stdin, stdout, wait_thr|
-      stdin.print(gr)
-      stdin.close
-      svg = stdout.read()
-      wait_thr.value
-      stdout.close()
-    end
-
-    svg = svg.sub(/<\?xml.*?\?>/m, "")
-    svg = svg.sub(/<!DOCTYPE.*?>/m, "")
+  def share
+    a = ApiClientAuthorization.create(scopes: sharing_scopes)
+    @search_sharing = search_scopes
+    render 'sharing_popup'
   end
 
-  def show
-    return super if !@object
-    @provenance = []
-    @output2job = {}
-    @output2colorindex = {}
-    @sourcedata = {params[:uuid] => {uuid: params[:uuid]}}
-    @protected = {}
-
-    colorindex = -1
-    any_hope_left = true
-    while any_hope_left
-      any_hope_left = false
-      Job.where(output: @sourcedata.keys).sort_by { |a| a.finished_at || a.created_at }.reverse.each do |job|
-        if !@output2colorindex[job.output]
-          any_hope_left = true
-          @output2colorindex[job.output] = (colorindex += 1) % 10
-          @provenance << {job: job, output: job.output}
-          @sourcedata.delete job.output
-          @output2job[job.output] = job
-          job.dependencies.each do |new_source_data|
-            unless @output2colorindex[new_source_data]
-              @sourcedata[new_source_data] = {uuid: new_source_data}
-            end
-          end
-        end
-      end
+  def unshare
+    @search_sharing = search_scopes
+    @search_sharing.each do |s|
+      s.destroy
     end
+    @search_sharing = search_scopes
+    render 'sharing_popup'
+  end
 
-    Link.where(head_uuid: @sourcedata.keys | @output2job.keys).each do |link|
-      if link.link_class == 'resources' and link.name == 'wants'
-        @protected[link.head_uuid] = true
-      end
-    end
-    Link.where(tail_uuid: @sourcedata.keys).each do |link|
-      if link.link_class == 'data_origin'
-        @sourcedata[link.tail_uuid][:data_origins] ||= []
-        @sourcedata[link.tail_uuid][:data_origins] << [link.name, link.head_kind, link.head_uuid]
+  protected
+
+  def find_usable_token(token_list)
+    # Iterate over every given token to make it the current token and
+    # yield the given block.
+    # If the block succeeds, return the token it used.
+    # Otherwise, render an error response based on the most specific
+    # error we encounter, and return nil.
+    most_specific_error = [401]
+    token_list.each do |api_token|
+      begin
+        using_specific_api_token(api_token) do
+          yield
+          return api_token
+        end
+      rescue ArvadosApiClient::ApiError => error
+        if error.api_status >= most_specific_error.first
+          most_specific_error = [error.api_status, error]
+        end
       end
     end
-    Collection.where(uuid: @sourcedata.keys).each do |collection|
-      if @sourcedata[collection.uuid]
-        @sourcedata[collection.uuid][:collection] = collection
-      end
+    case most_specific_error.shift
+    when 401, 403
+      redirect_to_login
+    when 404
+      render_not_found(*most_specific_error)
     end
+    return nil
+  end
 
-    @prov_svg = create_provenance_graph(@object.uuid)
+  def file_enumerator(opts)
+    FileStreamer.new opts
   end
 
-  protected
   class FileStreamer
+    include ArvadosApiClientHelper
     def initialize(opts={})
       @opts = opts
     end
     def each
       return unless @opts[:uuid] && @opts[:file]
-      env = Hash[ENV].
-        merge({
-                'ARVADOS_API_HOST' =>
-                $arvados_api_client.arvados_v1_base.
-                sub(/\/arvados\/v1/, '').
-                sub(/^https?:\/\//, ''),
-                'ARVADOS_API_TOKEN' =>
-                @opts[:arvados_api_token],
-                'ARVADOS_API_HOST_INSECURE' =>
-                Rails.configuration.arvados_insecure_https ? 'true' : 'false'
-              })
-      IO.popen([env, 'arv-get', "#{@opts[:uuid]}/#{@opts[:file]}"],
-               'rb') do |io|
-        while buf = io.read(2**20)
-          yield buf
+
+      env = Hash[ENV].dup
+
+      require 'uri'
+      u = URI.parse(arvados_api_client.arvados_v1_base)
+      env['ARVADOS_API_HOST'] = "#{u.host}:#{u.port}"
+      env['ARVADOS_API_TOKEN'] = @opts[:arvados_api_token]
+      env['ARVADOS_API_HOST_INSECURE'] = "true" if Rails.configuration.arvados_insecure_https
+
+      bytesleft = @opts[:maxbytes].andand.to_i || 2**16
+      io = IO.popen([env, 'arv-get', "#{@opts[:uuid]}/#{@opts[:file]}"], 'rb')
+      while bytesleft > 0 && (buf = io.read([bytesleft, 2**16].min)) != nil
+        # shrink the bytesleft count, if we were given a maximum byte
+        # count to read
+        if @opts.include? :maxbytes
+          bytesleft = bytesleft - buf.length
         end
+        yield buf
       end
+      io.close
+      # "If ios is opened by IO.popen, close sets $?."
+      # http://www.ruby-doc.org/core-2.1.3/IO.html#method-i-close
       Rails.logger.warn("#{@opts[:uuid]}/#{@opts[:file]}: #{$?}") if $? != 0
     end
   end