datahen 0.16.0 → 0.18.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 39397d5cb4e60a6d24cdec5bd979f543a23019b7c9b9dffe6140a204d330465c
-  data.tar.gz: 1db7c2b448179c2bc4b56e99428dfb4303cbb1451df032ec43cb5264f58935ec
+  metadata.gz: 0e1fcf7422236924fd818a1527337a6089cd444b1f35510b72fe140facbed7b0
+  data.tar.gz: 05be57d3e058ee9969d210ded0b1d043b388390d5f2ac834ece490691683f39d
 SHA512:
-  metadata.gz: 7058506211d537c8ea3c9a521625fd339b255f41188a70341cc04683ca1abc1fa7f19ed796026b5e07679bb2fd7e57f096d319fbe8a75ae6fb7fd59a704a9824
-  data.tar.gz: b8b60607cd27acbd654afe0816b0b7738871ca61c370fbafa747985d3723fec64f9b07c6078b34c1192db1d7b160682522a4901a7047c3d067860bc2b745b0b0
+  metadata.gz: 4e076509fa8a0fa7fa78406916530bfe2c1b6075ac1007baab43a447911d1c7d8e90bddd8a1438c339a4fadef4e05e629e4907f1f1bae3f4c1f283dba63c25c9
+  data.tar.gz: e7ceb1208c87cd75fa7202f55549c6b2f2ce24980f7642827aab6f721107ca8ddb59829b93742126e465d6930c6c5574de2d045d3a968a7f3a826bf099ee3c4b

data/lib/datahen/cli/scraper.rb

@@ -33,6 +33,8 @@ module Datahen
       option :multiple_jobs, type: :boolean, desc: 'Set true to enable multiple jobs. Default: false'
       option :max_job_count, type: :numeric, desc: 'Set a value to set max number of jobs available. Set -1 for unlimited. Default: 3'
       option :max_page_size, type: :numeric, desc: 'Set a value to set max page size when fetching a page. Set a value grather than 0 to set it as limit, 0 means any size. Default: 0'
+      option :enable_global_cache, type: :boolean, desc: 'Set true to enable page cache. Default: false'
+      option :retry_interval, type: :numeric, desc: 'Set a value to set retry time interval on seconds when refetching a page. Set a value grather than 0 to set it as new time to refetch, 0 means default time. Default: 0'
       def create(scraper_name, git_repository)
         # puts "options #{options}"
         client = Client::Scraper.new(options)
@@ -59,6 +61,8 @@ module Datahen
       option :multiple_jobs, type: :boolean, desc: 'Set true to enable multiple jobs. Default: false'
       option :max_job_count, type: :numeric, desc: 'Set a value to set max number of jobs available. Set -1 for unlimited. Default: 3'
       option :max_page_size, type: :numeric, desc: 'Set a value to set max page size when fetching a page. Set a value grather than 0 to set it as limit, 0 means any size. Default: 0'
+      option :enable_global_cache, type: :boolean, desc: 'Set true to enable page cache. Default: false'
+      option :retry_interval, type: :numeric, desc: 'Set a value to set retry time interval on seconds when refetching a page. Set a value grather than 0 to set it as new time to refetch, 0 means default time. Default: 0'
       def update(scraper_name)
         client = Client::Scraper.new(options)
         puts "#{client.update(scraper_name, options)}"
@@ -97,6 +101,7 @@ module Datahen
       option :proxy_type, desc: 'Set the Proxy type. Default: standard'
       option :vars, type: :string, banner: :JSON, desc: 'Set input vars. Must be in json format. i.e: [{"name":"foo", "value":"bar", "secret":false}] '
       option :max_page_size, type: :numeric, desc: 'Set a value to set max page size when fetching a page. Set a value grather than 0 to set it as limit, 0 means any size. Default: 0'
+      option :retry_interval, type: :numeric, desc: 'Set a value to set retry time interval on seconds when refetching a page. Set a value grather than 0 to set it as new time to refetch, 0 means default time. Default: 0'
       def start(scraper_name)
         client = Client::ScraperJob.new(options)
         puts "Starting a scrape job..."

data/lib/datahen/cli/scraper_job.rb

@@ -105,6 +105,8 @@ module Datahen
       option :profile, type: :string, desc: 'Set the profiles (comma separated) to apply to the job. Default: default'
       option :job, :aliases => :j, type: :numeric, desc: 'Set a specific job ID'
       option :max_page_size, type: :numeric, desc: 'Set a value to set max page size when fetching a page. Set a value grather than 0 to set it as limit, 0 means any size. Default: 0'
+      option :enable_global_cache, type: :boolean, desc: 'Set true to enable page cache. Default: false'
+      option :retry_interval, type: :numeric, desc: 'Set a value to set retry time interval on seconds when refetching a page. Set a value grather than 0 to set it as new time to refetch, 0 means default time. Default: 0'
       def update(scraper_name)
         if options[:job]
           client = Client::Job.new(options)

data/lib/datahen/cli/scraper_page.rb

@@ -46,6 +46,7 @@ module Datahen
       option :ua_type, :aliases => :u, desc: 'Set user agent type. Default: desktop'
       option :no_redirect, :aliases => :n, type: :boolean, desc: 'Set true to not follow redirect. Default: false'
       option :max_size, type: :numeric, desc: 'Set a value to set max page size when fetching a page. Set a value grather than 0 to set it as limit, 0 means any size. Default: 0'
+      option :retry_interval, type: :numeric, desc: 'Set a value to set retry time interval on seconds when refetching a page. Set a value grather than 0 to set it as new time to refetch, 0 means default time. Default: 0'
       def add(scraper_name, url)
         begin
           options[:headers] = JSON.parse(options[:headers]) if options[:headers]
@@ -80,6 +81,7 @@ module Datahen
       option :priority, type: :numeric, desc: 'Set fetch priority. The higher the value, the sooner the page gets fetched. Default: 0'
       option :vars, :aliases => :v, type: :string, desc: 'Set user-defined page variables. Must be in json format. i.e: {"Foo":"bar"}'
       option :max_size, type: :numeric, desc: 'Set a value to set max page size when fetching a page. Set a value grather than 0 to set it as limit, 0 means any size. Default: 0'
+      option :retry_interval, type: :numeric, desc: 'Set a value to set retry time interval on seconds when refetching a page. Set a value grather than 0 to set it as new time to refetch, 0 means default time. Default: 0'
       def update(scraper_name, gid)
         begin
           options[:vars] = JSON.parse(options[:vars]) if options[:vars]

data/lib/datahen/client/job.rb

@@ -22,6 +22,8 @@ module Datahen
         body[:proxy_type] = opts[:proxy_type] if opts[:proxy_type]
         body[:profile] = opts[:profile] if opts[:profile]
         body[:max_page_size] = opts[:max_page_size] if opts[:max_page_size]
+        body[:enable_global_cache] = opts[:enable_global_cache] if opts.has_key?("enable_global_cache") || opts.has_key?(:enable_global_cache)
+        body[:retry_interval] = opts[:retry_interval] if opts[:retry_interval]
         params = @options.merge({body: body.to_json})
 
         self.class.put("/jobs/#{job_id}", params)

data/lib/datahen/client/job_page.rb

@@ -16,6 +16,8 @@ module Datahen
         body[:priority] = opts[:priority] if opts[:priority]
         body[:vars] = opts[:vars] if opts[:vars]
         body[:max_size] = opts[:max_size] if opts[:max_size]
+        body[:enable_global_cache] = opts[:enable_global_cache] if opts.has_key?("enable_global_cache") || opts.has_key?(:enable_global_cache)
+        body[:retry_interval] = opts[:retry_interval] if opts[:retry_interval]
 
         params = @options.merge({body: body.to_json})
 
@@ -38,6 +40,8 @@ module Datahen
         body[:no_redirect] = opts[:no_redirect] if opts[:no_redirect]
         body[:cookie] = opts[:cookie] if opts[:cookie]
         body[:max_size] = opts[:max_size] if opts[:max_size]
+        body[:enable_global_cache] = opts[:enable_global_cache] if opts.has_key?("enable_global_cache") || opts.has_key?(:enable_global_cache)
+        body[:retry_interval] = opts[:retry_interval] if opts[:retry_interval]
 
         params = @options.merge({body: body.to_json})
 

data/lib/datahen/client/scraper.rb

@@ -29,6 +29,8 @@ module Datahen
         body[:multiple_jobs] = opts[:multiple_jobs] if opts[:multiple_jobs]
         body[:max_job_count] = opts[:max_job_count] if opts[:max_job_count]
         body[:max_page_size] = opts[:max_page_size] if opts[:max_page_size]
+        body[:enable_global_cache] = opts[:enable_global_cache] if opts.has_key?("enable_global_cache") || opts.has_key?(:enable_global_cache)
+        body[:retry_interval] = opts[:retry_interval] if opts[:retry_interval]
         params = @options.merge({body: body.to_json})
         self.class.post("/scrapers", params)
       end
@@ -51,6 +53,8 @@ module Datahen
         body[:multiple_jobs] = opts[:multiple_jobs] if opts.has_key?("multiple_jobs") || opts.has_key?(:multiple_jobs)
         body[:max_job_count] = opts[:max_job_count] if opts.has_key?("max_job_count") || opts.has_key?(:max_job_count)
         body[:max_page_size] = opts[:max_page_size] if opts.has_key?("max_page_size") || opts.has_key?(:max_page_size)
+        body[:enable_global_cache] = opts[:enable_global_cache] if opts.has_key?("enable_global_cache") || opts.has_key?(:enable_global_cache)
+        body[:retry_interval] = opts[:retry_interval] if opts[:retry_interval]
         params = @options.merge({body: body.to_json})
 
         self.class.put("/scrapers/#{scraper_name}", params)

data/lib/datahen/client/scraper_job.rb

@@ -12,6 +12,8 @@ module Datahen
         body[:browser_worker_count] = opts[:browsers] if opts[:browsers]
         body[:proxy_type] = opts[:proxy_type] if opts[:proxy_type]
         body[:max_page_size] = opts[:max_page_size] if opts[:max_page_size]
+        body[:enable_global_cache] = opts[:enable_global_cache] if opts.has_key?("enable_global_cache") || opts.has_key?(:enable_global_cache)
+        body[:retry_interval] = opts[:retry_interval] if opts[:retry_interval]
         if opts[:vars]
           if opts[:vars].is_a?(Array)
             body[:vars] = opts[:vars]
@@ -39,6 +41,8 @@ module Datahen
         body[:proxy_type] = opts[:proxy_type] if opts[:proxy_type]
         body[:profile] = opts[:profile] if opts[:profile]
         body[:max_page_size] = opts[:max_page_size] if opts[:max_page_size]
+        body[:enable_global_cache] = opts[:enable_global_cache] if opts.has_key?("enable_global_cache") || opts.has_key?(:enable_global_cache)
+        body[:retry_interval] = opts[:retry_interval] if opts[:retry_interval]
         params = @options.merge({body: body.to_json})
 
         self.class.put("/scrapers/#{scraper_name}/current_job", params)

data/lib/datahen/client/scraper_job_page.rb

@@ -15,6 +15,9 @@ module Datahen
         body[:page_type] = opts[:page_type] if opts[:page_type]
         body[:priority] = opts[:priority] if opts[:priority]
         body[:vars] = opts[:vars] if opts[:vars]
+        body[:max_size] = opts[:max_size] if opts[:max_size]
+        body[:enable_global_cache] = opts[:enable_global_cache] if opts.has_key?("enable_global_cache") || opts.has_key?(:enable_global_cache)
+        body[:retry_interval] = opts[:retry_interval] if opts[:retry_interval]
 
         params = @options.merge({body: body.to_json})
 
@@ -59,6 +62,9 @@ module Datahen
         body[:ua_type] = opts[:ua_type] if opts[:ua_type]
         body[:no_redirect] = opts[:no_redirect] if opts[:no_redirect]
         body[:cookie] = opts[:cookie] if opts[:cookie]
+        body[:max_size] = opts[:max_size] if opts[:max_size]
+        body[:enable_global_cache] = opts[:enable_global_cache] if opts.has_key?("enable_global_cache") || opts.has_key?(:enable_global_cache)
+        body[:retry_interval] = opts[:retry_interval] if opts[:retry_interval]
 
         params = @options.merge({body: body.to_json})
 

data/lib/datahen/scraper/batch_parser.rb

@@ -334,6 +334,7 @@ module Datahen
         Parallel.each(dequeue, in_threads: (worker_count)) do |page|
           parser_file = self.parsers[page['page_type']]
           begin
+            self.repeat_puts("Parsing page with GID #{page['gid']}")
             puts Datahen::Scraper::Parser.exec_parser_by_page(
               parser_file,
               page,
@@ -342,6 +343,7 @@ module Datahen
               nil,
               keep_outputs
             )
+            self.repeat_puts("Finish parsing page with GID #{page['gid']}")
           rescue Parallel::Kill => e
             puts "[Worker #{Parallel.worker_number}]: Someone tried to kill Parallel!!!"
           rescue Parallel::Break => e

data/lib/datahen/scraper/ruby_parser_executor.rb

@@ -8,7 +8,11 @@ module Datahen
       attr_accessor :refetch_self
       # Reparse self page flag.
       # @return [Boollean]
+      # @note It is stronger than #limbo_self flag.
       attr_accessor :reparse_self
+      # Limbo self page flag.
+      # @return [Boollean]
+      attr_accessor :limbo_self
 
       def initialize(options={})
         @filename = options.fetch(:filename) { raise "Filename is required"}
@@ -31,7 +35,8 @@ module Datahen
           :find_output,
           :find_outputs,
           :refetch,
-          :reparse
+          :reparse,
+          :limbo
         ].freeze
       end
 
@@ -104,7 +109,7 @@ module Datahen
 
       def refetch_page gid
         if save
-          Client::ScraperJobPage.new({gid: gid}).refetch_by_job(self.job_id)
+          Client::JobPage.new({gid: gid}).refetch(self.job_id)
           puts "Refetch page #{gid}"
         else
           puts "Would have refetch page #{gid}"
@@ -122,7 +127,7 @@ module Datahen
 
       def reparse_page gid
         if save
-          Client::ScraperJobPage.new({gid: gid}).reparse_by_job(self.job_id)
+          Client::JobPage.new({gid: gid}).reparse(self.job_id)
           puts "Reparse page #{gid}"
         else
           puts "Would have reparse page #{gid}"
@@ -138,6 +143,24 @@ module Datahen
         reparse_page page_gid
       end
 
+      def limbo_page gid
+        if save
+          Client::JobPage.new({gid: gid}).limbo(self.job_id)
+          puts "Limbo page #{gid}"
+        else
+          puts "Would have limbo page #{gid}"
+        end
+      end
+
+      def limbo page_gid
+        raise ArgumentError.new("page_gid needs to be a String.") unless page_gid.is_a?(String)
+        if page_gid == gid
+          self.limbo_self = true
+          raise Error::SafeTerminateError
+        end
+        limbo_page page_gid
+      end
+
       def eval_parser_script(save=false)
         update_parsing_starting_status
 
@@ -148,6 +171,7 @@ module Datahen
           page = init_page_vars(page)
           self.refetch_self = false
           self.reparse_self = false
+          self.limbo_self = false
 
           begin
             context = isolated_binding({
@@ -178,6 +202,8 @@ module Datahen
             refetch_page gid
           elsif reparse_self
             reparse_page gid
+          elsif limbo_self
+            limbo_page gid
           else
             update_parsing_done_status
           end

data/lib/datahen/version.rb

@@ -1,3 +1,3 @@
 module Datahen
-  VERSION = "0.16.0"
+  VERSION = "0.18.0"
 end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: datahen
 version: !ruby/object:Gem::Version
-  version: 0.16.0
+  version: 0.18.0
 platform: ruby
 authors:
 - Parama Danoesubroto
 autorequire:
 bindir: exe
 cert_chain: []
-date: 2021-07-22 00:00:00.000000000 Z
+date: 2021-09-09 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: thor