improve UTF-8 detection migration refs #6035

* use a single DJ instead of a DJ per applicable attachment
 * don't hold a transaction for any length of time
 * do direct updates, instead of through AR, to avoid 7 extra
   queries per update from before_save callbacks

test plan:
 * run the migration; it should run relatively quickly, and not
   have any errors

Change-Id: I6af6f0ddbcedbb69ad6afb3688d94feb32d0f201
Reviewed-on: https://gerrit.instructure.com/8905
Tested-by: Hudson <hudson@instructure.com>
Reviewed-by: Zach Wily <zach@instructure.com>

app/models/attachment.rb

@@ -127,9 +127,11 @@ class Attachment < ActiveRecord::Base
         end
         iconv.iconv(nil)
       end
-      self.update_attribute(:encoding, 'UTF-8')
+      self.encoding = 'UTF-8'
+      Attachment.update_all({:encoding => 'UTF-8'}, {:id => self.id})
     rescue Iconv::Failure
-      self.update_attribute(:encoding, '')
+      self.encoding = ''
+      Attachment.update_all({:encoding => ''}, {:id => self.id})
       return
     end
   end

db/migrate/20111024163214_detect_attachment_encoding.rb

@@ -0,0 +1,10 @@
+class DetectAttachmentEncoding < ActiveRecord::Migration
+  tag :postdeploy
+
+  def self.up
+    DataFixup::DetectAttachmentEncoding.send_later_if_production(:run)
+  end
+
+  def self.down
+  end
+end

db/migrate/20111024163214_schedule_attachment_encoding_detection.rb

@@ -1,12 +0,0 @@
-class ScheduleAttachmentEncodingDetection < ActiveRecord::Migration
-  def self.up
-    Attachment.transaction do
-      Attachment.find_each(:conditions => "encoding IS NULL AND content_type LIKE '%text%'") do |a|
-        a.send_later_if_production(:infer_encoding)
-      end
-    end
-  end
-
-  def self.down
-  end
-end

lib/data_fixup/detect_attachment_encoding.rb

@@ -0,0 +1,16 @@
+module DataFixup::DetectAttachmentEncoding
+  def self.run
+    begin
+      attachments = Attachment.find(:all, :conditions => "encoding IS NULL AND content_type LIKE '%text%'", :limit => 5000)
+      attachments.each do |a|
+        begin
+          a.infer_encoding
+        rescue
+          # some old attachments may have been cleaned off disk, but not out of the db
+          Rails.logger.warn "Unable to detect encoding for attachment #{a.id}: #{$!}"
+          Attachment.update_all({:encoding => ''}, {:id => a.id})
+        end
+      end
+    end until attachments.empty?
+  end
+end