1. 10 Aug, 2007 2 commits
  2. 12 Jul, 2007 2 commits
  3. 04 Jul, 2007 1 commit
  4. 26 Jun, 2007 1 commit
  5. 21 Apr, 2007 1 commit
  6. 11 Apr, 2007 1 commit
    • Yoshinori Okuji's avatar
      This big change optimizes the scheduling of active objects, · 1c300e5c
      Yoshinori Okuji authored
      and fix some bugs.
      
      The basic idea is to track a dependency graph to find executable
      messages quickly. This makes the activity system far more efficient,
      when you have many inter-dependent messages queued in the tables.
      
      Also, this obsoletes the time shifting in the schedulers,
      as executable messages can be found in a more efficient manner.
      So the activity parameter "at_date" should work expectedly.
      
      Now the API of validate methods in Activities return a
      list of message objects instead of a boolean value. Such
      a list contains messages that are depended upon by a given
      message.
      
      The validate method in Message accepts a new optional
      parameter, check_order_validation, to indicate whether
      order validation should be performed. The default behavior
      has not changed.
      
      getDependentMessageList is added to ActivityTool, Queue
      and Message. This method collects dependent message for
      a given message from all activities.
      
      There are some other subtle changes. Look at the diffs for
      more details.
      
      
      git-svn-id: https://svn.erp5.org/repos/public/erp5/trunk@14039 20353a03-c40f-0410-a6d1-a30d3c3de9de
      1c300e5c
  7. 29 Mar, 2007 2 commits
  8. 28 Mar, 2007 2 commits
  9. 26 Mar, 2007 2 commits
  10. 23 Mar, 2007 1 commit
    • Yoshinori Okuji's avatar
      Some optimization. Here are the details: · 1764ba15
      Yoshinori Okuji authored
      * SQLDict_assignMessage.zsql
      
      Avoid unnecessary assignments. Reduce message size.
      
      * SQLDict_setPriority.zsql
      
      Reduce message size.
      
      * SQLDict_readMessageList.zsql
      
      Remove unnecessary GROUP BY.
      Add uid into ORDER BY so that messages are enumerated
      mostly in the order of insertions, which potentially reduces
      chances that messages will be reordered later.
      
      * SQLDict_readMessage.zsql
      
      Use DISTINCT and GROUP BY, only if necessary.
      Add uid into ORDER BY for the same reason as above.
      
      * SQLDict_createMessageTable.zsql
      
      Specify NOT NULL whenever possible to reduce data size.
      Use SMALLINT instead of INT for processing_node to reduce data size.
      Remove useless indexes to reduce the cost of UPDATE.
      
      
      
      git-svn-id: https://svn.erp5.org/repos/public/erp5/trunk@13582 20353a03-c40f-0410-a6d1-a30d3c3de9de
      1764ba15
  11. 14 Mar, 2007 1 commit
  12. 02 Mar, 2007 1 commit
  13. 15 Sep, 2006 1 commit
  14. 25 Mar, 2006 1 commit
  15. 22 Feb, 2006 4 commits
  16. 10 Feb, 2006 1 commit
  17. 09 Feb, 2006 1 commit
  18. 23 Dec, 2005 1 commit
  19. 28 Nov, 2005 1 commit
  20. 06 Nov, 2005 1 commit
  21. 03 Nov, 2005 1 commit
  22. 28 Oct, 2005 1 commit
  23. 15 Sep, 2005 1 commit
  24. 27 Jun, 2005 1 commit
  25. 26 Apr, 2005 1 commit
  26. 06 Apr, 2005 1 commit
  27. 25 Jan, 2005 2 commits
  28. 13 Dec, 2004 2 commits
  29. 07 Dec, 2004 1 commit
  30. 28 Sep, 2004 1 commit