c2c-oerpscenario team mailing list archive

["Guewen Baconnier @ Camptocamp" <798732@xxxxxxxxxxxxxxxxxx>]

Hello,

Thanks for your explanation, but that's not the result I have with my code.
The only thing I modified in the view is the join between the table to use the ids instead of the dates. It should not modify the sum of attendances.

Also, we use that in our production at camptocamp and the attendances
sum is right.

openerp_prod_camptocamp=# select sheet_id, action, name from hr_attendance where sheet_id = 8077;
 sheet_id |  action  |        name         
----------+----------+---------------------
     8077 | sign_in  | 2011-09-20 07:06:19
     8077 | sign_out | 2011-09-20 12:42:35
     8077 | sign_in  | 2011-09-20 13:35:13
     8077 | sign_out | 2011-09-20 16:40:31
     8077 | sign_in  | 2011-09-21 08:21:10
     8077 | sign_out | 2011-09-21 11:25:59
     8077 | sign_in  | 2011-09-21 13:03:09
     8077 | sign_out | 2011-09-21 17:39:32
     8077 | sign_in  | 2011-09-22 08:19:19
     8077 | sign_out | 2011-09-22 13:02:35
     8077 | sign_in  | 2011-09-22 13:44:10

openerp_prod_camptocamp=# select * from hr_timesheet_sheet_sheet_day where sheet_id = 8077;
   id    |    name    | sheet_id | total_timesheet | total_attendance | total_difference 
---------+------------+----------+-----------------+------------------+------------------
  116977 | 2011-09-20 |     8077 |             8.7 | 8.68333333333333 |            -0.02
  117159 | 2011-09-21 |     8077 |             7.7 | 7.66666666666667 |            -0.03
 -117626 | 2011-09-22 |     8077 |               0 |             6.55 |             6.55

Do as you want, but as said before it would be a pity not to modify the
joins because the view take 1 second to display with the join on ids
against 20 seconds with the joins on dates and user_id. Anyway, when we
can do joins on FK and PK it is always better and it is the purpose of
my modification.

-- 
You received this bug notification because you are a member of C2C
OERPScenario, which is subscribed to OpenERP Project Group.
https://bugs.launchpad.net/bugs/798732

Title:
  hr_timesheet_sheet: Timesheets  big performance issues

Status in OpenERP Addons (modules):
  Fix Committed

Bug description:
  Hello,

  We are experiencing major problems of performance on timesheets.
  OpenERP 6.0.2
  Performance issues occur with the module hr_timesheet_sheet.

  We have ~ lines per tables :
   - hr_timesheet_sheet : 7'500
   - hr_analytic_timesheet : 110'000
   - hr_attendance : 110'000

  Opening the "Timesheets" menu takes a few minutes for the 80 first rows.
  Then open the timesheet form view takes nearly 30 seconds.
  Each action on a timesheet (Sign-in/out, fill a timesheet line, ..) takes nearly 30 seconds again.

  It's clear that the main cause of the problem is : hr_timesheet.sheet.day
  This view computes the indicators for the whole tables and only afterwards we get the rows that we need.

  The second cause of the problem is the fields sheet_id in
  hr.attendance and hr.analytic.timesheet which are function fields.

  The function fields total_attendance_day, total_timesheet_day,
  total_difference_day, total_attendance, total_timesheet,
  total_difference of hr_timesheet_sheet.sheet are based on this view,
  so that's why each action on the timesheet take 30 seconds, because
  the view is computed again and again.

  Here is the actions that I found to bring an acceptable situation:
   - Removed the "By Day" page on the timesheet view, instead I added a link which open a new view (this will still be slow but it can be acceptable as a first solution as you decide to open it...).
   - I replaced the methods to compute the hr_timesheet_sheet.sheet function fields by full python instead of a cr.execute() on the view.
   - There is still a bottleneck, on a search of sheet_id in hr.attendance and hr.analytic.timesheet, it executes each time a sql query. I added store=True to the function field. The only issue I see with that is in the case of a modification of the employee_id of the timesheet, we can add a trigger on this modification, but I don't see how we can search lines to update. What are your thoughts ? (Is it useful to change the employee of a timesheet after its creation, can we block this ?)

  This improves a lot the performances, but there is still a lot of room
  to improve them !

  Maybe we have to replace this function fields / one2many_mod with
  one2many fields ?

  Here is the link to my branch (I propose a merge.)
  https://code.launchpad.net/~gbaconnier-c2c/openobject-addons/6.0-hr_timesheet_sheet-performance-improvements

  Thanks
  Guewen

To manage notifications about this bug go to:
https://bugs.launchpad.net/openobject-addons/+bug/798732/+subscriptions