Taken from OpenERP wiki page , just in case you like me , may missed it!…

Open ERP objects can be created from `PostgreSQL views. The technique is as follows :

1. Declare your _columns dictionary. All fields must have the flag readonly=True.
2. Specify the parameter _auto=False to the Open ERP object, so no table corresponding to the _columns dictionnary is created automatically.
3. Add a method init(self, cr) that creates a `PostgreSQL View matching the fields declared in _columns.

Example The object report_crm_case_user follows this model.

report_crm_case_user(osv.osv):
      _name = "report.crm.case.user"
     _description = "Cases by user and section"
      _auto = False
      _columns = {
          'name': fields.date('Month', readonly=True),
          'user_id':fields.many2one('res.users', 'User', 

                readonly=True, relate=True),
          'section_id':fields.many2one('crm.case.section', 

               'Section', readonly=True, relate=True),
          'amount_revenue': fields.float('Est.Revenue', readonly=True),
          'amount_costs': fields.float('Est.Cost', readonly=True),
          'amount_revenue_prob': fields.float('Est. Rev*Prob.', readonly=True),
          'nbr': fields.integer('# of Cases', readonly=True),
          'probability': fields.float('Avg. Probability', readonly=True),
          'state': fields.selection(AVAILABLE_STATES, 'State', 

               size=16, readonly=True),
          'delay_close': fields.integer('Delay to close', readonly=True),
      }
      _order = 'name desc, user_id, section_id'
 
      def init(self, cr):
          cr.execute("""
                create or replace view report_crm_case_user as (
                    select
                        min(c.id) as id,
                        substring(c.create_date for 7)||'-01' as name,
                        c.state,
                        c.user_id,
                        c.section_id,
                        count(*) as nbr,
                        sum(planned_revenue) as amount_revenue,
                        sum(planned_cost) as amount_costs,
                        sum(planned_revenue*probability)::decimal(16,2) as 

                              amount_revenue_prob,
                        avg(probability)::decimal(16,2) as probability,
                        to_char(avg(date_closed-c.create_date), 

                              'DD"d" `HH24:MI:SS') as delay_close
                    from
                        crm_case c
                    group by substring(c.create_date for 7), 

                              c.state, c.user_id, c.section_id
           )""")
  report_crm_case_user()