[DOC] Updated the doc about auto_join: a bit more precise, and updated accordingly to last changes in implementation.

bzr revid: tde@openerp.com-20121205153449-68uel0ne0od3jl9i

doc/06_misc_auto_join.rst

@@ -5,11 +5,11 @@ Perfoming joins in select
 
 .. versionadded:: 7.0
 
-Starting with OpenERP 7.0, an ``_auto_join`` attribute is added on *many2one* and
+Starting with OpenERP 7.0, an ``auto_join`` attribute is added on *many2one* and
 *one2many* fields. The purpose is to allow the automatic generation of joins in
 select queries. This attribute is set to False by default, therefore not changing
-the default behavior of those fields. It is not recommended to use this attribute
+the default behavior. Please note that we consider this feature as still experimental
-unless you understand the limitations of the feature.
+and should be used only if you understand its limitations and targets.
 
 Without ``_auto_join``, the behavior of expression.parse() is the same as before.
 Leafs holding a path beginning with many2one or one2many fields perform a search
@@ -19,42 +19,53 @@ For example, if you have on res.partner a domain like ``[('bank_ids.name',
 performed :
 
 - 1 on res_partner_bank, with domain ``[('name', '=', 'foo')]``, that returns a
-  list of (res.partner.bank) bids
+  list of res.partner.bank ids (bids)
 - 1 on res_partner, with a domain ``['bank_ids', 'in', bids)]``, that returns a
-  list of (res.partner) pids
+  list of res.partner ids (pids)
 - 1 on res_partner, with a domain ``[('id', 'in', pids)]``
 
-When the _auto_join attribute is True,  it will perform a select on res_partner
+When the ``auto_join`` attribute is True on a relational field, the destination
-as well as on res_partner_bank.
+table will be joined to produce only one query.
 
-- the relational table will be accessed through an alias: ``'"res_partner_bank"
+- the relational table is accessed using an alias: ``'"res_partner_bank"
-  as res_partner__bank_ids``
+  as res_partner__bank_ids``. The alias is generated using the relational field
-- the relational table will have a join condition on the main table:
+  name. This allows to have multiple joins with different join conditions on the
+  same table, depending on the domain.
+- there is a join condition between the destination table and the main table:
   ``res_partner__bank_ids."partner_id"=res_partner."id"``
-- the condition will be written on the relational table:
+- the condition is then written on the relational table:
   ``res_partner__bank_ids."name" = 'foo'``
 
-This job is performed in expression.parse(). For leafs containing a path, it
+This manipulation is performed in expression.parse(). It checks leafs that
-checks whether the first item of the path is a *many2one* or *one2many* field
+contain a path, i.e. any domain containing a '.'. It then  checks whether the
-with the ``auto_join`` attribute set. If set, it adds a join query and recursively
+first item of the path is a *many2one* or *one2many* field with the ``auto_join``
-analyzes the remaining of the leaf, going back to the normal behavior when
+attribute set. If set, it adds a join query and recursively analyzes the
-not reaching an ``_auto_join`` field. The sql condition created from the leaf
+remaining of the leaf, using the same behavior. If the remaining path also holds
-will be updated to take into account the table aliases.
+a path with auto_join fields, it will add all tables and add every necessary
+join conditions.
 
-Chaining _auto_join allows to reduce the number of queries performed, and to
+Chaining joins allows to reduce the number of queries performed, and to avoid
-avoid having too long ``('id', 'in', ids)`` replacement leafs in domains.
+having too long equivalent leaf replacement in domains. Indeed, the internal
-However, severe limitations exist on this feature that limits its current use as
+queries produced by this behavior can be very costly, because they were generally
-of version 7.0. **This feature is therefore considered as experimental, and used
+select queries without limit that could lead to huge ('id', 'in', [...])
+leafs to analyze and execute.
+
+Some limitations exist on this feature that limits its current use as of version
+7.0. **This feature is therefore considered as experimental, and used
 to speedup some precise bottlenecks in OpenERP**.
 
 List of known issues and limitations:
 
-- using _auto_join bypasses the business logic; no name search is performed, only
+- using ``auto_join`` bypasses the business logic; no name search is performed,
-  direct matches between ids using join conditions
+  only direct matches between ids using join conditions
-- ir.rules are not taken into account when performing the _auto_join. 
+- ir.rules are not taken into account when analyzing and adding the join
-- support of active_test is not asserted
+  conditions
-- support of translation is not asserted
+
-- support of _auto_join leading to function fields
+List of already-supported corner cases :
+
+- one2many fields having a domain attribute. Static domains as well as dynamic
+  domain are supported
+- auto_join leading to functional searchable fields
 
 Typical use in OpenERP 7.0: