This allows onchange to recompute fields that depend on the inherited
field that has changed. The use-case we have is a model that inherits
from 'account.move'. On its form view, setting the (inherited) journal
field does not recompute the (inherited) company field.
When the field 'journal_id' is modified, the onchange should:
- assign the new value to move_id.journal_id (*)
- recompute move_id.company_id (related='journal_id.company_id')
- recompute company_id
This commit adds the missing part (*).
X-original-commit: 2e05d0a361aa06a34eecf64613821f2e92295298
Manually adding a line should not trigger the computation of the field,
just like modifying the line fields that occur in the field's domain.
In other words, the dependencies of the field should be limited to the
ones declared on field or its compute method.
closesodoo/odoo#71049
X-original-commit: 1c39814bd729cc08882f1545c7d88b0592e63f9a
Signed-off-by: Raphael Collet (rco) <rco@openerp.com>
The attributes field.depends and field.depends_context are problematic
for sharing fields across registries, because they depend on the model's
registry class, which may vary from one registry to another. In order
to make computed fields shareable, we have to move those values away
from fields.
For the same reason, field.recursive should not be inferred, because its
value may depend on the registry, although it is generally not the case.
Moreover, the flag recursive=True is set on a field when field triggers
are determined (on the registry). A compute method may be called before
the flag is set (if no update has been done yet), and that can lead to
incorrect computations.
This happened in test TestUsers2.test_reified_groups in module 'base'.
The user groups view was apparently determined without the flag being
set, and the view depends on the recursive field 'trans_implied_ids',
which was not correctly computed.
We thus force developers to be explicit about recursive computed fields.
The code now logs a warning when the flag is not set up properly.
With this commit, 2 changes are made to the way that the
fields.Selection.ondelete cleanup action works:
1) We manually delete ir.model.fields.selection **before** the deletion
of ir.model.fields purposefully to **avoid** SQL CASCADE deletes, as we
do not know which selections are deleted in cascade and thus we cannot
perform the corresponding ondelete cleanup action (which is implemented
within the ORM).
2) In some actions, namely 'set default' and 'set null', we write a
"safe" value to the records containing the Selection being deleted,
before this commit this would go through the ORM (records.write()) but
this is problematic if there's a write override for the record's model
that raises an error for the field being written to. With this commit,
we first try to go through the ORM but if there's a failure (because of
the raise in a write override) then we will bypass the ORM and set it
with SQL.
opw-2451126
closesodoo/odoo#67616
X-original-commit: f5c7e861ee3ca0cdeb6c2f06d227ada07f923ed0
Signed-off-by: Adrian Torres (adt) <adt@odoo.com>
Consider a constraint method with two fields: a normal field, and a
field with an inverse method. When calling write() with both fields,
the constraint method should be invoked once.
closesodoo/odoo#64433
X-original-commit: 25760348859f594abb796b5bc7b28f63e298cd53
Signed-off-by: Raphael Collet (rco) <rco@openerp.com>
Co-authored-by: Thibault Delavallée <tde@odoo.com>
This takes a former patch one step further.
Assume you recompute a field on two records, and the computation assigns
the first record but not the second one. Before this patch, the
recomputation process was accessing the field on both records. When
accessing the first record, both records are recomputed, and the value
of the first one is found in cache. When accessing the second record,
no recomputation is made (it has been done already), and the cache is
empty. The field is thus fetched from database, and this crashes because
of insufficient access rights.
This scenario was causing some obscure nondeterministic crashes in
tests. The nondeterminism comes from the choice of the environment for
flushing (it should not be in superuser mode); the order of records in
the set to compute; the order in which records are assigned in the
compute method; the fact that the compute method should assign some
records and some not.
The fix consists in adding a method that processes pending computations,
without doing anything special for unassigned records. The method is
used in field accessors (__get__ and mapped) and in method recompute().
The access error is gone because the recomputation no longer tries to
get the value of a field to compute.
closesodoo/odoo#64264
X-original-commit: b7676ec33172e6196b5fd1eda4df730ebf90eb6b
Signed-off-by: Raphael Collet (rco) <rco@openerp.com>
Fixes 3a6ac95 that ignores computation of recursive stored fields during
creation of records.
The problem happens when we have:
* a recursive stored computed field
* an `api.depends` of this field which is a relation to another record
When this "another record" is created and should recompute the recursive
field, it is ignored.
closesodoo/odoo#64117
X-original-commit: 9fe4fe404b73e9ab18ab6b3b913f075c64bb6954
Signed-off-by: Raphael Collet (rco) <rco@openerp.com>
Co-authored-by: Raphael Collet (rco) <rco@openerp.com>
This fixes an issue that occurs when marking fields to compute in the
following situation:
- it occurs before the actual modification,
- a relational field is marked to compute,
- the same field is traversed to inverse some dependency.
When the marking occurs after the modification, the traversal of the
field should actually recompute the field. However, if the marking
occurs before the modification, the inversion of dependencies must be
based on the current value of the field. This is the case with method
unlink(): we call method modified() to mark the fields that currently
depend on the records to be deleted, and the fields must be computed
only after the deletion!
X-original-commit: cd19c2d93db8911fe42802640ea32e5d87aadeec
When performing the onchange for the first time on a new record, if a
field has a default value, it could be two times in the `todo` field
list to compute.
closesodoo/odoo#64005
X-original-commit: cf2b0a505c117a50a91fa022bc1d6bf1eb6d9f77
Signed-off-by: Raphael Collet (rco) <rco@openerp.com>
When a form view is opened, the first call to `onchange` should always
compute fields, even if their dependencies have no default value. Make
sure it is the case for main records, and for records inside one2many
fields. The latter case was actually not working as expected.
closesodoo/odoo#63646
X-original-commit: 45422d56bce413b8577f1784e10dd22ede93c751
Signed-off-by: Raphael Collet (rco) <rco@openerp.com>
With this commit, a new ORM api decorator is introduced:
`api.ondelete(*, at_uninstall)`.
This decorator is to be applied to Model methods that check for specific
business conditions when attempting to unlink a record via the
interface.
E.g. trying to unlink a validated journal entry
This decorator allows this logic to exist outside the `BaseModel.unlink`
method and is automatically bypassed when in uninstall mode, this means
that during an uninstall any and all data related to a module can and
will be removed easily and cleanly while still being able to apply
business logic to manual deletion of records.
This feature opens the gates to solving a very big problem with
uninstalls: records and tables that remain in a database despite the
relevant module being uninstalled, because if an override to unlink
raises an error, the data will never be deleted from the database.
Henceforth, overrides of unlink shall solely be used for data-cleaning
purposes, i.e. deletion or modification of data that is related to the
one currently being deleted but cannot be automatically deleted because
there are no proper SQL relations.
In certain very specific, low-level scenarios an unlink may be
overridden to raise an error, but this should only be done if you know
what the fuck you're doing, most of the time you'll want to resort to
`@api.ondelete`.
Note that this new decorator includes a keyword-only, required argument
called `at_uninstall`, in most business cases this argument shall be
False as this argument dictates whether or not this method should be
executed in the `unlink` call during uninstall. It should only be set to
True if you are certain of all the implications which most likely means
that the records of the model in which this ondelete function is defined
will NOT be removed during uninstall, they will forever linger in the DB
until manual intervention, this in turn can mean a wide range of
undefined problems due to crap left on the database.
Following commits will replace any `unlink` overrides that raise
business errors by methods decorated with `api.ondelete`, another commit
will introduce a pylint checker that will raise a warning any time that
an unlink override raises an error.
This fixes the issue introduced by revision fd50ba9fb8
The way computed fields are invalidated depends on the order of the dict
passed as a parameter to method onchange(). This causes some unexpected
behavior when dealing with computed editable fields: the user loses the
value in that field because of the onchange.
Explanation: during the onchange, the modified field is cached by
record._update_cache(changed_values, validate=True)
If the field is a one2many, 'value' contains an update command for each
modified line. Because of 'validate=True', the assignment triggers
field computations on the lines, which are already handled by another
call to onchange(). In case anything on the parent model modifies some
line, the whole one2many values are returned to the user, and
accidentally recomputed fields show up in the interface.
closesodoo/odoo#62748
Signed-off-by: Raphael Collet (rco) <rco@openerp.com>
Consider a one2many field with a corresponding many2one field that is
computed. After modifying the many2one field's dependencies, the cached
value of the one2many field is inconsistent until the many2one field has
been recomputed. Therefore, one has to force the computation of the
many2one field before accessing the cached value of the one2many field.
closesodoo/odoo#59034
X-original-commit: b9201ebdd65eaabab9257cf97c58410681783fa6
Signed-off-by: Xavier Dollé (xdo) <xdo@odoo.com>
Signed-off-by: Raphael Collet (rco) <rco@openerp.com>
This is necessary when a field's dependencies are given by a callable.
This functionality was broken with the support for an explicit parameter
`depends` in the fields' definition, because the implementation could
not distinguish between the field parameter `depends` and the `depends`
evaluated from compute functions.
The fix consists in storing the field parameter `depends` into
`field._depends`, and the result of the setup into `field.depends`.
closesodoo/odoo#58144
X-original-commit: aebe9a76c9b35e84eda1f46bec3ef6da152b1539
Signed-off-by: Raphael Collet (rco) <rco@openerp.com>
Assume field B depends on A, and searching on B is rewritten as a domain
that mentions A. If A is modified, and then we search on B, one has to
flush A to the database before searching.
closesodoo/odoo#57963
X-original-commit: b242b7b1382ee39ebefa58c556931b7dae77f669
Signed-off-by: Raphael Collet (rco) <rco@openerp.com>
Non-stored readonly computed fields must be assigned by compute method.
Make the error message more explicit.
closesodoo/odoo#56269
Signed-off-by: Raphael Collet (rco) <rco@openerp.com>
Consider models `M` and `L`, with a one2many field `M.l_ids` with
inverse `L.m_id`. Consider also a computed field `L.foo` that depends
on `L.m_id`.
The field `L.foo` is not computed correctly by `onchange` on a new line
in the form view of `M`: its value is forced to `False` because it has
no default, and is not recomputed by `onchange` since the field `L.m_id`
is never triggered an `onchange`.
Modify the method `onchange` to not force fields without a default to
`False`, and return all fields.
Purpose
=======
Since the 13.0, a regression has been introduced in hr_timesheet.
Considering:
- The account.analytic.line model, representing a timesheet entry
- The project.task model, representing a task, on which we have the fields:
- timesheet_ids: One2many
- planned_hours: Computed (sudo), stored, representing the sum of the timesheet amounts
- An ir.rule restricting the timesheets CRUD to his own only.
A user can only see its own timesheets on a task, but the field "Planned Hours",
which is stored-compute_sudo, should take all the timesheet lines into account
However, when adding a new line and then recomputing the value, no existing line
from another user is binded on self, then the value is erased and saved on the
database.
Specification
=============
This commit introduces a test that illustrates that bad usage of a shared cache.
A correct way to fix this could be to avoid using the cache if an ir.rule is pointing
to the related model. That would force the recomputation in a real sudoed environment.
closesodoo/odoo#55408
Taskid: 2285924
X-original-commit: 50e229fe66b2a82717d4260287e18a875bbd26f5
Signed-off-by: Yannick Tivisse (yti) <yti@odoo.com>
When search is implemented on a given field, the domain returned by
`field.search()` must be normalized, otherwise its processing just
crashes (because of missing logic operators).
closesodoo/odoo#51927
X-original-commit: c83974ac5d797d38760a46c97468a74214218575
Signed-off-by: Xavier Dollé (xdo) <xdo@odoo.com>
Signed-off-by: Raphael Collet (rco) <rco@openerp.com>
If the original field is defined with a selection list, overrides with
method or methods names fall back on the list.
closesodoo/odoo#51232
X-original-commit: 272ba8d2238d1d1b87f47b114b55e6eed25314a8
Signed-off-by: Raphael Collet (rco) <rco@openerp.com>
When setting a many2one field, the corresponding one2many fields are
updated in cache. When such a one2many field has a domain, the
evaluation of the domain may require to fetch some fields from the
database. If the `towrite` cache has not been updated yet, the many2one
field is overridden by the database value.
Fix by setting the `towrite` cache before updating inverse fields.
closesodoo/odoo#50788
X-original-commit: c1ebfe05a66cfebc7ced36e25db5f6ff4bfb2d46
Signed-off-by: Raphael Collet (rco) <rco@openerp.com>
Consider a sales order with a single line. Edit the sales order: remove
the line, and add another line with the same subtotal. When saving, the
total of the sales order is 0.0.
Here is the explanation: the form view performs a `write` on the sales
order, and modifies the lines with a command `2` (remove line) and a
command `0` (create line).
After deleting the first order line, the cache is emptied, and a call to
`flush()` forces the recomputation of the total. The value is computed
to be 0, and assigned to the field. The assignment converts the value
for the cache, without prefetching the currency field (optimization),
and puts 0.0 in cache. The assignment then converts the value for the
database, which prefetches most fields on the sales order: the cache is
now inconsistent and contains the old value V, while the database is
then updated with 0.0.
After creating the new order line, the total is once again recomputed.
Its value is V, and because the cache also contains that value, no
update is performed to the database, which remains at 0.0!
The fix consists in avoiding the prefetching of fields when accessing
the currency field to round a monetary value.
opw-2223134
closesodoo/odoo#49741
X-original-commit: 048ea2f20a0a2fa6d629cbe262cbbe765248d36f
Signed-off-by: Raphael Collet (rco) <rco@openerp.com>
- Let A and B be two different modules.
- A defines a required Selection field F of model M and B extends
it through the `selection_add` argument.
- Create records of model M and have some of them have any of the
options introduced by module B selected for field F.
- Uninstall module B.
The result will be records of Model M with an option for field F that no
longer exists, this makes the registry inconsistent and prone to
crashing (it is sufficient to access the form view of such a record to
trigger a crash).
This commit introduces a mechanism similar to the `ondelete` argument
found in Many2one fields, the argument name is the same but it's
different both in behaviour and in implementation.
The `ondelete` mechanism for Selection fields is enforced for **any**
Selection field with required set to `True`, this means that the
developer is required to set a cleanup behaviour for when their module
is uninstalled. For possible cleanup options, see fields.Selection's
docstring.
As far as implementation goes, everything is implemented in Python
unlike with Many2one fields where the behaviour is delegated to
PostgreSQL.
The `ondelete` setting will be processed during
`ir.model.fields.selection.unlink()` to ensure that the registry is left
in an appropriate state after module uninstall.
Purpose of this commit is to let computed stored editable fields being
copied if their field class allows it.
Indeed the value of those fields is computed based on some triggers but
can also be updated manually by users. When copying a record, it makes
sense to consider that this value is what the user expects and allow its
copy, if the original field allows it. Either it was computed, and
copied value will be correct without having to call computation again
(well, provided all dependencies have been copied, too), or it was
updated and the copied value will be the one the user entered.
Without this fix, an edited field is not copied, and will be recomputed,
which may look like an inconsistent value.
Task ID 2209163
closesodoo/odoo#48383
X-original-commit: 4b274d3b4101fbae154a572cdf40d23838899773
Signed-off-by: Olivier Dony (odo) <odo@openerp.com>
Signed-off-by: Raphael Collet (rco) <rco@openerp.com>
Co-authored-by: Raphael Collet <rco@odoo.com>
Context:
Since 7593b887df an entry of ir.model.fields.selection is created
for each entry in a selection field. This allow a more modular
approach of tagging entries with an external id and translating
selections in the module that defined the selection (and not the one
that defined the field)
Problem:
When removing a selection, every record that used this selection is
updated with a 'set null'.
This was also intented on abstract models that do not have a table.
To reproduce:
1. create two models:
class ModelA(models.Abstract):
_name = 'abstract.a'
select = fields.Selection([
('foo', 'Foo'),
('bar', 'Bar'),
])
class ModelB(models.Model):
_name = 'model.b'
_inherit = 'abstract.a'
2. update code and removes the key 'foo'
3. update the module
-> error
psycopg2.errors.UndefinedTable: relation "abstract_a" does not exist
LINE 1: UPDATE "abstract_a" SET "select"=NULL WHERE "re...
When updating the values in db, only update on real models
closesodoo/odoo#48288
X-original-commit: 6b20a8241300ebfc185500eb64e210db755574a8
Signed-off-by: Martin Trigaux (mat) <mat@odoo.com>
Signed-off-by: Christophe Monniez (moc) <moc@odoo.com>
The use case is a cache miss of a stored field with compute on a new
record with origin. The field should be computed only when explicitly
triggered, i.e., when a dependency has been modified. Otherwise it
should be fetched from the origin record.
closesodoo/odoo#47353
Signed-off-by: Raphael Collet (rco) <rco@openerp.com>
Consider an x2many field `foo_ids` with `context={'active_test': False}`
in its definition, and a comodel with an active field. The value of the
field includes inactive records.
Now consider a search with a domain like `[('foo_ids.bar', op, value)]`.
The search should return all the records with corecords that satisfy the
domain `[('bar', op, value)]`, including inactive corecords, because the
field's context explicitly disables filtering on the active field.
closesodoo/odoo#43625
Signed-off-by: Raphael Collet (rco) <rco@openerp.com>
Purpose
=======
If check company is set on a field and if the user has no access
right to the field value (example: address_id on an expense sheet
could be set to a private res.partner, on an onchange method when
setting the employee), then the check_company mechanism will
raise an AccessError when trying the validate the companies on
the different records.
Specification
=============
As we only wish to validate the new record values and not the
access rights, the validation could be done as a superuser to
avoid unecessary errors.
closesodoo/odoo#43240
Taskid: 2170006
X-original-commit: 37a9b6c63dcbc268013fbe1a890cf9390eb8e223
Signed-off-by: Yannick Tivisse (yti) <yti@odoo.com>
Consider an x2many field with `context={'active_test': False}`. The
value of that field should always contain inactive records, as the
field's own context overrides the context of the current record.
closesodoo/odoo#42824
X-original-commit: a2fc37adc179fb8bfc11251137334f0a43c58135
Signed-off-by: Raphael Collet (rco) <rco@openerp.com>
Assume F and G are computed by the same method on a missing record R.
During recomputation of F on R, the compute method is called but fails
because R is missing. Both fields are re-marked to compute (because
computation failed), then F is discarded (because R is missing). Then
comes G's turn: G is accessed on R and the computation fails. Both
fields are re-marked to compute (because computation failed), then G is
discarded (because R is missing). Now F is marked again to compute: the
process never ends.
To avoid this situation, discard all fields to recompute on missing
records.
closesodoo/odoo#42234
X-original-commit: 78bf4dbaa1c4adfa1dac68a4d79b87800003ac05
Signed-off-by: Raphael Collet (rco) <rco@openerp.com>
Before this commit, when modifying a Many2one in a form view to add a
new record, it will create the new record, and calculate the inverse
and write it in all the records of the list. This will call the write
method in all records even if they weren't changed.
Now, the inverse is set to be modified only if it's different from the
current value.
opw-2091842
closesodoo/odoo#40258
X-original-commit: 417cee7f0fdb7ed7eb424178efb656b967fa0a5e
Signed-off-by: Jorge Pinna Puissant (jpp) <jpp@odoo.com>
Co-authored-by: Raphaël Collet <rco@odoo.com>
Raphael Collet