GitLab

Previously, setting a computed relation upon entity creation or
modification (using the ORM or an RQL query) would usually fail with an
operational error in the backend ("no such table"). However, under some
mysterious circumstances (like passing a string as value in cw_set for a
computed relation), the RQL to SQL transformation would simply drop the
clause.

To prevent this to happen, we add a check for computed relation before
adding a relation to an execution plan. This check raises a QueryError.
It happens in several places:

* in querier.InsertPlan.add_relation_def() (called from several places
  in ssplanner steps) for INSERT queries,
* in ssplanner.UpdateStep.execute() for SET queries and,
* in ssplanner.SSplanner.build_delete_plan() for DELETE queries.

Tests added to unittest_querier.py because unittest_sslplanner.py looked
inappropriate (it has only unit tests) and the former already had a
NonRegressionTC class.

This is probably only necessary for QuerierTC itself, move it there and drop
incantation from other derived classes to deactivate this feature.

This used to crash because of an attempt to access unexisting  table.

Closes #17113286

--HG--
branch : 3.25

so it's consistent with CubicWebTC and avoid access to a protected attribute
which has been exposed by the removal of the 'session' property a few csets
earlier.

which should be not necessary anymore, beside two remaining usage in our own
test suite.

This was not necessary anymore with promoted usage of the new data import API.
Turn repository's _type_extid_cache to _type_cache with only the entity's type
as key.

This introduces an backward incompatible change: entity_metas dict doesn't
contains anymore the extid key, but it doesn't seem used at all anywhere, so
this sounds acceptable.

Closes #15538317

Having TestServerConfiguration and alike in the test module namespace causes
pytest discovery errors, because it thinks it's a test class.

* update some copyright
* drop most __docformat__ declaration
* fix some flake8 warnings / errors

Specify module's __file__ instead of apphome. __file__ is saved on the class as
it will be used later (eg to find the proper pg cluster without relying on
shared dictionnary modifications).

This will require some tests update, but it should be less common in cubes than
in cubicweb itself.

Basically:

mkdir cubicweb
hg mv *.py -X setup.py cubicweb
hg mv dataimport devtools entities etwist ext hooks i18n misc schemas server skeleton sobjects test web wsgi cubicweb

Other changes:

* adjust path to cubicweb-ctl in devtools tests
* update setup.py to avoid importing __pkginfo__ (exec it instead),
  replace os.path.walk by os.walk and prepend `modname` here and there
* update tox.ini to account for new test locations
* update doc/conf.py so that it still finds __pkginfo__.py and CWDIR in
  doc/Makefile

It has been deprecated in a previous changeset and is now a no-op.

Related to #2919309.

Connection replaces ClientConnection everywhere.

Some notes:

* testlib: .client_cnx and .repo_cnx become aliases of .cnx (we might
  not want to tell people to update their tests again for just no real
  benefit, so we'll live with these aliases for a while)

* entity.as_rset must not be cached because we risk caching result
  sets having a Connection object as .req (helps unittest_breadcrumbs)

* entity._cw_dont_cache_attributes loses its repo/request special
  paths and only keeps its storage/bfss user (this helps
  unittest_wfobjs)

* moreover, entity.cw_instantiate and .cw_set stop overriding the
  attributes cache *after* the before_*_entity hooks have run, because
  there is no need to (it is now actually harmful to do it and
  unittest_hooks.test_html_tidy* tests remain green because of this)

* rset._build_entity sticks its .req onto the entity just fetched from
  the cache, because otherwise it might carry a _cw that is a
  Connection object where a Request is expected (helps
  unittest_views_actions)

* we get overall better cache usages (entity caches were split over
  Request + ClientConnection and Connection), hence the changes
  unittest_entity and unittest_wfobjs

* void the ecache when providing the cnx to a request object

  Having the entity cache pre-filled when we bind it to the request
  object hurts because these entities are bound to Connection objects,
  that lack e.g. `.form` or `.add_js` and crash the views subsystem.

  Thus, the unittest_testlib.test_error_raised test will are kept
  green.


Closes #3837233

If an attribute has more permissive security rules than the entity
type itself, we should be green and not deny action because of an
early global entity permission check (with the more restrictive
rules).

Only if one attribute with the entity-level permission rules is edited
will the global check be performed.

Note:

* the "if action == 'delete'" check at the entry of
  check_entity_attributes is a guard for a condition currently not
  happening in cubicweb itself (but application hooks could
  conceivably call this function with a 'delete' action)

Closes #3489895.

--HG--
branch : stable

0542a85fe667 replacing SQL OR by hooks for symmetric relations allows that.

This involve a single test value change for a border case: when querying a
symmetric relation without specifying the subject nor the object, you may get
some duplicated result. IMO this is fine to let the user explicitly use DISTINCT
or not and to remove the dedicated handling we had which didn't let any choice.

Related to #3259713

We now provide a more compact indexname, using the schema constraint
entity type and the position of the columns set in the entity type
unique constraints list.

This avoids a nasty name truncation issue.

The UniqueTogetherError object is made smarter: it computes the
rtypes, abstracting the underlying backend (pg/sqlserver vs sqlite).

The `user friendly` adapter is much simplified since there is no
longer any truncation issue.

Uses a new logilab.database version (ticket #151507) and a new yams
version (ticket #189299)

Closes #2514939

[jcr: disable hooks when temporarily dropping CWUniqueTogetherConstraint entities]

Don't override passed-in descriptions in ResultSet.__init__, even if
they're empty.

--HG--
branch : stable

It's more consistent, and avoids pylint warning
"Instance of 'tuple' has no 'append' member"

--HG--
branch : stable

They don't actually need a session, so let's make that clear.

--HG--
branch : stable

when some neged relation is involved. Closes #3058527.

--HG--
branch : stable

Related to #2919300 (almost done!)

since it now handles a connection to the system source only

Related to #2919300

[jcr: adjust 3.17.11 migration, fix a number of bugs in new
 ConnectionsSet implementation, fix
 source.{open,close}_source_connections]

_extract_eid_consts() implementation must consider arithmetic operators
when trying to associate rql variables to specific eids.

--HG--
branch : oldstable

We now:
* have CWComputedRelation in the bootstrap schema to store computed relations
* properly serialize/deserialize it
* test first if the database has been migrated and contains the related table

Related to #3546717
[jcr: adjust unittest_querier to pass with the added entity type]

They are unneeded since we dropped python 2.5 compatibility. This will help
Python 3.3 experimentation.

There is no self.transaction of TestCase. I assumed the test writer meant a
session. This seems to make the test pass.

--HG--
branch : stable

[test] use setUpClass rather than setUpModule to avoid inter-dependency pb with CubicWebTC sub-classes

--HG--
branch : stable

--HG--
branch : stable

[rql annotation] fix has_text_query detection (if coming from sub-query and if has_text on a column aliases. Closes #2275322

--HG--
branch : stable

Administrator should ask their users to reenter new password so they
benefit from the new encryption.

Also, new encryption is cross-platform compatible, eg you may now move an instance
from windows to linux painlessly

--HG--
branch : stable

[rql2sql] None for attributes in kwargs generate IS NULL, so should be considered in sql cache key. Closes #2116693

--HG--
branch : stable