GitLab

[ui] Implement outofcontext view for proper display of archive unit referenced by SEDAArchiveUnitRefId entities

by handling separatly case where archive unit holds a reference to another
archive unit from the case where it holds a full descriptive content.


This includes:

* a predicate to distinguish the two cases above

* some specific autoform_section / primaryview_section rtags for each case

* a tab for archive unit's sub-archive units (content only)

* custom breadcrumb adapter and incontext view for choice/seq elements that
  should not appear in the ui

* custom form and form renderer, for the same purpose

Only files modified:

* xsd.py
* entities/profile_generation.py
* i18n/{en,fr}.po

Related to #11447483.

Even if user has update permissions, sometimes we do not want to show edit or
delete icons in list views

This changed some csets ago.

Bad copy/paste. No need for calling parent __select__ in regid specific view.

And ``hide seda_archive_unit`` relation in relations section.

Note that the ``add archive unit`` link should not appear if there is already
an archive unit.

Related to #11447483.

Update a docstring that was copied from another one.

and move the XSDM_MAPPING singleton to the xsd module.

Quick & dirty but a bit more readable anyway.

It makes no sense to  specify them at profile creation time.

Generalizing the mecasism introduced for XSD profile generation of FileInfo
element, since we want the same "auto-cardinality" mecanism in this case.

And mark some of them private by prepending an '_'

Mostly because reledit doesn't support multi-valued relations.

introducing special display of its attributes.

Including a new view to display alternatives.

so handle this by a custom vocabulary function.

Deactivating it for all relations.

On our demo where this cube is in a CW_CUBES_PATH directory, I had to add a
__init__.py file in this directory and to put it in PYTHONPATH to avoid a crash
on start, because of the attempt to find __module__ of registrable instances.

This problem hasn't been reported to cubicweb nor lgc because the proper fix
would be to have cubes as standard python packages. Until then we'll live with
those monkey-patches.

algorithm is now directly attached to the binary data object. It's less
intuitive from a data model POV but enhance the UI.

This triggered a bug in profile XSD generation which is also fixed by this
commit.

* introduce a dedicated class (to be styled later)
* print 'no value specified' even for no value elements