Add static catalog vocabulary to support various widgets with relationsfields or uuid-fields

[pbauer]

No description provided.

[mister-roboto]

@pbauer thanks for creating this Pull Request and help improve Plone!

To ensure that these changes do not break other parts of Plone, the Plone test suite matrix needs to pass.

Whenever you feel that the pull request is ready to be tested, either start all jenkins jobs pull requests by yourself, or simply add a comment in this pull request stating:

@jenkins-plone-org please run jobs

With this simple comment all the jobs will be started automatically.

Happy hacking!

[alecpm]

It's using plone.memoize.view.memoize now. I've also added some tests and a smarter path method for the default title template.

[alecpm]

@jenkins-plone-org please run jobs

[jensens]

Remove the code that is no longer needed. Or did I miss something?

[jensens]

@jenkins-plone-org please run jobs

[ale-rt]

This import should go after import itertools.
Not a big deal, but imports are already sorted and would be nice to keep them sorted

[ale-rt]

If I am not wrong this might break in Python 2, where title is expected to be text. Please consider this before merging

[alecpm]

You're right, we probably need some special handling here for Python 2 where brain.Title is always an encoded string, but Term.title is always unicode. I think that can be fixed with a simple call to safe_unicode though.

[jensens]

This is the last blocker here before we can merge, right?

[ale-rt]

Yeah, wrapping this with safe_unicode seem to be the right thing to do

[pbauer]

@ale-rt @alecpm The change to use plone.memoize.view.memoize breaks this for me. In a edit-form with two different fields that use this feature with different queries I get the same options twice.
When I go back to 121c816 it works again and each field offers the expected options.

Here is the relevant excerpt from my schema:

    basisvariante = RelationChoice(
        title='Basisvariante',
        description=_('Leer lassen, wenn dies eine Basisvariante ist!'),
        vocabulary=StaticCatalogVocabulary(
            {
                'portal_type': ['product'],
                'base_variant': True,
                'path': '/Plone/produkte',
                'sort_on': 'id',
            },
            title_template='{brain.getId} - {brain.Title}',
        ),
        required=False,
    )
    directives.widget(
        'basisvariante',
        SelectFieldWidget,
    )

    sonderausstattung = RelationList(
        title='Sonderausstattung',
        default=[],
        value_type=RelationChoice(
            vocabulary=StaticCatalogVocabulary(
                {
                    'portal_type': 'accessory',
                    'path': '/Plone/accessories',
                    'sort_on': 'id',
                },
                title_template='{brain.getId} - {brain.Title}',
            )
        ),
        required=False,
        missing_value=[],
    )
    directives.widget(
        'sonderausstattung',
        AjaxSelectFieldWidget,
        vocabulary=StaticCatalogVocabulary(
            {
                'portal_type': 'accessory',
                'path': '/Plone/accessories',
                'sort_on': 'id',
            },
            title_template='{brain.getId} - {brain.Title}',
        ),
        pattern_options={
            'closeOnSelect': False,
            'minimumInputLength': 2,
            'ajax': {'quietMillis': 500},
        },
    )

[jensens]

@pbauer, indeed, view.memoize does take context physical path and function arguments into the cache key, both are the same in one request for brains if there are two fields. I would propose to use plone.memoize.ram and cache based on a custom cache key self.query, context.getPhysicalPath() and the catalogCounter (did I miss something?).

[alecpm]

The original implementation using request.memoize used the self.query as part of the cache key, and I think that's probably the simplest solution. The main goal of the caching was to avoid doing the same query multiple times in a request when doing field/widget validation using the __contains__ method. Fields tend to validate themselves a bunch of times, and doing a multiple catalog queries for each stored or input value gets expensive really fast. I think it's probably best if I restore the old caching behavior which was working fine. I'll do that and fix the unicode title issue later today.

[alecpm]

Sorry for the delay all. I've reverted the cache changes, added a call to safe_unicode, and cleaned up the import order. Should be ready to go.

[jensens]

@jenkins-plone-org please run jobs

[jensens]

Seems tests in Python 2.7 are still failing because of some unicode confusion. I would say, after fixing this just go and merge: https://jenkins.plone.org/job/pull-request-5.2/1959/testReport/junit/plone.app.vocabularies/catalog/StaticCatalogVocabulary/

[ale-rt]

Sorry for giving you a wrong hint about using plone.memoize!

[pbauer]

@jenkins-plone-org please run jobs

[pbauer]

@jenkins-plone-org please run jobs