@@contenttree-fetch slows down fetching indexes on brains for objects with large catalog indexes

[fredvd]

We ran into this issue on a site with a folder with 260 items that had a lot of words indexed in the SearchableText catalog. (Custom Dexterity contenttype with a file field containing a PDF that's indexed).

The related items behavior in plone.app.relationfield is using @@contenttree-fetch as well, default setup on Plone 4 with plone.app.contenttypes and trying to browse into the folder mentionned above takes between 30 and 40 seconds to render the listing.

Trying to profile and looking at a thread dump while the request is in progress there was always access to the method isBrainSelectable on PathSource at https://github.com/plone/plone.formwidget.contenttree/blob/master/plone/formwidget/contenttree/source.py#L161-L165 .

   def isBrainSelectable(self, brain):
        if brain is None:
            return False
        index_data = self.catalog.getIndexDataForRID(brain.getRID())
        return self.selectable_filter(brain, index_data)

These last two lines pull in a lot of index_data from the catalog, which takes a l a lot of time, especially when each content item has a lot of words.

Pull request #31 fixes the case when there are no criteria at all in CustomFilter to operate on, which makes fetching the index_data unnecessary in the first place. This fixes the related items behavior in plone.app.relationfield which doesn't use any filtering on the content type.

Plone 5 uses a different querying mechanism for plone.app.widgets related items, so it's not much use fixing this, just documented the behavior if someone else still runs into this.