[Grok-dev] catalog weirdness in functional tests

[Jan-Wijbrand Kolman]

Hi,


I'm building a simple application,that uses the catalog. In grok
there's now basically two ways for setting up a catalog within an
application:

1) use a grok.local_utility directive, like so:

  def setup_catalog(catalog):
      catalog['first_name'] = FieldIndex('first_name', IPerson)
      catalog['last_name'] = FieldIndex('last_name', IPerson)

  class Application(grok.Application, grok.Container):
      grok.local_utility(IntIds, provides=IIntIds)
      grok.local_utility(Catalog, provides=ICatalog, setup=setup_catalog)
    ...

2) The "newer" way, using grok.Indexes:

  class Application(grok.Application, grok.Container):
      ...

  class PatientIndexes(grok.Indexes):
      grok.context(IPerson) # that what gets indexed
      grok.site(Application)
      first_name = index.Text()
      last_name = index.Text()

Both work fine when actually running Zope. But something weird happens
when I try to test my application in a functional test. The relevant
test code looks like this:

  >>> from userforms.model import Application, Person
  >>> getRootFolder()['app'] = app = Application()
  >>> from zope.app.component.hooks import setSite
  >>> setSite(app)
  >>> from zope.app.catalog.interfaces import ICatalog
  >>> from zope.component import getUtility
  >>> catalog = getUtility(ICatalog)

In case the catalog is setup like in 1), this test passes. If I have a
catalog setup like in 2), the test fails with:

  ComponentLookupError: (
  <InterfaceClass zope.app.catalog.interfaces.ICatalog>, '')

Well, uhm, no clue here... Maybe anyone has an idea? (what I found
slightly worying, but maybe this is completely unrelated to this
issue, is that the ftests in grok that test the catalog by way of
grok.local_utility does not need any clean up, whereas the ftests for
grok.Indexes *do* need all kinds of clean up...)


kind regards,
jw
-- 
Jan-Wijbrand Kolman