Memoize vocabularies

[tomgross]

This has a big performance win for huge vocabularies.

[vincentfretin]

why don't you use super(OrderedSelectWidget, self).update() here? I think widget.HTMLSelectWidget.update is not called right now.

[vincentfretin]

Can you add another component to the key (not sure what) so that the cache is invalidated if you change the taxonomy ttw or by import?

[malthe]

Should we try and get this into 1.5?

[tomgross]

I'd love to have it but don't have time at the moment to make this work in a generic way. This is the cache function we use. We invalidate only on changes to keywords.

ONE_DAY = 60 * 60 * 24

def _cache_one_day(fun, self):
    catalog = api.portal.get_tool(name='portal_catalog')
    subjects = len(catalog._catalog.uniqueValuesFor('Subject'))
    key = '{0}{1}'.format(
        subjects,
        time() // ONE_DAY
    )
    return key

If someone wants to step in go ahead.

[malthe]

@petschki – would be nice to get this into 1.5.0 perhaps?

I think using the _p_mtime attribute would make it pretty trivial to implement in a generic way.

[petschki]

I’ll take a look at this tomorrow.

[petschki]

@malthe @vincentfretin well ... this does the trick ... can anybody test this with some huge sample vdex file or mail it to me, so I can test it?

[vincentfretin]

The changes looks good. I don't have huge vocabularies at hand. @andreesg may have that. See #36

[coveralls]

Coverage decreased (-0.1%) to 45.824% when pulling 7655c8e on tomgross-memoize_voc into 923b6ca on master.

[petschki]

I've tested this with a VDEX file with 3 languages and 1500 items each language ... so far no performance problems to me. I think we are ready to merge and release.