Provide a ++uuid++ traverser

[petri]

Useful for permalinks and URL shortcuts when there are long paths, etc. See discussion at https://community.plone.org/t/uuid-traverser-plone-app-uuid/4819

I wrote a draft traverser; see the uuid_traverser branch of this repository. It works, but only with published content so far.

[jaroel]

This should do the trick

def traverse(self, uuid, remaining):
    catalog = getToolByName(self.context, 'portal_catalog')
    brains = catalog.unrestrictedSearchresults(UID=uuid)
    if not brains:
        raise LocationError(self.context, uuid)
    return aq_inner(brains[0]._unrestrictedGetObject())

[petri]

Should we use NotFound, LocationError (as @jaroel above), or TraversalError? I think they all work, but I wonder if there's a difference that'd merit choosing one over another?

[jaroel]

NotFound is used in publication.
TraversalError is just an alias for LocationError: https://github.com/zopefoundation/zope.traversing/blob/master/src/zope/traversing/interfaces.py#L23

Looks like LocationError is just fine :)

[jaroel]

The interface specifies that we should raise LocationError: https://github.com/zopefoundation/zope.traversing/blob/master/src/zope/traversing/interfaces.py#L34

[jaroel]

@petri FYI: I think this would allow publication of non-published data. That doesn't seem that good :)

[petri]

Regarding the error raised, it seems TraversalError is also provided just for BBB now, so +1 for LocationError.

[petri]

@jaroel to my understanding, traversal (in case of namespace traversers) happens before security checks, does it not? So publishing would still be controlled by permissions.

[petri]

Shoud we "fix" the request so the end result of using our namespace traverser is as close to normal traversal as possible? That might mean changing variables such as request["ACTUAL_URL"], request["PATH_INFO"], request["PARENTS"] and so on. It seems to me that some of these must be changed, otherwise publishing (of even a normal page) can fail.

[jaroel]

@petri I'd guess the publisher takes care of that, just like with ++view++ traversing.
I'd fire up an instance and check if this is true :)

[petri]

@jaroel I did and it looks like the page is not published correctly without some request fixing. Taking a look at the published HTML, at least some of the paths to javascript now have ++uuid++ in them, which results in traversal errors.

[jaroel]

@petri maybe we should just redirect to the target url when publishing? Otherwise we'll get multiple urls for the same content.

[petri]

@jaroel there is already a @@redirect-to-uuidview for that. Fixing the request can mitigate the multiple urls issue, at least to a (perhaps sufficient) degree.

[petri]

I updated the uuid_traverser branch and it now appears to work as intended. There are probably some quirks, still, given the request["PARENTS"] remains 'unfixed'. But page viewing and editing seems to work fine, I see no difference compared to when accessing the page using normal traversal.

[jensens]

if ready, make it a pull request.