Hi everyone,
I have what's probably a loaded question about how Solr parses queries.
Basically, the client wants to restructure the way they handle
permissions. Instead of assigning specific permissions ids (as we've
done up until this point), they want to structure permissions by
adding a particular filter query depending on the products that the
user has purchased. In principle, it could save a lot of reindexing
time because when products change, it simply changes the filter
queries, instead of requiring reindexing of large portions of the
documents in the index.
However, I'm hitting a snag... Basically, here is the way we'd like
to structure the queries:
Documents can have zero or more tags in a particular field assigned to
them. All permissions are essentially tied to the tags in this
field. For each product a user has purchased, there are a certain
number of tags associated with it. Users should be able to see all
documents which match any of these tags, as well as all documents
which don't match *any* tag associated with a product. Basically,
this means that users can see any documents which match their specific
products, as well as any documents which aren't bound to any product
in particular (documents which are essentially open to anyone).
Let's say that the following tags represent all the restricted tags in
the system:
tag1
tag2
tag3
tag4
The current user is allowed access to documents with tags tag1 and
tag2, and documents which have none of these tags.
So, by my own logic, I'd make the query:
((my_field:tag1) OR (my_field:tag2) OR (-my_field:tag1 AND -
my_field:tag2 AND -my_field:tag3 AND -my_field:tag4))
Keep in mind that "my_field" can have other tags beyond tags 1-4 which
essentially have no impact on permissions.
Obviously, I'm writing here because this query doesn't work... at
all. I see according to the wikis that, apparently, - or "NOT"
*excludes* all documents which have the tag in question. So this is
just not the right syntax for what I'm doing.
What I can't seem to find is a syntax which says "return all documents
which don't match this query", or perhaps more importantly, "return
all documents which either match query1 or don't match query 2". I
tried using the "_query_" syntax but no luck there either.
Am I mistaken, or is there simply no way in SOLR to return the
documents which didn't match the specified query?
If not, I fear there's very little potential for optimization
here.... Reindexing all affected documents for a product every time a
product's parameters change is becoming very cumbersome - it can take
hours. We'd really like to find a way to avoid having to do that. If
anyone can help figure out a way to actually do this I'd be very
grateful.
Thanks!
--
Steve
