This existed mostly for livemarks, it's complex and should just die for now.

I think I may be able to remove another rock from FilterResultSet, and that would be another step towards async queries, let's see what I'll end up with.

So, the last vestiges of FilterResultSet() exist just to remove duplicate uris from tags and check if a node can be added to a query with search terms. The former will be fixable changing tags schema, the latter is not a bug, once the former is fixed FilterResultSet can be merged into EvaluateQueryForNode (or become an helper like EvaluateSearchTermsForNode). I wonder if I could subquery tags, so that the external query removes duplicates and completely kill FilterResultSet() now, would be fancy.

Comment on attachment 578053 [details] [diff] [review] patch v1.0 note that this depends on the livemarks changes, so it can't land before.

Comment on attachment 578053 [details] [diff] [review] patch v1.0 Review of attachment 578053 [details] [diff] [review]: ----------------------------------------------------------------- ::: toolkit/components/places/nsNavHistory.cpp @@ +4137,5 @@ > .Str(", h.url, page_title, tags, ") > .Str(nsPrintfCString(17, "0, 0, 0, 0, %d, 0)", > mozIPlacesAutoComplete::MATCH_ANYWHERE_UNMODIFIED).get()); > + // Serching by terms implicitly exclude queries. > + clause.Condition("NOT url BETWEEN 'place:' AND 'place;'"); self-comment: should be h.url

Comment on attachment 578053 [details] [diff] [review] patch v1.0 Review of attachment 578053 [details] [diff] [review]: ----------------------------------------------------------------- looks ok, r=me

fixes a subtle bug with excludeQueries that was failing a test on Try.

Comment on attachment 596000 [details] [diff] [review] patch v1.1 Review of attachment 596000 [details] [diff] [review]: ----------------------------------------------------------------- oops, I forgot I still need a SR here :) This options was useful only to filter out livemark children from queries, now that load-on-demand livemarks exist it is no more useful. The reason we can't just deprecate it is that its removal allows to speed up quite some queries, on the other side the places query strings just ignore unknown options, and this will just become one of those, so compatibility issues should be limited.

I don't expect any interesting breakage from this, btw adding the keyword.

and documentation may have some pointer to this

Docs updated: https://developer.mozilla.org/en/Places_query_URIs#Query_operators https://developer.mozilla.org/en/XPCOM_Interface_Reference/nsINavHistoryQueryOptions#Attributes Mentioned on Firefox 13 for developers.