k thanks gora that what I was thinking but thought I should ask as there might be some solution in solr it self...the last option fine I would go with that only.
thanks rohan On Thu, May 23, 2013 at 9:13 AM, Gora Mohanty <g...@mimirtech.com> wrote: > On 22 May 2013 19:37, Rohan Thakur <rohan.i...@gmail.com> wrote: > [...] > > this inital_boost is basically copy field of autosug but saved using > > different analysers taking whole sentence as single token and generating > > edge ngrams so that what I search on this field only term matching from > > first will match...and for any other infix term match I have autosug > > field.... > > > > so now what I want from this is to show the documents returned with > > initial_boost first and then the documents with autosug field sorted with > > pop field respectively (separately) and return the result... > > > > now from your suggestion I could do this using sort on multiple fields > by > > separating them by > > commas, as described under > > http://wiki.apache.org/solr/CommonQueryParameters#sort > > > > but for that I would require 1 field having value greater(all equal say > 2) > > for initial_boost field and smaller(all same say 1) for autosug field how > > can I do this? or is there some better solution.. > > Um, maybe it would help if you started out by explaining > what you are trying to do instead of changing the requirements > after a solution is proposed. > > The above design seems faulty. If the fields have the same > values, there is no way to sort them differently. Boosts are > not guaranteed to always promote documents in the > desired manner, and will not work with sorting in any case. > You can try to kludge things together in various ways, e.g., > * Add a prefix so that some fields always sort first, e.g., > prefix "aaa..." to a string field to try and ensure that it > sorts first > * Have not tried this, but it should be possible to use a > Solr function query with a function that sorts by the > field name: http://wiki.apache.org/solr/FunctionQuery > > As you seem to be using analysers on the fields, beware > that sorting will not work reliably on multi-valued or tokenised > fields. > > You could also index different fields to different Solr cores, > fetch the results for individual fields with separate queries, > and combine them as desired. This might be the cleanest, > if not the most efficient, solution. > > Regards, > Gora >
